Amicus Curiae Brief — Gonzales v. PLANNED PARENTHOOD FEDERATION OF AMERICA, INC.

Supreme Court brief2006

Ask Donna

What actually matters in this document.

Text

, = & oe,

IN THE OFFICE OF THE CLERK

Supreme Court of the United States

ALBERTO R. GONZALES, ATTORNEY GENERAL,

) Petitioner,

LEROY CARHART, ET AL.,

Respondents.

ALBERTO R. GONZALES, ATTORNEY GENERAL,

Petitioner,

—_—V—

PLANNED PARENTHOOD FEDERATION OF AMERICA, INC., ET AL.,

Respondents.

ON WRITS OF CERTIORARI TO THE UNITED STATES

COURT OF APPEALS FOR THE EIGHTH CIRCUIT AND THE

UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

BRIEF OF AMICI CURIAE STATISTICIANS

PROFESSOR GEORGE W. COBB, PROFESSOR MARY W. GRAY,

PROFESSOR NORMAN HENDERSON, PROFESSOR JOHN J. MCARDLE,

PROFESSOR JAMES TRUSSELL, AND

PROFESSOR JEFFREY A. WITMER

IN SUPPORT OF RESPONDENTS

MOLLY S. BOAST

Counsel of Record for the Amici

CHRISTIAN R. EVERDELL

AARON BLOOM

JOHN S. CRAIG

DEBEVOISE & PLIMPTON LLP

919 Third Avenue

New York, New York 10022

(212) 909-6000

Counsel for Amici Curiae

CE ne ae. iy ee ee ee ee eee oe

TABLE OF CONTENTS

Page

Fe Se ST vecntnsctrnersnvmustaersnemsmnacecenenanetienae i

DORR Ae FR I CP ccecensestnessssamncenemenercessonsnimmenenee iv

UTR e OR ARI T CUIAB cc cecccececevessvsscosevevevessconsscsossees l

INTRODUCTION AND SUMMARY OF

SITE sisicnstiicidiniiccaicopitninibdtnnenetninatanidiageniinpinniiineiiientieninis 3

SUMMARY OF THE CHASEN STUDY ..........0..cce cece 5

FD pscscincerensisitsinmintidesiinidiniiecidiigebipainininhpimaniianeundes 7

I. | Testimony That Intact D&X Patients In The

Chasen Study Showed An Increased Risk of

Subsequent Preterm Birth Is Inaccurate And

Should Be Disregarded..................ccrecersrsrseesesecsreersreeees 7

A. It Is Widely Accepted In The Scientific

Community That An Observed Difference

Between Two Groups Is Statistically

Significant Only When The P-value Equals

eR RE Ts Seicsinenitesanineyricssiesinbaieaendnnibiatiediiaptintmieniveteiiinns 8

B. The Chasen Study Found No Statistically

Significant Difference In The Incidence Of

Subsequent Preterm Birth Between The Intact

D&X Group And The D&E Group Based On A

Pn Ca a tditconiiiepinicinsiaiasighictnnibtcsisaniiinepitvnmtiidaiitil 1]

Page

C. Dr. Clark’s Statements About The Meaning Of

A P-value Of 0.30 Are Incorrect And Should

A SN IIID, sntrnciecnnnmsevenieaniviptncsniuntesiosiniatenddinabedin 14

D. Dr. Chasen’s Follow-up Study Suggests That

The Risk Of Subsequent Preterm Birth Was

Correlated To The Patient’s Medical Indication

For Abortion, Not To The Procedure She

NIT TUITTTTI cieneehetiices etapa itaae tactile 17

Il. Testimony That The Chasen Study Data

Showed “Trends” Toward Increased Risk In

The Intact D&X Group Is Unfounded And

Contradicted By Additional Data From Dr.

Chasen’s Follow-Up Study. .0.............ccsccsssseesseeeeenenees 20

A. The P-values In The Chasen Study Do Not

Indicate Even Borderline Statistical

III: icisesnictniiestitncasinnintnnctsdideiainineeneiianimeions 21

B. The Numerical Difference In The Incidence Of

Subsequent Preterm Birth Between The Intact

D&X And D&E Groups Decreased In Dr.

Chasen’s Follow-up Study. .0..........ccceeeseeeeeeeeeeees 24

III. Courts Require A P-value Of 0.05 Or Less, Or

Its Statistical Equivalent, Before Concluding

That An Observed Difference Is Statistically

RRL EARS SI es NTE I 26

EIU ‘occ hinnscnsvisennintiintaiitatatninanvennineniinitientitiniini 29

APPENDIX A

APPENDIX B

iii

Page

NNT Ni oRO Sa SO TOOE TRADE OO TOO UNEP DOTS A-|

Sncetieisiciehenpinieibibdedbibitinadtstatiiaiiiatiiinainactitaieitan dee iis B-1

TABLE OF AUTHORITIES

FEDERAL CASES

Benson v. Tocco, Inc.,

SED FDO CAD COSI CI, FOF Psccecsncsssicnscsaccnchsstnsenenisiiadls

Carhart v. Ashcroft,

331 F. Supp. 2d 805 (D. Neb. 2004) 0... ccceecceeeeeeneeeeee

Castaneda v. Partida,

GA Ge CEB Pcccsccsvincsnsensessenccinniiaiiieiidiiliatiia

Eldredge v. Carpenters 46 Northern California Counties

Joint Apprenticeship & Training Committee,

SOD Fe CIIO CHUNG. TOF Raccesnvinsescancstnisnsrcniogutoii oedns

Ford v. Seabold,

BSS FE GT GOR Ge, FoR ccwcontcenisecccstnvcsnsnctsensseivisiipiiian

Hazelwood School District v. United States,

GSD WB. BO CIT F Poceccsanscoccsccsntsnssscnvintenssisescinaniniinitia

National Abortion Federation v. Ashcroft,

330 F. Supp. 2d 436 (S.D.N.Y. 2004)... cccccecseeseeseeeeeees

National Abortion Federation v. Gonzales,

4ST FFG SFO CRBC, Be rccccessecsescsenssesssniivivnsincsnioniiiatie

Planned Parenthood Federation of America v. Ashcroft,

320 F. Supp. 2d 957 (N.D. Cal. 2004) 000.0... cecccceseeeeeeeeeenes

pescneaee 27

Page

Segar v. Smith,

Ss GEE WIED axccrccencsceccnsecevenesnesentneseevenenecntenss 28

Smith v. Xerox Corp.,

IEEE, (SIU arnesennsssesscesenssnnsanecsensensnatsemveenecsennense 27

MISCELLANEOUS

David H. Kaye & David A. Freedman, Reference Guide

on Statistics, in Federal Judicial Centcr, Reference

Manual on Scientific Evidence 83 (2d ed. 2000) ..........cc000006: 1

Stephen T. Chasen et al., Dilation and evacuation at > 20

weeks: comparison of operative techniques,

190 Am. J. Obstet. Gynecol. 1180 (2004) ..0.....ccccececececeee passim

Stephen T. Chasen et a/., Obstetric outcomes after

surgical abortion at > 20 weeks’ gestation,

193 Am. J. Obstet. Gynecol. 1161 (2005) .........:ccccceeeeeeeee passim

INTEREST OF AMICI CURIAE'

Amici curiae are professors at prominent universities and

colleges, as well as distinguished statisticians in their fields.

Even with the breadth of their professional interests, amici are

united in their belief that statistical analysis used in judicial

decision-making must rest on sound scientific pfinciples.

Amici are aware that the interpretation of scientific data often

plays a role in courts’ adjudication of facts and law. Absent

professional guidance, courts may misapply statistical

principles and render decisions based, at least in part, on

invalid or unsupportable assumptions.

Amici believe that portions of the record below reflect a

misapplication of sound statistical methodology. Misleading

and incorrect statements by some of the government’s expert

witnesses were relied upon by two of the district courts, and

have been perpetuated in briefs submitted to this Court by

Petitioner and supporting amici. Amici here believe that

correcting those misleading and incorrect assertions, and

providing an accurate interpretation of the data upon which

those assertions purport to rely, will contribute to the Court’s

understanding and help it reach a_ well-reasoned,

scientifically-principled decision in these highly important

appeals.

' Pursuant to Supreme Court Rule 37, amici curiae state that

this brief has not been authored in whole or in part by counsel for a

party in this case, and no entity other than the amici or their counsel

made a monetary contribution to the preparation or submission of

this bnef. Letters of consent to the filing of this brief have been

lodged with the Clerk.

Amici curiae include the following:

George W. Cobb, Ph.D.

Robert L. Rooke Professor of Mathematics and Statistics

Mount Holyoke College

Mar, W. Gray, Ph.D.

Professor, Department of Mathematics and Statistics

American University

Norman Henderson, Ph.D.

Professor (Emeritus), Department of Psychology _°

Oberlin College

John J. McArdle, Ph.D.

Professor, Department of Psychology

Director, Quantitative Area

University of Southern California

James Trussell, Ph.D.

Professor, Economics and Public Affairs

Director, Office of Population Research

Princeton University

Jeffrey A. Witmer, Ph.D.

Professor, Department of Mathematics

Oberlin College

INTRODUCTION AND SUMMARY OF ARGUMENT

The relative safety of intact dilation and extraction

(“intact D&X”) compared to dilation and evacuation by

disarticulation (“D&E”) is a critical question in the scientific

debate surrounding the Partial-Birth Abortion Ban Act of

2003, which all parties agree prohibits, at a minimum, intact

D&X. At the center of this debate is a peer-reviewed study

by Dr. Stephen T. Chasen and colleagues, which compared

the complication rates of intact D&X and D&E in a group of

383 women who received these procedures at The New York

Weill Cornell Medical Center. The study authors found the

comparative complication rates of the two procedures to be

similar, suggesting that women undergoing intact D&X face

no increased risk compared to those undergoing D&E.

Stephen T. Chasen ef al., Dilation and evacuation at > 20

weeks: comparison of operative techniques, 190 Am. J.

Obstet. Gynecol. 1180 (2004) (the “Chasen study”) (attached

as Appendix A).”

