Fuels and Fuel Additives; Waiver Decision/Circuit Court Remand

ENVIRONMENTAL PROTECTION AGENCY

[FRL-5027-1]

Fuels and Fuel Additives; Waiver Decision/Circuit Court Remand

AGENCY: Environmental Protection Agency (EPA).

ACTION: Notice

-----------------------------------------------------------------------

SUMMARY: On July 12, 1991, under section 211(f)(4) of the Clean Air Act

(Act), the Ethyl Corporation (Ethyl) requested a waiver to permit the

sale of its gasoline additive, methylcyclopentadienyl manganese

tricarbonyl (MMT), an octane enhancer commercially labeled by Ethyl as

HiTEC 3000, for use in unleaded gasoline. The Administrator of EPA

denied Ethyl's application for a waiver on January 8, 1992, based

primarily on concerns regarding the potential for increases in

hydrocarbon emissions resulting from MMT use. Ethyl subsequently sought

judicial review of that decision in the U.S. Court of Appeals for the

District of Columbia Circuit. Based on new emissions data developed and

submitted to EPA by Ethyl, EPA requested that the Court of Appeals

remand Ethyl's application to the Agency for further action.

On November 30, 1993, the Administrator of EPA found that Ethyl had

met its burden to demonstrate under section 211(f)(4) that approval of

its remanded application would not cause or contribute to a failure to

meet emission standards. Ethyl agreed to resubmit its application at

that time, thereby affording further time for the Agency to consider

the issue of potential health effects associated with use of MMT in

unleaded gasoline. Ethyl and EPA later agreed to further extend the

deadline for final action on Ethyl's application to July 13, 1994. The

Agency is today denying Ethyl's request for a waiver for HiTEC 3000

based on unresolved concerns regarding the potential impact of

manganese emissions resulting from MMT use on public health.

ADDRESSES: Copies of the information relative to this application are

available for inspection in public docket A-93-26, A-91-46 and A-90-16

at the Air Docket (LE-131) of the EPA, Room M-1500, 401 M Street, S.W.,

Washington, D.C. 20460, (202) 260-7548, between the hours of 8:30 a.m.

to noon and 1:30 p.m. to 3:30 p.m. weekdays. As provided in 40 CFR Part

2, a reasonable fee may be charged for copying services.

FOR FURTHER INFORMATION CONTACT: Joseph R. Sopata, Chemist, or James W.

Caldwell, Chief, Fuels Section, Field Operations Support Division

(6406J), U.S. Environmental Protection Agency, 401 M Street, S.W.,

Washington, D.C. 20460, (202) 260-2635.

SUPPLEMENTARY INFORMATION:

Index

I. Background

II. Statutory Framework

A. History of Statute

B. Two Stage Process

C. Consideration of Potential Health Effects

III. Method of Review

A. ``Causes or Contributes'' to Emission Standard Failure

B. Discretionary Review

IV. Analysis of Emissions Data

A. Description of Previous Test Programs

B. Comments on Vehicle Emissions Issues

C. Available Data Meet Previously Utilized Criteria

D. Data on Newer-Technology Vehicles Meet More Stringent

Criteria

E. Finding

V. The Onboard Diagnostics Issue

VI. Manganese Health Assessment

A. Introduction

B. Health Effects Assessment

1. Background

2. Earlier Assessments

3. 1993 Revised RfC

4. Alternative Approaches to Deriving RfCs

a. Conventional NOAEL- or LOAEL-Based Approach

b. NOSTASOT Approach

c. Benchmark Analyses

d. Bayesian Analyses

e. Summary of RfC Estimates

C. Exposure Assessment

1. Background

2. Additional Canadian Studies

3. The PTEAM Study

4. Estimated Mn Exposure Levels Associated with MMT

D. Risk Characterization

E. References

F. Comments on Health Assessment and EPA Response

VII. Fuel and Fuel Additive Registration and Research Needs

VIII. Other Issues

IX. Decision

I. Background

Section 211(f)(1)(A) of the Act makes it unlawful, effective March

31, 1977, for any manufacturer of a fuel or fuel additive to first

introduce into commerce, or to increase the concentration in use of,

any fuel or fuel additive for use in light-duty motor vehicles

manufactured after model year 1974 which is not substantially similar

to any fuel or fuel additive utilized in the certification of any model

year 1975, or subsequent model year, vehicle or engine under section

206 of the Act. An interpretive rule defining the term ``substantially

similar'' under section 211(f)(1)(A) was promulgated for unleaded

gasoline at 46 FR 38582 (July 28, 1981), and revised at 56 FR 5352

(February 11, 1991). Section 211(f)(1)(B) of the Act makes it unlawful,

effective November 15, 1990, for any manufacturer of a fuel or fuel

additive to first introduce into commerce, or to increase the

concentration in use of, any fuel or fuel additive for use by any

person in motor vehicles manufactured after model-year 1974 which is

not substantially similar to any fuel or fuel additive utilized in the

certification of any model year 1975, or subsequent model year, vehicle

or engine under section 206 of the Act. Thus, section 211(f)(1)(B)

expands the prohibitions of 211(f)(1)(A), which apply only to light-

duty vehicles.

Section 211(f)(4) of the Act provides that upon application by any

fuel or fuel additive manufacturer, the Administrator of EPA may waive

the prohibitions of section 211(f)(1) if the Administrator determines

that the applicant has established that such fuel or fuel additive will

not cause or contribute to a failure of any emission control device or

system (over the useful life of any vehicle in which such device or

system is used) to achieve compliance by the vehicle with the emissions

standards to which it has been certified pursuant to section 206 of the

Act. If the Administrator does not act to grant or deny a waiver within

180 days of receipt of the application, the statute provides that the

waiver shall be treated as granted. The subject of this notice is an

application by Ethyl under section 211(f)(4) of the Act for a waiver

for the fuel additive methylcyclopentadienyl manganese tricarbonyl

(MMT), commercially labeled by Ethyl as HiTEC 3000, to be blended in

unleaded gasoline resulting in a level of 0.03125 (1/32) gram per

gallon manganese (gpg Mn).

This Agency action is a reconsideration of Ethyl's fourth

application for a waiver for MMT. Ethyl's first application was

submitted on March 17, 1978 for concentrations of MMT resulting in 1/16

and 1/32 gpg Mn in unleaded gasoline. Ethyl's second application was

submitted on May 26, 1981 for concentrations of MMT resulting in 1/64

gpg Mn in unleaded gasoline. The Administrator denied these requests

for waivers due to concerns regarding increases in exhaust hydrocarbon

emissions resulting from MMT use. The decisions and justifications

thereof may be found in the September 18, 1978 Federal Register, 43 FR

41424, and the December 1, 1981 Federal Register, 46 FR 58630. Ethyl's

third application was submitted on May 9, 1990, for concentrations of

MMT resulting in a level of 0.3125 (1/32) gpg Mn in unleaded gasoline

(the same levels which are requested in the application which is the

subject of today's notice). Ethyl withdrew its third application on

November 1, 1990, before the deadline for the Administrator to make a

determination on the application. Because no determination had been

made at the time Ethyl withdrew that application, EPA accepted the

withdrawal and immediately terminated the proceeding without action on

the application.

Ethyl's fourth application was submitted on July 12, 1991. This

application was, from a practical standpoint, an extension of the third

application, the entire record of which was incorporated by Ethyl into

the current proceeding. On January 8, 1992, the Administrator of EPA

denied Ethyl's fourth application for a waiver (57 FR 2535, January 22,

1992). The application was denied based in part upon data submitted by

Ford Motor Company (Ford) which indicated that, for the model groups

tested by Ford and, for the conditions under which Ford tested its

vehicles, the increases in hydrocarbon exhaust emissions as a result of

the use of MMT were substantially greater than those observed in the

Ethyl test program. The Agency stated in its decision that a likely

factor which might account for the differences observed between the

Ethyl and Ford test programs was the severity of the driving cycle.

However, the Agency also concluded that other factors might be

responsible for the observed differences. In the denial decision, the

Agency stated that it had always accepted data from test programs which

``model'' the fleet in support of waiver applications, but that if an

interested party were to present data indicating that a potentially

significant subset of the fleet, not tested by the applicant, was

especially susceptible to the negative effects of the additive, the

Agency could reasonably require specific testing on representative

models of that sub-fleet.

In its decision, the Agency also stated that it believes it is

reasonable to consider the effect of a fuel on vehicles' ability to

meet future emissions standards. (The ``Tier I'' tailpipe standards

prescribed by section 202(g) of the Act began to take effect in model

year 1994, which began approximately in September 1993.\1\) Therefore,

regarding the Ford data mentioned above, the Agency stated in its

decision that the concerns raised by that data related to both current

and future standards.

---------------------------------------------------------------------------

\1\56 FR 25724-25790 (June 5, 1991).

---------------------------------------------------------------------------

Although not the basis of the 1992 denial, another important issue

arose during the consideration of Ethyl's third and fourth

applications. The Agency, as well as several commentors, expressed

concerns regarding the possible adverse health effects of an increase

in airborne manganese resulting from MMT use. These concerns were

centered around: (1) The known severe neurotoxic effects of high-level

exposure to manganese through inhalation, (2) the lack of data

regarding the chronic effects of low-level inhalation exposure to

manganese in humans, and (3) the lack of knowledge regarding potential

exposures due to MMT use. It was repeatedly pointed out by commenters

that neurotoxic damage could occur prior to the onset of overt

symptoms.

In those proceedings, Ethyl also submitted comments regarding

manganese emissions. Ethyl indicated that the manganese emissions

resulting from the use of MMT in unleaded gasoline would be so small as

to not materially affect human exposure to airborne manganese. In

support of its view, Ethyl submitted analyses and data on exposure

modeling and monitoring in both its 1990 and 1991 applications (and in

subsequent submissions associated with the remand discussed below).

(The issue of manganese emissions and public health is discussed in

more detail in Section VI of this document.)

During EPA's consideration of the 1990 Ethyl submission, EPA's

Office of Research and Development (ORD) conducted a manganese

inhalation risk assessment based on the available data which found that

because of ``the considerable uncertainties and data gaps in the

available information * * * it is not possible * * * to conclude

definitively that the increased use of MMT as a fuel additive will (or

will not) increase public health risk.''\2\ (EPA also investigated

potential hazards associated with water contamination resulting from

accidental spills or leakages of pure MMT and concluded that spills or

leaks, if they occurred, are likely to be contained and therefore would

not pose a human health risk due to groundwater contamination. However,

data available to EPA are insufficient to determine whether spills and

leaks could affect exposure to benthic organisms.)

---------------------------------------------------------------------------

\2\See ``Comments on the Use of Methylcyclopentadienyl Manganese

Tricarbonyl in Unleaded Gasoline'', Docket A-90-16.

---------------------------------------------------------------------------

Additionally, in order to obtain assistance in describing

information needed to improve its manganese health risk assessment (and

also to improve its environmental hazard identification of issues

associated with MMT itself), EPA, in conjunction with the National

Institute of Environmental Health Sciences, conducted a Manganese/MMT

Symposium and Workshop on March 12-15, 1991. The conference allowed the

Agency to solicit scientific information from invited extramural

scientists reflecting a wide range of scientific disciplines. Invited

participants included representatives of Ethyl Corporation, the

Environmental Defense Fund, the Centers for Disease Control, the U.S.

Food and Drug Administration and Environment Canada. A summary of the

workshop discussions was provided to each participant and the

information obtained from this meeting was also used by EPA to prepare

a report on prioritized research needed for improving its manganese

inhalation risk assessment.\3\

---------------------------------------------------------------------------

\3\Preuss, P.W. (1991) ORD Document on Information Needed to

Improve the Risk Characterization of Manganese Tetraoxide

(Mn3O4) and Methylcyclopentadienyl Manganese Tricarbonyl,

December 12, 1991 (memorandum to Richard Wilson). Washington, DC:

U.S. Environmental Protection Agency, Office of Research and

Development; December 16, 1991. For further information the reader

is referred to Air Docket A-93-26, II-A-16.

---------------------------------------------------------------------------

EPA raised the issue of potential health effects associated with

manganese exposure as a concern in its January 1992 denial, but did not

base its decision on this concern because the Agency concluded that the

uncertainties regarding hydrocarbon emissions increases prevented EPA

from making the requisite ``cause or contribute'' determination

concerning effects on regulated emissions.

On February 13, 1992, Ethyl filed a petition for review of the

January 8, 1992 waiver denial decision in the United States Court of

Appeals for the District of Columbia Circuit. EPA and Ethyl

subsequently entered discussions concerning a possible settlement of

the court case. In the context of those discussions, Ethyl submitted to

the Agency new emissions test data developed by Ethyl since the denial

decision.

Based on its inspection and analysis of the new Ethyl data, EPA

tentatively concluded that the data indicated that driving cycle did

not contribute significantly to MMT-induced increases in HC emissions.

(EPA's preliminary analysis was placed in docket A-92-41.) However, in

addition to addressing the issue of driving cycle, the Ethyl data

appeared to confirm the finding by Ford that 1991 Escorts experienced a

much higher MMT-induced HC increase than that observed in other models

tested (either in Ethyl's new program or in the original Ethyl test

program). The Agency remained concerned that these data might indicate

that certain engine and emissions control system configurations are

more vulnerable to a MMT-induced emissions increase irrespective of

driving cycle.

To facilitate further settlement discussions with Ethyl, EPA

decided to attempt to formulate an emission testing program intended to

address in a timely manner specific unresolved issues concerning the

effect of MMT on emissions: (1) whether other vehicles utilizing fuels

containing MMT are likely to experience increases in hydrocarbon

emissions similar to those observed in 1991 Ford Escorts; and (2)

whether fuels containing MMT have significant adverse effects on

emissions from vehicles utilizing the technologies most likely to be

employed to meet future standards. On October 28, 1992, EPA held a

public workshop to assist the Agency in its attempt to formulate such

an emission testing program (57 FR 44740, September 29, 1992). In

particular, EPA hoped to obtain information and assistance from

technical experts outside of the Agency concerning the test program

and, in view of the significance of any future waiver decision

concerning MMT for the auto industry and the general public, EPA was

interested in obtaining comments concerning a decisional framework

designed to address and resolve these issues. A proposed emission test

program developed by the Agency and presented at the public workshop,

was effectively adopted by Ethyl as its most recent vehicle emissions

test program involving the 1993 model fleet.

Although further settlement discussions between Ethyl and EPA were

held subsequent to the public workshop, the parties were not successful

in reaching a settlement. However, despite the failure of the parties

to reach agreement, EPA concluded that the Administrator's denial

decision should be reconsidered in light of the new emissions data

generated by Ethyl subsequent to the decision. Accordingly, EPA

requested that the United States Court of Appeals for the District of

Columbia remand the denial decision to EPA for reconsideration.

On April 6, 1993, the Court of Appeals issued a decision granting

the Agency's motion and remanding the case to the Agency to redetermine

within 180 days whether to grant or deny Ethyl's application. The

mandate implementing this judgement was transmitted to the Agency on

June 3, 1993. Pursuant to the court's remand decision, the Agency

published a notice indicating the commencement of a comment period (58

FR 35950, July 2, 1993). The Administrator's final decision on remand

was due within 180 days after the transmittal of the court's mandate,

or by November 30, 1993.

After the Court of Appeals granted the Agency's motion to remand

the denial decision concerning Ethyl's July 12, 1991 application, Ethyl

submitted to EPA a substantial amount of additional data on emission

testing with fuels containing MMT. (Specific aspects of these data are

discussed below in Section IV of this document).

During the course of the remand of Ethyl's waiver application, the

EPA Office of Research and Development (ORD) reviewed the available

data concerning the health effects associated with inhalation of

manganese as part of a process to revise the reference concentration

(RfC) for inhaled manganese.\4\ An inhalation reference concentration

is defined as an estimate (with uncertainty spanning perhaps an order

of magnitude) of a continuous inhalation exposure to the human

population (including sensitive subgroups) that is likely to be without

appreciable risk of deleterious non-cancer health effects during a

lifetime. The methodology for establishing an RfC accounts for

uncertainties and gaps in the health data base through the assignment

of uncertainty factors. In November, 1993, ORD completed preparation

and review of, and EPA released to Ethyl, a document identifying and

describing the rationale for a new inhalation RfC of 0.05 ug/m\3\ for

manganese and manganese compounds.

---------------------------------------------------------------------------

\4\In 1990, an inhalation reference concentration (RfC) for

manganese of 0.4 ug/m\3\ was verified and placed on IRIS. The

original RfC for manganese figured into a 1990 risk assessment of

MMT prepared by the EPA Office of Research and Development (ORD).

