Critiquing the economic argument for preemption
Regular Tortdeform commenter and Overlawyered star Ted Frank recently submitted an Amicus brief in the Wyeth case on behalf of several economists. I’ve reproduced the brief almost in its entirety, omitting only procedural information, headings, footnotes and citations to improve ease of reading. The whole brief is available here. The text of the brief is normal, my comments are in bold.
In approving the safety and efficacy, of prescription drugs, FDA faces the possibility of two types of error. First, it can approve a drug that should not have been approved because of safety problems. These errors are called Type I errors. Second, it can fail to approve a drug that should have been approved.
These errors are called Type II errors. Under well-established economic principles, the question whether FDA is more likely to commit Type I errors or Type II errors can be answered by focusing on the societal or institutional incentives that push FDA to err on one side or the other.
Repeated economic investigation of this question indicates that FDA incentives are skewed toward excessive caution in the regulation of drug development and the approval of new drugs, i.e., Type II errors.
This skewing can be best understood by considering the likely consequences to FDA’s image of a Type I versus a Type II error. When deciding whether the benefits of a proposed new drug exceed its risks, FDA staff know that if they commit a Type I error—the approval of a drug that turns out to be insufficiently safe once marketing begins—their error will become known. Because the harmful side-effects of the drug may be highly visible, a Type I error can and often does lead to impassioned criticism of the agency.
A Type I error also can and often does lead to injuries or death of large numbers of people. The brief suggests that FDA staff members are more concerned with public criticism than public health, a proposition I certainly hope is incorrect.
On the other hand, a Type II error—the failure to permit marketing of a drug that would in fact provide benefits in excess of harms—is typically known only by the relatively few persons who are intimately involved in developing the drug and are largely hidden from patients and the larger medical community who suffer the consequences of the error.
Yet the adverse public health impact of a failure to approve a beneficial drug may be even more severe than the approval of an insufficiently safe drug. Moreover, Type I errors in approving unreasonably unsafe drugs are often quickly corrected precisely because the public learns of the error. But the failure to approve a beneficial drug may go uncorrected for years, if at all. As a result, the net effect of the asymmetry in public knowledge and publicity is to bias even the best-intentioned FDA regulators towards excessive caution and delay in approving new drugs.
Let’s assume arguendo that the economists are right, and FDA regulators are biased towards excessive caution out of fear of public criticism. It’s a non sequitur that FDA preemption will alleviate this fear. Preemption won’t prevent the media, consumer groups, or legislators from criticizing the FDA for committing a Type I error. Instead, preemption will prevent the victims of that Type I error from receiving compensation for their injuries.
If anything, preemption may exacerbate Type I errors. Product liability lawsuits often force the FDA to correct a Type I error when it otherwise wouldn’t, or sooner than it otherwise would. For example, because of litigation over Paxil, GlaxoSmithKline agreed to publicize the results of all of its drug trials. Dr. Steven Nissen analyzed these documents and published a study linking Avandia to increased heart risks. As a direct result of the documents released because of litigation, the FDA put a black box warning on Avandia. If not for the litigation the economists hope to eliminate, the FDA would not have corrected its Type I error when it did, if ever.
When FDA approves a new drug, it also approves a detailed label to accompany the drug. The label contains indications (conditions to be treated), dosage, administration, and other details including warnings and contraindications, along with summaries of clinical trials and other data. These labels are designed for use by learned intermediary physicians (without whom prescription drugs may not be obtained), although others, including patients, may also make use of labels. As new information arrives about the risks and benefits of approved drugs, FDA continues to review the drug labels and effects labeling changes as appropriate. FDA employs specialized experts in the field to conduct this ongoing review and gathers voluminous post-marketing information from pharmaceutical firms, medical care providers, and other sources. (Citation omitted)
The “other sources” include documents revealed through litigation. The FDA doesn’t have subpoena power on its own, so it generally has to rely upon pharmaceuticals to voluntarily turn over all relevant drug information. In more than a few product liability lawsuits, plaintiffs’ lawyers have discovered important documents that pharmaceuticals didn’t turn over to the FDA.