The Chasen study was the subject of extensive expert

witness testimony in the trial court records in Nebraska, New

York, and California, where several of the government’s

experts incorrectly asserted, with varying degrees of

certitude, that portions of the data show that intact D&X may

be riskier than D&E. Amici believe those statements were

misleading because the numerical differences upon which

they were based are not “significant” as measured by

* Amici use the term “intact D&X,” as opposed to “intact D&E”

or simply “D&X,” to reflect the term Dr. Chasen used in his

published study. See Chasen, Dilation and evacuation, supra, at

1180-83.

accepted methods of statistical analysis. Those

misstatements were incorrectly relied upon by some of the

lower courts and now reappear in briefs submitted to this

Court. Amici believe it is critical to ensure that the Court has

a scientifically accurate understanding of the Chasen study

and the conclusions about the relative risk of intact D&X and

D&E that can, and cannot, validly be drawn from its data.

At the root of the government experts’ misstatements is

the meaning of the statistics! concept of probability value (or

“p-value”), which measures the statistical significance of

observed differences in outcomes among members of a group.

Amici therefore seek to do the following: (1) provide the

Court with an accurate description of the concept of p-value;

(2) explain why the relevant p-values in the Chasen study are

not statistically significant and therefore do not support the

assertions of the government's experts that intact D&X may

be more risky than D&E; and (3) highlight additional

statistical evidence from a follow-up study published by

Dr. Chasen that further undermines those experts’ assertions.

SUMMARY OF THE CHASEN STUDY

The objective of the Chasen study, published in the

peer-reviewed American Journal of Obstetrics and

Gynecology in 2004, was “to evaluate the relative safety of

dilation and evacuation and intact D&X in_ patients

undergoing surgical abortion late in the second trimester.”

Chasen, Dilation and evacuation, supra, at 1181. Prior to the

publication of the Chasen study, no data had been formally

published regarding the complication rates of intact D&X.

Dr. Chasen sought to collect data on the complications that

resulted from intact D&X and D&E in order to assess

whether intact D&X posed a greater health risk than D&E.

Dr. Chasen’s study sample included a group of 383

women who underwent either intact D&X or D&E at twenty

weeks’ gestation or greater at The New York Weill Cornell

Medical Center between June 1996 and June 2003.

Dr. Chasen retrospectively identified cases for inclusion in

the study sample by reviewing operative reports to determine

the extraction technique used. Of the 383 women included in

the study, 120 (31.3%) received intact D&X and 263 (68.7%)

received D&E. Dr. Chasen used this sample to assess

whether there was any difference between the two surgical

variations in the number or severity of operative and

post-operative complications. “Complications” was defined

in the study as “any situation requiring unplanned

intervention,” including “unplanned hospital admission,

repair of any genital tract lacerations, return to the operating

room for additional procedures, and blood transfusion.” /d.

at 1181.

Dr. Chasen also identified a smaller sample of 62

women from among the larger group of 383 who later

became pregnant and sought prenatal care at The New York

Weill Cornell Medical Center. Of those 62 women, 17 had

originally undergone intact D&X and 45 had originally

undergone D&E. Dr. Chasen used this sample to assess

whether there was any difference between the two surgical

variations in subsequent adverse pregnancy outcomes.

Specifically, Dr. Chasen examined the incidence of

spontaneous preterm birth, which occurs when a woman

unexpectedly delivers prematurely in the third trimester

before the fetus has been carried to term.

Dr. Chasen found that operative and post-operative

complications “occurred with similar frequency” in the two

patient groups—5.0% (6 out of 120) of the intact D&X group

compared to 4.9% (13 out of 263) of the D&E group. /d. at

1182. With regard to subsequent pregnancy outcomes,

Dr. Chasen found that 11.8% (2 out of 17) of the intact D&X

group and 4.4% (2 out of 45) of the D&E group experienced

spontaneous preterm birth in a later pregnancy. The

percentage difference in the incidence of subsequent preterm

birth, however, was not statistically significant when

measured by a p-value calculation, the established method of

assessing the statistical significance of different outcomes.

Specifically, the p-value calculation of 0.30 was far greater

than 0.05, the threshold at or below which an observed

difference is deemed statistically significant. Dr. Chasen

ultimately concluded, based on these data, that “intact D&X

appears to have similar complication rates as dilation and

evacuation” and that “[t]he observed complication rates and

subsequent obstetric outcomes appear comparable between

the 2 techniques.” /d. at 1183.

ARGUMENT

I. Testimony That Intact D&X Patients In The Chasen

Study Showed An Increased Risk of Subsequent

Preterm Birth Is Inaccurate And Should Be

Disregarded.

In his testimony about the Chasen study before Judge

Casey in the Southern District of New York, one of the

government's expert witnesses, Dr. Steven L. Clark, ascribed

a false significance to the numerical difference in the

incidence of subsequent preterm birth in the intact D&X

group versus the D&E group. Dr. Clark did this in two ways.

First, he asserted that patients who received intact D&X

showed a “threefold increased risk of premature birth,” N.Y.

Tr. 2386:4-2386:13, even though the numerical difference

between the two groups was not-statistically significant. One

cannot validly conclude from these data that patients

undergoing intact D&X have an increased risk of subsequent

preterm birth compared to patients undergoing D&E.

Second, Dr. Clark compounded his error by asserting that the

0.30 p-value in the Chasen study indicated a 70% chance that

there was in fact a true difference between intact D&X and

D&E in the likelihood of causing subsequent preterm births.

N.Y. Tr. 2425:15-2426:18. Such an assertion fundamentally

misstates the conclusions that can be drawn from a p-value

calculation. A p-value helps one evaluate whether an

observed difference is statistically significant; it does not

indicate the probability that there is a true difference between

two groups.

Dr. Clark's misstatements about the Chasen study data,

p-value calculations, and statistical significance persist in the

record before this Court. For example, although Judge Casey

struck down the Partial-Birth Abortion Ban Act, he

mistakenly endorsed Dr. Clark’s testimony in his opinion.

See Nat'l Abortion Fed’n (“NAF”) v. Ashcroft, 330 F. Supp.

2d 436, 476-77 (S.D.N.Y. 2004). The Solicitor General, in

his brief to this Court in Gonzales v. Carhart, reiterates Dr.

Clark’s misleading assertion that the Chasen study “showed a

nearly threefold increase in the risk of premature birth” in

women who had undergone intact D&X, without mentioning

the lack of statistical significance of those data. Petr.’s Br.

(Carhart) 38 n.12. Similarly, the amicus brief filed by the

American Association of Pro-Life Obstetricians and

Gynecologists (AAPLOG) ef al. cites Dr. Clark’s testimony

about the supposed “threefold increased risk of premature

birth” as evidence of the “possible dangers of D&X.” Br. of

Amici Curiae AAPLOG et al. 19-20.

Because these assertions disregard well-settled

Statistical principles for interpreting data, or worse,

fundamentally misstate those principles, amici seek to

provide the Court with an accurate explanation of the

meaning of a p-value calculation and to explain from a

Statistical standpoint why the Chasen study data do not

support the conclusion that intact D&X causes more

complications than D&E in subsequent pregnancies.

A. It Is Widely Accepted In The Scientific

Community That An Observed Difference

Between Two Groups Is Statistically Significant

Only When The P-value Equals 0.05 Or Less.

The Chasen study was designed to evaluate the relative

safety of intact D&X and D&E by examining data from a

sample of patients who received one or the other of these two

surgical variations. In any study that seeks to draw a

conclusion about the population as a whole from a smaller

sample, it is necessary to determine whether any differences

observed in the sample are “statistically significant.” If the

differences are not “statistically significant,” it is more likely

that they reflect variation due to chance. Only when

observed differences have been determined to be

“statistically significant” may these differences validly be

attributed to the general population. Statisticians perform

this analysis by calculating a probability value (or “p-value’’),

which accounts for factors such as a small sample size and a

small number of events that might skew the observed data

and suggest a false difference.

To use a straightforward example, suppose we wish to

compare how often a certain penny and a certain nickel turn

up “heads.” We begin by forming what is called a “null

hypothesis”"—a hypothesis that there is no difference

between the frequency with which the penny will turn up

heads and the frequency with which the nickel will turn up

heads. We then test the null hypothesis by flipping each coin

repeatedly and counting the number of times that each coin

turns up heads. Suppose we flip each coin 5 times and find

that the penny turns up heads in 4 of the 5 flips (or 80% of

the time) and the nickel turns up heads in 2 of the 5 flips (or

40% of the time). Even though the observed frequencies

clearly differ numerically, it would be a mistake to reject the

null hypothesis based on these limited data and conclude that

the penny is twice as likely as the nickel to turn up heads.

The number of flips is too small for us to question seriously

the null hypothesis. It is entirely plausible that we would

observe 4 out of 5 heads when tossing a “fair” penny and 2

out of 5 heads when tossing a “fair” nickel simply by chance.

In contrast, we would feel more confident about rejecting the

null hypothesis, and concluding that the penny and the nickel

are in fact different, if we flipped the coins 5,000 times each

and found that the penny turned up heads 4,000 times while

the nickel turned up heads 2,000 times.

Statisticians formally compare observed data with the

null hypothesis by calculating a p-value; ultimately, it is the

p-value that determines whether the null hypothesis will be

rejected. Statisticians calculate a p-value by first assuming

that the null hypothesis is true—that there is actually no

difference in the frequency with which the penny will turn up

heads and the frequency with which the nickel will turn up

heads. Given this assumption, a p-value then measures the

probability that we would see a difference in the sample

similar to or greater than the difference we actually observed.

To return to the coin flip example, a p-value measures the

likelihood that the difference we observed—that is, the

difference in the number of heads that turned up for the

penny versus the nickel—or an even greater difference,

would in fact occur, assuming there was no true difference

between them. In doing so, a p-value accounts for the

distortion caused by the small number of flips for each coin.

