CLINICAL TRIAL "EASY ESCAPE" CRITERIA CAN MAKE PLACEBO CONTROLS POSSIBLE IN LIFE-THREATENING SITUATIONS, FDA's TEMPLE SAYS: PROCARDIA EXAMPLE CITED

Clinical trials designed with "easy escapes" to move patients from placebo to treatment groups can minimize patient risk while allowing for scientifically sound studies, FDA's Drug Research & Review Director Robert Temple, MD, told a workshop on issues in clinical testing Sept. 10 in Washington. Citing ethical concerns and increasing reluctance by investigators to carry out placebo-controlled trials in conditions associated with morbidity, Temple described alternative study designs that "incorporated placebo treatments but that eliminated or minimized the exposure to patients of actual or preceived risks." Temple specifically cited a randomized withdrawal trial conducted by Pfizer for Procardia (nifedipine) for use in vasospastic angina, in which patients received nifedipine for two weeks, then were randomized into either a placebo or treatment group. "Early escape was built into the protocol in the event of increased duration, severity or frequency of angina suggesting an impending myocardial infarction, which included angina pain more than 30 minutes in duration, more intense pain than previously experienced or sudden increase in frequency or pain that interrupted normal activities or cardiovascular deterioration," he said. Squibb Capoten Study Shows That Placebo Controls Can Be Used "Even In Very Severe Hypertension During the trial, five patients from the placebo group withdrew early and entered the nifedipine group. The results, however, "were comparatively straightforward" and provided the basis for FDA approval of the drug, Temple said. "The study not only provided evidence of effectiveness in its own right, but suggested that (BRACKET)an earlier(BRACKET) open study was credible in supporting the drug effect, namely that the 85% response rate in the open study could not be attributed antirely to spontaneous improvement, since many patients worsened when the drug was stopped." Temple noted that the study "was conducted safely with one exception, one placebo patient may have had a myocardial infarction." Because the patient had angina for four hours before seeing a physician, Temple stressed that "it is critical that patients understand fully the early escape cristeria." The nifedipine study is one example of what Temple described as a "feel good" design. "I call these 'feel good' designs because they allow us to feel ethically comfortable while carrying out a scientifically sound study," he explained. Entitled "Current Issues in the Design, Conduct the Interpretation of Clinical Trials," the two-day workshop was organized by FDA and Natl. Institutes of Health officials and Professional Postgraduate Services. In his speech, Temple also cited a Squibb Capoten study to establish the effectiveness of captopril in hypertensive patients unresponsive to other therapies. He noted that patients underwent a triple-therapy lead-in before randomization into the captopril or placebo group, and had "various ways" of escaping into the captropril group based on blood pressure determinations. "Even in very severe hypertension, it is possible to do a reasonably well-controlled trials, complicated though it is," he asserted. Noting that the blood pressure of patients receiving placebos will sometimes drop in clinical trials, Temple reiterated the need for a control group in clinical trials. "Just switching people from an ineffective therapy is not very informative because many of them will respond to being in a study, to time or to heaven knows what," he said. Discussing the agency's position on positive-control designs, FDA Statistical Evaluation & Research Group Leader Gordon Pledger, PhD, stressed that the "'ideal' study design involves active ingredient, positive control and placebo." Because a positive control trial "offers no direct evidence of effectiveness of the test drug," Pledger said supplementary information is required for a determination of efficacy.In addition, Pledger asserted that such trials don't "go far enough because (BRACKET)they are(BRACKET) content to answer the question are A and B equivalent without addressing the possibility that neither A nor B will beat the placebo had a placebo group been included in the trials." Another area in which FDA often has difficulty in study review is cross-over trials, FDA's Statistical Evaluation & Research Branch Chief Satya Dubey, PhD, told the group. Common difficulties faced by the agency's reviewers in cross-over trials include "wrong designs, improper analyses, lack of wash-outs between treatments for no valid reasons, disregard for possible carry-over effects, inadequate number of persons in the trial for predicting true carryover effect, (BRACKET)and(BRACKET) lack of baseline measurements for each treatment period," he said. Dubey suggested that cross-over studies can be employed in relatively stable chronic diseases, for prophylactic drugs with relatively short half-lives and for relatively short treatment periods. The FDAer further outlined FDA guidelines for when such studies are acceptable. Cross-over Studies Recommended For Analgesics, Anti-anginals, Anticonvulsants, NSAIDs Under the guidelines, cross-over studies "may be considered as recommended for: analgesics, anti-anginals, anticonvulsants, anti-inflammatories and hypnotics," he said. Cross-over designs are "permitted" for: anti-anxiety agents, psychoactive drugs in children, and radiopharmaceuticals. Although such designs "may be considered to be permitted," they are discouraged for