FDA has long recognized that labels must give varying degrees of emphasis to specific items of information, and that labels should not contain all information of possible interest. FDA has for many years sought to modify and improve drug labels while avoiding the constant danger of overwarning.
If all potentially negative information were to be included, such as every known possible side-effect, two adverse consequences would follow. One is overwarning: the presence on labels of unfounded or disproportionately prominent warning information, which could deter or discourage drug use that would be beneficial after taking reasonable account of risks and benefits. The second is “clutter”: the presence of so much information that physicians would find it hard to distinguish important information from relatively unimportant information and might not even bother to peruse all the information.
FDA’s decisions about drug labels are strongly affected by the incentives faced by FDA staff. We have described how those incentives cause FDA to exercise excessive caution when approving new drugs. Similar reasoning applies to FDA decision making about what risk and benefit information to require in drug labels.
When deciding about the contents of new drug labels and changes in labels for approved drugs, FDA is once again faced with the possibility of both Type I and Type II errors, i.e., providing inadequate warnings or requiring warning language and contraindications that overstate risk and discourage beneficial use. And once again, the incentives facing FDA in making that decision cause FDA to err on the side of being overly cautious.
Note that the economists claim that it is the FDA, and not pharmaceuticals that decide what information to put on drug warning labels. This is Wyeth’s position. The economists later change their position and claim that pharmaceuticals may alter the warning labels on their own.
In deciding not to require a specific warning, for example, the FDA staff faces the prospect of vigorous, public criticism if patients are harmed by the drug, even if the more aggressive warning was not scientifically warranted or would have lead to disincentives in drug usage that outweighed any potential reduction in risks.
Far less criticism is likely to occur if the label warns too strongly about harms that in fact almost never occur, or if the label contraindicates certain uses that are likely to be more beneficial than harmful for patients. Patients would probably not know that they had foregone treatments that in fact would have been worth the true risks, and the victims of this FDA error accordingly will never call FDA to task. On the whole, the agency is likely to be far more risk averse than is justified by the actual balance of risks and benefits as known at the time.
Victims of any FDA error cannot call the agency to task; the FDA is immune from suit for any errors it makes. Again, assuming arguendo that the FDA is overly cautious in its labeling, the economists attribute that caution to the threat of “vigorous, public criticism.” The economists have just made a case that the First Amendment is to blame for Type II errors – not the tort system.
It’s strange that a brief authored by economists declines to mention the strong financial incentive the FDA has to commit Type II errors. In 1992, the Prescription Drug User Fee Act (PDUFA) was enacted to help speed the drug approval process. The PDUFA authorizes the FDA to charge pharmaceuticals various fees for reviewing drug submissions. The fee varies depending upon several factors, but is generally close to $1 million dollars. However, the FDA only collects the fee if it approves or denies the drug by a specific deadline. If the FDA hasn’t made a decision by the deadline, it loses the $1 million dollars. Remember, the FDA is so cash-strapped as to have made an unheard-of request to Congress for an additional $275 million in emergency funds. The FDA has also made a commitment to meet the deadline in 90% of applications.
When approaching the PDUFA deadline for a drug, the FDA has three choices if it isn’t confident that a drug is ready to be marketed. First, the FDA can choose to allow the product to be released anyway, and risk committing a Type I error. The consequences of this action of course include “vigorous, public criticism” but also may result in the injury or death of patients. The second option is to risk committing a Type II error by rejecting the drug and asking the manufacturer to submit supplemental information or conduct further studies. The consequences of this action to the FDA are small: The FDA gets to collect its fee, and as the economists note, there is little risk of public criticism.
The final option is to miss the deadline entirely and approve or deny the drug in its own time. The consequences for this action are actually the worst for the FDA. First, it will forfeit the million dollar fee. Next, it will risk criticism and investigation by Congress for failing to meet its 90% commitment. And of course, there’s no guarantee that even if the agency won’t make a Type I or II error after missing the deadline. Clearly, choosing to miss the deadline is the worst possible decision for FDA regulators.
