doi: 10.1161/CIRCULATIONAHA.105.604512
(Circulation. 2006;113:2253-2259.)
© 2006 American Heart Association, Inc.
Special Report |
Clinical Trials in the Wake of Vioxx
Requiring Statistically Extreme Evidence of Benefit to Ensure the Safety of New Drugs
From the Department of Statistics, University of Wisconsin, Madison.
Correspondence to Michelle D. Roth-Cline, BS, Department of Statistics, University of Wisconsin-Madison, K6 Box 4675, 600 Highland Ave, Madison WI 53792. E-mail mdrothcline{at}wisc.edu
Key Words: clinical trials • ethics • statistics
 |
Introduction |
|---|
 Top Introduction Two True PremisesThe Third PremiseThe Fourth PremiseToward a Coherent PolicyReferences |
|---|
Clinical trials have become the preeminent tool for investigating new diagnostic, therapeutic, and preventive treatment strategies. During these trials, periodic interim data analyses are conducted by data monitoring committees (DMCs) to ensure participant safety. If accumulated data indicate harm to participants, clear evidence of superior efficacy, or futility of collecting additional data, studies may be terminated before their planned conclusion.
Editorial p 2173
These termination decisions are invariably complex and involve important subjective and ethical elements.1 Of greatest concern when studies are terminated because of clear benefit is the possibility that the superior intervention has toxicities that remain undiscovered. Early termination entails the abbreviation of clinical experience that would permit assessment of the longer-term safety profile. However, study continuation when treatments are not equally effective also means that some participants will continue to receive a less effective therapy.
The tension between protecting both trial participants and future patients has been increasingly acute since rofecoxib (Vioxx [Merck]) and other cyclooxygenase-2 inhibitors were implicated in a greater risk of myocardial infarction. Thrombotic side effects of cyclooxygenase-2 inhibition were postulated as a result of the propensity for selective antagonists to decrease vascular prostacyclin production and possibly affect the balance between prothrombotic and antithrombotic eicosanoids.2 Little evidence of toxicity was evident in preapproval trials, however, and rofecoxib was approved in 1999 for relief of the signs and symptoms of osteoarthritis, the management of acute pain, and the treatment of primary dysmennorhea.3
The possibility of cardiac toxicity in human trials was recognized in the Vioxx Gastrointestinal Outcomes Research (VIGOR) trial.4 Meta-analyses, database studies, and prospective trials further corroborated cardiac toxicity,5–8 and Merck voluntarily withdrew rofecoxib from the US market on September 30, 2004,9 amid widespread criticism of both Merck and the Food and Drug Administration.10,11
In the wake of the withdrawal, drug safety has received increasing public attention. The need to ensure the reliability of safety information has prompted proposals to increase the length of preapproval trials to obtain longer-term safety information. One such strategy recently used by the planning committees for some definitive cardiovascular trials, in consultation with the Food and Drug Administration, is to require statistically extreme (eg, P<0.0001) evidence of benefit on several end points before considering early termination (personal communication, David DeMets, PhD, 2005). Typical stopping guidelines suggest termination when statistically conservative but more moderate (eg, P<0.005) evidence of benefit is attained on key end points, provided that other outcomes are favorable and consistent.12 Whereas as others have noted the possibility exists that early termination may lead to premature adoption of study findings,13 at issue here is the likelihood that toxicities would be observed with trial continuation. Although the utility of extending trials may seem self-evident, this analysis will show that obtaining toxicity information in this manner is less certain than it initially appears.
The argument for requiring statistically extreme evidence appears to be based on 4 premises: (1) If statistically extreme evidence of benefit were required for early termination, clinical trials would be longer; (2) if clinical trials were longer, additional toxicity data might be obtained; (3) if additional toxicity data were obtained as a result of requiring statistically extreme evidence of benefit, toxicities that had a bearing on future approvability might be observed that would otherwise remain hidden; and (4) the policy of requiring this additional toxicity data for the purpose of preventing the approval of therapies with unobserved toxicities is coherent.
In what follows, these premises are examined by considering the likely effects of imposing the requirement on the likelihood of approving drugs with hidden toxicities.
|
Two True Premises |
|---|
|
TopIntroductionTwo True PremisesThe Third PremiseThe Fourth PremiseToward a Coherent PolicyReferences |
|---|
(1) If statistically extreme evidence of benefit were required for early termination, clinical trials would be longer.
This first premise is prima facie true. Suppose a trial were conducted that compared 2 therapies intended to decrease mortality among patients with heart failure. Sample size is unaffected by the extreme monitoring boundary, because at the end of the study, the critical value for statistical significance remains at conventional levels. Suppose that at the time of interim analysis, patients on treatment A had a lower death rate than those on treatment B and that the probability value for benefit was 0.01. If more evidence of benefit were required (P=0.0001), additional evidence of benefit on treatment A would have to accumulate. Because of the time required to observe these additional outcomes, the length of the trial must increase, and termination would be delayed. Therefore, requiring statistically extreme evidence generally entails a longer trial, and the first premise is true.