Subsequently, in light of new information submitted by Ethyl and new

results from more recently published studies concerning manganese

inhalation health effects in workers, EPA reexamined the RfC for

manganese and revised it to a value of 0.05 ug/m\3\ in 1993. This

revised RfC for manganese was made available to Ethyl and placed on

IRIS in November 1993.

---------------------------------------------------------------------------

Ethyl subsequently provided to EPA a detailed critique of the

approach utilized to derive the revised manganese RfC. Among other

things, Ethyl argued that EPA used an inappropriate procedure to derive

the RfC from a study of occupational manganese exposures by Roels, et

al. (1992). Ethyl also argued that use of MMT would not result in

significant changes in background manganese exposures, and that the

favorable effects on public health resulting from changes in the

composition of gasoline when MMT is utilized would outweigh any

potential for adverse health effects. (Copies of documents describing

the revised RfC and of the Ethyl comments are available in the public

docket.)

As the deadline of November 30, 1993, for final action by EPA on

Ethyl's waiver application approached, EPA concluded that the extensive

data base on the emission effects of MMT assembled by Ethyl and others

during the consideration of the application was sufficient to permit a

decision concerning whether Ethyl had satisfied the statutory

requirement to show that use of MMT will not cause or contribute to

exceedence of emission standards. However, there had been insufficient

opportunity for public comment concerning the use of a revised

manganese inhalation RfC in assessing any risks that might be posed by

granting Ethyl's application. Ethyl argued that it had not been

afforded an adequate opportunity to study the derivation of the RfC and

to comment on its implications for Ethyl's application. While EPA

scientists did not necessarily agree with the specific technical

arguments concerning the revised RfC and other issues pertaining to

health effects made by Ethyl, EPA concluded that it might be useful to

review the revised RfC in light of further analyses of the available

data as well as the underlying data from occupational studies of

inhaled manganese if such data could be readily obtained. EPA also

concluded that it would be desirable in any case to have further

dialogue with Ethyl and other interested parties on issues related to

the health effects of manganese before EPA was to make a final decision

concerning Ethyl's waiver application.

As a result of these factors, Ethyl and EPA entered into

discussions concerning a possible extension of the deadline for a

decision. Ultimately, an agreement between Ethyl and EPA concerning

such an extension was implemented on November 30, 1993, and notice of

the agreement was published in the Federal Register on December 9, 1993

(58 FR 64761). The agreement provided for an extension of 180 days in

the deadline for final action by EPA on Ethyl's waiver application for

HiTEC 3000.\5\ EPA was thus required to take final action either

granting or denying Ethyl's resubmitted application by May 29, 1994.

For purposes of the resubmitted application, the EPA Administrator

determined that Ethyl had demonstrated, as required by section

211(f)(4), that use of HiTEC 3000 at the specified concentration will

not cause or contribute to a failure of any emission control device or

system (over the useful life of any vehicle in which such device or

system is used) to achieve compliance by the vehicle with the emission

standards with respect to which it has been certified.\6\ The Agency

stated clearly in the December 9, 1993 Federal Register notice that

this determination would not preclude any subsequent regulatory action

based on emission effects under Clean Air Act section 211(c) or any

other provision of the Clean Air Act in the event that the resubmitted

Ethyl waiver application were to be granted in the future. The Agency

also made it clear that this determination would not apply in the

context of any other new waiver application concerning HiTEC 3000 or

MMT which might be submitted in the future if EPA were to deny Ethyl's

resubmitted waiver application on other grounds. Further review of

Ethyl's application during this 180 day period focused in particular on

the issues relating to the potential health effects on public health if

EPA were to permit use of MMT as a fuel additive.

---------------------------------------------------------------------------

\5\To implement this agreement, Ethyl withdrew its July 12, 1991

waiver application, as remanded by the Court of Appeals, and

immediately resubmitted the application.

\6\As is explained in section IV of this document, this decision

was based primarily upon application of the previously used

statistical tests to the submitted emissions data. As is also

explained in section IV, the Agency believes that these tests may be

outdated and is considering a formal change in its method of

analysis of such data.

---------------------------------------------------------------------------

Ethyl and EPA both desired and intended to assure continuity

between the proceedings concerning the July 12, 1991 waiver

application, as remanded to EPA by the Court of Appeals, and Ethyl's

resubmitted waiver application. The entire administrative record

compiled by EPA in support of the original denial decision, as well as

all submissions to the public docket concerning the remanded

application, was incorporated in the record this final decision on the

resubmitted application. The docket number for the resubmitted waiver

application also remained the same.

The additional 180 days that were provided by Ethyl's agreement to

resubmit the waiver application were utilized by EPA to evaluate

remaining issues that may have been relevant to today's decision. In

particular, EPA continued to examine the effects on public health that

might be associated with approval of Ethyl's application. EPA

considered any additional underlying data concerning studies of

occupational manganese exposure that were obtained by or submitted to

EPA, as well as any additional data or information pertaining to the

health effects of manganese submitted by Ethyl or other interested

persons during the comment period. Any additional information that was

submitted was also considered in exploring alternative candidate RfC

estimates and their relationship to the verified revised RfC. EPA also

used the additional time provided by the extension to make a decision

on how the RfC should be utilized in assessing health effects that may

be associated with MMT use, evaluate potential exposure to manganese

compounds associated with MMT use, complete a risk assessment

concerning Ethyl's application, and decide what additional data, if

any, should be provided by Ethyl either before or after MMT is

introduced into the market.

On April 28, 1994, EPA provided Ethyl Corporation with a draft of

the revised risk assessment, which updated the 1991 ORD assessment and

incorporated further analyses performed during the 180-day extension

period. Subsequent to providing Ethyl with this draft, Ethyl provided

EPA with comments on the draft and some additional new data on ambient

manganese concentrations in several Canadian cities. In order to allow

the Agency time to consider this new data, Ethyl requested, and the

Agency agreed to, an extension of the decision deadline until July 13,

1994. An agreement implementing this extension was executed by EPA and

Ethyl counsel on May 24, 1994.

II. Statutory Framework

A. History of Statute

Congress first added section 211(f) to the Clean Air Act in 1977

based primarily on concerns that fuels or additives might damage

vehicle emission control devices. Thus, the original statute focused on

vehicles designed to use unleaded gasoline, prohibiting the general use

in fuels of materials not ``substantially similar'' to fuels used to

certify vehicles to emissions standards. Section 211(f) also provided

that the Administrator of EPA ``may waive the prohibitions * * * if he

determines that the applicant has established that such fuel or fuel

additive * * * will not cause or contribute to a failure of any

emission control device or system * * * to achieve compliance by the

vehicle with the emission standards with respect to which it has been

certified pursuant to section 206.''\7\ Additionally, the statute

provides that if the Administrator does not act to grant or deny the

waiver request within 180 days of receipt of the application, the

waiver request shall be treated as granted.

---------------------------------------------------------------------------

\7\Section 206 of the Act sets forth the certification

requirements with which vehicle manufacturers must comply in order

to introduce into commerce new model year motor vehicles. Under

Sec. 202 of the Act, standards for hydrocarbon (HC), carbon monoxide

(CO), and oxides of nitrogen (NOx) emissions for gasoline, gaseous

fuel, diesel and methanol-powered motor vehicles have been

established. For gasoline, gaseous fuel and diesel-powered motor

vehicles, standards have also been established for particulate

emissions.

---------------------------------------------------------------------------

Section 211(f) was initially interpreted by the Agency as applying

only to unleaded gasoline. In the 1990 Amendments, section 211(f)(1)

was broadly expanded to cover all other fuels and fuel additives,

including leaded gasoline, diesel fuel, and consumer additives.\8\ The

1990 Amendments also apply the provisions of this subsection to

vehicles other than lightduty vehicles. Section 211(f)(1)(B) of the Act

makes it unlawful, effective November 15, 1990, for any manufacturer of

a fuel or fuel additive to first introduce into commerce, or to

increase the concentration in use of, any fuel or fuel additive for use

by any person in motor vehicles manufactured after model year 1974

which is not substantially similar to any fuel or fuel additive

utilized in the certification of any model year 1975, or subsequent

model year, vehicle or engine under section 206 of the Act. Thus,

section 211(f)(1)(B) expands to all motor vehicles the fuel

prohibitions of the original section 211(f)(1) (now redesignated as

section 211(f)(1)(A)), which apply only to light-duty vehicles.\9\

---------------------------------------------------------------------------

\8\H.R. Rep. No. 490, Part 1, 101st Cong., 2d Sess. 313 (1990).

\9\An interpretive rule defining the term ``substantially

similar'' under section 211(f)(1)(A) was promulgated for unleaded

gasoline at 46 FR 38582 (July 28, 1981), and revised at 56 FR 5352

(February 11, 1991). An advance notice of proposed rulemaking

(ANPRM) has been published to begin the proces of promulgating an

interpretive rule to define the term ``substantially similar'' under

Sec. 211(f)(1)(B) for diesel fuel and diesel fuel additives. See 56

FR 24362 (May 30, 1991).

---------------------------------------------------------------------------

In adding section 211, the first focus of Congress was to prevent

the introduction of new additives which may prove harmful to emission

control devices but to allow for the introduction of such additives if

it could be demonstrated that they would not harm emission control

devices. Furthermore, in framing the statute such that the

Administrator was not required to grant a waiver, Congress provided

authority to the Administrator to take into account other

considerations associated with introduction of the new material into

commerce.

B. Two Stage Process

Section 211(f)(4) of the Act provides the legal authority for this

waiver decision.\10\ The Agency interprets section 211(f)(4) of the Act

as establishing a two stage process for the decision to grant or deny a

waiver application. The first stage of the process focuses solely on

whether a waiver applicant has met its burden to demonstrate that a

fuel does not cause or contribute to a failure to meet emission

standards. The second stage of the process reflects the discretionary

authority provided to the Agency by the statute.

---------------------------------------------------------------------------

\10\Section 211(f)(4) states that ``The Administrator, upon

application of any manufacturer of any fuel or fuel additive, may

waive the prohibitions established under paragraph (1) or (3) of

this subsection, or the limitation specified in paragraph (2) of

this subsection, if he determines that the applicant has established

that such fuel or fuel additive or a specified concentration

thereof, and the emission products of such fuel or additive or

specified concentration thereof, will not cause or contribute to a

failure of any emission control device or system (over the useful

life of any vehicle in which such device or system is used) to

achieve compliance by the vehicle with the emission standards with

respect to which it has been certified pursuant to section 206. If

the Administrator has not acted to grant or deny an application

under this paragraph within one hundred and eighty days of receipt

of such application, the waiver authorized by this paragraph shall

be treated as granted.''

---------------------------------------------------------------------------

In the first stage of the waiver process, the sole issue is whether

a fuel ``causes or contributes'' to an emission standard failure. The

waiver applicant bears the burden of demonstrating that a fuel will

neither cause nor contribute to an emission standard failure for any

regulated pollutant. Balancing of the emission effects of a fuel for

one pollutant against those for other pollutant(s) is not permissible

under the statutory language. For example, an applicant would not meet

its burden of proof if its testing of a fuel shows that it causes or

contributes to an emission standard failure for CO, even though testing

shows decreases in emissions of HC and NOx. If an applicant does not

meet its burden of demonstrating that the ``cause or contribute'' test

is met, the Agency cannot grant a waiver. If an applicant does meet its

burden, the Agency may then exercise its discretion to grant or to deny

a waiver in the second stage of the process.\11\

---------------------------------------------------------------------------

\11\Under the statute, if the Agency does not take action to

grant or deny a waiver application within 180 days of submittal, the

waiver is deemed granted.

---------------------------------------------------------------------------

The statute provides that the Agency ``may'' grant a waiver

application if the ``cause or contribute'' test is met, but does not

require such an action. The Agency may therefore choose not to grant a

waiver based on other issues (e.g., public health effects) that

indicate that it would not be in the public interest to do so. In this

second stage of the process, the Agency has a great deal of discretion

to determine which issues should be examined and to balance the

potential positive and negative impacts of a waiver. Such discretionary

authority is grounded not only in Congress' use of the term ``may''

rather than the term ``shall'' in section 211(f)(4), but also in the

goals and purposes of section 211 when read as a whole. The Agency does

not believe that Congress intended to require EPA to grant a waiver

under section 211(f)(4) when available information indicates that the

fuel would be potentially subject to regulatory control under section

211(c)(1) immediately upon issuance of the waiver. Similarly, EPA

believes that Congress did not intend to preclude a determination of

whether issuance of a waiver is consistent with other important goals

of the Act once it has been demonstrated that the mandatory ``cause or

contribute'' test has been met.

This does not mean that the Administrator has unfettered discretion

to deny a waiver application for any reason. The grounds for any denial

must not be arbitrary or capricious or constitute an abuse of

discretion. Thus, in using discretion to deny an application, the

Administrator must identify and explain the factors on which a

discretionary denial decision is based and must assure that the policy

adopted is consistent for all similarly situated waiver applicants.

C. Consideration of Potential Health Effects

Although the basis for a discretionary denial must be rational and

non-arbitrary, nothing in the statute limits the type of factors which

the Administrator may consider in deciding whether to deny an

application. Section 101(b)(1) states that one of the purposes of the

Act is to ``protect and enhance the quality of the Nation's air

resources so as to promote the public health and welfare and the

productive capacity of its population.'' Given this general goal of the

Act, certainly the potential effects on public health of vehicle

emissions would be a factor which the Administrator may reasonably

consider when utilizing the discretion which section 211(f)(4)

authorizes.

Furthermore, under sections 211(b)(2) and 211(e), the Administrator

must require the manufacturer of a fuel or additive to produce data

concerning potential health effects as a condition of, or a

prerequisite to, registration of the fuel or additive.\12\ Under

section 211(c)(1), the Administrator may, based on data collected under

sections 211(b) and 211(e) or otherwise available, issue regulations

controlling manufacture or sale of any fuel or fuel additive which the

Administrator finds ``may reasonably be anticipated to endanger the

public health or welfare.'' These provisions indicate that Congress

intended that the Administrator be concerned about the potential health

effects of fuels and fuel additives.

---------------------------------------------------------------------------

\12\Sections 211(a) and 211(b)(1) require the registration of

fuels and additives designated by the Administrator as a

precondition to introduction into commerce.

---------------------------------------------------------------------------

The fact that the Administrator may control fuels or fuel additives

which pose potential health effects under section 211(c)(1) does not

mean that the Administrator may not consider health effects as a factor

in deciding whether to grant a waiver under section 211(f)(4). Such a

construction of the statute would lead to absurd results, precluding

the Administrator from denying a waiver application and leading to

potential introduction of a fuel or additive into commerce, even in the

specific circumstances where the Administrator has concluded that there

are grounds for issuance of a proposed regulation prohibiting the fuel

or additive under section 211(c). However, although this reasoning

indicates that Congress could not have reasonably intended to

completely preclude the consideration of health effects under section

211(f)(4), this does not mean that section 211(c) limits the

circumstances in which the Administrator may consider potential health

effects as part of a waiver decision. Clearly, it was the intention of

Congress to treat fuels and fuel additives already registered and being

sold for a particular purpose differently than those which have not

already been introduced into commerce.

III. Method of Review

A. ``Causes or Contributes'' to Emission Standard Failure

Under section 211(f)(4) of the Act, twenty-three applications for

waivers of the section 211(f)(1) prohibitions have been received. Of

these, twenty-two applications have sought a waiver for additives for

unleaded gasoline. One, the most recent, sought a waiver of the section

211(f)(1)(B) prohibitions for an additive to diesel fuel.\13\ Of these

twenty-three applications, ten applications have been granted (some

with conditions attached), ten have been denied, and three were

withdrawn by the applicant prior to the Agency's decision.

---------------------------------------------------------------------------

\13\57 FR 45790 (October 5, 1992).

---------------------------------------------------------------------------

Section 211(f)(4) clearly places upon the waiver applicant the

burden of establishing that its fuel will not cause or contribute to

the failure of any vehicle to meet emission standards. Absent a

sufficient showing, the Administrator cannot make the required

determination and cannot grant the waiver. If interpreted literally,

however, this burden of proof imposed by the Act would be virtually

impossible for an applicant to meet, as it requires the proof of a

negative proposition: that no vehicle will fail to meet emission

standards to which it has been certified. Such a literal interpretation

could be construed as requiring the testing of every vehicle.