It is again surprising that the economists failed to discuss the financial incentives that pressure the FDA to commit a Type II error. What’s even more surprising is that the economists offer no evidence that preemption will reduce the likelihood that the FDA will commit Type I or Type II errors. Even if failure to warn lawsuits were completely preempted, the FDA will still face “vigorous, public criticism” if it makes a Type I error, and the FDA will still face little criticism if it commits a Type II error. And of course, preemption will do nothing to address the large financial incentive the FDA has to commit a Type II error rather than miss the PDUFA deadline.
While advocates of state tort law prescription drug litigation often argue that litigation requirements complement FDA requirements by countering the risk of FDA underwarning or underregulation, the economic analysis of FDA incentives in the drug and drug labeling approval process demonstrates that FDA is far more likely to err in the other direction. Accordingly, rather than creating requirements that cure FDA Type I errors, state tort law creates requirements that exacerbate FDA Type II errors by causing pharmaceutical manufacturers to hold back on seeking approval of new drugs or to add defensive labeling, including further undue contraindications or overwarnings that further squeeze out physician understanding of more significant labeled information.
Recall that earlier the economists argued that it was the FDA employs “specialized experts” to determine what information to include on a product warning label. That’s the position of Wyeth, the litigant that this brief is supposed to support. In its brief, Wyeth argues that it is not permitted to engage in defensive labeling:
“Had Wyeth unilaterally altered the labeling to change the warnings with respect to arterial blood exposure or to eliminate IV push as an approved method of administration, it would have been in violation of federal law and subject to enforcement action by FDA for unauthorized distribution or misbranding.” - Wyeth brief
The Pharmaceutical Research and Manufacturers of America also filed an Amicus brief, and it too supports the proposition that manufacturers may not engage in defensive labeling:
“[T]he actual freedom of manufacturers to unilaterally change the packet insert is minimal.” - Pharma brief
The U.S. Chamber of Commerce also agrees:
"...Wyeth was not at liberty to change the Phenergan label after its approval by the FDA. As the dissenting opinion explains, the applicable "regulation does not allow manufacturers to simply reassess and draw different conclusions regarding the same risks and benefits already balanced by the FDA." - Chamber of Commerce Brief
In their fervor to blame the tort system for Type II errors, the economists have actually advanced the antithesis of the legal argument for preemption in this case. If pharmaceuticals are in fact free to include “undue contraindications or overwarnings” that the FDA doesn’t approve, then Levine’s failure to warn claim is not preempted and Wyeth must lose.
In state tort lawsuits, juries necessarily focus on a highly specific personal tragedy rather than on societal trade-offs, giving more weight to the harm allegedly suffered by the plaintiff than to the benefits realized by past and future non-injured users of the drug, because benefited patients will play no role in the trial. Given the one-sided nature of their inquiry, there is little reason to expect the lay jurors’ labeling decisions to be superior to those of FDA in terms of balancing the risks and benefits of additional contraindications and warnings on the drug label. See Riegel v. Medtronic, Inc., 128 S. Ct. 999, 1008 (2008). Rather, the jury balance will be skewed in favor of more warnings. Of course, juries will not always decide for the plaintiff. But when they do, there is an excellent chance that the verdict will fault the manufacturer for failing to provide an overwarning or unwarranted contraindication or warning language that would have unduly cluttered the drug label, given that the label already reflects FDA’s excess caution.
The reason that jury decisions may be superior to FDA decisions is that jurors often have superior information upon which to base their decision. Thanks to the discovery process, jurors often see reports, memos, and emails that pharmaceuticals specifically chose not to send to the FDA, such as internal concerns over drug safety and efficacy. Perhaps this wouldn't be so if Congress chose to grant the FDA subpoena power.