(2) If clinical trials were longer, additional toxicity data might be obtained.
This second premise is also prima facie true. One of the disadvantages of early termination is that less long-term clinical experience with the new therapy is obtained. The probability of observing more adverse events (AEs) generally increases with patient exposure, whether the toxicity is based on cumulative dose or occurs at a steady state. So, combining the first 2 premises, if statistically extreme evidence of benefit were required for early termination, additional toxicity data might be obtained.
However, this apparently straightforward conclusion masks one obvious caveat. If preapproval trials are shorter than the necessary exposure period or do not administer a sufficient dose, toxicities would not be observed regardless of stopping criteria. The utility of toxicity data also may be limited by a lack of external validity or by inadequacies in data monitoring and analysis, but these limitations exist regardless of stopping criteria. Even if we assume that preapproval trials would capture all of these toxicities, the question remains of how much additional information would be gained by forgoing early termination.
The amount of toxicity information that could be obtained by delaying termination is unknown and unknowable a priori. However, phase III trials for life-threatening conditions often include mortality as an end point and are expected to follow up on a large number of participants for several months or even years. Early termination of these trials for benefit may reasonably be expected to shorten trial length by a year or more14–16 and thereby decrease the toxicity data obtained. So, using statistically extreme stopping criteria in these cases might increase toxicity information, subject to 2 additional caveats.
First, requiring highly significant evidence in some phase III trials would not have yielded additional toxicity information. Metoprolol CR/XL Randomised Intervention Trial in Congestive Heart Failure (MERIT-HF),15 Carvedilol Prospective Randomized Cumulative Survival (COPERNICUS),17 and Cardiac Insufficiency Bisoprolol Study II (CIBIS-II)16 are examples of trials in cardiology in which statistically extreme evidence of benefit was attained. In these trials, the evidence developed so rapidly between scheduled DMC meetings that termination before accumulation of highly significant evidence was unfeasible. For similarly efficacious therapies, statistically extreme evidence may be expected to develop regardless of initial stopping guidelines.
Second, one of the important considerations involved in the early termination of trials is previous clinical experience with the intervention. The ß-Blocker Heart Attack Trial (BHAT) is an example of a phase III study that was terminated for benefit with a nominal probability value on the primary outcome of 0.005,14 which is already substantially less than a nominal probability value of 0.05. If the observed mortality trend had been allowed to continue until the probability value reached 0.0001, the trial would have been an estimated 92% longer. Trial continuation in this case might have been expected to yield a corresponding increase in the possibility of discovering unknown toxicities.
However, propranolol was approved as an antihypertensive before BHAT, so the safety profile was already established in a population reasonably comparable to the targeted study population. The probability of detecting an unknown toxicity with trial continuation was probably small. Smaller phase III trials are sometimes terminated with moderate evidence of benefit when the safety profile is unknown, but in these cases, relatively few patient-years of follow-up would have been gained with trial continuation.18
As these examples illustrate, using statistically extreme stopping criteria may lead to additional toxicity information on longer trials of therapies for which the toxicity profile is not well established. For the requirement to be applicable, the evidence of benefit must develop gradually so that the DMC has the opportunity to examine the data before extreme evidence is attained. The next question is whether toxicities with a bearing on future approvability would be observed under these conditions.
|
The Third Premise |
|---|
|
TopIntroductionTwo True PremisesThe Third PremiseThe Fourth PremiseToward a Coherent PolicyReferences |
|---|
(3) If additional toxicity data were obtained as a result of requiring statistically extreme evidence of benefit, toxicities that had a bearing on future approvability might be observed that would otherwise remain hidden.
For the toxicity of an agent to have a bearing on future approvability, the toxicity must be serious enough to alter the risk-benefit profile substantially and must be detectable during clinical trials. For a serious toxicity to be detected, 1 of 2 things must occur: Either the agent must be associated with a toxicity that is rare in the population under study, or it must substantially increase the frequency of a common toxicity relative to its comparator. If the study population were adults >60 years of age, a few cases of a rare AE such as rhabdomyolysis might be enough to raise safety concerns. If the AE were myocardial infarctions, only a substantial imbalance in the number of events between treatment groups might be considered indicative of toxicity. Examples of both scenarios are considered below.