antacids, anti-arrthmics, antidepressants, antidiarrheals, bronchodilators, gastric secetory depressants, G.I. motility modifers and laxatives, he said. Cross-over designs are "not recommended" for: anti-infectives, osteoporosis, dental caries agents, general or local anesthetics or peridontal disease agents. In addition, Temple discussed what he termed "enrichment" of a study population. Such enrichment may consist of limiting enrollment in a trial to patients who are more likely to respond to treatment, he said. "We have no compunctions about doing this if we have a pathophysiologically reasonable basis for it," he said. However, when patients are "excluded for other reasons, we get nervous about the question of generalizability." FDA Biometrics Division staffer Mary Johnson, PhD, recommended analyzing "all randomized patients as well as evaluable patients" in her discussion of dropouts and exclusions. She noted that she would "tend to question conclusions from a trial with a high exclusion rate if the results in the subset of evaluable patients differed markedly from those based on all patients as they were randomized." Johnson cited an NDA submission which compared "drug A" to propranolol in the treatment of hypertension to illustrate the "dramatic impact patient exclusion can have on the conclusions of a study." The protocol allowed for the withdrawal of patients after five weeks of titration if a response was not achieved. Of the 288 patients randomized, 128 dropped out during the double blind phase, and 32 completed the study but were found to be protocol violators, leaving 128 evaluable patients. Using evaluable patient data, the "sponsor found no significant difference between drug groups in average blood pressure reduction." However, "analysis which included all evaluable patients in the titration period indicated that drug A was significantly less effective than propranolol." Johnson concluded that in the case of "drug A" the analysis of evaluable patients was misleading in evaluating "the efficacy of the drug in the general hypertensive patient population for which it would be prescribed." She stated that "for this purpose results of the 'intent to treat' analysis led to an entirely different conclusion, not only by increasing the power of statistical tests but also by more accurately representing the differences between the two drugs." The FDAer also noted that endpoint analysis could be problematic, "particularly when the pattern of withdrawals over time differs between treatment groups." She cited an NDA submission which compared "drug X" to aspirin in the treatment of rheumatoid arthritis. The protocol allowed for dropouts due to insufficient efficacy and adverse reactions. Patients in the study had tried aspirin and presumably found it unsatisfactory. The study had "extremely high dropout rates" with a greater early withdrawal for aspirin patients. Johnson said that "the sponsor chose to evaluate drug effects based entirely on endpoint analysis." Differences between drugs were significant in favor of "drug X." Johnson explained that "because of the baseline symptom flair in arthritis trials, the difference in the pattern of withdrawals over time . . . will tend to bias results against the drugs with early dropout." Johnson suggested that patients were able to deduce which treatment they were taking because of their familiarity with aspirin's side effects. Johnson illustrated the "early dropout bias" by showing that "endpoint results were not consistent with results at particular time periods. CDC Official Criticizes Pooling; Heart, Lung Institute's Friedman Points Out Positives Johnson recommends avoiding "dropout and exclusion problems in the first place through careful study design." She stated that "it may be worth including in the protocol specific instructions against randomizing patients who are not likely to tolerate or who may be unable to receive any of the treatment schedules." In addition, she suggested independent group monitoring of protocol adherence of the participating clinics. Johnson also recommended that the "criteria for withdrawing patients should be strictly defined." She said that this "would lend some degree of credibility to methods of analysis in which the reasons for a patient's dropping out have a bearing on drug efficacy." She also noted that "a high withdrawal rate will seriously limit the interpretation of a study if reasonable methods to incorporate the dropouts markedly influenced your results." David Hall, PhD, Centers for Disease Control, discussed his objections to pooling in the analysis of clinical trials. Hall stated that a study "mechanically combined with other studies loses its persuasiveness and in effect becomes uninterpretable." The CDC staffer asserted that "pooling methods are usually used when studies have failed to provide a convincing basis for the conclusion." He suggested disecting and analyzing such studies "rather than attempt(BRACKET)ing(BRACKET) to obscure failings and address(BRACKET)ing(BRACKET) a secondary cause of failure of sample size." Lawrence Friedman, MD, Natl. Heart, Lung & Blood Institute, discussed the positive aspects of pooling at the symposium. Friedman, speaking for the institute's Salim Yusuf, MD/PhD, asserted that "combining data of the same endpoint from all relevant trials . . . is not to obtain more significant P values but to minimize the effect of chance and obtain a more reliable answer." He concluded that although pooling is not perfect it "is better than emphasizing the most or least promising trial or trials chosen by some arbitrary post hoc criteria."