The impact of this added layer of tort liability overdeterrence leads to numerous adverse consequences that are contrary to public wellbeing, such as:
1. Limiting drug availability: Because prescription drug product liability lawsuits often involve allegations of substantial harm, plaintiff verdicts not uncommonly involve sizable awards for pain and suffering damages, punitive damages, or both. Experience has shown that pharmaceutical firms often treat a large damages verdict as a predictor of more such verdicts to come. Cite omitted
While estimates of the cost of liability for pharmaceuticals are few, liability costs are not trivial. For example, a report prepared by the Council of Economic Advisors found that in 2000, liability costs across all U.S. industries were $180 billion, or roughly 1.8 percent of GDP.5 In the area of drugs and medical devices, Richard Manning identified liability costs for the diphtheria-pertussis-tetanus vaccine by comparing changes in the price of the diphtheriatetanus vaccine, estimating that at their peak, expected liability costs accounted for roughly 90 percent of the vaccine’s price. In related work, Manning found that differences in product liability regimes can explain much of the difference in Canadian and U.S. prices on drugs.
The total liability costs of all U.S. industry is of no value in determining whether pharmaceuticals bear an undue burden. But let’s put that $180 billion in perspective. Wal-Mart’s total sales last year were nearly twice that amount at $348 billion. When a single U.S. company’s sales are almost twice the entire liability costs of “all U.S. industries,” does that really suggest that the tort system is out of control?
The expected costs of product liability litigation are added to a firm’s cost-benefit analysis in the marketing of drug products and can lead to the unavailability of drug products, both because some patients may no longer be able to afford treatment and because the expected litigation costs may drive products off the market entirely. For example, childhood vaccine manufacturers quickly raised prices and in many cases exited the market in the wake of a few adverse verdicts in the 1980s. The anti-nausea drug Bendectin was withdrawn from the market in 1983 after a small number of adverse verdicts (which were subsequently reversed), despite FDA insistence that the drug did not cause the birth defects that gave rise to litigation. See Paul H. Rubin, John E. Calfee, and Mark F. Grady, BMW vs Gore: Mitigating the Punitive Economics of Punitive Damages, Sup. Ct. Econ. Rev. vol. 5, pp. 179, 194 (1997); see also American Medical Association, Report of the Board of Trustees, Impact Of Product Liability On The Development Of New Medical Technologies 1, 79 (1988) (“AMA Board Report”) (“Certain older technologies have been removed from the market, not because of sound scientific evidence indicating lack of safety or efficacy 6 Richard M. Manning, but because product liability suits have exposed manufacturers to unacceptable financial risks.”).
The "reform" movement always alleges that but for the tort system, a slew of new miracle drugs would save us all from the ravages of disease. Yet the best example they can come up with is Benedictin, an anti-nausea drug taken off the market after a few bogus lawsuits. So why isn't Benedictin back on the market in the U.S.? Probably because Benedictin is made of two ingredients: Vitamin B-6 and doxylamine succlinate. You know where to find Vitamin B-6. But where to find doxylamine succlinate? Buy some Unisom or Nyquil - it's the active ingredient in both. 25mg in the former and 6.25 mg in the latter, plus varying amounts in about a dozen knock-offs. Now, why on Earth would a pharmaceutical spend the millions of dollars to bring Benedictin back to market when it (a) wouldn't receive patent protection and (b) can be exactly duplicated by taking a cheap vitamin pill and a Unisom?
If fear of lawsuits were what was keeping Benedictin off the market, its active ingredient wouldn't be found in over a dozen cold & flu remedies. Benedictin is hardly evidence that lawsuits are keeping medication off the market.
2. Disincentives for Research & Development:
The costs imposed by product liability litigation have an impact not only on the availability of drugs already on the market but on the pipeline for new drugs, because pharmaceutical companies necessarily factor these costs into decisions whether to incur the significant expense required to research and develop drug products and bring them to the market. Where the level of risk is high, the risk of state tort law liability is inversely related to investment in research and development activity. Firms invest less in drugs that promise to provide substantial clinical benefits but also have significant side-effects (such as many drugs for cancer and multiple sclerosis). Thus, particularly to the extent companies are faced with state tort liability risks that are largely divorced from the FDA balancing of risks and benefits in drug and drug labeling approval, the financial payoff from researching these drugs and bringing them to market is reduced, causing fewer such drugs to be made available to patients.