It is generally accepted within the scientific community

that an observed difference between two outcomes is

significant only if the p-value is less than or equal to 0.05.°

A p-value at or below the 0.05 threshold indicates that if

there is no true difference in the population, the likelihood of

observing the difference we observed in our sample is

small—only 5%. If the probability of observing such a

* A p-value of 0.05 is the broadest cut-off for statistical

significance recognized by the scientific community. Designation of

more restrictive p-value thresholds of 0.01 or even 0.001 as

indicators of stronger statistical significance is also generally

accepted.

difference is that low or lower, it is very unlikely that the

observed difference can be explained by chance. At that

point, statisticians feel comfortable rejecting the null

hypothesis and concluding that there is a “statistically

significant” difference between the two outcomes. A p-value

greater than 0.05, on the other hand, indicates sufftcient

possibility that the observed difference was due to chance

that it would’ be inappropriate to reject the null hypothesis

and conclude that the difference was significant. This

statistical convention is well-settled and uncontroversial. See

David H. Kaye & David A. Freedman, Reference Guide on

Statistics, in Fed. Judicial Ctr., Reference Manual on

Scientific Evidence 83, 121-125 (2d ed. 2000). Indeed, the

government’s experts, including Dr. Clark, did not contest

this point in their testimony. See N.Y. Tr. 2425:15-2425:17

(Clark); N.Y. Tr. 2197:8-2197:10 (Sprang); Neb. Tr.

1721:18-1722:8 (Lockwood).

In the context of the Chasen study, p-value calculations

help to assess whether any differences between intact D&”

and D&E observed in the study sample were likely due to

chance or were significant enough for one validly to conclude

that a similar difference would likely occur in the general

population of all women undergoing intact D&X and D&E.

Only a p-value of 0.05 or less would support the latter

conclusion.

B. The Chasen Study Found No Statistically

Significant Difference In The Incidence Of

Subsequent Preterm Birth Between The Intact

D&X Group And The D&E Group Based On A

P-value Of 0.30.

Because the Chasen study data pertaining to differences

in subsequent preterm birth rates between the intact D&X

and D&E groups is not statistically significant, it is

inaccurate to conclude, as Dr. Clark did, that intact D&X

showed a “threefold increased risk” in subsequent preterm

birth compared to D&E. N.Y. Tr. 2386:10. Dr. Clark

focused solely on the numerical difference in the percentage

of patients who experienced subsequent preterm births in the

intact D&X group compared to the D&E group. He did not

take into account the effect of the small number of events and

the relatively small sample size that, in this case, renders the

numerical difference statistically inconsequential. In sum,

Dr. Clark’s assessment of the data disregards the importance

of p-value.

To be sure, the raw numbers show that the patients who

received intact D&X experienced a higher rate of subsequent

preterm births than those who received D&E. Out of a total

of four preterm births, two occurred in each patient group.

This translates to an event rate of 11.8% (2 out of 17

patients) in the intact D&X group versus 4.4% (2 out of 45

patients) in the D&E group, a roughly threefold difference.

The numerical comparison of these event rates, however, is -

only the beginning of the analysis. To draw any valid

scientific conclusion from these data about the relative risk of

preterm birth from undergoing intact D&X versus D&E, it is

necessary to evaluate how likely it is that the observed

difference between the two groups is an _ accurate

representation of what would occur in the broader population

of women undergoing these procedures. In other words, one

must determine whether the threefold numerical difference

observed in the Chasen study would likely replicate itself in a

larger sample, or whether, as in the coin flip example, the

observed difference is easily explained by chance.

Dr. Chasen appropriately considered this question in his

study by calculating a p-value to determine whether the

observed difference in preterm births was significant. The

resulting p-value of 0.30 falls well above the recognized

threshold for statistical significance of 0.05 or less, indicating

that there is enough of a possibility that the difference was

caused by chance that it would be invalid to conclude that

there is increased risk for the intact D&X group. Dr. Chasen

correctly determined that the difference was insignificant and

thus concluded that the “subsequent obstetric outcomes

appear comparable between the 2 techniques.” Chasen,

Dilation and evacuation, supra, at 1183.

Dr. Charles Lockwood, the Chairman of the

Department of Obstetrics, Gynecology and Reproductive

Sciences at Yale University and one of the government's

expert witnesses, agreed that these data lacked significance

and could not validly support a conclusion that intact D&X

subjected women to increased risk in later pregnancies. With

regard to the difference in subsequent preterm births, Dr.

Lockwood testified: .

It’s not statistically significant, so | wouldn’t. . .

put an enormous amount of weight on it... .

[T}here were additional risk factors in the D&X

group for subsequent premature delivery. .. .

{FJormally as a clinician-researcher, I wouldn't

draw any conclusions from it.

Neb. Tr. 1721:23-1722:4; see also Cal. Tr. 971:10-971:13

(testimony of government's expert witness, Dr. Watson A.

Bowes, Jr., that the Chasen study “certainly does not Suggest

that D&E where the fetus is removed intact is any less safe

than where the [fetus] is disarticulated”’).

The conclusions of Dr. Chasen and Dr. Lockwood that

the data do not indicate an increased risk of subsequent

preterm birth in the intact D&X group are based on an

accurate understanding of p-value and are consistent with

accepted methods of statistical analysis. These conclusions

were also endorsed by the district courts in Nebraska and

California. See Carhart v. Ashcroft, 331 F. Supp. 2d 805,

1022-23 & n.150 (D. Neb. 2004) (crediting the testimony of

Dr. Lockwood over that of Dr. Clark about the risk of

subsequent preterm birth); Planned Parenthood Fed'n of Am.

v. Ashcroft, 320 F. Supp. 2d 957, 1001 (N.D. Cal. 2004)

(finding that the Chasen study does not support a conclusion

that intact D&X may lead to an increased risk of subsequent

preterm birth).

Amici urge this Court to disregard the testimony of Dr.

Clark about the relative risk of preterm birth in subsequent

pregnancies for intact D&X and D&E, and similarly to reject

any arguments by the Solicitor General or supporting amici

based on that testimony.

C. Dr. Clark’s Statements About The Meaning Of A

P-value Of 0.30 Are Incorrect And Should Be

Disregarded.

In addition to drawing a statistically invalid conclusion

about the Chasen study data, Dr. Clark made two inaccurate

statements about the meaning of p-value in his testimony that

are not in accord with accepted principles in the scientific

community.

First, Dr. Clark implied that p-values of 0.30 and 0.05

were substantively similar. In his testimony about the

difference in the incidence of spontaneous preterm birth in

the two patient groups, Dr. Clark conceded that the result

would be statistically significant only if the p-value were less

than 0.05, but then erroneously suggested that the 0.30

p-value in the Chasen study was roughly equivalent to a p-

value of 0.05:

Well, I would say that if the P value is less

than .05 then there is absolutely no question.

Usually we don’t accept something as absolute

scientific truth unless the chances are less than

five percent. Jn this particular case it’s just

stretching it a little bit to say 30 percent

chance, but you’ re roughly correct.

N.Y. Tr. 2426:2-2426:8 (emphasis added). Contrary to Dr.

Clark’s statement, a p-value of 0.30 does not even approach

the level of statistical significance indicated by a p-value of

0.05. A p-value of 0.30 means that, if intact D&X and D&E

truly posed the same risk of subsequent preterm birth and we

conducted the Chasen study multiple times, we would expect

to observe the same (or greater) difference in subsequent

preterm births that we observed in the Chasen study 30% of

the time. Given that we would observe the same (or greater)

difference relatively frequently—almost one third of the

time—it would be inappropriate to reject our null hypothesis

that intact D&X and D&E are similar. Only when the

likelihood of observing that difference is small—5% percent

or less—can we validly conclude that the difference we

observed is significant. Dr. Clark’s assertion that a p-value

of 0.30 can be “stretched” to equate to a p-value of 0.05 is

simply incorrect.

Second, Dr. Clark erroneously testified that | minus

0.30, or 0.70 (70%), could measure whether there was an

actual difference between intact D&X and D&E in the risk of

subsequent preterm birth:

{T}here is a 30 percent chance this occurred by

chance and a 70 percent chance that it in fact is

a true, meaningful, increased risk . . . .

N.Y. Tr. 2429:20-2429:22: see also id. at 2426:12-2426:13

(“[T}here is a 70 percent chance that in fact this is a real

difference.”). This misstatement was subsequently adopted

in the district court’s opinion in NAF. See NAF, 330 F. Supp.

2d at 476-77.

Dr. Clark’s statements fundamentally misconstrue the

meaning of p-value. A p-value is not the probability that

there is no real difference between two groups, and one

minus a p-value is not the probability that there is an actual

difference between two groups. P-values are calculated

assuming that there is no real difference between the two

groups. A p-value can only tell us the likelihood that an

observed difference between two groups would be repeated

in subsequent studies, assuming the null hypothesis were

true. To use the Chasen study as an example, a p-value of

0.30 means that, if the Chasen study were rerun multiple

times, and there were in fact no true difference in subsequent

obstetric outcomes between intact D&X and D&E, then one

would expect to see differences as large or larger than the

one observed in the Chasen study in 30% of the subsequent

trials and differences smaller than the one observed in the

Chasen study in 70% of the subsequent trials. N.Y. Tr.

2683:5-2683:19 (Howell). The 0.30 p-value does not

measure the percentage likelihood that there is a “real

difference” between intact D&X and D&E that caused the

difference in subsequent preterm birth rates.

Dr. Clark's testimony that the Chasen study indicates a

70% chance of increased risk of subsequent premature birth

following intact D&X is therefore incorrect and should be

rejected.

D. Dr. Chasen’s Follow-up Study Suggests That The

Risk Of Subsequent Preterm Birth Was

Correlated To The Patient’s Medical Indication

For Abortion, Not To The Procedure She

Received.

Dr. Chasen’s follow-up study, published in 2005 in the

American Journal of Obstetrics and Gynecology, provides

further evidence that the higher percentage of subsequent

preterm births observed in the intact D&X group was not

caused by any heightened risk of the procedure itself, but

instead was related to the medical condition of the woman

seeking the initial abortion. Stephen T. Chasen ef al.,

Obstetric outcomes after surgical abortion at > 20 weeks’

gestation, 193 Am. J. Obstet. Gynecol. 1161 (2005) (the

“Chasen follow-up study”) (attached as Appendix B).

The Chasen follow-up study examined data from the

same patient sample of 383 women in the original Chasen

study, but had a different objective. The Chasen follow-up

study sought to “identify risk factors for spontaneous preterm

birth” and, as a secondary goal, to assess whether intact D&X

or D&E was “associated with higher rates of spontaneous

preterm birth in future pregnancies.” /d. at 1162. With over

a year of additional follow-up data available since the

publication of the original study, the number of subsequent

pregnancies among the 383 women increased from 62 to

120.* Of these 120 subsequent pregnancies, 102 reached the

* Amici in support of Petitioner imply that the addition of 58

pregnancies from the time of the original Chasen study to the time

third trimester and could be evaluated for subsequent preterm

birth rates. Within that sample, spontaneous preterm births

occurred in only 7 of the pregnancies, 3 in the intact D&X

group and 4 in the D&E group.

