Oral argument — Painters & Allied Trades District Council 82 Healt v. Takeda Pharmaceutical Company Limited

And we'll hear your argument next in Painters & Allied Trades District Council 82 Health Care Fund against Takeda Pharmaceutical Company Limited. Mr. Wisner, yes, please. I apologize, Your Honor, Mr. Wisner. Oh, I'm sorry, I misread. So it's, yeah, Mr. Wick is appellant. Sorry. Thanks for responding. Please proceed. I'm Robert Wick for the defendants, and I would like to reserve four minutes for rebuttal, please. The district court's class certification order should be reversed for four reasons. First, the court failed to conduct a rigorous analysis of the expert evidence it relied on to satisfy the predominance requirement. Second, the district court certified a class even though the plaintiffs have no common proof capable of sifting uninjured class members from allegedly injured ones. Third, the district court certified a class despite finding that it remains an open question whether individualized causation questions will overwhelm common ones. And finally, superiority and Lilly-specific issues also warrant reversal. Turning to the first of those issues, the plaintiffs offered estimates from their expert, Dr. Commoner, that 56% of ACTOS prescriptions were excess prescriptions and 98.5% of the class paid for an excess prescription. Those estimates were the linchpin of the district court's predominance finding, but the court never subjected those estimates to the rigorous analysis required by Rule 23. Instead, the district court simply took those estimates, quote, at face value, close quote. The plaintiffs say it's enough that the district court conducted a Daubert analysis and found Commoner's evidence admissible. But this court has said at least three times that a Daubert analysis is not enough, that even if a district court finds expert evidence admissible under Daubert, the district court still has to conduct a rigorous analysis of the evidence, it has to examine whether the expert relied on unsupported assumptions, and it has to resolve disputes on those questions. Go ahead, please. Do you think it could ever be appropriate to certify a class of this nature? I mean, if you had a — I mean, I understand, and I mentioned, we'll talk about a lot of your objections to the econometrics of Mr. Commoner, but if you had a better, in your view, econometric analysis, and the district court looked at it and said, you know, I find that this is valid and I agree with it, could a class like that go forward, or do you think there's a more fundamental problem here with proving reliance in an aggregate way? Well, two things. First, if the district court actually did the rigorous analysis, which didn't happen here, and found that the expert evidence withstood the rigorous analysis, then we would have a case much closer to Olian, in which case this Court affirmed class certification because the district court did the rigorous analysis it was supposed to do. There's a further problem here that I don't think is curable, which is that even if you do take Commoner's 98.5 estimate at face value, it still doesn't prove predominance because they have no common proof capable of sifting the 98.5 percent from the 1.5 percent. That makes this case just like the Asikal case in the First Circuit. There, the plaintiffs offered evidence that 90 percent of the class was injured, but the plaintiffs didn't have any way to separate the 90 percent from the 10 percent, short of combing through the entire class in individual inquiries. Well, likewise here, even if you take 98.5 percent at face value, which is giving it much more credit than it's due, the plaintiffs don't have any way of separating the 98.5 percent from the 1.5 percent, short of combing through the whole class in individual inquiries, which destroys predominance. The district court seemed to emphasize that the 1.5 percent just wasn't enough to overcome the, quote, mountain of evidence that the plaintiffs presented at the class stage, the certification stage. Is there some de minimis argument to be made here? Is it enough if it's 1 percent or 0.5 percent to defeat certification? It depends, Your Honor, on whether you can find the 1 percent or the 1.5 percent without combing through 100 percent of the class. The problem here is not that 1.5 percent is automatically too large. The problem here is that you have to comb through 100 percent of the class to find the 1.5 percent. And under ASICOL, Rail Freight, and only in footnote 13, that destroys predominance. The second problem with the district court's reliance on the 1.5 percent problem is that it just took commoners' estimates at face value. And it shouldn't have taken the estimates at face value. It should have scrutinized whether they relied on unsupported assumptions. And there are at least three. So how do we, as a matter of appellate review, examine rigor? I mean, obviously, the district court has to do a rigorous analysis, and we've reversed when it doesn't appear the district court did any analysis, just that we have competing experts, and so