Letter on the Ability of Drug Companies to Add Safety Warnings to Product Labels

March 17, 2008

Re: Supplemental Applications Proposing Labeling Changes for Approved Drugs, Biologics, and Medical Devices (Docket No. 2008N-0021)

Division of Dockets Management (HFA-305)
Food and Drug Administration
5630 Fishers Lane
Room 1061
Rockville, MD 20852

To whom it may concern:

If one thing was clear from the debate over the Food and Drug Administration Amendments Act (FDAAA) of 2007, it was that the FDA had failed to satisfy the public’s thirst for prompt warnings about emerging issues in pharmaceutical safety.  Accordingly, the FDAAA provided the FDA with new authority to impose certain drug labeling changes. 

The legislation came in the wake of a series of drug safety debacles.  Prominent among these was the Vioxx (rofecoxib) case, in which the sponsor was able to string the FDA along for at least 14 months as it haggled with the FDA about appropriate changes in the drug’s label that would reflect the drug’s propensity to cause heart attacks and strokes; in the interim patients and consumers were kept in the dark.  Similarly, it took over a year from when the FDA first obtained information connecting Avandia (rosiglitazone) to heart attacks before the agency changed the drug’s label. 

An Institute of Medicine Report on drug safety released before the passage of the FDAAA had concluded that, “FDA and the pharmaceutical industry do not consistently demonstrate accountability and transparency to the public by communicating safety concerns in a timely and effective fashion.”[1] The agency has, at least to some extent, taken these criticisms to heart, having released 14 safety alerts in a four-month period, more than it did in some entire years.[2]

Yet, in the midst of the public outcry for more information, the FDA is proposing to restrict the conditions under which certain safety label changes can be made.[3]  At issue are the so-called changes being effected (CBE) supplements, under which the sponsor can, under certain circumstances, make changes to a drug’s label unilaterally.  (The FDA retains the authority to subsequently overrule any changes made.)  The FDA proposes that such label changes be based only upon information that (1) demonstrates a “causal association” between the drug and the adverse event and (2) represents “newly acquired information.”

First, the proposed regulations’ insistence on a causal relationship between the drug and adverse events should not be implemented.  At present, the conditions under which a CBE supplement can be implemented are not defined in the regulations.  The FDA proposes limiting CBEs to instances in which there is evidence of a “causal association” to conform to changes made voluntarily under 314.70(c) to those mandatory labeling updates required by 21 CFR 201.57(c).

As the regulations appear to accept, causality is not simply a factor that is or is not present.  Rather, there is a continuum of causality ranging from weak evidence to suggestive evidence to near-ironclad evidence.  In microbiology, for example, a causal relationship is assumed only when the extremely stringent Koch’s Postulates have been satisfied.  In pharmacoepidemiology, the closest to that standard is the randomized, controlled trial, but even the implications of these trials are often hotly debated.  Moreover, randomized, controlled trials are not often performed primarily to assess safety, in part because of the rarity of many adverse reactions and in part because of sponsors’ aversion to the potential bad news.  The majority of drugs removed from the market are therefore withdrawn on the basis of spontaneously reported adverse events, not on the basis of clinical trials.  These adverse events gather greater steam as evidence of causality to the extent that the adverse event is unusual in its clinical characteristics, its background rate is low, it is rendered plausible by animal or basic science evidence and there are no alternative explanations for the event.  Replication in other epidemiological studies and the presence of a dose-response relationship also strengthen the inference of causality.  The presence and strength of these factors combine to produce an assessment of the overall strength of the causal relationship.