The Chasen follow-up study identified two factors that

appeared to contribute to the risk of subsequent preterm birth

in the study population, neither of which related to the

abortion procedure the woman had previously undergone:

(1) whether the patient had received an abortion due to

preterm premature rupture of membranes (PPROM) or

spontaneous cervical dilation, medical conditions already

known to carry a heightened risk for spontaneous preterm

birth in subsequent pregnancies, and (2) whether the patient’s

subsequent pregnancy involved multiple fetuses, such as

twins or triplets, which typically deliver prematurely. /d. at

1162-63.

Dr. Chasen found that 27.3% (3 out of 11) of the

women in the study population who underwent abortion due

to PPROM or spontaneous cervical dilation experienced

subsequent preterm births, compared to 4.4% (4 out of 91) of

of the follow-up study is suspect. See Br. of Amici Curiae

AAPLOG ef al. 25 n.91 (the Chasen follow-up study used the

dataset from the original study, “but somehow [found] twice as

many prenatal patients”). This contention has no basis. The original

Chasen study contained data on 62 subsequent pregnancies in 62

women; in other words, one subsequent pregnancy per woman. The

follow-up study, on the other hand, contained data from 120

pregnancies in 89 women; in other words, it included women who

had multiple subsequent pregnancies. Hence, while the number of

subsequent pregnancies in the sample nearly doubled, the number of

prenatal patients did not.

the women who underwent abortion for other reasons. That

difference was statistically significant with a p-value of 0.03.

Dr. Chasen also found that women who received an abortion

due to PPROM or spontaneous cervical dilation made up a

significantly greater percentage of the patients who had

received intact D&X compared to those who received D&E

(25% versus 4.3%, p-value = .001), and that the rates of

subsequent preterm birth among patients who had undergone

the earlier abortion for other medical reasons “were nearly

identical” among those who had undergone intact D&X and

those who had undergone D&E (4.2% versus 4.5%). Id. at

1163.

From these data, Dr. Chasen concluded that “[tJhe

association between surgical abortion and _ subsequent

spontaneous preterm birth in these women is related to the

indication for abortion, rather than the abortion procedure

itself’ and that “[sJurgical abortion with either variation of

D&E late in the second trimester should not be considered a

risk factor for subsequent spontaneous preterm birth.” /d.

Although Dr. Chasen conceded that the sample size of his

follow-up study was too small to rule out the possibility that

either intact D&X or D&E may increase the risk of

subsequent preterm birth, the study provides concrete data

supporting an alternative explanation for the numerical

difference in the incidence of preterm birth between the

intact D&X and D&E groups.

Amici believe the results of the-follow-up study support

Dr. Chasen’s explanation for the difference in preterm birth

rates observed between the intact D&X and D&E groups.

The follow-up study’s statistical evidence is certainly more

scientifically compelling than the speculative assertion

offered by Dr. Clark in his testimony that the numerical

difference may have resulted from the greater cervical

dilation that occurs with intact D&X. N.Y. Tr. 2386:12. The

district court in Carhart, which considered both explanations

in connection with the original Chasen study, concurred. See

Carhart, 331 F. Supp. 2d at 963-64, 1022-23; see also Cal.

Tr. 971:10-971:13 (Bowes) (agreeing that the Chasen study

does not suggest that intact D&X is “any less safe” than

D&E).

Accordingly, this Court should not credit Dr. Clark’s

speculative causation hypothesis or the arguments based on

“that speculation put forth by the Solicitor General and

supporting amici. See Petr.’s Br. (Carhart) 37-38 n.12; Br.

of Amici Curiae AAPLOG et al. 19-20.

II. Testimony That The Chasen Study Data Showed

“Trends” Toward Increased Risk In The Intact

D&X Group Is Unfounded And Contradicted By

Additional Data From Dr. Chasen’s Follow-Up

Study.

Unlike Dr. Clark, the government’s other expert

witnesses did not assert that the numerical increase in the

incidence of subsequent preterm birth in the intact D&X

group suggested an actual difference between intact D&X

- and D&E. Instead, Dr. M. LeRoy Sprang and other

government experts made a similarly invalid claim by

testifying that the numerical increase, while not statistically

significant, represented a “trend” that was cause for concern.

Neb. Tr. 1158:8-1159:10 (Sprang); N.Y. Tr. 2122:8-2122:21

(Sprang); Cal. Tr. 1106:9-1107:15 (Sprang); N.Y. Tr.

2547:4-2547:8 (Cook). Dr. Sprang also mistakenly identified

another purported “trend” in the data on cervical lacerations

and all genital trae: lacerations, which showed a roughly

threefold and twofold numerical increase, respectively, in the

—

intact D&X group. Neb. Tr. 1159:15-1160:2; N.Y. Tr.

2122:22-2123:16, 2124:22-2125:8; Cal. Tr. 1103:21-1104:6.

According to Dr. Sprang, both of these “trends” “would

support that D&X has more risk than D&E.” Neb. Tr.

1160:8-1160:11.

Dr. Sprang’s assertions, like those of Dr. Clark, are not

grounded in accepted principles of statistical analysis and

should be disregarded, as should the portions of Judge

Straub’s dissent in NAF and the government's supporting

amicus briefs that credit Dr. Sprang’s testimony. See Nat’]

Abortion Fed’n v. Gonzales, 437 F.3d 278, 309 (2d Cir.

2006) (Straub, J., dissenting); Br. of Amici Curiae

Congressman Ron Paul ef al. 9-10; Br. of Amici Curiae

AAPLOG et al. 20 n.73.

A. The P-values In The Chasen Study Do Not

Indicate Even Borderline Statistical Significance.

As explained above, statisticians require a p-value of

0.05 or less before concluding that an observed difference is

Statistically significant. As p-value gets closer to 0.05, it

becomes more difficult to attribute the observed difference to

chance. Where the p-value falls just shy of that mark,

Statisticians may assign a borderline or marginal significance

to the observed difference. Generally, only a p-value falling

between 0.05 and 0.10 qualifies as marginally significant.

Thus, if the p-value approaches 0.05, but docs not meet or

fall below that threshold, statisticians will not reject the null

hypothesis by concluding that the observed difference is

Statistically significant, but they might determine that the null

hypothesis should be tested further. Outside this p-value

range, there is still enough likelihood that the difference was

caused by chance that it would be invalid to conclude that the

observed difference is even marginally significant, or, to use

Dr. Sprang’s terminology, that there is a “trend.”

In the Chasen study, the observed difference in the

incidence of spontaneous preterm birth between the intact

D&X and D&E groups was 11.8% versus 4.4%, with a

p-value of 0.30. That measure falls well outside the range of

marginal significance. Dr. Chasen did not calculate p-values

for the operative and post-operative complications he

measured, which included cervical lacerations and all genital

tract lacerations (cervical, perineal, and labial), but he did

report the number of those complications that occurred in

each treatment group. Dr. Chasen found that 3 patients in the

intact D&X group and 2 patients in the D&E group, or 2.5%

versus 0.8%, experienced cervical lacerations during the

procedure. Similarly, 4 patients in both the intact D&X and

D&E groups, or 3.3% versus 1.5%, experienced genital tract

lacerations. Chasen, Dilation and evacuation, supra, at

1182. The p-values for these differences are 0.18 and 0.26,

respectively. Both fall well outside the 0.05 to 0.10 range

and therefore do not support a finding of even marginal

statistical significance.

Dr. Sprang’s own testimony identified the very reason

why it is inappropriate to draw any conclusions from these

* As a technical matter, the use of the term “trend” in this

context is inappropriate. The statistical term “trend” generally refers

to changes that occur over time; for example, an increase in the

unemployment rate each year for 10 years accurately can be termed

a trend. In contrast, the Chasen study contains no indication that the

rate of complications from D&E or imtact D&X increases or

decreases over time. A true test of trend in results requires a

different statistical test than was performed in this study.

numbers: “[W]hen you look at [the data] because the

numbers were so small it’s hard to draw meaningful

conclusions.” N.Y. Tr. 2121:16-2121:17. It is precisely

because so few of these events occurred in the study, a factor

that the p-value calculations of 0.30, 0.18 and 0.26 take into

account, that it is invalid to conclude that these differences

are Statistically significant or even marginally significant.

That is, there are not enough data in the study on spontaneous

preterm births, cervical lacerations, or genital tract

lacerations to conclude for any of these events that a “trend”

exists in the intact D&X group. Indeed, the government’s

own experts could not agree on which of these differences

indicated a “trend” and which did not. Compare Neb. Tr.

1158:8-1160:11; N.Y. Tr. 2122:8-2125:8 (Sprang) (finding

“trends” in all three events), with N.Y. Tr. 2419: 15-2419:24,

2416:4-2416:13 (Clark) (stating that intact D&X posed

“significant hazards to future pregnancies,” but that intact

D&X and D&E showed “no difference in any short-term

outcome”), and Neb. Tr. 1719:16-1722:8 (Lockwood)

(testifying that he “wouldn’t draw any conclusions” from the

difference in spontaneous preterm births and that “it would

be extraordinarily unlikely that these two procedures have

markedly different occurrences in the rate of . . . short-term

complications.”).°

° Furthermore, all three of the serious complications in the

Chasen study that required admission of the patient to the surgical

intensive care unit—one amniotic fluid embolus, one pulmonary

embolus, and one uterine perforation—occurred in the D&E group.

See Chasen, Dilation and evacuation, supra, at 1182. By

Dr. Sprang’s logic, these data would indicate a “trend” against D&E

for all serious complications, despite the obvious paucity of data.

Amici do not mean to imply that there would be no

benefit in further examining the complications or subsequent

obstetric outcomes of intact D&X and D&E. The data from

the Chasen study taken alone, however, simply do not

suggest—even marginally—that there is an increased risk of

operative or post-operative complications, or subsequent

preterm birth, among women who receive intact D&X,

compared to those who receive D&E. Amici therefore urge

this Court to disregard the unfounded testimony of

Dr. Sprang and others on the statistical meaning of the

Chasen study data.