they're admissible, and that's the end of it. But here you do have analysis. How do we, as appellate judges, analyze rigor? I can give you two answers. First, here we don't have any analysis other than a Daubert analysis, and that is reversible error under Ellis v. Costco. We know a mere Daubert analysis is not enough. That's Ellis v. Costco. That's only in footnote 9. Secondly, rigorous analysis. Did he actually say, I'm only doing a Daubert analysis, or did he actually do? Well, he says explicitly at ER 22 and at footnote 88 that I take commoners' estimates at face value. And then if you read the body of the opinion, you won't find any rigorous analysis of three unsupported assumptions. The second answer to your question, Judge Thomas, is we know at a minimum that a rigorous analysis requires investigating whether the expert evidence relies on unsupported assumptions, and we have to resolve disputes on that. That, among other things, is Olean at footnote 9 and Olean at 683. There are three unsupported assumptions here that are not analyzed at all, let alone rigorously. The first is commoners' assumption that all nonrefill prescriptions are statistically independent of each other. The second is his assumption that 100 percent of the decline in Actos's market share from 1999 to 2013 is attributable to the 2010 bladder cancer disclosures and that no other market factors contributed. And the third is his assumption that all excess prescriptions, all of them, would have been replaced by cheaper prescriptions in the but-for world. Commoner admits that all — that he has no basis for any one of those three unsupported assumptions, which makes this a clean and easy case. As to the statistically independent prescription assumption, the plaintiffs can't apply the 98.5 percent estimate to the class unless and until they prove that each class member paid for five statistically independent prescriptions. Well, that would require thousands of individual inquiries in and of itself. Commoner simply assumed away those thousands of individual inquiries by simply assuming that all nonrefill prescriptions are statistically independent of each other. But he then admitted at his deposition he had no basis for that assumption. He testified at his deposition he doesn't know whether prescriptions written by the same doctor or prescriptions subject to the same plan formulary are actually independent of each other. That's ER 319 to 321, which makes this an easy case. It's not rigorous analysis when you ignore an unsupported assumption that even the plaintiff's own expert admits he has no basis for. Secondly — And did you — did you introduce evidence that — I can't remember if this is in the Hughes Declaration or somewhere else — that the plan formularies vary from plan to plan in a way that one might expect to result in different distributions of prescriptions among different plans? Is that right? Yes. We introduced evidence that access prescriptions are not going to be randomly distributed. They're going to be nonrandomly distributed if they were written by the same doctor, even if they're influenced by the same plan demographics. And that evidence is at ER 368 to 369 and ER 398 to 401, and it was never controverted by Commoner or the plaintiffs. Well, the district court does analyze the issue of random distribution of the prescriptions. What's the flaw in that analysis? No. He doesn't analyze whether the prescription — the access prescriptions are randomly distributed. All he says is that there's data that can tell you which prescriptions are refill and nonrefill prescriptions. That's the only thing he said. But our whole point is not all nonrefill prescriptions are statistically independent of each other, and Commoner concedes he doesn't know whether all nonrefill prescriptions are statistically independent of each other. As to the 56 percent assumption, to get to the core estimate that 56 percent of excess — of Actos prescriptions were excess, what Commoner did was start with Actos's market share in December of 2013, which he estimated was about 4.7 percent. He then compared Actos's December 2013 market share to its average market share over the 11-year class period. And he found that, on average, it was about 56 percent lower in December 2013 than it was during the class period. He then made a raw assumption, a whale of an assumption, that all of that 56 percent decline was attributable to one and only one thing, which is the updated bladder cancer disclosures of 2010 and 11. But he then admitted at his deposition even he doesn't believe that whale of an assumption. He testified at ER 265 to 266, not all of the decline in Actos's market share as of December 2013 was attributable to the bladder cancer disclosures. That is a devastating admission because he flat-out admits that even he doesn't believe the whale of an assumption at the heart of his 56 percent estimate. Zyprexa and Sargent's are directly on point in a nearly identical factoring factual setting. They say that kind of raw, unsupported assumption cannot establish predominance or satisfy Rule 