Recognizing that Congress contemplated a workable waiver provision, EPA

has previously indicated that reliable statistical sampling and fleet

testing protocols may be used to demonstrate that a fuel under

consideration would not cause or contribute to a significant failure to

meet emission standards by vehicles in the national fleet.\14\

---------------------------------------------------------------------------

\14\See Waiver Decision on Tertiary Butyl Alcohol (``TBA''), 44

FR 10530 (February 2, 1979).

---------------------------------------------------------------------------

To determine whether a waiver applicant has established that the

proposed fuel will not cause or contribute to vehicles failing emission

standards, EPA reviews all the material in the public docket, including

the data submitted with the application and public comments on the

application, and analyzes the data to ascertain the fuel's emission

effects. The analysis concentrates on four major areas of concern--

exhaust emissions, evaporative emissions, materials' compatibility, and

driveability--and evaluates the data under statistical methods

appropriate to the various types of emission effects. Emission data are

analyzed according to the effects that a fuel is predicted to have on

emissions over time. If the fuel is predicted to have only an

instantaneous effect on emissions (that is, the emission effects of the

fuel are immediate and remain constant throughout the life of the

vehicle when operating on the waiver fuel), then ``back-to-back''

emissions testing will suffice.\15\

---------------------------------------------------------------------------

\15\Back-to-back emission testing involves testing a vehicle on

a base fuel (i.e., a gasoline which meets specifications for

certification fuel or is representative of a typically available

commercial gasoline), then testing that same vehicle on the fuel for

which the waiver is requested. The difference in emission levels is

attributed to the waiver fuel.

---------------------------------------------------------------------------

Unlike materials traditionally allowed in unleaded gasoline,

metallics, such as MMT, produce non-gaseous combustion products, some

of which may be deposited in the parts of the vehicle that come in

contact with the combustion products of the burned fuel. These areas of

the vehicle include the combustion chamber, the catalyst, the oxygen

sensor, and all parts of the exhaust system.\16\ Since these materials

build up over time,\17\ it has been traditionally accepted that the

emissions effects of such additives occur over time as miles are

accumulated, and that the method of deposition suggests that the

effects are permanent. If the fuel is predicted to have such a long-

term deteriorative effect, durability testing over the useful life of

the vehicle,\18\ in addition to back-to-back testing, is

appropriate.\19\

---------------------------------------------------------------------------

\16\Automakers and catalyst manufacturers point out that, since

catalysts are designed with a honeycomb structure in order to

maximize contact between engine combustion gases and catalyst

materials, if channels within the honeycomb become blocked, the

catalyst is less able to break down the exhaust gases. Furthermore,

although the mechanisms associated with manganese deposits have not

been completely described, catalyst manufacturers suggest that the

mere disposition of manganese (without blockage of channels) would

hinder the catalytic activity of the catalyst. Ethyl, however,

believes that the manganese deposition on the catalyst does not

hinder its activity.

\17\Reply Comments of Ethyl Corporation in Support of the HiTEC

3000 Waiver Application, August 10, 1990, 28.

\18\The ``useful life'' of a 1993 or earlier model year light-

duty vehicle (LDV) (i.e., the amount of time or mileage accumulation

through which the LDV must meet the standards to which it has been

certified) is 50,000 miles or five years, whichever occurs first

(Sec. 202(d)). The 1990 Amendments extended the useful life of LDVs

to 100,000 miles or ten years, beginning with 1994 model year

vehicles. The amendments also tightened emissions standards for 40

percent of a vehicle manufacturer's LDV and light-duty truck (LDT)

sales in model year 1994, 80 percent in model year 1995 and for all

vehicles after model year 1995 (Sec. 202(g)). The useful life for

heavy-duty vehicles and engines is generally 120,000 miles or eleven

years.

\19\Durability testing over the useful life of the vehicle has

involved testing two identical sets of vehicles for 50,000 miles (in

the case of pre-1994 standards for LDVs), one set using the base

fuel and the other using the waiver fuel. Each vehicle is tested for

emissions at 5,000 mile intervals. This is essentially the same

testing pattern which has been required for certification of a new

motor vehicle under Sec. 206 of the Act. As noted above, under the

1990 Amendments, the useful life of LDVs has been extended to

100,000 miles beginning with the 1994 model year when more stringent

emissions standards took effect (see Sec. 202 (d) and (g)).

---------------------------------------------------------------------------

In addition to emissions data, EPA also reviews data on fuel

composition and specifications, both to fully characterize a proposed

fuel, and to determine whether that fuel would cause or contribute to a

failure of vehicles to comply with their emission standards. Such a

failure often can be predicted from characterization data. For example,

volatility specifications of the fuel could demonstrate a tendency for

high evaporative emissions. Similarly, data on materials compatibility

could show potential failure of fuel systems, emission related parts,

and/or emission control parts from use of the fuel. Such failures could

result in greater emissions. Likewise, fuel characteristics that could

cause significant driveability problems could result in tampering with

emission controls and, thus, increased emissions.

One issue raised previously in the context of Ethyl's present

application was whether Ethyl was required to show that MMT will not

cause or contribute to noncompliance with emission standards by

vehicles certified to ``future'' emission standards (i.e., 1994 model

year standards, which were not in effect at the time of the waiver

application), as well as vehicles certified to ``current'' standards

(i.e., standards in effect at the time of the waiver application).

Ethyl believes that the statute only requires it to establish that MMT

will not cause or contribute to the failure of vehicles to meet current

emission standards. For the reasons outlined in the Agency's January

1992 waiver decision, EPA disagrees with this reading of the statute

and continues to believe that it is appropriate to consider the effects

of an additive on vehicles' ability to meet more stringent future

standards under circumstances similar to these.\20\ (See 57 FR 2537-8

January 22, 1992.)

---------------------------------------------------------------------------

\20\EPA also considered effects on compliance with future

standards in a previous MMT decision. See 43 FR 41424 (September 18,

1978), In Re Application for MMT Waiver.

---------------------------------------------------------------------------

In the past, EPA has analyzed both instantaneous emission effects

and durability effects using statistical tests to determine if the fuel

additive will cause a ``significant'' number of vehicles to fail

emissions tests.\21\ Generally speaking, these tests have focused on

the portion of the fleet that will actually fail emission standards as

a result of using the fuel or additive.\22\ Thus, the tests used to

date by the Agency primarily consider only the ``cause'' language in

the statute and do not consider the portion of the statute which

requires that the applicant must also show that the fuel or additive

will not ``contribute'' to the non-compliance of vehicles with emission

standards.

---------------------------------------------------------------------------

\21\For a detailed description of the statistical tests which

have been used in the past for instantaneous effects see ``Decision

Document'', Texas Methanol Waiver Decision, U.S. EPA Air Docket

Number EN-87-06, and for those used for durability effects, see 43

FR 41426.

\22\In fact the primary criteria allows for the failure of some

portion of the fleet as a result of use of the fuel or additive.

---------------------------------------------------------------------------

The Agency believes that its present statistical tests and criteria

do not give adequate weight to the requirement in Section 211(f)(4)

that an applicant demonstrate that a fuel will not ``contribute'' to an

emission standard failure.\23\ This is of particular significance in

light of the Clean Air Amendments of 1990, which evidence a strong

Congressional concern that more needs to be done to ensure that people

are not exposed to unhealthy levels of airborne pollution. EPA is

presently reviewing alternative criteria and statistical methodologies

for determining whether use of a fuel or fuel additive will ``cause or

contribute'' to emission exceedances. The Agency expects to initiate a

rulemaking in the near future which will propose more appropriate

criteria and statistical methodologies for reviewing waiver

applications and will afford formal notice to future applicants of the

Agency's intention to adopt revised criteria and methodologies.

---------------------------------------------------------------------------

\23\In fact, the Agency raised questions about the

appropriateness of these previously used approaches in its original

decision on Ethyl's 1991 MMT waiver application. See 57 FR 2535,

2537 and 2538 (January 22, 1992).

---------------------------------------------------------------------------

As explained below, the Agency has concluded that it would not be

appropriate to utilize new criteria and statistical tests concerning

which Ethyl received no prior notice in evaluating Ethyl's application.

However, in the event that Ethyl reapplies in the future for a section

211(f)(4) waiver to allow the use of MMT or any other additive, that

application will be evaluated in accordance with any new fuel waiver

criteria in effect at that time.

B. Discretionary Review

As discussed in part II of this decision, above, the Agency

believes that the use of the term ``may'' in Section 211(f)(4) of the

Act affords the Administrator broad discretion to consider other

factors in deciding whether to grant a waiver, once a waiver applicant

has demonstrated that a fuel or fuel additive will not cause or

contribute to an emission standard failure. This construction of

section 211(f)(4) is also consistent with the other provisions of and

the general purposes underlying the Clean Air Act. Although the

Administrator has not relied on this discretionary authority to deny a

waiver in the past, certain general principles should guide the

Administrator's exercise of such authority.

Although the discretion of the Administrator to consider other

factors in making a waiver decision is broad, it is not unfettered. To

assure that any decision based on factors other than emission standard

failures is not arbitrary and is based on a proper record, the

applicant and other interested persons should be afforded proper notice

of any additional factors to be considered by the Administrator and an

opportunity to comment or submit information concerning those factors.

Any decision based on the discretionary authority of the Administrator

to consider other factors should include an explanation of the factors

which were considered and the relation of those factors to the

decision. Moreover, any policy adopted as part of a decision to deny a

waiver on a discretionary basis should be applied consistently to all

similarly situated applicants.

Protection of the public health is a major goal of both the Clean

Air Act in general and the section 211 fuels provisions in particular.

Accordingly, the Agency believes that when a waiver is sought for a

fuel or fuel additive and there are unresolved concerns regarding the

potential impact of that fuel or fuel additive on public health,

potential health effects can and should be examined as part of the

waiver process. As part of this examination of the potential health

effects of a fuel or additive, the Agency should review any relevant

studies or analyses of which it is aware or which are brought to its

attention by the waiver applicant or by commenters on the waiver

application.

In addition to potential health effects, the Agency may consider

other factors as appropriate in deciding whether it would be in the

public interest to grant a waiver. In particular, the Agency may

consider whether a waiver would be consistent with the objectives of

the Clean Air Act. In each instance, the factors considered and relied

upon should be clearly identified.

IV. Analysis of Emissions Data

A. Description of Previous Test Programs

In support of its request, Ethyl conducted an extensive test

program to determine the effect of MMT on the ability of vehicles to

comply with current and future emission standards. It also considered

the impact of MMT on nonregulated vehicle emissions, urban smog or

ozone, refinery emissions, and crude oil use. Ethyl claimed that its

test results established that MMT would not cause or contribute to

exceedences of current or future emission standards. It also claimed

that MMT use would result in other benefits consistent with Clean Air

Act goals.

In 1988, Ethyl assembled a test fleet of 48 light-duty vehicles,

composed of eight different model types (six Buick Centurys (2.5

liter), six Buick Centurys (2.8 liter), six Buick Centurys (3.8 liter),

six Chevrolet Cavaliers (2.0 liter), six Ford Escorts (1.9 liter), six

Ford Tauruses (3.0 liter), six Ford Crown Victorias (5.0 liter) and six

Dodge Dynastys (3.0 liter)) that together represented a broad spectrum

of then current (1988) technology vehicles. To accumulate mileage,

Ethyl utilized the ``Alternative Mileage Accumulation Cycle'' (AMA)

which is a standard procedure utilized to accumulate mileage for

certification purposes.\24\ It utilized two laboratories to measure

each vehicle's exhaust emissions of the regulated pollutants (HC,

oxides of nitrogen (NOx) and carbon monoxide (CO)) at 5,000-mile

intervals up to 75,000 miles in the case of most vehicles and up to

100,000 miles in the case of several.\25\ It also tested a number of

these vehicles for evaporative HC, particulate and manganese emissions,

materials compatibility, driveability and catalyst durability.

---------------------------------------------------------------------------

\24\A driving cycle is a description of how to drive a vehicle

to accumulate mileage, including such things as a what percentage of

driving should be done at what speed and what the overall average

speed should be. The AMA cycle is described in EPA Mobile Source

Advisory Circular 37-A, (See Docket A-91-46) and is essentially

prescribed for use by manufacturers to accumulate mileage for

certification of vehicles (See 40 CFR 86.092-26). A driving cycle is

used so that test vehicles accumulate mileage in a manner that is

supposedly representative of in-use vehicles. The emissions of a

test vehicle that has accumulated mileage according to a driving

cycle representative of in-use vehicles are more likely to be

representative of in-use vehicles' emissions. There are actually

three alternative cycles associated with the AMA; however, the

average speeds of the three alternatives are very similar, ranging

from 29.9 mph to 30.72 mph.

\25\The ``useful life'' of model year 1993 and earlier light-

duty vehicles (LDV's) is 50,000 miles or five years, whichever

occurs first (section 202(d)). However, the Clean Air Act Amendments

of 1990 extended the useful life of LDV's to 100,000 miles or ten

years, beginning with 1994 model year vehicles. For the standards

that begin to take effect in model year 1994, section 207(c)

provides for intermediate in-use standards for several years.

---------------------------------------------------------------------------

Ethyl analyzed the data collected using EPA's previously used

statistical tests (43 FR 41424, September 18, 1978) and additional

tests developed by its consultants to further characterize the data.

Its analysis indicated that, on average, MMT at the requested

concentration would result in a 0.018 gpm increase in HC emissions and

decreases in NOx and CO emissions. The analyses further indicated

that, when EPA's previously used tests are applied, the increase in HC

emissions would not cause or contribute to vehicles' failure to meet

the current HC emission standard. The results of Ethyl's testing for

materials compatibility, driveability and catalyst durability also

indicated that MMT would have no significant adverse effects on

vehicles' ability to meet current emission standards under average

driving conditions. On that basis, Ethyl claimed that it had made its

statutorily required showing.

Ethyl also submitted data on the catalyst efficiency of the

vehicles which it tested. Ethyl performed back-pressure tests\26\ on

all its vehicle fleet except one model group after accumulation of

75,000 miles. Back-pressure tests were also performed on a pair of Ford

Crown Victorias, one operated on MMT-fuel and one on clear fuel, at

speeds higher than those used in Ethyl's 48-vehicle test program.\27\

The results of these tests indicated that back-pressure was not

significantly different in the MMT vehicles when compared to the clear

fuel vehicles. Ethyl also operated two 5.7 liter Corvettes at extremely

high speeds (100 mph) for 25,000 miles, one using MMT fuel and one

using clear fuel. Although similar in magnitude, the back pressure for

the MMT vehicle was slightly higher than that for the clear vehicle.

Ethyl also presented catalyst efficiency\28\ data based on engine-out

emissions of its fleet and based on ``slave engine'' testing\29\ for

half of its fleet. Results of the slave engine testing indicated no

statistically significant difference between the catalyst efficiencies

for the MMT vehicle components when compared with the clear vehicle

components. Finally, four Chevrolet Corsicas were operated to 100,000

miles, two utilizing MMT fuel and two with clear fuel. The purpose of

this testing was to investigate MMT's effect on the catalyst for a

longer mileage interval than the 75,000 miles over which most of

Ethyl's fleet had been driven. Catalyst efficiencies of the MMT

vehicles were not significantly different when compared to the clear

fuel vehicles.

---------------------------------------------------------------------------

\26\Back pressure tests are used to determine if significant

plugging has occurred in a vehicle's catalyst. The total pressure

ahead of the catalyst is back pressure. This pressure is a measure

of constriction in flow through the exhaust system caused by flow of

the exhaust through the emissions control system and the noise-

reducing components of the vehicle. If plugging has occurred in a

vehicle, the total pressure ahead of its catalyst, the back

pressure, should be greater than expected (e.g., greater than a

matching control vehicle).

\27\In this program the maximum speed was 65 mph for the first

25,000 miles and 80 mph for an additional 10,000 miles.

\28\Catalyst efficiency is a measure of what fraction of the

emissions entering the catalyst are actually removed (or catalyzed)

by the catalyst.

\29\``Slave engine'' testing is the testing of vehicle

components on a single engine which is not in a vehicle. In this

case, catalyst efficiencies between control and MMT vehicles were

investigated using exhaust gases from this single engine which were

routed through the removed catalysts. This would likely result in a

more accurate analysis of catalyst efficiency, since one possible

confounding factor, vehicle to vehicle variability, would be

eliminated.