Cerivastatin was an HMG Co-A reductase inhibitor (statin) that was withdrawn from the market in August 2001 because of high reporting of rhabdomyolysis associated with its use. A database study of claims from large health plans indicated that to observe 1 case of hospitalized rhabdomyolysis, 1873 patients needed to be treated with cerivastatin for 1 year.19 If hundreds of additional person-years of exposure could be gained by forgoing early termination as suggested previously, we might believe that the probability of observing 1 or more cases of rhabdomyolysis is reasonably high and might therefore influence approvability.
Although in some circumstances additional AEs may be observed by applying the requirement for extreme evidence, these situations are limited primarily by 2 factors. Trials that necessitate stopping criteria for benefit are usually very large and long term, with major morbidity or mortality as outcomes. The majority of preapproval trials are intended to demonstrate short-term outcomes like cholesterol reduction or pain relief, so only a few hundred participants need to be treated with a short course of therapy. To make the extreme evidence requirement generally applicable in preapproval studies, these shorter trials would need to be redesigned to include long-term outcomes, an option that will be discussed shortly.
Although short-term pivotal phase IIb/III trials like those for cerivastatin often have open-label extension studies that follow up on participants for about a year,20 the requirement for statistically extreme evidence again is rarely applicable because early termination for benefit in extension studies is rare. All patients frequently receive the study intervention to maximize patient exposure, which prevents a long-term assessment of relative efficacy. The initial stage of a phase III study is optimized to assess efficacy, and a controlled extension study would probably not be pursued if efficacy were established in the initial stage. Placebo-controlled extension studies have occasionally been pursued for therapies such as nonsteroidal antiinflammatory drugs,21 but usually because the placebo response rate is high. In these cases, the difference in apparent efficacy between groups would not be expected to become highly significant.
Even if requiring studies with a thousand or more person-years of follow-up were desirable for all investigational drugs, a closer examination of the 17 Food and Drug Administration–approved drugs withdrawn from the market since 1990 reveals that only some have primary indications that clearly support longer-term phase III studies (Table). Antiarrhythmics,22 antihypertensives,15 and statins23 have all been tested in longer-term phase III trials with major morbidity or mortality as outcomes. Although antibiotics have been used in rare trials that necessitate longer-term treatment,24 these studies are not representative of typical use patterns and have not been used in phase III. In fact, shorter-term trials for which the extreme evidence requirement would not be applicable are characteristic of phase III studies for most of the indications in the Table.
|
Supposing all preapproval trials were long enough to sustain the extreme evidence requirement, a second factor may limit the toxicity information lost by early termination. Person-years of drug exposure are calculated by multiplying the number of persons by the follow-up time measured in years. However, the toxicities observed in an additional 1000 person-years of follow-up are not necessarily equal to the toxicities observed by following up 1000 patients for 1 year. For example, the information obtained by following up 500 people for 2 years is likely to be different from the information obtained by treating 4000 people for 3 months, although the person-time is the same in each case. Toxicity information gained by person-years of follow-up acquired in different ways is equal only if the probability of an AE per unit time is constant over the course of the study. In clinical trials, this assumption is probably unrealistic. Adding to the uncertainty, participant entry is staggered, so there would be no uniformity to the individual follow-up periods that were lost. Although requiring statistically extreme evidence may entail additional person-years of exposure, it is less clear that it would entail more information about a particular toxicity.
Even if additional person-years of exposure would invariably result in relevant toxicity information, the exposure necessary to influence drug approval is substantial, as illustrated in the Figure. This figure was generated by assuming a dichotomous response variable and computing a fixed number of AEs using a standard binomial probability distribution. Specifically, the calculations assume a 90% probability of detecting at least 1 to 5 or more AEs occurring at rates ranging from 1:10 to 1:10 000. When each estimate is multiplied by 1 person-year and the simple assumption noted earlier is made, the estimates are also equal to the number of person-years of additional follow-up necessary to detect AEs 90% of the time at each rate. A more complicated model could be generated using an exponential function for time to events, but the conclusion would be the same: An untenably large number of participants are required to detect very rare AEs.
|
As noted earlier, a database study suggests that 1873 patients need to be treated with cerivastatin for 1 year to observe 1 case of rhabdomyolysis.19 From the above calculations, &4300 participants would have to be treated for 1 year to detect at least 1 case, 10 000 to detect at least 3 cases, and 15 000 to observe at least 5 cases. Only 1 statin trial was large enough to detect these outcomes.23 Although the incidence of rhabdomyolysis is greater with combined statin-fibrate therapy than with monotherapy, participants taking the combination in preapproval trials were at most a subgroup of those in the treated arm. In either case, statin-induced rhabdomyolysis would probably not have been observed in preapproval trials frequently enough to affect approvability.