23. I'm at four minutes. May I reserve the remainder? Yes, you may. Thank you. Thank you, Mr. Wick. And now we'll hear from Mr. Wisner. Good morning. Brent Wisner on behalf of the Apelli Painters' Health Fund, as well as the National Certified Class. May it please the Court. The district court had a big job in front of it. He had a lot of evidence, an MDL with multiple rulings, a full panoply of discovery that had been conducted, and he did his job. He systematically went through that evidence. He assessed its reliability, its admissibility, and then after that decided whether or not it was persuasive in finding that a predominance of the evidence, the standard that is required, met the elements of class certification. There is no evidence that that decision was illogical, irrational, or unconnected to the record, and this court gives district courts, rightly so, broad discretion on how to assess these elements. This court did exactly what he's supposed to. My friend here started his argument saying that there was no rigorous analysis. I don't think that's a plausible reading of this district court decision. Sure, he did a Daubert analysis, and he concluded that it was reliable and admissible. But once it's admissible, right, once it's established that on average, depending on the month and year, 56 percent of them were fraudulently induced. Once that's established as an admissible opinion, which he did under Daubert, and it's not subject to this Court's review, the question then becomes, putting aside the merits of whether that, in fact, is true or not, is that show that it, as the Olean case said, capable of showing that individual issues are resolvable with common evidence. And that's precisely what he does in his order at length and in detail. I mean, Your Honor, Judge Rosenthal, you specifically pointed out that he specifically has a section titled distribution across the TPPs to systematically think about and weigh the evidence about how these things could be distributed. So the thing that you have to prove, not for class certification, but on the merits, is that doctors relied on the false statements or omissions, right, in writing prescriptions. And so to get class certification, you need to show that that question of reliance can be determined with common proof, right? To some extent. I think focusing on reliance is a bit myopic. And the reason why I say that is because the RICO statute actually says we have to show economic injury by reason of the fraud. And I mean, the Court considered this in Bridge, right, which I have some familiarity with. And reliance is, like, you need to show causation, but as a practical matter, if nobody relies on the fraud, there's not going to be causation. Sure. And the reason why I say it's myopic is only because there's other potential reliances, right? I mean, consumers can be prescribed a drug and decide not to take it. And then the third-party payer never purchases it. A formulary could change the status of the drug. There's a whole milieu of things that happen because of the bladder cancer warning that directly and indirectly affects the ultimate injury here, which is the purchase of a fraudulently induced prescription. And so I hesitate to say we have to just look at doctors. I think we have all of this. And that's why the econometrics is so powerful, right? Well, okay. So, all right. But I think we're agreeing that you need to show, whether it's reliance by doctors or something else, you need to show that the fraud had some effect on somebody's And to get class certification, you need to show that you can determine that question with common proof. And you can do that with common order, but only if we think that common order is basically right. If we think that common order's analysis is flawed, then it doesn't show causation on a class-wide basis, and you're back to a series of individualized inquiries, right? Well, I would point out two things. I think you're right, but I think the question is not whether or not this court believes Dr. Commoner was right. I think the question is, did the district court abuse its discretion in concluding that Dr. Commoner was right? Well, but I'm not, and that gets to the question of did the court really conclude that? I mean, your friend on the other side mentions the face value language, but that sentence goes on to say, you know, takes the report at face value and does not prejudge its accuracy. Correct. Isn't making a judgment about its accuracy what the district court needed to do at this stage? No, Your Honor. And Olien specifically addresses this. Quote, a district court is limited to resolving whether the evidence establishes that a common question is capable of class-wide resolution, not whether the evidence in fact establishes that plaintiffs would win at trial. So can I follow up on Judge Miller's good questions? Sure. Reliance is the reason, the individual nature of the reliance inquiry is the reason the district court rejected certification of the class of the direct, the patients who were obviously making decisions based on what their doctors