---------------------------------------------------------------------------

Ford presented original test data which Ford said supported its

contention that actual in-use MMT-induced HC emissions increases are

potentially far greater than those reported by Ethyl.\30\ Ford

conducted testing on a more limited scale utilizing eight vehicles,

representing two model groups, run for 105,000 miles. Ford chose two

model groups which were representative of its newest technology

vehicles at the time. One (the Explorer) represented a technology that

Ford believed may be especially prone to exhibit a buildup of

manganese, due to significantly higher operating temperatures and loads

than those of passenger cars. The other model group, the Escorts, had

close-coupled catalysts, a design which is being incorporated into many

new vehicles in order to meet tighter emissions standards. Like Ethyl,

Ford operated part of its test fleet on clear fuel and part on fuel

containing 1/32 gpg MMT. However, Ford's test program differed from

Ethyl's program in several ways. When accumulating mileage, Ford

utilized a commercial gasoline which contained all of the additives

(detergents, etc.) typically found in such fuels. Ethyl utilized a very

high quality test fuel with tight specifications and no additives.

(Although used for actual emissions testing purposes, Ethyl's fuel

would not be allowed for mileage accumulation when certifying vehicles

since it is not representative of in-use fuel.) When accumulating

mileage, Ford utilized what it called its ``durability cycle'' which it

had previously developed. Compared to the AMA cycle used by Ethyl,

Ford's driving cycle had a higher average speed (54 miles per hour

(mph) versus 30 mph), and a higher percentage of high speed

driving.\31\ (As previously mentioned, Ethyl utilized the AMA cycle

used for certification purposes.) Additionally, in the Ford program,

vehicles were tested for emissions at five mileage intervals (5,000,

20,000, 55,000, 85,000\32\ and 105,000 miles) and six emissions tests

were done at each testing interval. Ethyl, by comparison, conducted

testing every 5,000 miles to 75,000 miles (15 intervals) and utilized

two emissions tests at each interval.\33\ Ford's test vehicles showed

an elevation of HC emissions with MMT that was substantially greater

than the 0.018 gpm reported by Ethyl from its test program.

---------------------------------------------------------------------------

\30\EPA's emissions testing lab and Ford's lab routinely undergo

correlation testing and the data indicate that correlation is good

between the labs. (See memorandum, with attached data, from Martin

E. Reineman, EPA Manager of Correlation and Engineering Services,

Office of Mobile Sources, January 3, 1992, Docket A-91-46.)

\31\Ford indicated that drivers who accumulated mileage in its

test program were asked to follow posted speed limits. Ford

indicated that the cycle consisted of 5% city driving (25 to 45

mph), 5% gravel or off road driving (25 to 45 mph), 20% rural

driving (45 to 55 mph), and 70% highway driving (65 mph). Posted

speed limits are shown in parentheses. By way of comparison, the AMA

cycle consists of 16.1% of driving at 30 mph, 22.6 at 35 mph, 20.9

at 40 mph, 6.4 at 45 mph, 17% at variable speed and one of the three

following options: 16.7% at 50 mph or 16.5% at 55 mph or 8.6% and

7.9% at 55 mph and 70 mph, respectively.

\32\In fact, only two of the four Escorts were tested at 85,000

miles.

\33\Although Ethyl conducted additional emissions tests at some

mileage intervals when the initial two tests showed high variation,

these additional tests were not used in Ethyl's analysis of its

data.

---------------------------------------------------------------------------

Toyota also submitted data on a single vehicle which was operated

for 30,000 miles on MMT-containing fuel after which the oxygen sensor

and catalyst were replaced with new components and then driven on fuel

not containing MMT for 30,000 miles. Toyota also used a driving cycle

with an average speed (41.7 mph) higher than that used by Ethyl for

mileage accumulation and used fuel with what Toyota believed was a

relatively high trace level of lead than that usually found in unleaded

gasoline (0.0045 gpg lead) and oil with a relatively high phosphorus

level (0.13 weight percent). Toyota referred to this test procedure as

the ``Toyota 9-Laps'' and presented evidence which it said suggested

that the catalyst degradation seen by vehicles using the Toyota 9-Lap

test was very similar to in-use catalysts tested by Toyota. Hence,

Toyota suggested, these ``adjustments'' made in creating the Toyota 9-

Lap make the testing of a vehicle more consistent with what would

happen in actual in-use driving. Toyota's data indicated an HC level

after the first 30,000 miles of vehicle use (on MMT fuel) about 0.1 gpm

higher than the same vehicle after the vehicle was driven for a second

30,000 mile interval with a new catalyst and oxygen sensor. Toyota also

submitted data indicating that the efficiency at which the catalyst was

operating for the MMT-exposed components was less than that for the

non-MMT exposed components.

Some time after EPA's January 8, 1992 denial decision, EPA and

Ethyl entered into discussions concerning a possible settlement of the

court case which Ethyl had filed. In the context of these discussions,

Ethyl submitted to the Agency new data it had developed since the

denial decision. Ethyl tested six 1991 Escorts, using both the

relatively high-speed driving pattern similar to that utilized by Ford

in its testing of 1991 Escorts (the Ford cycle) and, also, after

changing emissions system components (catalyst and oxygen sensor), the

driving cycle used by Ethyl in the original test program (EPA's

durability certification cycle also known as the AMA). Half of the

vehicles utilized MMT-containing fuel and half were run on clear fuel

(fuel not containing MMT). Ethyl also performed some catalyst

efficiency tests on these vehicles utilizing a ``slave engine.''

Ethyl also tested six 1988 Escorts which were used in its original

test program driven on the AMA cycle. In the new program, after

replacing the catalyst and oxygen sensor, Ethyl continued mileage

accumulation, from 75,000 to 100,000 miles, utilizing the Ford cycle.

Likewise, Ethyl tested six 1988 Buicks from its original fleet

accumulating mileage (100,000 to 115,000 miles) using the Ford cycle

but without replacing any components. Ethyl also accumulated mileage on

seven pairs of 1992 vehicles (four Crown Victorias, Six Buick Regals

and four Ford Mustangs) in test programs covering from 45,000 to

100,000 miles beyond break-in with and without MMT, using the Ford

cycle.

Based on its inspection and analysis of the new Ethyl data, the

Agency ultimately concluded that Ethyl's program had demonstrated

driving cycle does not contribute significantly to MMT-induced

increases in hydrocarbon emissions. However, in addition to addressing

the issue of driving cycle, the Ethyl data appeared to confirm the

finding by Ford that 1991 Escorts experienced a much higher MMT-induced

HC increase than that observed in other models tested (either in

Ethyl's 1992 fleet or in the original 1988 Ethyl fleet). The Agency was

concerned that these data could indicate that certain engine and

emissions control system configurations were more vulnerable to an MMT-

induced emissions increase irrespective of driving cycle.

To further assist the Agency in developing a test program, EPA held

a workshop in October of 1992 and presented a proposed test program

which could address in a timely manner specific unresolved issues

concerning the effect of MMT on emissions: (1) Whether other vehicles

utilizing fuels containing MMT are likely to experience increases in

hydrocarbon emissions similar to those observed in 1991 Ford Escorts;

and (2) whether fuels containing MMT have significant adverse effects

on emissions from vehicles utilizing the technologies most likely to be

employed to meet future standards.

Ultimately the court case was not settled; however, the test

program presented by the Agency at the workshop was largely adopted by

Ethyl and is the basis of its most recent test program involving the

1993 fleet. These vehicles (with the previously mentioned 1992

vehicles) comprise Ethyl's most recent dataset.\34\

---------------------------------------------------------------------------

\34\On May 25, 1993, and on subsequent dates, Ethyl provided

summaries of the 1992/93 test data to EPA staff and these have been

placed in public docket A-93-26.

---------------------------------------------------------------------------

Ethyl accumulated mileage on three 1992 model year vehicles (four

Crown Victorias to 100,000 test miles,\35\ six Buick Regals to 65,000

test miles and four Ford Mustangs to 45,000 test miles) and six 1993

model year vehicles (six Toyota Camrys to 85,000 test miles, six

Oldsmobile Achievas to 65,000 test miles, six Dodge Shadows to 55,000

test miles, six TLEV Ford Escorts to 85,000 test miles, six Honda

Civics to 80,000 test miles and four 49-state Ford Escorts to 30,000

test miles) with and without MMT. The driving cycles used for these

vehicles were an intermediate driving cycle of 45 mph on average for

the 1993 model year vehicles, an average 55 mph driving cycle (i.e.,

the Ford Cycle) for all mileage accumulation on the 1992 Ford Mustangs

and for the initial 45,000 miles of operation on the 1992 Crown

Victorias and Buick Regals and an average driving cycle of 45 mph was

utilized for these two models thereafter.

---------------------------------------------------------------------------

\35\As referred to here, ``test miles'' indicates mileage

accumulated after break-in (break-in mileages vary among these

models) and during which some vehicles were run on fuel containing

MMT while control vehicles were run on clear fuel.

---------------------------------------------------------------------------

B. Comments on Vehicle Emissions Issues

EPA provided an opportunity for the public to submit written

comments.\36\ Many comments were received from a wide variety of

interests, including refiners, automakers, emission control

manufacturers, states committees, environmental and public interest

groups and private citizens. Taken together, the comments touched on

every aspect of Ethyl's application. The following is a summary of the

comments.

---------------------------------------------------------------------------

\36\As mentioned previously, the comments received concerning

Ethyl's remanded waiver application are available in public docket

A-93-26.

---------------------------------------------------------------------------

Four automakers (Ford Motor Company (Ford), General Motors

Corporation (GM), Toyota Technical Center, U.S.A., Inc. (Toyota), and

Chrysler Motors Corporation (Chrysler)), the American Automobile

Manufacturers Association (AAMA), and the Manufacturers of Emission

Controls Association (MECA) all recommended denial of Ethyl's request

and expressed several concerns with regard to the addition of MMT to

unleaded gasoline. First, they noted that the use of MMT will cause an

increase in HC emissions. Most indicated that the more stringent

emissions standards that began taking effect in model year 1994 will

make any increase in HC emissions particularly troublesome. Further,

they stated that newer technology vehicles will likely be equipped with

catalysts which are nearer the engine (more ``closely coupled'') and

that such close coupling, they stated, results in higher catalyst

temperatures that may make the catalyst more prone to the deposition of

manganese. These commenters indicated that deposition of manganese

compounds on the surface of the catalyst would impair the catalytic

breakdown of emissions from the engine, thereby decreasing catalyst

effectiveness. Additionally, they were concerned that MMT, even at the

1/32 gpg Mn concentration requested, would plug catalysts and thus

reduce the surface area of the catalyst available to break down

emissions from the engine, especially in the case of vehicles operated

under driving conditions which result in higher temperatures such as

heavy load or high speed. Under such conditions, it was pointed out,

the vehicle may be more prone to deposition of manganese.

Ethyl indicated that the assertions that it must ``conclusively''

demonstrate the absence of negative effects is not required by the

section 211(f)(4) standard. Ethyl believes that it need only

demonstrate, by a preponderance of evidence, that the additive will not

cause or contribute to the failure of emission control devices to

comply with applicable emission standards, and, further, it believes

that it has made this showing. Ethyl also stated that the EPA test

program proposed at its October 1992 workshop involving the

accumulation of 65,000 test miles, would be sufficient for purposes of

gauging the effect of MMT on emissions. Ethyl commented that it

followed this proposal in the 1992/93 test fleet, although mileage

accumulation has continued beyond 65,000 miles for three of the eight

model year vehicles tested without new emission results different from

the trends established through 65,000 miles.

With respect to the automakers' concerns about effect of MMT on

newer emission technology such as close-coupled catalysts, Ethyl

indicated that the use of the 1993 Transitional Low Emission Vehicle

(TLEV) Honda Civic in its most recent test program was intended so as

to introduce a vehicle which has the most physically possible close-

coupled emission technology (i.e., one connected directly to the

exhaust manifold). Despite such close-coupling, Ethyl indicated that

the differences in hydrocarbon emissions between clear and MMT-fueled

1993 TLEV Honda Civics was minimal. Ethyl also indicated that this

concern about close-coupled catalysts completely ignores that Ethyl

tested two 1988 models and three 1993 models equipped with close-

coupled catalysts without showing any significant adverse effects on

emissions.

Toyota submitted data on catalysts and oxygen sensors from in-use

customer vehicles from Canada where MMT is used as a fuel additive.

Toyota believes that these catalysts and oxygen sensors indicate that

exhaust emissions of hydrocarbons and carbon monoxide are higher from

catalysts/oxygen systems collected in Canada than comparable catalyst/

oxygen systems from U.S. vehicles. Also Toyota submitted photographs of

a catalyst taken from a high mileage Canadian Hilux pickup truck which

showed plugging of the catalyst passages.

Ethyl's response to Toyota's catalyst/oxygen system data is that it

is not clear from the description of the Toyota test results precisely

what can be concluded from the test program. Ethyl stated that, without

a detailed vehicle history, there is no basis to conclude that MMT had

an effect on the catalyst/oxygen system data.

Chrysler submitted data on the analysis of four catalysts, which

was completed by Johnson Matthey Incorporated (JMI) at Chrysler's

request, that had various degrees of manganese deposition from Canadian

vehicles exposed to MMT in the fuel. It indicated that the results of

the analysis demonstrate that the washcoat of both the partially

plugged catalysts and unaffected catalysts exhibit a clear layer of

``densified'' washcoat containing large quantities of manganese oxides.

Chrysler believes that the JMI report supports its concern that

manganese oxides can fill the catalyst pores, thereby covering precious

metal sites or decreasing wash coat surface area, consequently

eventually decreasing catalyst activity. Regarding the automakers'

concerns about the Additive's effect on emissions system components,

such as exhaust oxygen sensors, exhaust gas recirculation valves,

catalysts and oxygen sensors, Ethyl stated that it has already provided

extensive data showing that the Additive does not adversely effect any

of these emission system components.\37\ (Ethyl's test programs are

discussed in the previous section.)

---------------------------------------------------------------------------

\37\Public Docket A-92-41, No. IV-D-3 (summarizing Ethyl's

emission control component testing.)

---------------------------------------------------------------------------

Nineteen small refiners including the National Petroleum Refiners

Association all recommended approval. They concurred in Ethyl's

assessment of the economic benefits and reduced refinery and vehicle

emissions that would accrue from the replacement of octane obtained

through higher-severity refining with octane obtained from MMT. Several

emphasized that MMT would be especially helpful to small refiners since

octane enhancement from MMT requires less capital investment than other

means of increasing octane. Many refiners also pointed out that

refinery operations at lower severity would result in decreased

aromatic and benzene emissions from vehicles and increased yield for

each barrel of crude oil refined.

C. Available Data Meet Previously Used Criteria

The criteria and statistical tests previously used by EPA to

examine durability waiver applications were used only once by the

Agency prior to Ethyl's 1990 application for the use of MMT.\38\ These

tests include a variety of approaches to durability data designed to

determine whether the additive causes increases in regulated pollutants

and, if so, whether those increases bring about failure of vehicles in

the fleet to meet the standards to which they were certified. While EPA

has some concerns regarding the appropriateness of these criteria and

tests for current conditions, the Agency does not intend in this action

to hold Ethyl to any new criteria and/or tests that are not currently

in place. Accordingly, the following discussion is addressed primarily

to the results of applying the most critical of the previously used EPA

tests.\39\

---------------------------------------------------------------------------

\38\The test were used in EPA's examination of Ethyl's 1978

application. For a description of these tests, see EPA's decision on

the application at 43 FR 41424, September 18, 1978.

\39\EPA has carefully reviewed Ethyl's application of the test

to these data in various combinations and has concluded that the

tests were conscientiously and accurately applied. This review

focused particularly upon the application of the ``integrated

emissions test'', the ``cause or contribute'' test, and the overall

sign test.

---------------------------------------------------------------------------

The earliest set of test results under consideration here (tests of

1988 vehicles submitted with Ethyl's 1990 application) exhibit the most

pronounced MMT-caused emissions increases of the data generated by the

applicant (about 0.02 gm/mi\40\, but these increases fall substantially

short of failure on the determinative ``cause or contribute'' test\41\

(3 of 8 vehicle models tested fail for HC and 4 of 8 models fail for

CO, while 7 of the 8 models tested are required to fail before the

additive fails this overall test on either pollutant).

---------------------------------------------------------------------------

\40\Determined by integration of emissions test results gathered

over the full range of mileage supplied by the applicant.

\41\Ethyl's consultant, Systems Applications, Inc., describes

this test on page 19 of a report that was included as Appendix 2A in

Ethyl's May 9, 1990 application for waiver. This co-called ``cause

or contribute'' test, really addresses the question of whether the

additive ``causes'' a failure to meet the certified standard for a

regulated pollutant for each model group and then looks to see if

enough model groups failed the test to warrant the conclusion with

high confidence that more than half of the models are caused to fail

by operation on the additive.