B. The Numerical Difference In The Incidence Of

Subsequent Preterm Birth Between The Intact

D&X And D&E Groups Decreased In Dr.

Chasen’s Follow-up Study.

Dr. Chasen’s follow-up study offers additional evidence

that the numerical differences observed in the original

Chasen study are not indicative of true differences in the

subsequent obstetric outcomes of intact D&X and D&E, but

are simply the consequence of small numbers. As noted

above, the follow-up study reported data from an additional

58 subsequent pregnancies from within the original study

population of 383 women. Chasen, Obstetric outcomes,

supra, at 1162. These additional data nearly doubled the

number of subsequent pregnancies that could be used to

measure the incidence of spontaneous preterm birth, up from

62 in the original Chasen study to 120 in the follow-up study.

Dr. Chasen found that, with the increased sample size, the

difference in the rate of subsequent preterm births between

the intact D&X and D&E groups narrowed considerably

from 11.8% (2 out of 17) versus 4.4% (2 out of 45) in the

original study to 9.4% (3 out of 32) versus 5.7% (4 out of 70)

in the follow-up study, down from a roughly threefold

numerical difference to a difference of just over one and one

half. /d. at 1163. Similarly, the p-value for this difference

more than doubled from 0.30 in the original study to 0.68 in

the follow-up study, departing even further from the 0.05

threshold for statistical significance. /d.

The Chasen follow-up study illustrates perfectly why it

is invalid to conclude that there is an actual difference

between two outcomes when the p-value does not equal or

fall below 0.05. The data from the follow-up study suggests

that the difference in spontaneous preterm births observed in

the original Chasen study was the result of chance

fluctuations due to small numbers. Adding more subsequent

pregnancies to the sample, similar to adding more coin flips

to our example, caused the observed difference to decrease,

the p-value to increase, and the so-called “trend” to

evaporate. As a result of the Chasen follow-up study, there is

now even less reason to credit the claims of Dr. Sprang.’

’ Amici disagree with the brief of the AAPLOG amici, which

argues that “because the D&X data were excluded when D&X was

performed in the presence of [PPROM] and/or advanced cervical

dilation,” no valid conclusions about preterm birth can be drawn

from the Chasen follow-up study. Br. of Amici Curiae AAPLOG

etal. 25 n.91. The 9.4% and 5.7% figures reported in the study

represent the rates of a// preterm births in intact D&X and D&E

groups, respectively, including patients in those groups who

received an abortion due to PPROM and/or advanced cervical

dilation.

Ill. Courts Require A P-value Of 0.05 Or Less,

Or Its Statistical Equivalent, Before Concluding

That An Observed Difference Is Statistically

Significant. .

Courts analyzing statistical evidence have widely

adopted the principle that a p-value of 0.05 or lessor a

roughly equivalent measure of at least 2-3 standard

deviations between an expected outcome and an observed

outcome, is the appropriate gauge for determining whether

observed differences amiong groups are _ statistically

Significant, as opposed to the result of mere chance.

Accordingly, courts have required a p-value calculation of

0.05 or less, or a minimum of 2-3 standard deviations, before

crediting a claim of actual difference between two groups.”

For example, in analyzing claims of racial

discrimination in the grand jury selection process in

Castaneda v. Partida, 430 U.S. 482 (1977), this Court relied

upon the principle that statistical significance is achieved

where a difference of greater than two or three standard

deviations is observed between the expected outcome and the

* Standard deviations can be used to measure the degree to

which an observed result differs from the expected result, assuming

the null hypothesis were true. Generally speaking, a p-value of 0.05

is equivalent to 1.96 standard deviations. For simplicity’s sake, it is

commonly said that a p-value of 0.05 is analogous to 2 standard

deviations and a p-value of 0.01 is analogous to 3 standard

deviations. As such, an observed outcome of greater than two or

three standard deviations from the expected outcome is generally

required for a finding of statistical significance.

observed outcome. See id. at 497 n.17.” In Castaneda, a

difference of twelve standard deviations between the racial

composition of a randomly selected grand jury pool and the

pool at issue was held sufficient to establish the defendant's

prima facie case of discrimination. See id. at 496; see also

Hazelwood Sch. Dist. v. United States, 433 U.S. 299, 309

n.14, 311 n.17 (1977) (analyzing racial composition of

workforce and repeating Castaneda principle that a

difference of greater than 2—3 standard deviations would cast

doubt on hypothesis that workforce was hired randomly

without regard to race).

The courts of appeal take a similar approach in

analyzing the meaning of statistical evidence in a variety of

contexts. See, e.g., Smith v. Xerox Corp., 196 F.3d 358, 366-

67 (2d Cir. 1999) (requiring p-value of 0.05 or less,

equivalent to 2 standard deviations, to establish statistical

significance of data supporting disparate impact claims);

Benson v. Tocco, Inc., 113 F.3d 1203, 1209 (11th Cir. 1997)

(observing that standard deviation of 3.04 between rate of

termination for employees under age 40 and those over age

40, and standard deviation of 2.66 between rate of

* Amici note that whereas the Castaneda court discussed the

significance of the 2—3 standard deviation threshold in the context

of a binomial distribution of data, the Chasen data are based on

hypergeometric distributions. Although equating standard

deviations to p-values for data based on hypergeometric

distributions is more complicated than for a binomial or a normal

distribution, particularly when the sample size is relatively small,

the principle that an outcome of 2-3 standard deviations is roughly

equivalent to a p-value of 0.05 or less remains the same, subject to

minor fluctuations depending on the distribution and symmetry of

the data.

termination for employees under age 50 and those over age

50, would support an inference of discriminatory

termination); Ford v. Seabold, 841 F.2d 677, 688-89 & n.12

(6th Cir: 1988) (rejecting statistical evidence relating to

composition of grand jury pool because “probability [that

number of nonwhites in jury pools would deviate from the

population] did not reach the 0.05 level, the level which most

Statisticians . . . deem statistically significant” and observing

that “[t]he two or three standard deviation benchmark applied

in Castaneda is essentially equivalent to probability values of

0.05 and 0.01 respectively”); Eldredge v. Carpenters 46 N.

Cal. Counties Joint Apprenticeship & Training Comm., 833

F.2d 1334, 1340 n.8 (9th Cir. 1987) (holding that level of

significance of 0.045 in percentage difference of men and

women admitted to apprenticeship program gave rise to

inference of discrimination); Segar v. Smith, 738 F.2d 1249,

1282-83 (D.C. Cir. 1984) (observing that “social scientists

usually accept a study that achieves statistical significance at

the .05 level’).

Accordingly, this Court should not hesitate to conclude

that the p-value calculations from both Chasen studies of

0.68, 0.30, 0.26, and 0.18—all well above the 0.05

threshold—fail to demonstrate a statistically significant, or

even a marginally significant, difference in the rate of

spontaneous preterm births or cervical/genital tract

lacerations between the intact D&X and D&E groups. Any

suggestion that the numerical differences between the two

groups signifies something more meaningful than mere

chance finds no support in fundamental principles of

Statistical analysis, and is at odds with courts’ established

method of interpreting statistical evidence.

CONCLUSION

The Chasen study and its follow-up are critical pieces

of scientific evidence concerning the safety of intact D&X.

If the Court chooses to consider the Chasen data in its

constitutional analysis of the Partial-Birth Abortion Ban Act

of 2003, amici believe those data should be evaluated

objectively and in accordance with widely accepted

principles of statistical analysis. Straightforward p-value

calculations indicate that there is no evidence in the Chasen

data upon which to base a conclusion that intact D&X is

more dangerous than D&E. The Court should not be

misguided by the unfounded and speculative assertions to the

contrary of Dr. Clark, Dr. Sprang, and other government

experts. For the foregoing reasons, amici urge this Court to

disregard those misstatements, along with the arguments of

Petitioner and supporting amici that perpetuate them, and

instead to evaluate the results of the Chasen data using

principles of interpretation that form the standard relied upon

by the scientific community.

Respectfully submitted.

MOLLY S. BOAST,

Counsel of Record for the Amici

CHRISTIAN R. EVERDELL

AARON BLOOM

JOHN S. CRAIG

DEBEVOISE & PLIMPTON LLP

919 Third Avenue

New York, NY 10022

(212) 909-6000

Counsel for Amici Curiae

August 10, 2006

A-|

APPENDIX A

American Journal of Obstetrics and Gynecology

(2004) 190, 1180-3

American Journal of

Obstetrics and Gynecology

www.elsevier.com/locate/ajog

EDITORS’ CHOICE

Dilation and evacuation at 220 weeks: Com-

parison of operative techniques

Stephen T. Chasen, MD,* Robin B. Kalish, MD,

Meruka Gup a, Jane E. Kaufman, MD, William K.

Rashbaum, MD, Frank A. Chervenak, MD |

Department of Obstetrics and Gynecology, Weill

Medical College of Cornell University, New York, NY

Received for publication October 16, 2003; revised

December 19, 2003; accepted December 22, 2003

KEY WORDS

Abortion

Dilation and evacuation

Objective: The objective of this study is to compare the

relative safety of 2 techniques for surgical abortion late

in the second trimester.

* Reprint requests: Stephen T. Chasen, MD, 525 E 68th St, Room

J130, New York, NY 10021.

E-mail: stchascn@med.cornell.edu

reserved. doi: 10.1016/j.ajog.2003.12.034

A-2

Study design: Retrospective review of patients who

underwent surgical abortion at 220 weeks’ gestation at

our hospital from June 1996 through June 2003. Records

were reviewed to determine whether the technique used

was dilation and evacuation or intact dilation and extrac-

tion. Subsequent pregnancies at our hospital were iden-

tified, and obstetric outcomes were recorded. Categorical

data were compared with Fisher exact test and xy’ anal-

ysis. Continuous data were compared with Mann-Whit-

ney U test.

Results: Three hundred eighty-three patients met inclu-

sion criteria. Intact dilation and extraction was per-

formed in 120 cases, and dilation and evacuation was

used in 263. Intact dilation and extraction was associated

with higher parity, later gestational age, and more pre-

operative cervical dilation. There was no difference in

procedure time or estimated blood loss in the 2 groups.