The economists also fail to point out one of the fundamental economic decisions that go into new drug research: Expected sales. The market for cancer and multiple sclerosis drugs is much smaller than the market for erectile dysfunction drugs, or arthritis drugs, or weight loss drugs, or a variety of other drugs that treat more common illnesses. The harsh economic reality is that firms invest less in drugs that promise to provide substantial benefits to a small number of people than drugs that promise to provide marginal benefits to a large number of people. A pill that cures baldness will be far more profitable than one that cures multiple sclerosis.
3. Loss of FDA control over drug labeling:
Even a single costly loss in a state tort lawsuit alleging failure to warn imposes a legal standard that effectively requires a manufacturer to make changes in its labeling to avoid anticipated costs of future litigation. Although it theoretically might be possible to change labels only for drugs sold in the state in which the lawsuit was brought (assuming such a state-by-state approach was legal [Wyeth and the FDA contend that it isn’t.]), this seems unlikely. For drugs used in doctors’ offices, a state-by-state labeling arrangement would require considerable control over the actions of wholesalers, who are normally free to resell their supply more or less where they please. [Such an arrangement would probably violate the Commerce clause of the Constitution, too.]It would also be hard to prevent patients from using the drug in another state after purchase. Finally, advertising and marketing, which involve risk communication, would become far more complex. Thus it seems likely that the label would be changed nationwide, not just in the state in which the adverse verdict occurred. As a result, a single state tort judgment could effectively wrest control of nationwide drug labeling requirements from FDA, thus depriving the medical community in all states of the benefits of FDA expert determinations on proper and balanced drug warnings.
The economists misunderstand the extent of FDA regulation – the FDA also regulates the content of advertisement and marketing materials. If, as the FDA and Wyeth contend, a pharmaceutical may not add warnings to a drug label without FDA approval, neither can the pharmaceutical add warnings to a drug advertisement.
4. Defensive labeling:
Firms that suffer adverse verdicts would reasonably attempt to predict other warnings which, if added to the label, might prevent costly litigation in the future. Their competitors, who can be expected to pay close attention to litigation and its outcomes, would perform similar analyses on their own drugs. With no way to know exactly what would be required through future litigation, firms are likely to add new warning language and contraindications and create overly crowded labels that seek to anticipate a wide variety of potential plaintiff allegations. Much of this new warning information would have greater prominence than would be justified by the balance of risks and benefits (precisely because FDA declined to require such information notwithstanding its tendency toward excessive caution). The effect would be to discourage beneficial use of drugs whose labels contain these litigation-induced contraindications and warnings.
As the FDA and others have long recognized, faced with state tort liability regardless of FDA approval of specific warning language, manufacturers may seek to supply warnings about virtually all possible harms that might form the basis for a lawsuit. For example, a series of Wall Street Journal articles published in 2005 noted that the three erectile dysfunction drugs on the market each carried labels more than 20 pages long. This litigation-defensive manufacturer tendency to clutter labels with risk information exacerbates FDA’s tendency toward excess caution in drug labeling.
Perhaps the economists meant to file this brief on behalf of Levine, because they repeatedly contradict Wyeth's position that it legally cannot "clutter labels." The United States filed a brief in this case as well, and here’s what it had to say about manufacturers cluttering drug labels to avoid litigation:
“Thus, any changes to a drug’s labeling without prior FDA approval still must be the subject of a supplemental application, which FDA can approve or reject, and must be based on material new information—not information that was previously available to FDA, nor even cumulative new information that does not add materially to the information that was previously available to the agency. As the dissent explained, Section 314.70(c) does not “allow manufacturers to simply reassess and draw different conclusions regarding the same risks and benefits already balanced by the FDA.” - United States brief
Thus, under Wyeth and the FDA’s interpretation of the law, there is only one circumstance in which a pharmaceutical may change its label if it loses a product liability lawsuit: If the pharmaceutical is forced to produce documents in the litigation that it previously withheld from the FDA and the FDA then approves changes based upon those documents. In such an instance, litigation would actually have corrected a Type I error. (Although it’s terribly unfair to attribute an error to the FDA under those circumstances.)