---------------------------------------------------------------------------

When the larger body of all available and appropriate\42\ long-term

emissions data on High-Tech 3000 is evaluated using these previously

used EPA tests, the conclusion is that these increases (averaging 0.02

gm/mi for HC) bring about failure of the ``cause or contribute'' test

in only 4 (for HC) or 5 (for CO) of the 19 model groups tested by the

applicant and others. Failure of that test\43\ must occur in at least

13 of the 19 model groups examined before the additive is deemed to

have failed the overall test with 90% confidence. Fourteen of 19 must

fail before the test is failed at the 95% confidence level.\44\

---------------------------------------------------------------------------

\42\Appropriate data are considered to be those collected with

Federal Test Procedure (FTP) testing using an experimental design

with a control group and no obvious sources of bias. The data

referred to here include the eight 1988 models tested by Ethyl, the

two 1991 models tested by Ford, and the eight 1992 and 1993 models

tested by Ethyl.

\43\In order for a model to fail the test, emissions from the

additive-fueled vehicles must be sufficiently high that then percent

of the represented fleet of that model group using the additive is

predicted to exceed the standard beofre the end of its useful life.

The control vehicles must reach this failure rate at a higher

mileage than the additive-fueled vehicles.

\44\These ``confidence levels'' correspond, respectively, to the

0.10 and 0.05 significance levels. The significance level is the

probability that a decision to reject the null hypothesis (and find

an increase) will be a result of sampling error and thus be

incorrect.

---------------------------------------------------------------------------

If the newer technology 1992 and 1993 vehicles tested by the

applicant are examined in isolation from the earlier test programs, the

data (with an average HC effect of 0.002 gm/mi) pass the historical

tests even more easily than is the case for the data combinations

examined above. None of the nine models failed the ``cause or

contribute'' test for hydrocarbons and only one failed for carbon

monoxide. Seven of nine models would have to fail for the additive to

fail the overall sign test at the 90 percent confidence level and eight

would have to fail for 95 percent confidence.

The overall conclusion from the above analysis, then, is that

Ethyl's additive passes the most critical of the historical tests with

a comfortable margin.

D. Data on Newer-Technology Vehicles Meet More Stringent Criteria

Notwithstanding the Agency's conclusion that it would not be

appropriate to require Ethyl to satisfy new statistical tests

concerning which it has not been given prior notice and the Agency's

decision to evaluate Ethyl's application primarily according to the

previously utilized statistical tests, the Agency nevertheless

considers its existing tests and the criteria that they implement to be

obsolete under current conditions.\45\

---------------------------------------------------------------------------

\45\The tests are extremely conservative in that they place most

of the burden of proof on the Agency rather than on the applicant.

The ``cause or contribute'' test is failed by an engine family only

under circumstances where emissions from the family are so high that

an additive--caused increase in some pollutant pushes more than ten

percent of the vehicle fleet into violation of the standard.

Moreover, the final sign test that is applied to the model-specific

results is failed by the additive only when it may be concluded with

high confidence that more than half of the models in the represented

population would fail the model-specific test. In practical

situations with relatively small samples, this sign test permits a

high percentage of the models in the sample to fail before the

additive is declared to have failed the test. These tests, then, may

permit the granting of waivers in the face of substantively

significant emissions increments attributable to an additive--

increments that would tend to offset the benefits from an

increasingly stringent regulatory program aimed at bringing the

nation's most serious air quality problems under control. Agency

concerns with these tests were addressed previously in its decision

on Ethyl's 1990 application (57 FR 2535, January 22, 1992) and in

(58 FR 64761, December 9, 1993).

---------------------------------------------------------------------------

Therefore, EPA has gone beyond the historical tests to examine

Ethyl's data on the use of the additive with newer technology vehicles

under more stringent criteria of the sort that seem to be warranted by

current conditions. For this analysis, EPA chose to examine the

additive's performance against the most stringent of the possible

criteria--a requirement that the additive cause no statistically

significant increase in emissions.

If one uses a one-sided null hypothesis that the additive causes no

increase in HC emissions, one may employ various statistical tests to

examine the credibility of that hypothesis in light of the test

results. One such test is the computer-intensive ``permutation test''

in a form called an ``approximate randomization'' test.\46\ Application

of this test to the full mileage range of HC emissions data from

Ethyl's tests of 1992 and 1993 vehicles results in a failure to discern

any ``real'' emissions increase at all--that is, no increase that we

may not reasonably attribute to sampling error rather than to an

additive effect on HC in the sampled vehicle population.\47\

---------------------------------------------------------------------------

\46\The permutation test is built around the idea that, if the

null hypothesis is correct, the increase due to the additive in the

sample is only one member of a distribution of all possible such

increases computed from assignments of vehicle emissions to fuel

groups within models. Only if the fuel-related increase from the

sample is an extremely unusual result in theis distribution of

possible increases is it reasonable to reject the null hypothesis

and conclude that the additive actually brought about an increase in

emissions. The way that this method works in practice is that, on

each iteration of the computer program, the computer randomly

rearranges the fuel group assignments among the emission results

within each model group separately. The emission values assigned

(for that iteration) to the additive fuel group are then summed over

the entire sample to form the test statistic. This process is

repeated a very large number of times (one million in this case) and

the resulting test statistics are tabulated. Only if the same test

statistic, as computed from the empirical sample, exceeds a pre-

determined percentage of the simulated test statistics may we reject

the ``no-difference'' hypothesis and conclude with the necessary

degree of certainty that an increase has occurred.

\47\This conclusion holds even when the test is performed at the

0.10 significance level used in conducting the statistical testing

on the data from Ethyl's 1978 application. It is important to note

that the original Ethyl test fleet of 1988 model year vehicles that

are now older than those representing the newest Ethyl data set did

not fare as well and, as mentioned previously, do demonstrate

statistically significant increases in HC emissions.

---------------------------------------------------------------------------

E. Finding

Based on all of the information then available concerning the

potential effect of use of MMT in unleaded gasoline on regulated

emissions, as submitted by Ethyl and others, the Administrator of EPA

determined on November 30, 1994 that, ``Ethyl has satisfied its burden

under Clean Air Act 211(f)(4) to establish that use of HiTEC 3000 at

the specified concentration will not cause or contribute to a failure

of any emission control device or system (over the useful life of any

vehicle in which such device or system is used) to achieve compliance

by the vehicle with the emission standards with respect to which it has

been certified.'' The basis for this determination was described

briefly in the Administrator's November 30, 1994 notice, and has been

reviewed in detail above.

The November 30, 1994 determination was specific to Ethyl's present

waiver application. As the Administrator made clear in the notice

announcing the determination, it does not apply to any new application

concerning either HiTEC 3000 or MMT in the event that this decision to

deny Ethyl's application on the basis of concerns regarding potential

health effects is upheld in any subsequent judicial review. Although

Ethyl may be able to sustain its burden under Section 211(f)(4) in the

context of any future waiver application, any such application must

include satisfactory data addressing the effect on vehicles in

production at that time and will be evaluated according to the

statistical methods and criteria for evaluation of waiver applications

in effect at that time.

V. The Onboard Diagnostics Issue

Prior to the Administrator's November 30, 1994 finding concerning

emission effects, three auto manufacturers, Ford, General Motors (GM),

and Chrysler, and the American Automobile Manufacturers Association

(AAMA), all commented on concerns about the impact of the oxidative

products of MMT on onboard diagnostic (OBD II) systems employing

before-catalyst and after-catalyst oxygen sensors.\48\

---------------------------------------------------------------------------

\48\An Onboard Diagnostic System, with the present generation

commonly known as OBD-II, monitors the activity of an automobile's

emission control system, primarily the catalytic converter, and

alerts the driver via a dashboard light in the event of a

malfunction. Put simply, this aspect of the OBD system functions by

utilizing devices before and after the catalyst which ``sense'' the

presence of oxygen. If the catalyst is functioning properly, it will

absorb a certain amount of oxygen and a specified decrease in oxygen

content in the exhaust gases can be determined by comparing the

oxygen ``sensed'' before and after the catalyst. If the catalyst is

functioning improperly, oxygen storage by the catalyst is impaired

and a drop in exhaust gas oxygen after the catalyst beyond the

proper range is ``sensed'' by the OBD system.

---------------------------------------------------------------------------

GM concerns regarding the OBD II system were two-fold. Its first

concern was that since it is known that manganese oxide has the ability

to store oxygen, a potential problem could occur with dual oxygen

sensor systems. GM stated that, with manganese oxide covering the

catalyst and the oxygen sensors, a false oxygen storage capacity of the

catalyst could be indicated by the OBD II system, which could then

indicate that the catalyst was still working properly while the

opposite could be true. GM's second concern was that the catalyst would

act as a ``filter'' and manganese oxide from MMT combustion passing

through the exhaust system would coat the before-catalyst oxygen sensor

and after-catalyst oxygen sensors unevenly, thus causing the OBD II

system to malfunction. GM also stated that in the 1994 model year GM

planned to market two engine families equipped with OBD II systems

employing before- and after-catalyst oxygen sensors.

With respect to the automakers' concern that use of MMT would

adversely affect operation of the OBD II system, on July 15, 1993,

Ethyl submitted data which it believed demonstrated that this concern

has no basis.\49\ Ethyl stated that no production vehicles were then

equipped with OBD-II systems and that the primary hardware approach

being considered by the automobile manufacturers involves the use of

exhaust gas oxygen (EGO) sensors before and after the catalytic

converter to monitor converter efficiency. Ethyl further commented that

test data generated by Ethyl showed that use of the additive would have

no adverse effect on either the hardware component of these planned

ODB-II systems (i.e., the oxygen sensors), or on the catalytic

converter itself. Ethyl noted that, ``[s]ince these future systems are

currently under development, it is impossible to consider the long term

effects of MMT on these systems.''

---------------------------------------------------------------------------

\49\See Public Docket A-93-26, Number II-D-8, Appendix 5.

---------------------------------------------------------------------------

On November 4, 1994, only 26 days prior to the mandatory date for a

decision on Ethyl's waiver application, Ford Motor Company submitted a

report describing bench testing\50\ of catalysts, in which Ford

measured the oxygen storage capacity of catalysts which had been

deliberately degraded and then exposed to the emissions from MMT-

containing fuel. Ford's conclusion based on these tests was that the

exhaust gas oxygen (EGO) sensors would be affected by the deposition of

manganese oxides associated with MMT use, thus sending incorrect

signals to the diagnostic control system in the vehicle.

---------------------------------------------------------------------------

\50\Bench testing means the testing of components during which

time the components are not actually in the vehicle. The details of

this testing can be found in Document II-D-56 in Docket A-93-26. (An

incomplete preliminary report of this information was submitted to

the Agency in Document II-D-38, Docket A-93-26.)

---------------------------------------------------------------------------

Although the Agency regarded the concerns expressed by Ford in its

November 4, 1994 submission regarding the effect of MMT use on OBD

systems as potentially very important, based on the very limited

analysis which could be undertaken prior to the November 30, 1994

deadline for a decision concerning Ethyl's application, the Agency

concluded that the limited bench testing submitted by Ford did not

allow a conclusion concerning the likelihood that a significant impact

would actually occur during vehicle operation. In addition, the Agency

had several questions regarding the procedures involved in the Ford

testing which could not be resolved within the available time. The

November 30, 1994 notice announcing the Administrator's determination

concerning emission effects made it clear that EPA was concerned about

this issue and would retain the authority to take appropriate action in

the future pursuant to Clean Air Act Section 211(c).

EPA met with staff of Ford in February of 1994 in order to discuss

Ford's concerns raised in its November 4, 1993 submission. Ford

generally expressed the same concerns as had been expressed by GM (and

discussed above).\51\ According to Ford, its testing showed that

combustion of gasoline containing MMT deposits a layer of manganese

oxide on top of the catalyst washcoat and that this causes the EGO

Sensor to measure a lower oxygen level, thereby indicating a higher

oxygen storage capacity than that which would be indicated by the

catalyst without MMT. As a result, a malfunctioning catalyst might not

be detected. Ford expressed particular concern because it had just

introduced three 1994 model year vehicle families employing OBD-II,

whereas the other automakers will not have systems out until the 1996

model year.

---------------------------------------------------------------------------

\51\Ford's concerns are discussed in more detail in a memo to

docket A-93-26, with an attachment submitted to the Agency entitled

``Section 211(c) Impacts of MMT''.

---------------------------------------------------------------------------

Recently, on May 3, 1993, Ford submitted additional information

which the Agency is currently reviewing. This new information appears

to provide further evidence to substantiate the concerns expressed by

Ford regarding the impact of MMT use on OBD systems. Unlike the

previously submitted Ford data, the new data address an actual

production vehicle fitted with a failed catalyst and the effect use of

MMT had on the OBD system's ability to detect failure of the catalyst.

The Agency is continuing to investigate the question of the

potential impact of use of MMT in unleaded gasoline on OBD systems. If

after further investigation EPA concludes that the concerns expressed

by the vehicle manufacturers are warranted, EPA intends to initiate an

appropriate rulemaking under Section 211(c).

VI. Manganese Health Assessment\52\

---------------------------------------------------------------------------

\52\The assessment presented here is taken from ``Reevaluation

of Inhalation Health Risks Associated with Methylcyclopentadienyl

Manganese Tricarbonyl (MMT) in Gasoline'' (United States

Environmental Protection Agency, 1994b) which can be found in the

docket in its entirety.

---------------------------------------------------------------------------

A. Introduction

In 1990, the EPA Office of Research and Development (ORD) assessed

the potential health risks associated with the use of

methylcyclopentadienyl manganese tricarbonyl (MMT) as an additive in

unleaded gasoline (U.S. Environmental Protection Agency, 1990).\53\

Later, ORD (Preuss, 1991) reaffirmed its assessment after considering a

resubmitted waiver application for MMT from Ethyl Corporation. As

identified in earlier ORD evaluations (U.S. Environmental Protection

Agency, 1990; Preuss, 1991), a key issue is the potential health risk

associated with inhalation exposure to manganese tetroxide

(Mn3O4), which is the primary by-product resulting from the

combustion of MMT in gasoline. New information on manganese (Mn) health

effects and exposure is incorporated in this revised risk assessment.

(United States Environmental Protection Agency, 1994b)

---------------------------------------------------------------------------

\53\The many references in this section of the decision dealing

with manganese health effects are referred to in parentheses and

listed at the end of this section in subsection E.

---------------------------------------------------------------------------

This reevaluation has four components: (1) a health effects

assessment, (2) an exposure assessment, (3) a risk characterization

relating the first two, and (4) a summary and conclusions. This

evaluation summarizes earlier ORD assessments and incorporates

information from certain other major new reports and analyses.\54\

---------------------------------------------------------------------------

\54\The reader is referred to the appendices of the full EPA/ORD

reevaluation for more detailed background information. This report,

``Reevaluation of Inhalation Health Risks Associated with

Methylcyclopentadienyl Manganese Tricarbonyl (MMT) in Gasoline'',

can be found in its entirety, including the appendices, in docket A-

91-46. Appendix A presents dose-response analyses, Appendix B

presents an exposure assessment, and Appendix C contains the current

verified Mn inhalation reference concentration (RfC) as it appears

in the U.S. EPA Integrated Risk Information System (IRIS, 1993).

---------------------------------------------------------------------------

B. Health Effects Assessment

1. Background

The toxicity of Mn varies according to the route of exposure. By

ingestion, Mn has relatively low toxicity at typical exposure levels

due in part to a low rate of absorption from the gastrointestinal tract

and in part to efficient regulation by homeostatic mechanisms.

Manganese is considered a nutritionally essential trace element and is

required for certain enzymes important for normal functioning of the

central nervous system and other body organs. However, by inhalation,

Mn has been known since the early 1800s to be toxic to workers. It

should be noted that Mn occupational studies predominantly (and

sometimes exclusively) involve men. Neurobehavioral, respiratory, and

reproductive effects are the primary features of excessive occupational

exposure to Mn. Manganism is characterized by various psychiatric and

movement disorders, with some general resemblance to Parkinson's

disease in terms of difficulties in the fine control of some movements,

lack of facial expression, and involvement of underlying

neuroanatomical and neurochemical factors. Neurobehavioral effects of

Mn intoxication are generally more clinically prominent than

respiratory or reproductive effects. However, respiratory effects

(e.g., pneumonitis) and reproductive dysfunction (e.g., reduced libido)

are also frequently reported features of occupational Mn intoxication.