Complications occurred in 19 cases (5.0%), and occurred

with similar frequency in the 2 groups. We identified 62

subsequent pregnancies. There were no second-trimester

miscarriages. Spontaneous preterm birth occurred in 2 of

17 (11.8%) pregnancies in the intact dilation and extrac-

tion group, compared with 2 of 45 (4.4%) in the dilation

and evacuation group (P = .30).

Conclusion: Outcomes appear similar between patients

undergoing dilation and evacuation ard intact dilation

and extraction after 20 weeks’ gestation. Subsequent

obstetric outcomes are similar between the 2 groups. The

technique for surgical abortion should be determined by

the physician on the basis of intraoperative factors.

In the United States, induced abortion late in the sec-

ond trimester is uncommon. In 1999, 9643 abortions

A-3

were performed at 221 weeks’ gestation, representing

only 1.5% of total abortions reported.' Dilation and

evacuation is the most common method used for second

trimester abortion! and is considered the safest abortion

technique in the second trimester.’*

Dilation and evacuation involves preoperative dilation

of the cervix with osmotic dilators, such as)laminaria,

which are placed in the cervix for 1 or more days before

operative evacuation. All published reports of dilation

and evacuation have described the use of grasping for-

ceps to remove the fetus and placenta. With the use of

forceps, fetal parts are grasped, and the fetus and pla-

centa are disarticulated as they are removed. Low com-

plication rates for this procedure have been reported.**

A variant of dilation and evacuation can be performed

when sufficient cervical dilation is present. In these

cases, the fetus is delivered via breech extraction. If the

fetus is not in the breech presentation, internal podalic

version may be performed. In most cases, after delivery

of the body, the fetal head will become lodged in the

cervix, and cranial decompression with suction must be

performed to complete delivery. This procedure has been

referred to as intact dilation and extraction, or D&X.

There are no published data regarding the frequency or

complication rate of this procedure. Despite this, some

have stated that this procedure poses serious maternal

risks and is less safe than dilation and evacuation.®’

The objective of this study is to evaluate the relative

safety of dilation and evacuation and intact D&X in

patients undergoing surgical abortion late in the second

trimester.

Materials and methods

This study was a review of patients who underwent sur-

gical abortion at 220 weeks’ gestation at the New York

A-4

Weill Cornell Medical Center from June 1996 to June

2003. All procedures included in this study were per-

formed by | of 2 physicians (S.T.C. and W.K.R) who are

skilled in both techniques, dilation and evacuation and

intact D&X. Institutional Review Board approval was

obtained for this study.

Gestational age was confirmed by ultrasound in all

cases. All cases 224 weeks’ gestation were performed

because of fetal demise. Dilation of the cervix was

achieved with insertion of laminaria. Laminaria were

placed preoperatively on 2 consecutive days, unless the

cervix was sufficiently dilated at the initial examination.

Women who went into labor after laminaria placement

and delivered without dilation and evacuation or intact

D&X were not included in this study.

The surgical technique was determined when the

patient was examined under anesthesia, and was based

on cervical dilation and fetal position. Both physicians

performing these procedures used similar criteria in

determining the optimal surgical technique. All cases

were considered complete when, after evacuation of the

fetus and placenta, hemostasis was apparent after sharp

and suction curettage. Blood loss was estimated by the

surgeon. Procedures were performed in an ambulatory

setting, unless patients had been hospitalized before the

procedure for medical or obstetric conditions. The type

of anesthesia used for the procedure, general, regional,

Or intravenous sedation, was at the discretion of the

anesthesiologist. Intraoperative ultrasound guidance was

not routinely used.

Operative reports were reviewed to determine the

technique used in each case. If the fetus was delivered

intact in the breech presentation to the level of the

umbilicus or higher, the procedure was considered an

intact D&X, whether the entire fetus was removed intact

A-5

or decompression of the head was required. In some

cases, the presenting fetal head was well applied to the

cervix and was initially decompressed with suction, fol-

lowed by intact delivery of the fetus. These cases were

also considered intact D&Xs, as disarticulation with

forceps was not required. All other cases, which were

performed with multiple insertions of forceps, were cat-

egorized as dilation and evacuation.

Complications included any situation requiring

unplanned intervention. These included unplanned hos-

pital admission, repair of any genital tract lacerations,

return to the operating room for additional procedures,

and blood transfusion. Subsequent pregnancies at our

hospital were identified from a review of the medical

records. We did not obtain follow-up on patients rou-

tinely, so subsequent pregnancies with prenatal care

obtained elsewhere were not included in this analysis.

Thirty-eight patients in this series, 26 of whom under-

went dilation and evacuation and 12 of whom underwent

intact D&X, were included in a prior publication

describing obstetric outcome after second trimester abor-

tion.®

Demographic characteristics and outcomes were com-

pared on basis of operative technique. Categorical vari-

ables were analyzed with the use of 2-tailed Fisher exact

test and x? analysis where appropriate. Mann-Whitney

U test was used to analyze continuous variables, which

were not normally distributed. Analysis was performed

with SPSS 11.0 (Chicago, Ill). A P value <.05 was con-

sidered significant.

Results

Three hundred eighty-three patients met inclusion cri-

teria. Intact D&X was performed in 120 cases (31.3%),

and dilation and evacuation was used in 263 (68.7%).

A-6

Demographic characteristics of the 2 groups are listed in

Table I. Intact D&X was associated with later gestational

age, higher parity, and younger maternal age.

Indications for surgical abortion are listed in Table II.

Preterm cervical dilation and/or preterm premature rup-

ture of membranes (PPROM) were more common in

those who underwent intact D&X. Abnormal fetal kary-

otype was more common in those who underwent dila-

tion and evacuation.

Intraoperative variables are listed in Table lll. Over-

all, laminaria were used in 96.1% of cases. Laminaria

were less likely to be used in patients undergoing intact

D&X. All 15 patients in whom laminaria were not used

had preterm cervical dilation. Preoperative cervical dila-

tion was greater in those who underwent intact D&X.

Table I Demographic characteristics of patients under-

going dilation and evacuation

D&E D& X

(n= 263) (n=120) P-value

Median maternal age 34 y 32 y .O1*

(range) (16-45) (12-43)

Median parity 0 i .04*

(range) (0-7) (0-5)

Median gestational 21 wks 23 wks <.001*

age atD&E (20-27) (20-25)

(range)

Prior vaginal delivery 81 (30.8%) 50 (41.7%) .05!

Prior cesarean delivery 49 (18.6%) 26 (21.7%) .49'

Multifetal pregnancy 13(4.9%) 8(6.7%) .48'

D & E, Dilation and evacuation.

* Mann Whitney U test;

t Fisher exact test.

A-7

Table II Indication for dilation and evacuation*

Indication D&E D& X

forD&E (n = 263) (n=120) #£=Pvalue'

Abnormal fetal

karyotype 112 (42.6%) 33(27.5%) .005

Structural fetal

abnormality 96 (36.5%) 47 (39.2%) .65

Intrauterine fetal

demise 27 (10.3%) 14(11.7%) .72

Premature cervical

dilation/PPROM 14 (5.3%) 20(16.7%) .00l1

Other 22 (8.4%) 13(10.8%) .45

* Some patients had more than | indication;

T Fisher exact test.

No differences were noted in procedure time or esti-

mated blood loss between the 2 groups.

In the entire group of 383 patients, complications

occurred in 19 cases (5.0%). Complications occurred

with similar frequency in the dilation and evacuation and

intact D&X groups (4.9% vs 5.0%; P > .999). The 6

complications occurring in the intact D&X group

included 4 genital tract lacerations (3 cervical mucosal

lacerations and 1 perineal laceration) noted and suc-

cessfully repaired at the time of surgery; 1 case of exces-

sive bleeding requiring the return to the operating room

on the day of surgery for curettage; and | patient read-

mitted because of heavy bleeding on postoperative day

13 who underwent curettage for retained placenta.

Complications in patients undergoing dilation and

evacuation included 4 genital tract lacerations (2 cervi-

cal lacerations and 2 labial lacerations) noted and suc-

cessfully repaired at the time of surgery; 2 cases of

A-8

excessive bleeding requiring the return to the operating

room on the day of surgery (for curettage in | case, and

repair of cervical lacerations in | case); 2 postoperative

admissions for intravenous antibiotics because of

endometritis; | admission for observation after excessive

bleeding; and | admission because of severe nausea from

general anesthesia.

Table III Intraoperative vaiiables

D&E D & X

(n = 263) (n = 120) P value

Laminaria 259 (98.5%) 109 (90.8%) .001*

Median

preoperative 3cm(0-6) Scm(2-10) .00OIT —

cervical dilation

(range)

Anesthesia

General 92 (35.05%) 53 44.2%) .22#

Regional 9 (3.4%) 3 (2.5%)

Local/conscious 162 (61.6%) 64 (53.3%)

sedation

Median procedure 22 min 22 min 541

time (range) (6-60) (6-45)

Median estimated |

blood loss 100 mL 100 mL .38t

(range) (40-1500) (20-1200)

* Fisher exact test;

t Mann-Whitney U test:

+ Chi-square analysis.

Three patients, all of whom underwent dilation and

evacuation, had complications requiring admission to the

surgical intensive care unit. One patient, with a fetal

A-9

demise, had an amniotic fluid embolus with dissemi-

nated intravascular coagulation requiring transfusion of

blood and clotting factors. One patient was diagnosed

with sepsis and pulmonary embolus. One patient had a

uterine perforation at the site of a cesarean delivery scar

and required exploratory laparotomy and blood trans-

fusion.

Forty-five women (17.1%) in the dilation and evacu-

ation group and 17 women (15.0%) in the intact D&X

group had a subsequent pregnancy and received care at

our medical center. There were no second-trimester

spontaneous abortions in either group. Spontaneous

preterm birth occurred in 2 of 17 (11.8%) pregnancies in

the intact D&X group compared with 2 of 45 (4.4%) in

the dilation and evacuation group (P = .30). Both spon-

taneous preterm births in the intact D&X group occurred

in women at high risk for prematurity: one woman, who

underwent intact D&X caused by PPROM at 23 weeks’

gestation, subsequently delivered at 32 weeks, and the

other underwent intact D&X at 23 weeks’ gestation

because of cervical incompetence with advanced cervi-

cal dilation, and subsequently delivered at 35 weeks.