Other drugs withdrawn since 1990 also had AEs that were too rare or incidental to be detectable even with longer-term clinical trials and thus would not have been detected regardless of stopping criteria. Bromfenac-associated hepatotoxicity occurred in only 1 in 20 000 patients taking the drug for >10 days.25 About 50 reports of any serious AE relating to temafloxacin occurred in the 6 months after the drug was marketed, although perhaps 500 000 patients received the drug.26 Even if only 1 in 5 of these very serious adverse effects were reported, the incidence of AEs would still only have been 1 in 2000. Terfenadine, mibefradil, astemizole, and cisapride were all recalled primarily because of drug interaction–induced toxicity rather than because monotherapy with the drug was unsafe.
We now turn to the scenario concerning the detection of an increased relative frequency of a common toxicity. Rofecoxib-associated myocardial infarctions are consistent with this profile. As with the statin example, trial length and design considerations would obviate application of the requirement for shorter trials. In larger trials, the difference in the rate of AEs between arms of the trial would have to be unexpected, reasonably large, and practically detectable. As another example illustrates, this is not always the case.
The antibiotic grepafloxacin was withdrawn in 1999 because of ventricular arrhythmias stemming from a prolonged QT interval, even though clinical trials of grepafloxacin did not exclude patients at risk for QT interval prolongation.27 QT prolongation is a common side effect of approved drugs in many classes, including other fluroquinolones.28,29 Although effects vary between drugs, a summary of animal QT interval data related to fluroquinolone use suggests that grepafloxacin was comparable to other marketed fluroquinolones.27 Some QT prolongation in grepafloxacin trials might have been expected and probably allowable. Finally, assessments of the mean change in QTc intervals in clinical studies suffer from a number of technical limitations. Even if preapproval trials had been long enough to use statistically extreme stopping criteria, it is still not clear that grepafloxacin toxicity sufficient to bar approval would have been detected.
In longer-term phase III studies that include major morbidity or mortality as outcomes, it is more probable that early termination for benefit could reduce the probability of detecting toxicities with a bearing on approvability. However, DMCs look carefully at accrued data before trials are terminated and, regardless of early termination guidelines, would probably not recommend discontinuation for benefit if the efficacious study intervention were also associated with an increased risk of AEs. Therefore, the requirement for extreme evidence would be important only if treatment-associated toxicity were absent or minor at the time statistically moderate evidence of benefit was realized.
Slightly different questions arise when the reason for extending the trial is to obtain a more precise statistical estimate of an anticipated risk. In a trial of a new anticoagulant, for example, it might be important to characterize the risk of uncontrolled bleeding. The underlying assumption relating to approvability decisions is that a precise estimate of the risk would afford an accurate risk-benefit assessment and that an accurate assessment would prevent approval of a drug with side effects sufficient to warrant later withdrawal.
The veracity of the assumption is more uncertain than is initially evident. For the requirement to yield valuable information, the trial must be large and the toxicity relatively frequent but not frequent enough to warrant termination for harm. Of the drugs that have been withdrawn since 1990, only 2 had primary indications that would clearly support long-term phase III studies and had a toxicity that was relatively frequent. The requirement per se would not have prevented approval in either case. Encainide proved so toxic in the Cardiac Arrhythmia Suppression Trial that the trial was terminated for harm.22 Similarly, flosequinan was withdrawn in 1993 after the interim results of the PROFILE trial indicated that patients taking the drug had an increased risk of hospitalization or death.30 Termination criteria for benefit in these cases would have been immaterial. Therefore, by definition, the relatively imprecise estimates obtained in trials to date have been adequate to ensure that drugs such as the anticoagulant in the above example were rightly approved. It is unclear why future trials merit a different standard.
Even if precise estimates of risk were necessary, a well-known limitation of clinical trials is the generalizability of results. Issues such as compliance, comorbidities, genetics, concomitant medications, age, and socioeconomic status may mean that trial volunteers are different from nonvolunteers.31–33 Some cardiovascular studies have suggested comparability between trial results and outcomes in clinical practice,34,35 whereas others found substantial variation.31,36–38 Systematic differences may exist in the quality of care received by participants in clinical trials compared with patients in general practice,38,39 and physicians’ adherence to evidence-based practice guidelines also may be uneven.40,41 For all these reasons, risk estimates generated in a clinical trial may be expected to vary in a different population. Although trial results may indicate the general acceptability of the risk-benefit profile, a mere increase in the precision of the estimate would not necessarily make it robust.
Finally, the uncorroborated results of a single long-term trial that suggests the toxicity of a new drug, although worrisome, might not impede approval unless the evidence of harm was unequivocal or the risk-benefit profile was clearly unacceptable relative to other treatment options. The interim results of the Adenomatous Polyp Prevention on VIOXX (APPROVe) trial that prompted the rofecoxib withdrawal9 appeared momentous in part because prospective trials, database studies, and meta-analyses had previously suggested an increased risk of cardiac events in a drug marketed for pain relief.