The available evidence is inadequate to determine whether or not Mn is

carcinogenic; some reports suggest that it may even be protective

against cancer. Based on this mixed but insufficient evidence, EPA has

placed Mn in a Group D weight-of-evidence category, which signifies

that it is not classifiable as to human carcinogenicity. Given these

features of Mn toxicity, the health assessment focuses on the potential

for chronic noncancer effects.

Various epidemiological studies of male workers exposed to Mn at

average levels below the current American Conference of Governmental

Industrial Hygienists Threshold Limit Value (TLV) (5 mg/m\3\)\55\ have

shown neurobehavioral, reproductive, and respiratory effects, both by

objective testing methods and by workers' self-reported symptoms on

questionnaires. Neurobehavioral effects generally have reflected

disturbances in the control of hand movements (e.g., tremor, reduced

hand steadiness) and/or the speed of movement (e.g., longer reaction

time, slower finger-tapping speed). Reproductive effects have included

a decrease in the number of children born to Mn-exposed workers

(compared to matched controls) and various self-reported symptoms of

sexual dysfunction. In recent studies at low to moderate occupational

exposure levels, respiratory effects have been reflected primarily in

self-reported symptoms of respiratory tract illnesses rather than in

differences between objective pulmonary function measurements in Mn-

exposed and control workers. However, the lack of studies using more

sensitive investigational methods and the existence of some limited

evidence from an epidemiological study of school children raise a

degree of concern about pulmonary function effects in relation to lower

level Mn exposure.

---------------------------------------------------------------------------

\55\The American Conference of Governmental Industrial

Hygienists (1992) has given notice of intent to lower the TLV to 0.2

mg/m\3\.

---------------------------------------------------------------------------

The precise mechanisms of Mn neurotoxicity are not well understood,

but it appears that Mn can affect several different aspects of central

nervous system (CNS) function and structure. Some experimental evidence

suggests that the mechanisms of Mn toxicity may depend on the oxidation

state of Mn. However, both the trivalent form (Mn3+) and the

divalent form (Mn2+) have been demonstrated to be neurotoxic.\56\

Also, both forms of Mn can cross the blood-brain barrier, although

research suggests that Mn3+ is predominantly transported bound to

the protein transferrin (Aschner and Gannon, 1994), whereas Mn2+

may enter the brain independently of such a transport mechanism (Murphy

et al., 1991). Unlike ingested Mn, inhaled Mn is transported directly

from the respiratory system to the vicinity of the brain before its

first pass by the liver. Depending on the form of Mn inhaled, its

conversion to other oxidation states (e.g., oxidation of Mn2+ to

Mn3+ or reduction of Mn4+ to Mn3+), and its ability to

enter the brain (through a protein transport mechanism or otherwise),

it is quite possible that a significant fraction of even small amounts

of inhaled Mn would be able to reach target sites in the CNS. Thus, the

apparently greater toxicity of inhaled versus ingested Mn may reflect

important pharmacodynamic and pharmacokinetic differences of Mn that

enters the body by different routes. A more definitive understanding of

these issues will require more empirical information.

---------------------------------------------------------------------------

\56\Various elements can exist in more than one form of charged

atom, depending on the number of negatively charged and positively

charged particles contained in the atom. Manganese is one such

element where, depending on the number of charged particles

associated with the atom, the atom may have a net charge of two or

three ``plus'' charges resulting in a ``divalent'' or ``trivalent''

form, respectively.

---------------------------------------------------------------------------

2. Earlier Assessments

Earlier ORD health assessments have been based on the RfC, which is

defined as an estimate (with uncertainty spanning about an order of

magnitude) of a continuous inhalation exposure level for the human

population (including sensitive subpopulations) that is likely to be

without appreciable risk of deleterious noncancer effects during a

lifetime. The basic procedure for derivation of an RfC entails

identifying a no-observed-adverse- effect level (NOAEL) and a lowest-

observed-adverse-effect level (LOAEL) from a ``principal'' study,

generally defined as the available study that best defines the highest

NOAEL or lowest LOAEL for the most sensitive endpoint affected by a

chemical. When an investigation of occupationally exposed humans is the

principal study (as in the case of the Mn RfC), the NOAEL or LOAEL is

adjusted for differences in ventilation rates and exposure durations

between the occupational exposure scenario (10 m\3\ air breathed per 8-

h workday, 5 days/week) and the ``general public'' scenario (20 m\3\

air breathed per 24-h day, 7 days/week). The adjusted NOAEL or LOAEL is

then divided by uncertainty factors and a modifying factor. In the case

of the original (1990) RfC for Mn, uncertainty factors of 10 each were

used for extrapolating from a healthy worker population to the general

population (including sensitive subpopulations) and for extrapolating

from a LOAEL to a NOAEL. Also, an uncertainty factor of 3

(approximately one-half of 10 on a log scale) was used for

extrapolating from subchronic to chronic exposure. A modifying factor

of 3 was used because of statements by the authors of the principal

study (Roels et al., 1987) that past exposure levels of workers in the

subject study were probably lower than those measured at the time the

study was conducted. The resulting RfC of 0.4 g Mn/m\3\ was

used for the earlier ORD risk assessment (U.S. Environmental Protection

Agency, 1990) and was entered on EPA's IRIS computer database of human

health risk and regulatory information in December 1990.

3. 1993 Revised RfC

The original RfC for Mn was revised, in part, because newer

information supplied in conjunction with the resubmittal of the MMT

waiver application by Ethyl indicated that the workers' exposure levels

in the principal study had probably not increased over time, and thus

the modifying factor could be ``eliminated'' (i.e., set equal to 1).

Another reason for revising the original RfC was that more recent

studies (Roels et al., 1992; Mergler et al., 1994) provided additional

evidence of health effects in workers at relatively low airborne

concentrations of Mn.

Independently of their earlier study of Mn-exposed workers (Roels

et al., 1987), Roels et al. (1992) conducted a cross-sectional study of

neurobehavioral and other endpoints in another group of workers from a

different factory-namely, 92 male alkaline-battery plant workers

exposed to manganese dioxide (MnO2) dust--who were compared to a

matched control group of 101 male workers without industrial Mn

exposure. The geometric mean occupational-lifetime integrated

respirable dust concentration was 793 g Mn/m\3\ x years

(range: 40 to 4,433). The equivalent value for total dust was 3,505

g Mn/m\3\ x years (range: 191 to 27,465). The authors noted

that the monitored concentrations were representative of the usual

exposures of the workers because work practices had not changed during

the last 15 years of the plant's operation. No data on particle size or

chemical purity were provided in the report by Roels et al. (1992), but

based on information provided by Roels et al. (1992) and Roels (1993),

the median cut point for the respirable dust fraction was 5 m

aerodynamic diameter. The respirable fraction is more representative of

the toxicologically significant particles (i.e., the smaller particles

that are inhaled and deposit predominantly in the lower respiratory

tract). Total dust measurements comprised the respirable dust as well

as larger particles that deposit predominantly in the nose and throat

region (via nasal breathing) and would be cleared more rapidly from the

respiratory tract than the smaller particles retained in the lower

regions. Therefore, the respirable dust measurements were considered to

be a more accurate indicator of exposure in relation to the observed

health effects.

Manganese-exposed workers in the 1992 study by Roels et al.

performed significantly worse than matched controls on several measures

of neurobehavioral function, particularly visual reaction time, eye-

hand coordination, and hand steadiness. Similar neurobehavioral

impairments were also found in the earlier study by Roels et al. (1987)

of a different occupational population exposed to mixed Mn oxides and

salts at approximately the same levels of total dust (respirable dust

was not measured). In addition, a recent study in Canada by Mergler et

al. (1994) indicated that, among other effects, performance on tests of

the ability to make rapid alternating hand movements, to maintain hand

steadiness, and to perform other aspects of fine motor control was

significantly worse, compared to matched controls, in workers who were

exposed to even lower concentrations of respirable dust (35 g

Mn/m\3\ at the time of the study). If Mergler et al. had included

information on integrated past exposure levels (which they have since

provided to ORD in a preliminary form not yet submitted for

publication), their study would have provided a fivefold lower LOAEL

for the derivation of the RfC. In addition, reports of a Swedish study

of Mn-exposed steel workers (Iregren, 1990; Wennberg et al., 1991,

1992) provided compelling evidence of comparable neurobehavioral

impairments, including slower reaction time and finger-tapping speed.

The median total dust concentration in the Swedish study was 140

g Mn/m\3\, with respirable dust reported as constituting 20 to

80% of individual workers' total dust exposures. Thus, the LOAEL from

this study would be somewhat lower than that from Roels et al. (1992),

but the less fully characterized exposure histories in the Swedish

study made it more appropriate as a supporting (rather than principal)

study for deriving the Mn RfC.

Taken together, the above epidemiological studies provide a

consistent pattern of evidence indicating that neurotoxicity is

associated with low-level occupational Mn exposure. The fact that speed

and coordination of motor function are especially impaired is

particularly noteworthy, given its consistency with other

epidemiological, clinical, and experimental animal evidence of higher

concentration Mn intoxication.

Differences among these studies in the duration of workers'

exposure to Mn raise another issue of relevance to this discussion. In

the Roels et al. (1992) study, the mean period of exposure was 5.3

years (range: 0.2 to 17.7 years). In the other studies, the mean

durations of exposure were longer: 7.1 years in Roels et al. (1987),

9.9 years in Iregren (1990), and 16.7 years in Mergler et al. (1994).

The indications of lower LOAELs in the Canadian and Swedish studies

suggest that neurobehavioral effects might occur at lower

concentrations of Mn if the exposure periods were longer. In addition,

the age of the workers may be an important factor in interpreting these

findings. The oldest worker in the Roels et al. (1992) study was less

than 50 years old; also, the average age in that study was only 31.3

years, versus 34.3 years in Roels et al. (1987), 43.4 years in Mergler

et al. (1994), and 46.4 years in Iregren (1990). These points suggest

that longer exposure and/or testing later in life might result in the

detection of effects at lower concentrations than is possible after

shorter periods of exposure and/or in younger workers. On the other

hand, it is also evident from these studies that a much shorter period

than a full lifetime of occupational Mn exposure may be sufficient to

induce Mn neurotoxicity.

As Roels et al. (1992) and other investigators have noted, a

threshold for the neurotoxic effects of Mn has not been reported in the

epidemiological literature. Therefore, instead of a NOAEL, a LOAEL was

obtained from the study by Roels et al. (1992) by dividing the

geometric mean integrated respirable dust concentration (793 g

Mn/m3 x years) by the average period of worker exposure (5.3

years) to eliminate time (in years) from the time-weighted average,

thereby yielding a LOAEL of 150 g Mn/m3. (The geometric

mean concentration was used to represent the average exposure because

the workers' exposure measurements were log-normally distributed, and

the arithmetic mean exposure period was used because it was the only

value reported by Roels et al. (1992).) The workplace-based LOAEL of

150 g Mn/m3 was then adjusted for nonoccupational

lifetime exposure by multiplying it by (1) the quotient of 10 m3/

day divided by 20 m3/day (for worker versus nonworker ventilation

rates) and (2) the quotient of 5 days divided by 7 days (for work week

versus full week). The resulting adjusted LOAEL, labeled the human

equivalent concentration (HEC), was 50 g Mn/m3, which was

then divided by a total uncertainty factor of 1,000 to yield an RfC of

0.05 g/m3. The total uncertainty factor of 1,000

incorporated the following factors: 10 to protect sensitive

individuals; 10 for using a LOAEL in lieu of a NOAEL; and a composite

factor of 10 for database limitations reflecting the less-than-chronic

periods of exposure and the lack of reproductive and developmental

toxicity data, as well as potential but unquantified differences in the

toxicity of different forms of Mn. A modifying factor was not used

(i.e., it was set equal to 1).

Each RfC is assigned an overall rating of low, medium, or high

confidence level, based on two subsidiary confidence ratings reflecting

the quality of the evidence from the principal studies and the quality

of the overall database for the chemical in question, respectively. The

revised Mn RfC was assigned a medium level of confidence. The evidence

for the neurobehavioral effects of low-level Mn exposure by inhalation

was compelling and consistent across several well-conducted studies.

However, the limited duration of exposure and the lack of a NOAEL for

neurotoxicity in any of the principal or supporting studies prevented

assigning a confidence level greater than medium. Also, the lack of

definitive data on the concentration-response relationship and on the

potential reproductive and developmental toxicity of inhaled Mn limited

the degree of confidence in the database to a medium rating. Virtually

all of the human health evidence is based on healthy, adult male

workers. No known studies have investigated human female reproductive

function, and even though male worker reproductive function is known to

be affected by Mn exposure, it has not received adequate investigation.

The limited available information concerning the developmental toxicity

of inhaled Mn suggests the possibility that prenatal exposure of

laboratory rodents to MnO2 (via the air supplied to the pregnant

mother) may depress neurobehavioral activity in neonatal rats and that

continued postnatal exposure of the pups may intensify this depression.

In addition, several studies have demonstrated alterations in

neurochemical (dopamine) levels in young mice and rats exposed during

early postnatal development to Mn via other routes. Thus, the potential

for developmental toxicity due to Mn exposure exists. The

concentrations and durations of exposure sufficient to induce such

effects are not known. Although adequate epidemiological studies of

children and the elderly have not been conducted, it is known that

certain populations, such as children, pregnant women, elderly persons,

iron- or calcium-deficient individuals, and individuals with liver

impairment, may have an increased potential for excessive Mn body

burdens due to increased absorption or altered clearance mechanisms.

Another concern raised by the lack of studies involving longer

periods of exposure and/or older subjects is that the compensatory or

reserve capacity of certain neurological mechanisms may be stressed by

Mn exposure earlier in life, with manifestations of impairments only

becoming evident much later, perhaps at a geriatric stage. One reason

for the latter concern is that Parkinson's disease is typically a

geriatric disease in which symptoms are only seen when the loss of

brain cells that produce dopamine (which is also apparently involved in

Mn toxicity) reaches 80% or more. Indeed, some neurologists think that

a long latency period of perhaps several decades may precede various

parkinsonian syndromes. These points lead to a concern that if Mn

reduces the compensatory or reserve capacity of the nervous system,

parkinsonian-type effects might occur earlier in life than they would

otherwise. Thus, several questions remain to be answered before higher

confidence in the accuracy of the RfC can be achieved.

The two studies of Roels et al. (1992, 1987) were considered

coprincipal studies for the derivation of the revised RfC, with

supporting evidence in the reports of Mergler et al. (1994), Iregren

(1990), and Wennberg et al. (1991, 1992). Given the fact that these

studies involved exposure to various oxides and salts of Mn, the RfC is

designated as applying to Mn and Mn compounds (including

Mn3O4). The previous RfC of 0.4 g Mn/m3 applied

to Mn only, due to undifferentiated forms of Mn in the principal study.

Given that different forms of metals may have different toxic

properties (due to different oxidation states, different solubilities,

and possibly other factors), it is likely that different compounds of

Mn vary in toxicity. However, sufficient data on the comparative

toxicity of various compounds of Mn are not available to judge the

relative toxicity of Mn3O4 specifically.

As noted above, Mn affects multiple organ systems, including the

respiratory and reproductive systems as well as the CNS. However,

because the only available evidence suggests that the CNS is the most

sensitive target for Mn toxicity, neurobehavioral endpoints were the

focus of the RfC derivation. Although other types of effects remain a

concern, it is presumed, based on the limited data now available, that

protecting against neurotoxicity provides protection against these

other, apparently less sensitive endpoints.

In revising the RfC for Mn, a draft version was subjected to peer

review by external experts (from academic and non-EPA governmental

institutions) as well as internal experts. Following this peer review,

a further-revised version was submitted to and verified by an EPA-wide

RfD/RfC work group in September 1993. The current RfC for Mn was made

available through IRIS in early November 1993 through two mechanisms. A

special notice beginning November 1 in the news section of EPA's

internal IRIS2 database announced the availability of a hard copy of

the text to EPA requesters who contacted the Risk Information Hotline;

also, the text was obtainable through the National Library of

Medicine's publicly accessible on-line computer database, TOXNET,

beginning November 10, 1993. It also became available on line via the

EPA IRIS database beginning December 1, 1993.\57\

---------------------------------------------------------------------------

\57\The complete text of the revised RfC as it exists on IRIS2

may be found in Appendix C in the docket.