5. Problems exacerbated further in state tort cases alleging missing contraindications:
The concerns discussed above based upon state tort law requirements for added warnings take on particular prominence where, as in this case, a plaintiff is arguing that a drug should have contained a contraindication not required by FDA. Normally, the risk-benefit balancing process for prescription drugs operates at two points: first, when FDA decides what warnings to include on the label and how those warnings are organized, and second, when physicians take account of label warnings as they decide what drugs to prescribe. If failure-to-warn lawsuits cause unfounded or excessive warnings to be placed on the label, physicians can still exercise their usual role in balancing risks and benefits, albeit with less accurate information than would otherwise be available.
Now the economists state that the FDA “decides what warnings to include on the label” but just one sentence later revert to the assumption that litigation will dictate the content of warning labels. The FDA is never a party to product liability lawsuits. A jury verdict is not binding upon the FDA in any way. Nor does a jury verdict force a pharmaceutical to change its label after a verdict. It is the position of the FDA that only it can determine when to modify a warning label, and it will not revisit that decision because of an adverse jury verdict. The FDA and Wyeth argue that if a manufacturer wants to add a warning label after a jury verdict, it would have to submit new information to the FDA in support of that warning label.
Contraindications work differently. Physicians are likely to view contraindications as outright bans, because to prescribe in the face of a labeled contraindication is to court a malpractice lawsuit and punitive damages if anything goes wrong. Contraindications therefore largely replace, rather than supplement, the usual balancing of risks and benefits.
Even if we assume that contraindications do “largely replace, rather than supplement” the risk benefit balance, preemption won’t change how physicians treat contraindications.
Again, there is no reason to expect juries in tort liability trials to perform a better balancing act than FDA. Juries will tend to impose new contraindications, which would prevent physicians from taking due account of comparative risks and benefits in the highly fact-specific situations in which physicians often making prescribing decisions. As the dynamics of litigation play out, the problem will probably become worse. A contraindication imposed by a jury in one state will likely be translated by pharmaceutical firms into a nationwide contraindication because of the practical inability to limit litigation exposure by single-state label changes. If juries in certain states or regions are especially inclined to find fault with drug labels and impose contraindications, physicians in other states will likely face the same contraindications as label changes are implemented nationwide. Then patients who would have benefited from the contraindicated use will be denied those benefits even if the expected benefit greatly exceeds the likelihood of harm. In extreme cases, manufacturers may choose to remove useful drugs from the market, as happened with many childhood vaccines in the 1980s.
Once more, the economists contradict the FDA and Wyeth, both of whom argue very strenuously that neither jurors nor manufacturers can unilaterally add new contraindications to a warning label. Perhaps Levine will cede time at oral argument to let the economists argue that Wyeth and the FDA are wrong.
The question whether state tort litigation can complement public safety by imposing requirements in excess of those imposed by FDA necessarily depends in part on whether FDA regulation itself is insufficiently or overly cautious. Because FDA is faced with incentives that lead it to stake out overly cautious positions on drug approval and drug labeling, state tort litigation imposing additional requirements leads to a further departure away from the most socially beneficial outcome.
Given this background, it is clear that the public health would only be improved if state tort lawsuits like the one below were held preempted. For the reasons set forth above, the economist Amici, Messrs. Calfee, Berndt, Hahn, Philipson, Rubin, and Viscusi, respectfully submit that the decision of the Supreme Court of Vermont should be reversed.
If anything is clear, it's that preemption won't solve and may in fact exacerbate many of the problems the economists complain of. Even if the FDA is predisposed to commit Type II errors, preemption won't eliminate the incentives for it to do so. It especially won't eliminate the large financial incentive for the agency to commit a Type II error, and in fact the economists don't even mention the financial incentive.