In summary, requiring statistically extreme evidence would increase the likelihood of observing toxicity data with a bearing on approvability only in select circumstances. The trial must be long enough to permit interim analysis. Early termination would have to substantially reduce the number of toxicities observed. For the requirement to yield useful information, a rare toxicity must be sufficiently common with drug treatment to be observed in the period after the DMC would have otherwise terminated the trial. Alternatively, a greater number of events in the treated arm would have to occur during the same period. Any imbalance in the number of AEs also would have to be small enough when moderate evidence of benefit was achieved that the DMC would recommend termination in the absence of the extreme evidence requirement. Finally, to fulfill the third premise, the toxicity information observed as a result of the requirement must be sufficiently compelling to impede approval.
|
The Fourth Premise |
|---|
|
TopIntroductionTwo True PremisesThe Third PremiseThe Fourth PremiseToward a Coherent PolicyReferences |
|---|
(4) The policy of requiring statistically extreme evidence of benefit for the purpose of preventing the approval of therapies with unobserved toxicities is coherent.
The availability of alternatives and risk-benefit determinations lie at the heart of the decision to market drugs. Despite previous knowledge of drug interactions leading to cardiac arrhythmias, for example, terfenadine was withdrawn only after fexofenadine, the active metabolite, became available. Astemizole was explicitly withdrawn because of the availability of other prescription antihistamines.42 Although &1 in 1000 people taking alosetron is expected to have serious constipation and 1 in 350 to have ischemic colitis, it was reintroduced to the market in 2002 because of the paucity of other effective therapies for irritable bowel syndrome.43 Therefore, even if serious AEs were observed because of the extreme evidence requirement, those toxicities would affect approvability only pending an assessment of the overall risk-benefit profile.
Additionally, lengthening all trials to impose this requirement would entail a delay in dissemination of the vast majority of therapies that are helpful, because only &2.5% of approved drugs are later withdrawn.44 From a purely utilitarian standpoint, more people would be harmed by delaying dissemination of all drugs for a year than by releasing a small number of drugs with undesirable risk-benefit profiles. Previously recalled drugs have been from varying classes, for varying indications, with toxicities that may be undetectable regardless of any realistic increase in the length of preapproval trials.
Finally, there are issues to consider even if the imposition of the requirement were limited to studies that are currently long enough to sustain multiple interim analyses. Trial continuation in the face of a substantial difference in apparent efficacy entails that some participants would continue to receive a therapy that is likely to be inferior. Many trials in which questions of termination arise also have serious or irreversible outcomes as end points. For additional evidence to accumulate, participants on the less effective regimen would necessarily continue to suffer those outcomes at a higher rate.
|
Toward a Coherent Policy |
|---|
|
TopIntroductionTwo True PremisesThe Third PremiseThe Fourth PremiseToward a Coherent PolicyReferences |
|---|
For reasons of generalizability noted earlier, the range of AEs observed in a clinical trial is likely to be artificially small. Policy discussions must therefore extend to the optimization of active postmarketing surveillance efforts, which are clearly necessary to capture AEs that are not likely to be observed in clinical trials. In the wake of the rofecoxib withdrawal, these complex issues have been discussed substantively.45,46 At issue here is what stopping criteria provide the optimal marriage of safety information and proof of efficacy given the inherent limitations of preapproval trials.
The concern about unduly foreshortening clinical trials is not new. Peto47 stopping boundaries proposed in 1976 require probability values for early termination in the range of 0.001. Most commonly used boundaries require highly significant evidence for termination early in the trial,12 both to guard against termination in the face of fluctuating early trends and to ensure some minimum of follow-up data. Adequate follow-up was a particular concern in MERIT because of substantial questions about the safety of ß-blockers in patients with heart failure. The trial was designed with 2 separate criteria that were met before termination: crossing of a conservative upper boundary for benefit and a prespecified requirement that at least 50% of the total number of expected deaths were observed.48
Because the stated purpose of the requirement for extreme evidence is to ensure adequate follow-up, the most direct approach is to add a second criterion to early termination guidelines in lieu of requiring extreme evidence, as was done in MERIT. This additional criterion could be customized to suit the needs of an individual trial and might be based on follow-up time, number of events, or even some limit on the toxicities observed. A second termination criterion for the hypothetical anticoagulant trial noted earlier might stipulate that the rate of major bleeds was within a clinically acceptable range, analogous to a noninferiority design. Other benefits of the second criterion approach include increased transparency of the study aims and enhanced flexibility to terminate the trial on the basis of a range of safety and efficacy considerations.