---------------------------------------------------------------------------

4. Alternative Approaches to Deriving RfCs

After the revised RfC for Mn became available to the public, Ethyl

Corporation and other interested parties submitted comments on the RfC

and issues related to it. One of the primary comments concerned the

availability of various statistical techniques for deriving a NOAEL

from the study by Roels et al. (1992) and/or from supplementary data

for that study provided to ORD by Roels (1993). In response to Ethyl

Corporation's request that EPA consider alternative approaches to

analyzing these data and deriving an RfC for Mn, further analyses of

the subject data were undertaken using a variety of statistical

methods. These approaches may be identified as (1) conventional NOAEL-

or LOAEL- based analyses, (2) ``no statistical significance of trend''

(NOSTASOT) analyses of the type described by Tukey et al. (1985), (3)

benchmark dose analyses of the type described by Crump (1984), and (4)

Bayesian analyses of the type described by Jarabek and Hasselblad

(1991). These analyses and their results\58\ yield several possible RfC

estimates, so designated because the current and only verified RfC for

Mn is that which has been verified by the EPA-wide RfD/RfC work group

and entered on IRIS. It must be emphasized that the RfC estimates

developed for the purpose of this risk assessment do not represent a

revision of the current verified RfC for Mn. Reexamination of the

current Mn RfC, and any decision to revise or reaffirm the current RfC,

will be under the purview of the EPA-wide RfD/RfC work group at some

future date.

---------------------------------------------------------------------------

\58\See ``Reevaluation of Inhalation Health Risks Associated

with Methycycloentadienyl Manganese Tricarbonyl (NMT) in Gasoline'',

Appendix A, Docket A-91-46.

---------------------------------------------------------------------------

A fundamental issue pertaining to all of the approaches presented

here is the selection of a measure of exposure. Roels et al. (1992)

described two measures of respirable dust, the occupational lifetime

respirable dust concentration (LIRD), expressed as g/m3 x

years, and the current concentration of respirable dust (CRD),

expressed as g/m3. The CRD concentration was measured at

the time the study was conducted by Roels et al. and refers to a

representative concentration measured for the type of job performed by

a worker (e.g., electrician, maintenance worker). The LIRD value for

each worker was a cumulative exposure measure derived by adding the CRD

values over the worker's entire period of employment. If a worker

changed jobs within the plant during his period of employment, the CRD

for each job held was multiplied by the number of years the worker

performed that job. Thus, if more than one job classification was

worked, the worker's LIRD was the sum of the products of CRD multiplied

by years of performance of the respective jobs. However, if a worker

held only one job classification, his LIRD was simply equal to his CRD

multiplied by the number of years employed. Another measure of exposure

may be derived from LIRD by dividing an individual worker's LIRD value

by his total number of years of employment. The latter measure,

designated as the average concentration of respirable dust (ACRD),

reflects a worker's time-weighted cumulative exposure level but removes

years from the unit of measurement of LIRD and is expressed as

g/m3. Although Roels et al. (1992) did not refer to ACRD,

this value could be calculated for each individual and for the entire

cohort by using the unpublished data provided to ORD by Roels (1993).

For reasons to be discussed later, ACRD offers advantages for certain

analyses and, unless otherwise noted, is the exposure measure used in

the alternative RfC estimates discussed here.

a. Conventional NOAEL- or LOAEL-Based Approach. The conventional

method, and only method used thus far by EPA, to derive an RfC has been

to identify a NOAEL or LOAEL from a study and divide that concentration

by uncertainty factors, as described above for the Mn RfC. In the case

of the study by Roels et al. (1992), the geometric mean LIRD

concentration of the Mn-exposed workers was used as a LOAEL. Roels et

al. (1992) also performed an exposure-response analysis of their data

by grouping the exposed workers into three exposure categories and

comparing the prevalence of abnormal neurobehavioral scores for each of

the three groups to those of controls. As indicated in the summary

sheet for the Mn RfC (see Appendix C), the results of this exposure-

response analysis were not used in deriving the revised Mn RfC because

the reported analysis did not correct for multiple comparisons.

However, ORD's analyses of additional data provided by Roels (1993)

suggest a possible RfC estimate of 0.03 g/m3 (versus the

current RfC of 0.05 g/m3), if a one-tailed test of

statistical significance is accepted (see Appendix A, ``Reevaluation of

Inhalation Health Risks Associated with Methyl- cyclopentadienyl

Manganese Tricarbonyl (MMT) in Gasoline'' as referenced in the

reference section below, hereafter referred to as Appendix A.). Because

it was based on an exposure-response analysis, this RfC estimate is

labeled as such in Figure 1, which could not be reproduced in the

Federal Register. It is available by calling the person listed in the

FOR FURTHER INFORMATION CONTACT section of this notice. It is also

available in docket A-93-26, item number II-A-17, page 12 (see ADDRESS

section of this notice for docket location).

b. NOSTASOT Approach. Another approach to analyzing dose-response

data makes use of a procedure known as NOSTASOT, described by Tukey et

al. (1985). In essence, the procedure applies a trend test sequentially

to determine the highest noneffective dose of a series of doses by

eliminating one dose at a time. In this manner, the dose level at which

the response is not significantly different from controls is determined

to be the NOSTASOT dose, which could therefore be considered a NOAEL.

Applied to Roels' (1993) epidemiologic data by beginning with the

highest individual ACRD exposure and moving downward (i.e., a ``top-

down'' approach), the procedure yielded a NOSTASOT of 285 g/

m3 for eye-hand coordination (see Table A-4, Appendix A). This

approach implies that once nonsignificance is reached, further

application of trend tests to lower dose groups would also yield

nonsignificance. However, this was not the case with Roels' (1993)

epidemiologic data, and thus it was important to determine not simply

the highest NOAEL but the highest NOAEL below the lowest LOAEL. By this

``bottom-up'' approach, the highest nonsignificant exposure below the

lowest statistically significant exposure was 21 g/m3,

for visual reaction time. Using the latter value as a NOAEL and a total

uncertainty factor of 100 (the same as that used for the current Mn

RfC, except omitting a factor of 10 for extrapolating from a LOAEL to a

NOAEL), one would obtain a value of 0.07 g/m3 for an RfC

estimate (Figure 1). Disparities in the NOSTASOTs obtained for various

endpoints by the top-down and bottom-up approaches raise questions

about the suitability of this technique for deriving a NOAEL from the

data of Roels (1993).

c. Benchmark Analyses. Another approach to deriving an RfC estimate

is the benchmark dose (BMD) approach, which has been described by Crump

(1984) and others (e.g., Kimmel and Gaylor, 1988; Faustman et al.,

1994; Allen et al., 1994). A BMD is an estimate of the dose (the term

dose is used interchangeably here with concentration, although the

latter is more appropriate for inhalation exposure) that will produce a

specified effect (e.g., a 10% increase in the prevalence of abnormal

scores on a neurobehavioral test in the case of the study by Roels et

al. (1992)). The BMD is calculated by fitting a mathematical model to

the available data and obtaining a maximum likelihood estimate of the

dose associated with a specified increase in response (typically 10, 5,

or 1%). A lower confidence limit is then calculated for the BMD

(usually the 95th percentile), and the result is denoted as a benchmark

dose level (BMDL), which has been proposed as a substitute for a NOAEL

in deriving RfDs or RfCs (Crump, 1984; Barnes et al., 1994). Subscripts

designate the effect level (10, 5, or 1%) for which the BMDL has been

calculated, as in BMDL10, BMDL5, or BMDL1.

A large number of mathematical models could be used for deriving

BMDLs, but six frequently used models have been selected for the

present exercise (as discussed in Appendix A). In applying these models

to the dataset provided by Roels (1993), it appears that the models fit

the CRD and ACRD data better than the LIRD data. (As explained in

Appendix A, it made little difference whether the LIRD values were

obtained from the group data provided in the report by Roels et al.

(1992) or from the individual exposure data supplied by Roels (1993),

so the latter LIRD data were used here.) In principle, LIRD is superior

to CRD as a measure of long-term or cumulative exposure. One reason for

the difference in goodness of fit between LIRD and either CRD or ACRD

is that two workers with low LIRD values had abnormal eye-hand

coordination responses (exceeding the 95th percentile of control

scores). These two subjects appear to have had rather short exposure

durations (0.3 and 0.4 years) and moderately high CRD values (201

g/m\3\ each). Thus, these two data points suggest an LIRD

exposure-response relationship that is better fit by a supralinear

curve with a power term 1, EC5, and EC10,

respectively. One guide to the choice of an effect level is that the

resulting BMD (before calculating the lower confidence limit) is

preferably near or within the range of observed exposure concentrations

(cf. Barnes et al., 1994). Because the BMD for EC1 falls outside

this range of observed concentrations, the primary focus in this

discussion is devoted to the BMDL5 and the BMDL10.

It should be kept in mind that the BMDL represents the lower 95th

percent confidence interval for the effective concentration in

question, and therefore the BMDL probably inherently reflects some

degree of conservatism. However, the degree of conservatism obviously

varies with the effective concentration for different percentage effect

levels and with the nature of the effect (e.g., severe versus moderate

impairment). For the purposes of this assessment, if one treats the

BMDL10 derived from the dichotomized (quantal) data of Roels as if

it were a minimal (less severe) LOAEL and the BMDL5 as if it were

a NOAEL, uncertainty factors of 3 and 1, respectively, would be

warranted. On this basis, as shown in Figure 1 and in Table A-39 of

Appendix A for the quantal linear model using ACRD, an RfC estimate of

0.09 g/m\3\ would be obtained by using the quantal BMDL10

as if it were a LOAEL and a total uncertainty factor of 300 (10 for

intraspecies sensitivity, 10 for database limitations, and 3 for a

minimal severity LOAEL). Similarly, the quantal BMDL5 would yield

an RfC estimate of 0.1 g/m\3\, based on a total uncertainty

factor of 100 (10 for intraspecies sensitivity, 10 for database

limitations, and 1 for a NOAEL). As applied here, the benchmark

approach yields candidate RfC estimates of 0.09 to 0.1 g/m3.

d. Bayesian Analyses. Another approach to deriving a substitute for

a conventional LOAEL or NOAEL, which bears some resemblance to the BMD

approach just described, is known as the Bayesian approach (see

Appendix A). In essence, the Bayesian approach yields a distribution of

concentrations (rather than a point estimate) associated with a

specified effect. Some features of the BMD approach are common to the

Bayesian approach: a mathematical model must be fit to the data, an

effect level must be selected, and a confidence bound on the estimated

concentration associated with a given effect level must be calculated

(although the calculation procedures are different). If these choices

are consistent with those for the BMD approach, the results are quite

similar. By the Bayesian analysis, for a 10% increase in abnormal eye-

hand coordination scores, the lower 90% credible set limit (roughly

equivalent to the quantal BMDL 95% confidence limit\59\ based on the

estimated median concentration obtained with the quantal linear model

is 73 g/m\3\. Adjusting this value to a human equivalent

concentration (HEC) and treating the result (26 g/m\3\) as if

it were a LOAEL(HEC), one may divide by a total uncertainty factor of

300 (10 for intraspecies sensitivity, 10 for database limitations, and

3 for a minimal severity LOAEL) and obtain an RfC estimate of 0.09

g/m\3\. Similarly, the 5% effect level yields an RfC estimate

of 0.1 g/m\3\, based on a total uncertainty factor of 100.

Thus, as applied here, the Bayesian approach yields candidate RfC

estimates of 0.09 to 0.1 g/m\3\, essentially identical to the

results of the benchmark analysis (Figure 1).

---------------------------------------------------------------------------

\59\U.S. Environmental Protection Agency. (1994a) ``Reevaluation

of inhalation health risks associated with methylcyclopentadienyl

manganese tricarbonyl 9MMT) in gasoline.'' Washington, DC: Office of

Research and Development; EPA report no. 600/R-94/062. For further

information see Air Docket A-93-26, II-A-12.

---------------------------------------------------------------------------

One advantage of the Bayesian approach is that it lends itself well

to using continuous as well as dichotomous data. Although Roels et al.

(1992) did not provide individual continuous data (i.e., actual raw

scores instead of designations of normal/abnormal) on the performance

of the workers in their study, they did report mean differences and

standard deviations. With this information, it is possible to estimate

the concentration at which certain effect levels would occur based on

the Bayesian posterior distribution. For example, a 10% increase in the

proportion of subjects with abnormal scores would be associated with a

median concentration of 112 g/m\3\, which has a lower 90%

credible set limit of 90 g/m\3\. Adjusting the latter value as

if it were a LOAEL(HEC) yields a concentration of 32 g/m\3\

and an RfC estimate value of 0.1 g/m\3\, based on a total

uncertainty factor of 300 (10 for intraspecies sensitivity, 10 for

database limitations, and 3 for a minimal severity LOAEL). Note that

these calculations based on continuous data essentially approximate the

quantal BMD and Bayesian calculations for a 10% effect level based on

dichotomous data (see Figure 1). Similar calculations for the actually

observed difference (i.e., 13%) between the Mn-exposed and control

workers in the Roels et al. (1992) study yield an RfC estimate of 0.2

g/m\3\, based on a total uncertainty factor of 300 (including

a factor of 3 for a minimal severity LOAEL). Calculating the

concentration associated with the difference between the exposed and

control mean values that just achieves statistical significance (a 4%

difference in this case) also results in a candidate RfC value of 0.2

g/m\3\, based on a total uncertainty factor of 100

(eliminating the minimal LOAEL factor of 3). Thus, as applied here, the

Bayesian analyses of continuous data yield candidate RfC estimates of

0.1 to 0.2 g/m\3\.

e. Summary of RfC Estimates. Figure 1 displays the current,

verified RfC along with over 100 possible Mn RfC estimates based on

various exposure measures, models, effects measures, and uncertainty

factors. Not all of these RfC estimates are equally plausible or worthy

of consideration in assessing the potential health risks associated

with Mn inhalation exposure due to MMT usage. As discussed above, some

combinations of the three exposure measures and six mathematical models

fit one another better than other combinations. Based primarily on

considerations of cumulative dose toxicity, statistical goodness-of-

fit, and parsimony, ACRD and the quantal linear model appear to achieve

the best results in this respect. Given the similarities of the

benchmark and Bayesian analytic results using ACRD and the quantal

linear model, little distinction can be made between the two analytic

approaches in the present application. As for the results obtained for

different effect levels, using a severity uncertainty factor of 3 with

a 10% effect level (for either benchmark or Bayesian analyses) is

essentially equivalent to using a severity UF of 1 with a 5% effect

level. Note that the terms LOAEL and NOAEL do not actually correspond

to the results for 10% and 5% effect levels, and therefore neither is

preferable to the other in the sense that a NOAEL is preferable to a

LOAEL in deriving an RfC. Therefore, benchmark and Bayesian results for

10% and 5% effect levels (using ACRD with the quantal linear model) are

regarded as equally worthy of consideration here. These particular

analyses yield Mn RfC estimates of 0.09 to 0.1 g/m\3\.

In general, continuous response data are preferred to dichotomized

data, primarily because they provide more information and avoid the

basically arbitrary division of effect measurements into categories

(e.g., normal versus abnormal). The Bayesian analysis based on mean

differences between exposed and control groups offers some of the

advantages of using continuous data, in that the reported means and

standard deviations (from Roels et al., 1992) provide a basis for

estimating the distribution of continuous response measures. However,

this use of continuous data is not immune to certain common problems,

such as the issue of statistical power associated with studies of

limited size, for the approaches calculating observed or just-

statistically significant differences. Also, whereas the dichotomous

data analyses yield more precision in estimating the effective

concentration associated with a somewhat imprecise response variable,

the continuous data analyses offer the opposite trade off (i.e., more

precision in the response variable but less in the exposure estimate).

Nevertheless, the continuous data analyses appear to merit

consideration as well as the analyses based on dichotomous data. By the

Bayesian analyses of continuous data, Mn RfC estimates of approximately

0.1 to 0.2 g/m\3\ are obtained.

Based on the available data and on decisions and assumptions

involved in analyses of these data, the leading candidate estimates for

an alternative Mn RfC appear to fall in a range of approximately 0.09

to 0.2 g/m\3\. (Ethyl Corporation (1994) has proposed an

alternative RfC estimate based on a BMDL10 value of 87 g/

m\3\. Treating this value as essentially a NOAEL (thereby eliminating

an uncertainty factor for use of a LOAEL), Ethyl Corporation divided

the adjusted NOAEL(HEC) by a single uncertainty factor of 10 for

sensitive subpopulations to derive a Mn RfC estimate of 3 g/

m\3\.)\60\

---------------------------------------------------------------------------

\60\The RfC listed here is not simply the BMDL10 of 87

g/m\3\ reduced by the uncertainty factor of 10 because the

BMDL10 must first be adjusted to produce an adjusted NOAEL(HEC)

(i.e., to go from an occupational exposure scenario to a scenario

for the general public) prior to reduction by the uncertainty

factor.