Comment

On the basis of this series of 383 patients, surgical abor-

tion late in the second trimester is a safe procedure.

There were low rates of complications in procedures per-

formed with either technique. As described in a prior

publication,® we did not see attigh rate of spontaneous

preterm birth in subsequent pregnancies in those who

received obstetric care at our medical center.

Because this study was retrospective, we cannot sfate

that one technique is superior to the other. Procedures

performed with intact D&X occurred at later gestational

ages and with more preoperative cervical dilation.

A-10

Although more advanced gestation might be expected to

increase complication rates, more preoperative cervical

dilation might be expected to facilitate these procedures.

Differences between maternal age, gestational age, and

indication for surgical abortion were noted between the

2 groups. The difference in gestational age may be due

to changes in the cervix that occur as pregnancy pro-

gresses. With advancing gestation, the cervix may be

more likely to respond to laminaria placement with

greater dilation. Intact D&X requires more dilation, as

the intact fetus cannot pass through a minimally dilated

cervix. In addition, more women in the intact D&X

group underwent abortion caused by premature cervical

dilation or PPROM. The incidence of these conditions

increases with advancing gestation, and the spontaneous

cervical dilation occurring in these cases probably facil-

itated intact D&X.

The difference in maternal age is likely because of the

higher rate of fetal aneuploidy in the dilation and evac-

uation group. Amniocentesis is typically performed at 16

to 18 weeks, and adva.uced maternal age is the most

common indication. Most cases of fetal aneuploidy are

identified by 20 weeks. This could explain why those in

the dilation and evacuation group were older, and under-

went abortion at earlier gestational ages.

Our approach of performing-intact D&X when possi-

ble is intended to minimize the use of forceps in

extracting the fetus. We believe that the use of forceps to

grasp the fetus can cause inadvertent trauma to the uter-

ine wall. At these gestational ages, evacuation of a fetus

can require multiple insertions of forceps, and intact

D&X avoids this. Though we believe our low compli-

cation rate validates our approach, we acknowledge that

the retrospective nature of this study precludes us from

A-11

concluding with certainty that intact D&X prevented

adverse outcomes.

Another important limitation is the relatively small

number of patients receiving prenatal care at our hospital

in subsequent pregnancies. Most patients returned to

their referring obstetrician for future obstetric care, and

we were unable to assess subsequent pregnancy out-

comes of patients delivered at other institutions. We

believe that significant bias as a result would be

unlikely, though we cannot be certain that outcomes in

pregnancies followed at our hospital are representative

of outcomes in all subsequent pregnancies. Though we

are reassured by the low number of complications in

subsequent pregnancies in both groups, we acknowledge

our lack of power to conclude that subsequent pregnancy

outcomes are not different.

Some have stated that intact dilation and extraction

poses serious risks to the health of a woman beyond the

risks associated with dilation and evacuation. Such puta-

tive risks include higher rates of cervical incompetence,

uterine rupture, abruption, amniotic fluid embolism, and

trauma to the uterus.®’ We are not aware of any pub-

lished data supporting these statements. In our patients,

the overall rate of complications was comparable

between those undergoing dilation and evacuation and

intact D&X. No patient undergoing intact D&X experi-

enced uterine rupture, amniotic fluid embolism, or

required blood transfusion. Because our approach is to

perform intact D&X when possible on the basis of cer-

vical dilation and fetal position, it is unlikely that intact

D&X could have been performed in these patients under-

going dilation and evacuation who experienced severe

complications.

In conclusion, our data affirm that abortion after 20

weeks’ gestation with intact D&X appears to have sim-

o> &.

A-12

ilar complication rates as dilation and evacuation when

performed by experienced physicians. The observed

complication rates and subsequent obstetric outcomes

appear comparable between the 2 techniques. In accor-

dance with the American College of Obstetricians and

Gynecologists statement of policy,’ our data supports

that the most appropriate technique for surgical evacu-

ation of pregnancy after 20 weeks’ gestation should be

based on intraoperative factors. Attempts to regulate

intact D&X on the basis of concern for maternal well-

being cannot be supported by available evidence.

References

1. Elam-Evans LD, Strauss LT, Herndon J, Parker WY, Whitehead

S, Berg CJ. Abortion surveillance—United States, 1999.

MMWR Morb Mortal Wkly Rep 2002;51:1-28.

2. Grimes DA, Schutz KF, Cates W, Tyler CW. Mid-trimester abor-

tion by dilation and evacuation: a safe and practical alternative.

N Engl J Med 1977;296:1141-5.

3. Grimes DA, Hulka JF, McCutchen ME. Midtrimester abortion

by dilation and evacuation versus intra-amniotic instillation of

prostaglandin F2: a randomized clinical trial. Am J Obstet

Gynecol 1980;137:785-90.

4. Schneider D, Halperin R, Langer R, Caspi E, Bukovsky I. Abor-

tion at 18-22 weeks by laminaria dilation and evacuation.

Obstet Gynecol 1996;88:412-4.

5. Autry AM, Hayes EC, Jacobson GF. Kirby RS. A comparison of

medical induction and dilation and evacuation for second-

rimester abortion. Am J Obstet Gynecol 2002;187:393-7.

v. Partial Birth Abortion Act of 2002, HR 760, US Congress

(2003).

7. Sprang ML, Neecrhof MG. Rationale for banning abortions late

in pregnancy. JAMA 1998:280:744-7.

8. Kalish RB, Chasen ST, Rosenzweig LB, Rashbaum WK, Cher-

venak FA. Impact of midtrimester dilation and evacuation on

A-13

subsequent pregnancy outcome. Am J Obstet Gynecol 2002;187:

882-5.

American College of Obstetricians and Gynecologists. State-

ment on so-called “partial birth abortion” laws. Washington,

DC: The College: 2002.

“nid

ty a em: Pie

ae ee

iff

B-]

APPENDIX B

American Journal of Obstetrics and Gynecology (2005)

193, 1161-4

American Journal of

Obstetrics & Gynecology

www.ajog.org

Obstetric outcomes after surgical abortion at

220 weeks’ gestation

Stephen T. Chasen,** Robin B. Kalish," Meruka Gupta,”

Jane Kaufman,” Frank A. Chervenak*

Division of Maternal-Fetal Medicine,* Department of

Obstetrics and Gynecology,” Weill Medical College of

Cornell University, New York, NY

Received for publication March 1, 2005; revised May 9,

2005; accepted May 25, 2005

KEY WORDS

Dilation and evacuation

Spontaneous preterm birth

Presented at the Twenty-Fifth Annual meeting of the Socicty for

Maternal Fetal Medicine, February 7-12, 2005, Reno, Nev.

* Reprint requests: Stephen T. Chasen, MD, Weill Medical Col-

lege of Cornell University, 525 East 68th Street, J-130, New York,

NY 10021.

E-mail: stchasen@med.cornell.edu

reserved. doi: 10.1016/j.ajog.2005.05.078

B-2

Objective: The purpose of this study was to describe

obstetric outcomes after surgical abortion at >20 weeks,

and to identify risk factors for subsequent spontaneous

preterm birth.

Study design: Patients who had surgical abortion at 220

weeks’ gestation from 1996 to 2003 and received sub-

sequent prenatal care at The New York Weill Cornell

Medical Center were identified. Indication for abortion,

operative technique, and subsequent pregnancy outcomes

were reviewed. Student f test, Fisher exact test, and

Mann-Whitney U were used where appropriate.

Results: One hundred and twenty pregnancies in 89

women were identified. Thirteen (10.8%) ended with

early miscarriage, and 5 were electively terminated. Of

the remaining 102 pregnancies, 7 ended with sponta-

neous preterm birth. Those who experienced preterm

birth were more likely to have undergone abortion due to

cervical dilation and/or preterm premature rupture of

membranes (PPROM) (27.3% vs 4.4%; P = .03). Those

with a multifetal pregnancy in the subsequent pregnancy

were more likely to have preterm birth (75.0% vs 4.3%;

P < .001). In patients who underwent dilation and evac-

uation (D&E) for reasons other than cervical dilatiogs——_—

and/or PPROM, rates of spontaneous preterm birth were

identical between those who had intact dilation and

extraction (D&X) and D&E using forceps (4.2% vs

4.5%: P= 1.0).

Conclusion: In those who have undergone D&E at 220

weeks, only a history of midtrimester cervical dilation

and/or PPROM or a current multifetal pregnancy were

associated with spontaneous preterm birth.

B-3

Dilation and evacuation (D&E) is considered the

safest method of midtrimester abortion.'? Although D&E

has been categorized as a risk factor for future sponta-

neous preterm birth, published studies have not docu-

mented such an association.** Relatively few second

trimester abortions are performed beyond 20 weeks’ ges-

tation; in the United States, only 1.4% of abortions were

performed beyond 20 weeks in 2001.° Because these pro-

cedures typically require more cervical dilatation, it is

plausible that D&E late in the second trimester couid be

a risk factor for future preterm birth.

Most D&Es involve disarticulation of the fetus with

forceps. A variation of D&E, “intact dilation and extrac-

tion,” or “intact D&X,” has been described.° Intact D&X

involves extraction of the largely intact fetus through the

dilated cervix, and generally requires more cervical dila-

tion. Complication rates for intact D&X are similar to

those for D&E using forceps, although intact D&X is

usually done later in pregnancy.’ It has been suggested,

in the absence of supportive data, that intact D&X in

particular may increase the risk of subsequent preterm

birth.*?

Our objective was to describe the rate of and identify

risk factors for spontaneous preterm birth in women who

have undergone D&E at 220 weeks’ gestation. A sec-

ondary objective was to determine whether operative

technique is associated with higher rates of spontaneous

preterm birth in future pregnancies.

Material and methods

Patients who underwent dilation and evacuation at our

hospital at 220 weeks’ gestation from 1996 to 2003, and

who received subsequent prenatal care at The New York

Weill Cornell Medical Center were identified by search-

ing computerized hospital records using the appropriate

B-4

procedure codes. Gestational age had been confirmed by

ultrasound in all cases. Medical records were reviewed

to collect data about the terminated pregnancies, includ-

ing the technique of surgical abortion, and the outcomes

of subsequent pregnancies. This study was approved by

the Institutional Review Board of our institution.