Limited circumstances might exist in which allowing trials to progress beyond usual significance levels is reasonable, as when outcome measures trend in opposing directions. In the Prostate Cancer Prevention Trial, for example, finasteride significantly reduced the incidence of prostate cancer, but a greater proportion of men who developed cancer had high-grade disease.49 Notably, even in this complex scenario, the trial was terminated with probability values of 0.001. Additional evidence probably would not have changed treatment recommendations. Although there may be trials in which requiring statistically extreme evidence could provide toxicity information with a bearing on approvability, these situations are more limited than is initially evident, and widespread application of the requirement may have undesirable consequences.50
|
Acknowledgments |
|---|
The author gratefully acknowledges thesis advisors David L. DeMets, PhD, and Norman Fost, MD, MPH, for their helpful comments and suggestions on the manuscript, as well as Thomas Cook, PhD, for his guidance on statistical questions.
Disclosures
None.
|
References |
|---|
|
TopIntroductionTwo True PremisesThe Third PremiseThe Fourth PremiseToward a Coherent PolicyReferences |
|---|
1. Friedman L, Furberg C, DeMets D. Fundamentals of Clinical Trials. 3rd ed. New York, NY: Springer-Verlag; 1998: 247.
2. Title LM, Giddens K, McInerney MM, McQueen MJ, Nassar BA. Effect of cyclooxygenase-2 inhibition with rofecoxib on endothelial dysfunction and inflammatory markers in patients with coronary artery disease. J Am Coll Cardiol. 2003; 42: 1747–1753.
3. CDER Approval Package Vioxx (Rofecoxib) Tablets. 1999. Available at: http://www.fda.gov/cder/foi/nda/99/021042_52_vioxx.htm. Accessed February 4, 2006.
4. Bombardier C, Laine L, Reicin A, Shapiro D, Burgos-Vargas R, Davis B, Day R, Ferraz MB, Hawkey CJ, Hochberg MC, Kvien TK, Schnitzer TJ. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis: VIGOR Study Group. N Engl J Med. 2000; 343: 1520–1528.
5. Graham DJ. Risk of acute myocardial infarction and sudden cardiac death in patients treated with COX-2 selective and non-selective NSAIDs. 2004. Available at: http://www.fda.gov/cder/drug/infopage/vioxx/vioxxgraham.pdf. Accessed February 4, 2006.
6. Juni P, Nartey L, Reichenbach S, Sterchi R, Dieppe PA, Egger M. Risk of cardiovascular events and rofecoxib: cumulative meta-analysis. Lancet. 2004; 364: 2021–2029.[CrossRef][Medline] [Order article via Infotrieve]
7. Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events associated with selective COX-2 inhibitors. JAMA. 2001; 286: 954–959.
8. Maugh TH, Gellene D. Arthritis drug Vioxx pulled; risk of heart attacks is cited. Los Angeles Times. Home ed. October 1, 2004: A1.
9. Merck announces voluntary worldwide withdrawal of VIOXX®. 2004. Available at: http://www.vioxx.com/rofecoxib/vioxx/consumer/press_release_09302004.jsp. Accessed February 4, 2004.
10. Topol EJ. Failing the public health: rofecoxib, Merck, and the FDA. N Engl J Med. 2004; 351: 1707–1709.
11. Horton R. Vioxx, the implosion of Merck, and aftershocks at the FDA. Lancet. 2004; 364: 1995–1996.[CrossRef][Medline] [Order article via Infotrieve]
12. Friedman L, Furberg C, DeMets D. Fundamentals of Clinical Trials. 3rd ed. New York, NY: Springer-Verlag; 1998: 263–265.
13. Montori VM, Devereaux PJ, Adhikari NK, Burns KE, Eggert CH, Briel M, Lacchetti C, Leung TW, Darling E, Bryant DM, Bucher HC, Schunemann HJ, Meade MO, Cook DJ, Erwin PJ, Sood A, Sood R, Lo B, Thompson CA, Zhou Q, Mills E, Guyatt GH. Randomized trials stopped early for benefit: a systematic review. JAMA. 2005; 294: 2203–2209.
14. A randomized trial of propranolol in patients with acute myocardial infarction, I: mortality results. JAMA. 1982; 247: 1707–1714.
15. Effect of metoprolol CR/XL in chronic heart failure: Metoprolol CR/XL Randomised Intervention Trial in Congestive Heart Failure (MERIT-HF). Lancet. 1999; 353: 2001–2007.[CrossRef][Medline] [Order article via Infotrieve]
16. The Cardiac Insufficiency Bisoprolol Study II (CIBIS-II): a randomised trial. Lancet. 1999; 353: 9–13.[CrossRef][Medline] [Order article via Infotrieve]
17. Packer M, Fowler MB, Roecker EB, Coats AJ, Katus HA, Krum H, Mohacsi P, Rouleau JL, Tendera M, Staiger C, Holcslaw TL, Amann-Zalan I, DeMets DL. Effect of carvedilol on the morbidity of patients with severe chronic heart failure: results of the Carvedilol Prospective Randomized Cumulative Survival (COPERNICUS) study. Circulation. 2002; 106: 2194–2099.