---------------------------------------------------------------------------

C. Exposure Assessment

1. Background

Very limited data have been available by which to estimate

potential Mn personal exposure levels likely to be associated with the

use of MMT as an additive in unleaded gasoline. For example, after the

completion of ORD's 1990 exposure assessment for Mn (U.S. Environmental

Protection Agency, 1990), Ethyl Corporation provided EPA a brief report

of a personal monitoring study as part of Ethyl Corporation's

resubmittal of a waiver application for MMT. The study focused on 6

taxi drivers and 17 office workers in Toronto, ON, where the allowable

MMT concentration in gasoline is \1/16\ (0.062) g Mn/gal. (In the

Toronto study, the actual concentration was reported as \1/26\ (0.039)

g Mn/gal, which is only slightly greater than the \1/32\ (0.031) g Mn/

gal concentration proposed for the United States. As confirmed by

Kirshenblatt (1993), MMT concentrations in Canadian gasoline average

well below the allowable limit there.) In comments on Ethyl's

resubmittal, ORD considered the Toronto data in conjunction with

results from independent field studies of personal exposures to carbon

monoxide to develop a revised Mn exposure assessment (Preuss, 1991). A

key element of the 1991 ORD assessment was the assumption that taxi

drivers (six of whom were monitored in Toronto within a 2-week period)

were members of a high-exposure cluster reflecting the upper 4% of the

population in a model based on the carbon monoxide field studies. The

result of the 1991 assessment was an estimate that 4% of the general

public might be exposed to Mn at approximately 0.09 g/m\3\,

although this estimate had an undetermined amount of uncertainty due to

the inadequacies of the available data.

2. Additional Canadian Studies

Since the 1991 ORD assessment, additional personal exposure studies

have been completed in Montreal and Toronto (described in Appendix B,

``Reevaluation of Inhalation Health Risks Associated with Methyl

cyclopentadienyl Manganese Tricarbonyl (MMT) in Gasoline'' as

referenced in the reference section below, hereafter referred to as

Appendix B.) As shown in Figure B-10 of Appendix B, the average

concentrations reported in the Canadian studies vary by as much as an

order of magnitude for small groups (5 to 19 persons each) of garage

mechanics, taxi drivers, and office workers. The highest average Mn

personal exposure level was 0.25 g/m\3\ for Montreal garage

mechanics while at work; the other averages ranged from 0.002 to 0.035

g/m\3\ for various particle size fractions. Although it is

impossible to extrapolate the results of these studies to the

distribution of Mn exposure levels for the general population, it does

appear that there is a general relationship between personal exposure

levels of Mn and proximity to vehicular emissions of combusted MMT.

Thus, populations living near high traffic-volume areas such as inner

cities and expressways would probably tend to experience higher Mn

exposure levels in relation to MMT usage.

Some of the limitations of the Canadian studies with respect to

development of a quantitative exposure assessment are reviewed in

detail in Appendix B and may be summarized briefly as follows.

The studies did not have adequate sample sizes and did

not sample according to a probabilistic statistical design that

would help ensure the representativeness of the sampled individuals.

The sampling periods were relatively short, 1 to 2

weeks at most. Meteorological and other factors that would be

expected to influence ambient measurements over relatively short

periods of time were not characterized.

Because the studies did not use ambient monitors

collocated with reference monitors (such as the dichotomous samplers

used by Canadian agencies), it is difficult to relate data from the

studies to larger databases from the government monitoring networks.

Because the studies did not use identical monitors to

measure personal exposure levels and outdoor ambient levels, it is

difficult to distinguish between personal exposures and ambient

levels or to relate one to the other.

Quality assurance and certain other important

methodological details are not fully provided in the available

reports.

Because of the substantial limitations of the above exposure

studies, no quantitative assessment of personal exposures to Mn in a

Canadian population is possible at present.

3. The PTEAM Study

The only published study that has used a probability-based

representative sampling design for evaluating exposure levels of Mn in

a general population is the Particle Total Exposure Assessment

Methodology (PTEAM) study, which was conducted in Riverside, CA, over a

7-week period in the fall of 1990 (Pellizzari et al., 1992). This study

used personal and stationary monitors to measure Mn concentrations

indoors and outdoors. The personal samplers collected PM10, and

the stationary samplers collected PM2.5 as well as PM10 (see

glossary of terms in Attachment B-6 to Appendix B). Of the 139,000

nonsmoking residents age 10 years and older in Riverside, 178

individuals were selected through a stratified sampling plan to

represent the general population and were monitored over two 12-hour

periods (daytime and nighttime). More than 2,750 particle samples were

collected. Quality assurance and other procedures are summarized in

Appendix B and are described extensively elsewhere (e.g., Pellizzari et

al., 1992; Clayton et al., 1993; Thomas et al., 1993). The PTEAM study

has been presented in various peer-reviewed publications and discussed

in several scientific forums (see Attachment B-5 to Appendix B). It

represents the best available information on an actual distribution of

general population exposures to Mn. It also provides valuable

information on potential Mn exposure associated with MMT use, because

MMT was used in leaded gasoline in California prior to and during the

period of the PTEAM study.

4. Estimated Mn Exposure Levels Associated with MMT Usage

As noted above, the substantial limitations of the available

Canadian exposure studies make them unsuitable for estimating

population exposure levels of Mn in relation to MMT usage. In addition,

ambient monitoring data typically underestimate and may be uncorrelated

with personal exposure levels of automotive-source pollutants.

Therefore, of the currently existing published evidence pertaining to

Mn exposure levels in relation to MMT usage, only the PTEAM Riverside

study (Pellizzari et al., 1992) provides a reasonable basis for

estimating potential future exposure levels in relation to a scenario

where 100% of unleaded gasoline contains \1/32\ g Mn/gal as proposed by

Ethyl Corporation.

In the PTEAM study, measurements of personal exposure levels of

PM10 Mn indicated that approximately half of the population in

Riverside in the 1990 study period had 24-hour personal exposures to

PM10 Mn above 0.035 g/m\3\, with the highest 1% of the

population having exposures above 0.223 g/m\3\ PM10 Mn.

However, given the use of PM5 Mn exposure measurements in the

study of Roels et al. (1992), it would be preferable to consider a

population distribution of personal exposure levels of PM5 Mn. Due

to limitations in the available data, the exposure assessment in

Appendix B focuses on estimated personal exposure levels for PM4

Mn, not PM5. Although the difference is probably small, PM4

levels are an underestimate of PM5 levels. The derivation of the

projected exposure estimates involved several steps, which may be

summarized as follows.

The automotive and nonautomotive contributions to particulate Mn

exposures in the PTEAM study were estimated using data from Lyons et

al. (1993), who reported particle size distributions up to PM4 of

selected trace metals, including Mn, at two locations near Riverside in

the winter and summer of 1989. They attributed most of the PM4 Mn

to automotive sources. Based on their findings and data from other

sources, it is possible to estimate that 69% of the PM2.5 fraction

of PM4 Mn they measured was derived from automotive sources

(namely the combustion of MMT in motor vehicle fuel, as then allowed in

leaded gasoline in California) and that 31% was derived from paved road

dust (mostly earth crustal material). Next, the PTEAM Mn measurements

from stationary indoor monitors (SIMs) were used to estimate personal

exposure levels by adjusting the SIM PM2.5 Mn data to reflect the

typically higher levels of all elements measured by personal exposure

monitors (PEMs). This adjustment was made in two ways, either by the

PEM:SIM ratio obtained for Mn or by the ratio obtained for lead (Pb),

another element related to automotive fuel usage. These two methods of

adjusting the SIM data to PEM values resulted in two projected

distributions, as will be described below. The next step in the

derivation procedure involved adjusting the PM2.5 personal

exposure estimates to reflect PM in the size range from 2.5 to 4

m (based again on data from Lyons et al., 1993).

With these estimates of PM4 Mn personal exposure levels due to

automotive sources, it was then possible to project from the situation

in Riverside around the time of the PTEAM study (when leaded-MMT

gasoline constituted about 14% of the gasoline sold and contained an

average of 0.048 g Mn/gal) to a future scenario that assumes 100% of

the unleaded gasoline contains MMT at \1/32\ (0.031) g Mn/gal. This

aspect of the derivation is described in detail in Attachment B-4 to

Appendix B. In essence, a factor was calculated to reflect the

estimated increase in MMT usage between 1990 and 1995 (i.e., the first

full year in the near future). This projection factor assumed an

increase of 1% per year in gasoline usage and no difference in the Mn

emission rate (grams Mn emitted per gram Mn in fuel combusted) for

noncatalyst vehicles using leaded-MMT gasoline in 1990 versus catalyst

vehicles using unleaded-MMT gasoline in 1995.

Next, the nonautomotive contribution to PM4 Mn was estimated

and added to the estimated automotive contribution to obtain the

projected personal exposure levels of total PM4 Mn. Assuming the

estimated PM4 Mn distribution has the same form as the PM10

Mn distribution from PTEAM (approximately lognormal with equal

geometric standard deviations), the ratio of the PM4:PM10

arithmetic mean personal exposure levels yields a scaling factor that

can be applied to the PM10 distribution to obtain the PM4

distributions. Because of the alternative bases for adjusting the SIM

PM2.5 data for personal exposures (as noted above), two different

scaling factors were multiplied by the PM10 distribution, thereby

producing a higher and a lower estimate of the distribution of 24-hour

average PM4 Mn personal exposure levels.

In addition, because long-term exposures are likely to have less

variance than 24-hour exposures, it is appropriate to adjust the

distributions of 24-hour average exposure levels to better reflect

longer periods of exposure. Of various methods that may be used for

this purpose (Wallace et al., 1994), two approaches were applied to

adjust the geometric standard deviations of the two projected exposure

distributions, based on data from either the PTEAM study or the smaller

pilot study that preceded the PTEAM Riverside study. These alternative

methods were applied to the two estimated distributions of 24-hour

average exposures to yield the distributions of long-term average

PM4 Mn personal exposure levels depicted as lines 1 and 2 in

Figure 2 (only the highest and lowest of the four resulting estimates

are shown). It must be emphasized that these two distributions do not

represent upper and lower bounds, because even higher or lower

estimates could be produced by alternative assumptions and adjustments

of the data. Moreover, if data were available for another time of the

year (e.g., spring in addition to fall), the estimates would not be

season-specific and could possibly be much higher or much lower.

Nevertheless, given the limited available data, lines 1 and 2 in Figure

2 represent two reasonable estimates of the projected long-term

(autumnal) personal exposure levels of PM4 Mn in relation to MMT

usage at \1/32\ g Mn/gal in 100% of unleaded gasoline.

By examination of the logarithmic-probability plot of long-term

personal exposure levels of Mn, it is estimated that half of the

population would be exposed to PM4 Mn levels of more than

approximately 0.045 to 0.050 g/m\3\. Also, based on the two

projection estimates, approximately 5 to 10% of the population would

have personal exposure levels around 0.1 g/m\3\ PM4 Mn or

higher. The highest 1% would be predicted to have PM4 Mn exposure

levels above 0.15 g/m\3\. It should be noted that these

projections refer specifically to Riverside, CA, with a population of

more than 139,000 persons. However, in many significant respects (e.g.,

meteorology and traffic volume), Riverside is reasonably representative

of the greater metropolitan area of Los Angeles, which has a total

population of over 14.5 million persons. The exposure projection

estimates for Riverside imply the possibility that hundreds of

thousands of persons in the Los Angeles area alone could be exposed to

PM4 Mn levels exceeding 0.1 g/m\3\. To the extent that

any other U.S. cities (e.g., in the Southwest) share some degree of

resemblance in meteorology, vehicle miles traveled (VMT), and possibly

other characteristics of relevance to automotive Mn levels, the

estimated exposure levels for Riverside could be pertinent, at least

qualitatively, to other locales or portions of locales as well.

Similarities and differences in point-source contributions to Mn

exposure would also figure into comparisons with other communities. The

presence of a major point source or sources of Mn in a community (which

was not a factor in Riverside) would add some increment to the level of

Mn exposure experienced by the persons in that community. Although

these Riverside estimates cannot be applied quantitatively to any other

U.S. metropolitan areas, the total population of the U.S. counties with

VMT levels greater than that of Riverside (apart from the four counties

Los Angeles comprises) is approximately 15 million persons. Possibly,

then, several hundreds of thousands of persons could be exposed to

PM4 Mn levels of approximately 0.1 g/m\3\ or higher if

MMT were used in 100% of the unleaded gasoline in all of these areas.

However, it must be emphasized that because of the limited available

data, a great deal of uncertainty surrounds such estimates. The actual

exposure levels could be much higher or lower.

BILLING CODE 6560-50-P

TN17AU94.000

BILLING CODE 6560-50-C

D. Risk Characterization

To assess the public health risk associated with the use of MMT in

gasoline in the United States, the available qualitative and

quantitative health effects information on Mn must be related to the

available exposure information. From the standpoint of a qualitative

hazard identification, the available evidence amply demonstrates that

inhaled Mn is toxic to the nervous system, the respiratory system, and

the male reproductive system. The toxicity of Mn by different routes of

exposure has been demonstrated by numerous medical reports and

epidemiological and experimental studies. However, available data do

not allow quantitative estimation of the relative toxicological potency

of different Mn compounds or permit quantitative route-to-route

extrapolations for predicting the effects of Mn3O4.

The focus of the above health assessment discussion has been on the

RfC and the types of risks associated with chronic Mn exposures

because, for the most part, acute effect levels appear to be

considerably higher than the highest projected exposure levels.

However, the issue of less-than-chronic exposures does arise with

respect to the potential for developmental toxicity. It is widely

recognized that the human CNS develops over a period of several years,

prenatally and postnatally, and can be vulnerable to long-term or

irreversible effects if damage occurs during certain ``critical

stages'' of development. Recent evidence from ongoing longitudinal

studies of children indicates that lead (Pb) exposure (measured as

blood Pb level) around 2 years of age in particular is associated with

reduced cognitive performance at 4 to 10 years of age. Such evidence

raises the concern that exposure to another neurotoxic metal such as Mn

during part of early development might also be capable of inducing

permanent or irreversible damage to the developing CNS. Moreover, the

ramifications of such damage might extend to other important functions,

such as reproduction.

Children may also be at higher risk in terms of exposure because of

biomedical and metabolic differences at a young age (greater uptake and

retention) and/or because of the longer duration of their exposure over

a lifetime. Over time, small impairments in neurobehavioral function

may accumulate. For this reason, the elderly, whose neurobehavioral

function may already be compromised by normal aging processes and

possibly by other disease states (e.g., parkinsonism or preclinical

parkinsonism), also represent a special population of concern. The

ability of the elderly or other subpopulations to compensate for such

declines in neurobehavioral function may be overwhelmed eventually by

additional, albeit possibly quite small, insults due to Mn. If so, the

effect could be manifested as a more severe or earlier onset of

declining function in senescence, with consequent implications for

increased societal health-care costs.

Special subpopulations at increased risk may be defined not only in

terms of their biological susceptibility, as exemplified above by the

young and the elderly, but also by their increased risk of exposure. In

this respect, inner city residents and others who live near high

traffic areas such as expressways (e.g., low-income and minority

communities) would possibly have a disproportionate likelihood of

higher Mn exposure levels due to their closer proximity to vehicular

emissions.

The nature of the neurobehavioral effects observed in occupational

studies such as Roels et al. (1992) should be understood as effects

that probably would not be treated by, or even be readily evident to, a

clinical physician. Nonetheless, they are significant from a public

health standpoint when considered in terms of population effects. This

concept is illustrated by the well documented findings on low-level Pb

neurotoxicity in children, where changes of as little as 1 or 2 points

in IQ have been repeatedly demonstrated in several independent,

prospective, epidemiological studies in recent years. Such changes

could not be reliably demonstrated either in a clinical setting or in

earlier cross-sectional epidemiological studies; yet they are now well

established to be ``real'' and significant from a public health

standpoint. With regard to the reductions in neurobehavioral function

observed in various epidemiological studies of Mn-exposed workers,

these studies independently converge on findings of impaired motor

function (e.g., reductions in eye-hand coordination, slower hand or

finger movements, and less control of fine movement). As recently

expressed in a document prepared by the Subcommittee for Risk

Assessment of the Federal Coordinating Council for Science, Engineering

and Technology (Federal Register, 1993), an adverse effect can include

``b

This text is long and has been trimmed here. Open the source document for the complete record.