By contrast, pharmaceutical companies often treat causal inference in the safety arena in binary fashion: It either is or (more commonly) is not present.  It is standard industry public relations to greet the announcement of any risk suggested by adverse event reports with the proclamation “These reports do not prove a causal relationship.”[4],[5],[6]  Even risks clearly demonstrated in randomized, controlled trials are denied (see e.g., suicidality and selective serotonin reuptake inhibitors, Avandia and Vioxx).  It seems likely, therefore, that, under the guise of industry’s devotion to stringent epidemiological principles, this “causal association” standard will be used to delay needed label changes that might provide doctors and patients with early warning of a drug hazard.[7]

Furthermore, the criteria under section 201.57(c) for mandatory updates to certain information (e.g., contraindications, warnings and precautions) in the drug label is a substantially different matter from the conditions under which a company may permissibly update safety information without FDA pre-approval.  The proposed rule offers no reason why these two regulations should be based on the same standard, and from a public health perspective, they need not be.  To require that labeling updates must be made when there is “reasonable evidence of a causal relationship” [201.57(c)] but to allow companies greater flexibility to decide whether an immediate update is warranted based on evidence of less certain causality [314.70] as the current rules provide is a logical scheme and need not be modified.  Moreover, section 201.57(c) does not apply to drugs approved before 2001 or to devices, and yet the amended section 314.70(c) rule would use its standard for those products as well.

Second, with respect to the proposed rule’s requirement for “newly acquired information,” such information might not include the failure of a previous warning to adequately reduce the incidence of a previously identified adverse event.  Although, to our knowledge, not the focus of a CBE, the Rezulin (troglitazone) case illustrates this issue well.  The FDA required a series of changes to the drug label, each requiring more stringent monitoring of patients for hepatotoxicity.  Evidence later published by the FDA demonstrated that these warnings had little impact on clinician practices.[8]  Were a company to learn that prior labeling changes had been unsuccessful, would this qualify as “newly acquired information”?  We believe it should.  In general, companies learn of adverse events (and thus of the impact of label changes) before the FDA does (as much as a year before in some circumstances) and so are well placed to assess the impact of label changes expeditiously and to make label changes accordingly.

The FDA claims that these new proposals simply “codify the agency’s longstanding view[s].”  We do not recall the FDA taking this view until the past couple of years.  Even if the proposal did reflect a longstanding view, one questions why the agency would expend the time and effort necessary to complete such a rulemaking in the midst of ongoing concerns about the agency’s ability to, among a slew of other items on its plate, ensure the safety and efficacy of medications and prevent the importation of tainted pharmaceutical products from abroad.  The agency has failed to demonstrate a need for revising this regulation, nor has it been able to point to a single case in which lack of clarity over the CBE supplement requirements has interfered with the process of changing a drug’s label or adversely affected public health.  We urge you to withdraw this ill-considered proposal.

Yours sincerely,

Peter Lurie, M.D., M.P.H.
Deputy Director

Sidney M. Wolfe, M.D.
Health Research Group at Public Citizen

[1] Institute of Medicine. The Future of Drug Safety: Promoting and Protecting the Health of the Public. National Academy of Sciences. September 2006.

[2] Kritz FL. Why all the drug alerts? Step-up signals FDA bid for renewed trust. Washington Post, February 26, 2008, p. HE01.

[3] Our comments generally cite regulations and provide examples for drugs, but our comments apply equally to biologics and medical devices.

[4] Stolberg SG. Heart drug [Posicor] withdrawn as evidence shows it could be lethal. New York Times, June 9, 1998.

[5] Willman D. Drug Lotronex pulled over safety fears. Los Angeles Times, November 29, 2000, p. A1.

[6] Appleby J. What do you believe when drug messages conflict? USA Today, December 27, 2004.

[7] Our concern is further amplified by the manner that the Proposed Rule’s Summary and Preamble characterize the proposed regulation.   While 21 CFR 201.57(c) typically refers to the need for “reasonable evidence” of a causal association before the safety information can be included, the Summary and Preamble instead refer to the need for “sufficient evidence of a causal association.”

[8] Graham DJ, Drinkard CR, Shatin D, Tsong Y, Burgess MJ. Liver enzyme monitoring in patients treated with troglitazone. Journal of the American Medical Association 2001; 286:831-3.