18. Fischl MA, Richman DD, Grieco MH, Gottlieb MS, Volberding PA, Laskin OL, Leedom JM, Groopman JE, Mildvan D, Schooley RT. The efficacy of azidothymidine (AZT) in the treatment of patients with AIDS and AIDS-related complex: a double-blind, placebo-controlled trial. N Engl J Med. 1987; 317: 185–191.[Abstract]
19. Graham DJ, Staffa JA, Shatin D, Andrade SE, Schech SD, La Grenade L, Gurwitz JH, Chan KA, Goodman MJ, Platt R. Incidence of hospitalized rhabdomyolysis in patients treated with lipid-lowering drugs. JAMA. 2004; 292: 2585–2590.
20. Davignon J, Hanefeld M, Nakaya N, Hunninghake DB, Insull W Jr, Ose L. Clinical efficacy and safety of cerivastatin: summary of pivotal phase IIb/III studies. Am J Cardiol. 1998; 82: 32J–39J.[CrossRef][Medline] [Order article via Infotrieve]
21. Medical Officer Review, Bromfenac New Drug Application 20-535. US Food and Drug Administration. 1995.
22. Preliminary report: effect of encainide and flecainide on mortality in a randomized trial of arrhythmia suppression after myocardial infarction: the Cardiac Arrhythmia Suppression Trial (CAST) Investigators. N Engl J Med. 1989; 321: 406–412.[Abstract]
23. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet. 1994; 344: 1383–1389.[CrossRef][Medline] [Order article via Infotrieve]
24. Cannon CP, Braunwald E, McCabe CH, Grayston JT, Muhlestein B, Giugliano RP, Cairns R, Skene AM. Antibiotic treatment of Chlamydia pneumoniae after acute coronary syndrome. N Engl J Med. 2005; 352: 1646–1654.
25. Friedman MA, Woodcock J, Lumpkin MM, Shuren JE, Hass AE, Thompson LJ. The safety of newly approved medicines: do recent market removals mean there is a problem? JAMA. 1999; 281: 1728–1734.
26. Omniflox recalled: antibiotic. FDA Newsletter. September 1992.
27. Owens RC Jr. Risk assessment for antimicrobial agent-induced QTc interval prolongation and torsades de pointes. Pharmacotherapy. 2001; 21: 301–319.[CrossRef][Medline] [Order article via Infotrieve]
28. Viskin S. Long QT syndromes and torsade de pointes. Lancet. 1999; 354: 1625–1633.[CrossRef][Medline] [Order article via Infotrieve]
29. Moss AJ. The QT interval and torsade de pointes. Drug Saf. 1999; 21 (suppl 1): 5–10;discussion 81–87.[Medline] [Order article via Infotrieve]
30. Heart drug withdrawn: Boots Pharmaceuticals’ Manoplax. FDA Consumer. Washington, DC: US Government Printing Office; October 1993.
31. Smith P, Arnesen H. Mortality in non-consenters in a post-myocardial infarction trial. J Intern Med. 1990; 228: 253–256.[Medline] [Order article via Infotrieve]
32. Zarin DA, Young JL, West JC. Challenges to evidence-based medicine: a comparison of patients and treatments in randomized controlled trials with patients and treatments in a practice research network. Soc Psychiatry Psychiatr Epidemiol. 2005; 40: 27–35.[CrossRef][Medline] [Order article via Infotrieve]
33. Heiat A, Gross CP, Krumholz HM. Representation of the elderly, women, and minorities in heart failure clinical trials. Arch Intern Med. 2002; 162: 1682–1688.
34. Ong AT, Hoye A, Aoki J, van Mieghem CA, Rodriguez Granillo GA, Sonnenschein K, Regar E, McFadden EP, Sianos G, van der Giessen WJ, de Jaegere PP, de Feyter P, van Domburg RT, Serruys PW. Thirty-day incidence and six-month clinical outcome of thrombotic stent occlusion after bare-metal, sirolimus, or paclitaxel stent implantation. J Am Coll Cardiol. 2005; 45: 947–953.
35. Chan PS, Hayward RA. Mortality reduction by implantable cardioverter-defibrillators in high-risk patients with heart failure, ischemic heart disease, and new-onset ventricular arrhythmia: an effectiveness study. J Am Coll Cardiol. 2005; 45: 1474–1481.