told them and their own decisions. But the third-party payers' act of reliance, paying for the prescriptions that were issued, is derivative of these individual decisions. That is correct, Your Honor. So you have to be able to make a meaningful distinction between third-party reliance and first-party reliance in terms of the common proof that would be available at a trial of a disputed class. Not a class certified for settlement, but for trial. I agree. So how do you reconcile the fact that these are derivative decisions, but one could not be certified for class determination, and one could, the other could not? I think it's a reflection, candidly, of the district court's rigorous analysis, because we had the goods, so to speak, on the third-party payer class, but we did not have the goods on the consumer class. And what I mean by that is we had data, right? We have the claims data, and we're able to apply it. And to be clear, Dr. Kaminer conducted a regression. I mean, it was not just taking a market share. That was to get his benchmarks. That's correct. And then he extrapolated. But the regression accounts for all the potential market variables, so it doesn't just take Well, no, no, no. I mean, the regression was done on a single time series from, I think, 2010 to 2013, and accounted for a handful of variables that do indeed seem like the most plausible variables to think of. But as Hughes points out, there's a lot of other things happening in the world at that time, a lot of other things relevant to the drug, and it's hard to infer a causation from just looking at one time series. Well, to be clear, he actually looked at the pre-time pre-render regression, Model 1. Model 2 was the post-series. Model 3 was then a more narrowed and extrapolated one. And then Model 4 was actually the entire time period from beginning to end, which showed in every single instance that there was a statistically significant coefficient to the effect of the bladder cancer risk warning. So if you look at the rebuttal report, a lot of these concerns are squarely addressed by Dr. Commoner, and the district court considered that in his Daubert ruling as well, candidly, in his certification ruling. And so, but going back to Your Honor's question, which I think is an important one, right, this is where the Narantan cases are so important, because there they dealt with do you need to have individual doctor testimony to establish causation? And if you have a reliable, which is not really at issue here, right, because that order is not on appeal, if you have a reliable model, and this is the important part, that was just one piece of the mountain, we have the defendant's own studies, their own analysis of prescription behavior showing that over 85% of physicians would stop prescribing it if there in fact was a bladder cancer warning, and that was done contemporaneously with the fraud in 2002. So we have a mountain, we have peer-reviewed studies that use the same method and came to the same conclusion on different data sets. This isn't a situation where it's one, you know, outlandish econometrician coming up with, you know, stats can prove anything. What happens if this goes to trial, and, you know, the jury, I guess, would be told, like, would be asked, you know, do you think that Dr. Commoner is reliable? And if they decide that he is not, would that mean that you lose, then, across the board? Probably. But I don't see why that's true, though, because if he's reliable, if his methodology is determined to be unreliable, it doesn't mean that nobody was influenced by the fraud. It just means that we can't tell who was influenced by the fraud and who was. Well, I think a competent trial defense attorney would say, they have the burden of proof, they have to prove fraud actually caused it, and their own guy's own model is not appropriate, and therefore they haven't met their burden, and the jury would find against us. I mean, I think that's how it would play out. But that's not, I mean, that's not really a determination of an across-the-board common conclusion that there was no effect of the fraud. Well, I agree, but I think the question at class certification is not whether or not we would lose, right, but whether or not we can win, right, because that's, is it capable of establishing on a class-wide basis? And so if we assume the jury believes the 56 percent, as applied individually, and believes the information and the evidence, this mountain that we've discussed, they could survive a summary judgment or directed verdict that there was, in fact, but-for causation. And if that can be done purely with common evidence, then I believe Olien says the Court is not abusing its discretion in finding predominance. But the district court did take into account that there were going to be affirmative defenses that were individual in nature, and there were going to be defenses to the causation theory that were individual in nature. It simply said, well, I've only got two depositions saying that, and that just is outweighed. Is that, is the weighing the appropriate approach here? I