All patients underwent serial laminaria insertion

before surgical abortion, unless spontaneous cervical

dilation had already occurred. Laminaria were placed

preoperatively on 2 consecutive days, and operative

evacuation was performed on the third day. At each

insertion, as many laminaria as could be accommodated

by the cervix were placed. Patients received doxycycline

the days of laminaria insertion and the day of operative

evacuation for antibiotic prophylaxis. The operative

technique was categorized as “D&E” if the procedure

required disarticulation of the fetus with forceps. The

operative technique was categorized as “intact D&X” if

the fetus was delivered largely intact without disarticu-

lation with forceps.’ Spontaneous preterm birth was

defined as delivery at <37 weeks’ gestation resulting

from preterm labor or preterm premature rupture of

membranes (PPROM).

Fisher exact test was used to compare categorical out-

comes. Studeni ¢ test and Mann-Whitney U were used to

compare continuous data. Statistical analysis was per-

formed using Statistical Package for the Social Sciences

(SPSS Release 11.0, SPSS, Inc, Chicago, Ill) software.

Results

There were 120 pregnancies in 89 women included in the

study. These pregnancies were identified from a group of

383 patients who underwent surgical abortion at 220

weeks’ gestation from 1996 to 2003. The outcomes of 62

of these pregnancies were briefly summarized in a pre-

B-5

vious publication that focused on operative techniques of

surgical abortion late in the second trimester.’

The median maternal age at the time of surgical abor-

tion was 33 years (interquartile range 29-37 years), and

the median interval between abortion and the estimated

date of confinement in a subsequent pregnancy was 18

months (interquartile range 14-33 months). The median

gestational age at abortion was 21 weeks (interquartile

range 20-23 weeks).

First trimester spontaneous abortion occurred in 13

(10.8%) pregnancies. Five pregnancies (4.2%) were elec-

tively terminated. There were no cases of second

trimester pregnancy loss. The remaining 102 pregnancies

occurred in 80 women. There were 93 deliveries at term.

One woman with lupus nephritis underwent surgical

abortion at 24 weeks’ gestation because of fetal demise.

Her 2 subsequent pregnancies were delivered at 31

weeks and 28 weeks because of HELLP syndrome.

Spontaneous preterm birth occurred in the remaining

7 pregnancies (6.9%), which are described in Table I.

There were no significant differences in maternal age

or gestational age at time of abortion with regard to the

likelihood of subsequent spontaneous preterm birth. The

association between indication for surgical abortion and

subsequent spontaneous preterm birth can be seen in

Table II. Of those undergoing abortion due to PPROM or

spontaneous cervical dilation, 27.3% experienced spon-

taneous preterm birth in a subsequent pregnancy, com-

pared with 4.4% of those undergoing abortion for other

indications (P = .03; OR 8.2 [1.5-43.0]).

Those with a multifeta’ pregnancy after surgical abor-

tion had a 75% incidence of sportaneous preterm birth,

compared with 4.3% of those with a singleton pregnancy

(P < .001; OR 70.5 [5.9-837.2]). In 87 singleton preg-

nancies in which surgical abortion was done for reasons

B-6

yopdatay syoomgg XC NPI] SYIOM €7Z Aprojdnouy L

UIM] SYIIM OF qyd SYIIM OZ Apiojdnouy 9

U0}d[BUIS SYIOM OF avd SYIIM 17 WOudd S

UTI] SUIS syoom cg XC 1910] SYIIM HC GANl/NWOUdd v

uoIa[ SUIS syoom7e = =—- KPC 19810] SYIOM CC WOudd t

Ulm SYIOM | avd SYIOM 17 Apiojdnouy Z

UOJIIBUIS SYooM OF ayxd $y9OM OZ Apiojdnouy I

sasnjaj jo ~ AIDALPOP 1x9 1B anbiuyso} uoIIogR ye = UOTIOGP jeoigins = aseD

JoquUINN 358 [RUOTIEISIH DatyRiadQ ade [RUOTIEISIH Io} UONROIpU]

SYIIM OTS 1B UOHIOge jeoisins Ja1je Yq wiaiaid snoourjuods yum saroueusaig | e1q®L

B-7

other than PPROM or cervical dilation, spontaneous

preterm birth occurred in only | (1.1%).

The operative technique used was D&E in 70 cases

and intact D&X in 32. There was no significant differ-

ence in the rate of spontaneous preterm birth between

those who had undergone D&E versus intact D&X (5.7%

vs. 9.4%; P = .68). PPROM and/or spontaneous cervical

dilation were more common indications in those who had

undergone intact D&X (25% vs 4.3%; P = .001). In the

91 patients who underwent abortion for other indica-

tions, the rates of subsequent spontaneous preterm birth

were nearly identical between those who had undergone

D&E and intact D&X (4.5% vs 4.2%; P= 1.0).

Comment

D&E is the most common procedure used to terminate a

pregnancy in the second trimester in the United States.

Lower complication rates have been described for D&E

compared with medical induction.'? D&E may also be

preferable to some patients because it is typically done

on an Ouipatient basis, and women do not have to endure

labor. When Grimes et al attempted to perform a ran-

domized clinical trial to compare D&E with medical

induction using misoprostol and mifepristone, most

women did not consent to randomization and opted to

undergo D&E instead.'? While most women may regard

D&E as preferable to induction, many might reconsider

if D&E were found to be a risk factor for subsequent

preterm birth.

Based on our data, only those who have undergone

surgical abortion late in the second trimester because of

PPROM and/or premature cervical dilation are at high

risk for spontaneous preterm birth in subsequent preg-

nancies. The high rate of subsequent spontaneous

preterm birth seen in our patients with PPROM or cer-

B-8

vical dilation before surgical abortion (27.3%) is com-

parable to published data describing the rate of sponta-

neous preterm birth after midtrimester loss.'' The

association between surgical abortion and subsequent

spontaneous preterm birth in these women is related to

the indication for abortion, rather than the abortion pro-

cedure itself. The other significant association between

multifetal pregnancy and preterm birth is obviously not

Surprising.

Table II Indication for abortion and subsequent

spontaneous preterm birth

No spontaneous Spontaneous

Indication for preterm birth preterm birth P

surgical abortion (n=95) (n =7) value

Fetal abnormality 69 (72.6%) 4 (57.1%) 0.40

Intrauterine fetal 12 (12.6%) 0 (0%) 1.0

demise

PPROM or -

spontaneous 8 (8.4%) 3 (42.9%) 0.03

cervical dilation

Other 6 (6.3%) 0 (0%) 1.0

Fisher exact test used for statistical comparison.

The rate of spontaneous preterm birth in singleton

pregnancies after surgical abortion for indications other

than PPROM and/or cervical dilation was only 1.1%. No

patient who underwent surgical abortion using intact

D&X for indications other than PPROM or cervical dila-

tion delivered prematurely, with the exception of a

triplet pregnancy delivered at 36 weeks. Our data do not

support the notion that achieving advanced cervical dila-

tion with laminaria is a risk factor for preterm birth. In

fact, we have found that in women undergoing D&E

B-9

from 14 to 24 weeks, the risk of subsequent preterm

birth was inversely correlated with the degree of cervi-

cal dilation achieved before surgical abortion.’

Because of the relatively small number of patients

studied, we cannot exclude the possibility that surgical

abortion late in the second trimester may increase the

risk of spontaneous preterm birth. In our hospital, the

rate of spontaneous preterm birth in singleton pregnan-

cies is approximately 5%. To have an 80% likelihood of

detecting a 50% increase in the rate of spontaneous

preterm birth to 7.5% at a significance level of 0.05,-we

would need 539 patients. In this study with 102 preg-

nancies, our power to detect this magnitude of an

increase in spontaneous preterm birth was only 34%.

In conclusion, surgical abortion at 220 weeks’ gesta-

tion is not associated with a high risk for spontaneous

preterm birth in future pregnancies. Surgical abortion

with either variation of D&E late in the second trimester

Should not be considered a risk factor for subsequent

spontaneous preterm birth.

References

1. Stubblefield PG, Carr-Ellis S. Borgatta L. Methods for induced

abortion. Obstet Gy yee! 2004:104:174-85.

2. Autry AM, Hayes EC, Jacobson GF, Kirby RS. A comparison of

medical induction and dilation and evacuation for second-

trimester abortion. Am J Obstet Gynecol 2002;187:393-7.

3. Kalish RB, Chasen ST, Rosenzweig LB, Rashbaum WK, Cher-

venak FA. Impact of midtrimester dilation and evacuation on

subsequent pregnancy outcome. Am J Obstet Gynecol 2002;

187:882-5.

4. Schneider D, Halperin R, Langer R, Caspi E, Bukovsky I. Abor-

tion at 18-22 weeks by laminaria dilation and evacuation.

Obstet Gynecol 1996:88:412-4.

11.

B-10

Strauss LT, Herndon J, Chang J, Parker WY, Levy DA, Bowens

SB. et al. CDC. Abortion surveillance—United States, 2001.

MMWR Surveill Summ 2004;53:1-32.

Haskell WM, Easterling TR, Lichtenberg ES. Surgical abortion

after the first trimester. In: Paul M, Lichtenberg ES, Borgatta L,

Grimes DA, Stubblefield PG, editors. A clinician’s guide to

medical and surgical abortion. New York (NY): Churchill Liv-

ingston; 1999. p. 123-54.

Chasen ST, Kalish RB, Gupta M, Kaufman JE, Rashbaum WK,

Chervenak FA. Dilation an‘ evacuation at > or =20 weeks: com-

parison of operative techniques. Am J Obstet Gynecol

2004; 190:1180-3.

Partial Birth Abortion Act of 2002, HR 760, US Congress

(2003).

Sprang ML, Neerhof MG. Rationale for banning abortions late

in pregnancy. JAMA 1998;280:744-7.

. Grimes DA, Smith MS, Witham AD. Mifepristone and miso-

prostol versus dilation and evacuation for midtrimester abor-

tion: a pilot randomised controlled trial. BJOG 2004;

111:148-53.

Goldenberg RL, Mayberry SK, Copper RL, Dubard MB, Hauth

JC. Pregnancy outcome following a second-trimester loss.

Obstet Gynecol 1993;81:444-6.

This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.