36. Bahit MC, Cannon CP, Antman EM, Murphy SA, Gibson CM, McCabe CH, Braunwald E. Direct comparison of characteristics, treatment, and outcomes of patients enrolled versus patients not enrolled in a clinical trial at centers participating in the TIMI 9 Trial and TIMI 9 Registry. Am Heart J. 2003; 145: 109–117.[CrossRef][Medline] [Order article via Infotrieve]
37. Dauerman HL, Ryan TJ Jr, Piper WD, Kellett MA, Shubrooks SJ, Robb JF, Hearne MJ, Watkins MW, Hettleman BD, Silver MT, Niles NW, Malenka DJ. Outcomes of percutaneous coronary intervention among elderly patients in cardiogenic shock: a multicenter, decade-long experience. J Invasive Cardiol. 2003; 15: 380–384.[Medline] [Order article via Infotrieve]
38. Brown N, Melville M, Gray D, Young T, Skene AM, Wilcox RG, Hampton JR. Relevance of clinical trial results in myocardial infarction to medical practice: comparison of four year outcome in participants of a thrombolytic trial, patients receiving routine thrombolysis, and those deemed ineligible for thrombolysis. Heart. 1999; 81: 598–602.
39. Kandzari DE, Roe MT, Chen AY, Lytle BL, Pollack CV Jr, Harrington RA, Ohman EM, Gibler WB, Peterson ED. Influence of clinical trial enrollment on the quality of care and outcomes for patients with non-ST-segment elevation acute coronary syndromes. Am Heart J. 2005; 149: 474–481.[CrossRef][Medline] [Order article via Infotrieve]
40. Dzewaltowski DA, Glasgow RE, Klesges LM, Estabrooks PA, Brock E. RE-AIM: evidence-based standards and a Web resource to improve translation of research into practice. Ann Behav Med. 2004; 28: 75–80.[CrossRef][Medline] [Order article via Infotrieve]
41. Katz DA, Muehlenbruch DR, Brown RL, Fiore MC, Baker TB. Effectiveness of implementing the agency for healthcare research and quality smoking cessation clinical practice guideline: a randomized, controlled trial. J Natl Cancer Inst. 2004; 96: 594–603.
42. Janssen Pharmaceutica announces the withdrawal of hismanal from the market. FDA Talk Paper. 1999. Available at: http://www.fda.gov/bbs/topics/ANSWERS/ANS00961.html. Accessed February 4, 2006.
43. Lotronex Tablets (alosetron hydrochloride): questions and answers. Available at: http://www.fda.gov/cder/drug/infopage/lotronex/lotronex-qa_0602.htm#gsk. Accessed February 4, 2006.
44. CDER report to the nation: 2002. Available at: http://www.fda.gov/cder/reports/rtn/2002/rtn2002–3.htm. Accessed February 4, 2006.
45. Ray WA, Stein CM. Reform of drug regulation: beyond an independent drug-safety board. N Engl J Med. 2006; 354: 194–201.
46. Harris G. FDA to create an advisory board on drug safety. The New York Times. February 16, 2005: A1.
47. Peto R, Pike MC, Armitage P, Breslow NE, Cox DR, Howard SV, Mantel N, McPherson K, Peto J, Smith PG. Design and analysis of randomized clinical trials requiring prolonged observation of each patient: I: introduction and design. Br J Cancer. 1976; 34: 585–612.[Medline] [Order article via Infotrieve]
48. Feyzi J, Julian D, Wikstrand J, Wedel H. Data monitoring experience in the Metoprolol CR/XL Randomized Intervention Trial in Chronic Heart Failure (MERIT-HF) potentially high-risk treatment in high risk patients. In: Data Monitoring in Clinical Trials: A Case Studies Approach. New York, NY: Springer Science Business Media; 2005.
49. Thompson IM, Goodman PJ, Tangen CM, Lucia MS, Miller GJ, Ford LG, Lieber MM, Cespedes RD, Atkins JN, Lippman SM, Carlin SM, Ryan A, Szczepanek CM, Crowley JJ, Coltman CA Jr. The influence of finasteride on the development of prostate cancer. N Engl J Med. 2003; 349: 215–224.
50. Wysowski DK, Swartz L. Adverse drug event surveillance and drug withdrawals in the United States, 1969-2002. Arch Intern Med. 2005; 165: 1363–1369.
This article has been cited by other articles:
 |
|
 |
|
  C. D. Furberg, A. A. Levin, P. A. Gross, R. S. Shapiro, and B. L. Strom The FDA and drug safety: a proposal for sweeping changes. Arch Intern Med, October 9, 2006; 166(18): 1938 - 1942. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Avorn Evaluating Drug Effects in the Post-Vioxx World: There Must Be a Better Way Circulation, May 9, 2006; 113(18): 2173 - 2176. [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||