believe so. And I think that's literally what this Court held in the Vann decision, that, you know, whether or not individual defenses are going to emerge is a burden carried by the defendant. And failure to present that evidence, you can't, you can't undo a class certification based on, quote, mere speculation. The two depositions we have, candidly, don't even support the defense. Both doctors testified that they prescribed less lactose because of the warning. And they weren't even related to the certified class. They related to the California claim. So, I mean, there is literally zero evidence of individual issues. Now, the Court has left the door open, and that's, that's the function of class certification being an interlocutory order, right? The Court can readdress class certification as we get closer to trial. And prior to the Court taking this appeal, the Court said we are going to have discovery for a year and give the defendants an opportunity to see if they can undo the class cert, and we're prepared to do that. Of course, in Vann, the Court does say that whether it's two or 18 examples that the defendant produces, that could be enough to show individualized issues. I agree. They produced zero. That's the problem. The two prescribers, Your Honor, aren't even part of the class. They're from a different class. And the two prescribers, if you look at the record, clearly testify that their use of lactose decreased. And it makes sense, right? One of the warnings was a contraindication for lactose use with people who are male who smoke. So unless they're going to go against the FDA's admonition not to use a drug with certain people, by definition, there is going to be a reduction of some sort. The defendants may very well get there, but that's not the case before this Court, and that wasn't the case before the District Court. One of the things that troubled me about the opinion was the absence of a significant analysis of manageability for trial. Sure. And Rule 23 requires that. Sure. But in most cases that I've been familiar with, there has been a detailed trial plan submitted. I did not see that in the record. Have I missed it? No, you did not, Your Honor. We did not submit a detailed trial plan. And maybe that was— Then how did you show manageability other than the District Court's very brief statement that this is going to be more efficient than having individualized cases? Sure. During the argument— That's not manageability. During the argument and during the briefing, we spelled out to the Court how we intended to prove our case, and that's why the Court mentions, you know, the length of trial. There is a record here of us discussing this issue. It came down to a question of, would it be possible? And if we are able to show all this common evidence, and currently right now there is very little individual opposite evidence, as it stands right now, it would be, as the Court said, a nightmare of a trial, but it's still a trial that can be done. And when you compare that to what it would take to try potentially thousands of individual cases that would never be worth trying individually, it becomes pretty obvious that the Court had considered that and didn't abuse its discretion in deciding to proceed forward as a class at this time. I don't believe the rigorous analysis that we're concerned with on manageability applies to manageability in the same way. I mean, the Court has broad discretion on deciding how it wants to proceed with a trial. And if the, you know, the Court would like to remand this to the Court to come up with a detailed trial plan, we can, but we will do that in the upcoming year as we proceed to trial as soon as the stay is lifted in this case. I think the concern that animated Rule 23-F is that the decision on certification is going to provide such a pressure on the defendants that you'll never get there. The need to settle will overwhelm further development. I can't comment on whether or not I believe that to be true or not. I don't want to violate the confidentiality and candor with opposing counsel. But I don't think that's really an issue here, if I'm just being honest. I think this case will proceed to trial. I think it's one of the few that will. I mean, they tried a bunch of cases in the personal injury cases before they decided to resolve it on a larger basis, and I think that will happen here. I think fundamentally, and I only have a few seconds left, but fundamentally, this case is a case about fraud, right? And we can prove, and the District Court over the MDL held as much, that there is a clear, concerted effort by these defendants to sell this drug and hide this risk. They exposed millions of people and thousands of third-party payers made payments for those prescriptions they never would have. Could you address, I mean, there's some debate over whether 1.5 percent is an accurate measure of the uninjured class members, but I think everyone seems to agree that at least some of them are. What would be the process for identifying which ones were uninjured? I appreciate you asking that question because this is something I wanted to address. In any class, there's always an individual question of whether or not an individual is a class member, right? That doesn't defeat class certification. The way you go about determining class membership, if it's based on objective and easily done criteria, then that's sufficient. Here, it's based purely on claims data, which everyone agrees is readily available for all third-party payers, and we can sort that data very quickly and very easily, as both experts did and the District Court noted, on which ones have at least five new prescriptions. Now, there's very, very few, very, very few, and the Court noted this, of third-party payers that only have five, right? We're talking about a very small portion of people, of third-party payers that could even be in this bucket. The vast majority of them have hundreds, if not thousands, of new prescriptions, and the likelihood of one of those being uninjured is beyond minuscule. But for that very small subset, we do have an ability to run the data and ascertain those people that don't meet the class and therefore are unlikely to be injured. If there's individual issues for that very small subset, that doesn't defeat certification, as the Court held in Olien, quote, that the defendant might attempt to pick off the occasional class member here or there through individual rebuttal does not cause individual questions to predominate. Any further questions? Thank you so much for your time. I have four points. First, on this question of whether the evidence is capable of resolving injury and causation on a class-wide basis, what Olien says, among other things, is that evidence is not capable of resolving a question on a class-wide basis if it doesn't survive a rigorous analysis, and it doesn't survive a rigorous analysis if it rests on unsupported assumptions. And there are at least three of them here that I detailed previously. Another way of looking at it is that if commoner's evidence doesn't survive a rigorous analysis, then the question of injury and causation are not common questions. They're individual questions, for the reasons Judge Miller was getting at in his question. If there's a defect in commoner's common proof, that doesn't prove nobody in the class was injured. All it proves is that we revert to the traditional mode of looking at individualized evidence. There's an admission that tens of millions of prescriptions still would have been written in the but-for world. So if we're right that commoner rests on unsupported assumptions, which even he admits, then we revert to the traditional mode of individualized proof and individualized causation questions overwhelm common ones. The Supreme Court endorsed exactly the logic of the questions you were asking, Judge Miller, at pages 473 and 474 of Amgen, and the Second Circuit applied the exact same logic in the same factual setting in Zyprexa and in Sargent's. Secondly, with respect to this regression, I want to emphasize it is not a regression that generates this estimate that 56% of prescriptions are excess prescriptions. What commoner does with that regression is estimate something we already knew. He estimates the real-world market share of Actos in December of 2013. The regression doesn't spit out 56%. What spits out 56% is taking that December 2013 market share, comparing it to market share in the class period, and then making a raw assumption, a whale of assumption, of an assumption that 100% of the decline in market share is attributable to bladder cancer disclosures. Commoner himself admitted at his deposition he doesn't believe that assumption. The third point, how are they going to identify the missing 1.5%? He says the class definition does it. The class definition doesn't do it. He's clear in his brief. You get into the class if you just pay for five non-refill prescriptions. Well, non-refill isn't enough to get you to the math of this 98.5% calculation. To get to 98.5%, you need five statistically independent prescriptions. So even if you satisfy the class definition, we're going to have to have thousands of individual inquiries into how many statistically independent prescriptions you paid for before they could even try to apply the 98.5% as common proof. Fourth and finally, on the manageability issues that Your Honor was asking about, there was no finding that this case is manageable as a class action. All we have is a statement that individual actions would be bad. That doesn't answer the threshold question of whether this case is manageable and whether this case can be tried as a class action. And those manageability issues are especially pronounced for Lilly. For Lilly, it stopped promoting Actos in 2006. So for all post-2006 prescriptions, there's going to have to be another set of individual inquiries as to whether it was Lilly's pre-2006 conduct or some other post-2006 event that generated any particular post-2006 prescription. If there are no further questions, I thank the Court. Thank you. We thank both counsel for their helpful arguments. The case is submitted, and we are adjourned.