Stratification of randomization is not required for a pre-specified subgroup analysis.

Published literature and regulatory agency guidance documents provide conflicting recommendations as to whether a pre-specified subgroup analysis also requires for its validity that the study employ randomization that is stratified on subgroup membership. This is an important issue, as subgroup analyses are often required to demonstrate efficacy in the development of drugs with a companion diagnostic. Here, it is shown, for typical randomization methods, that the fraction of patients in the subgroup given experimental treatment matches, on average, the target fraction in the entire study. Also, mean covariate values are balanced, on average, between treatment arms in the subgroup, and it is argued that the variance in covariate imbalance between treatment arms in the subgroup is at worst only slightly increased versus a subgroup-stratified randomization method. Finally, in an analysis of variance setting, a least-squares treatment effect estimator within the subgroup is shown to be unbiased whether or not the randomization is stratified on subgroup membership. Thus, a requirement that a study be stratified on subgroup membership would place an artificial roadblock to innovation and the goals of personalized healthcare.

Tumor burden modeling versus progression-free survival for phase II decision making.

Randomized Phase II oncology trial endpoints for decision making include both progression-free survival (PFS) and change in tumor burden as measured by the sum of longest diameters (SLD) of the target lesions. In addition to observed SLD changes, tumor shrinkage and growth parameters can be estimated from the patient-specific SLD profile over time. The ability of these SLD analyses to identify an active drug is contrasted with that of a PFS analysis through the simulation of Phase II trials via resampling from each of 6 large, Phase II and III trials, 5 of which were positive and one negative. From each simulated Phase II trial, a P value was obtained from 4 analyses-a log-rank test on PFS, a Wilcoxon rank-sum test on the minimum observed percentage change from baseline in SLD, and 2 nonlinear, mixed-effects model analyses of the SLD profiles. All 4 analyses led to approximately uniformly distributed P values in the negative trial. The PFS analysis was the best or nearly the best analysis in the other 5 trials. In only one of the positive studies did the modeling analysis outperform the analysis of the minimum SLD. In conclusion, for the decision to start a Phase III trial based on the results of a randomized Phase II trial of an oncology drug, PFS appears to be a better endpoint than does SLD, whether analyzed through simple SLD endpoints, such as the minimum percentage change from baseline, or through the modeling of the SLD time course to estimate tumor dynamics.

Dynamic randomization and a randomization model for clinical trials data.

Randomization models are useful in supporting the validity of linear model analyses applied to data from a clinical trial that employed randomization via permuted blocks. Here, a randomization model for clinical trials data with arbitrary randomization methodology is developed, with treatment effect estimators and standard error estimators valid from a randomization perspective. A central limit theorem for the treatment effect estimator is also derived. As with permuted-blocks randomization, a typical linear model analysis provides results similar to the randomization model results when, roughly, unit effects display no pattern over time. A key requirement for the randomization inference is that the unconditional probability that any patient receives active treatment is constant across patients; when this probability condition is violated, the treatment effect estimator is biased from a randomization perspective. Most randomization methods for balanced, 1 to 1, treatment allocation satisfy this condition. However, many dynamic randomization methods for planned unbalanced treatment allocation, like 2 to 1, do not satisfy this constant probability condition, and these methods should be avoided.

Inefficiency of randomization methods that balance on stratum margins and improvements with permuted blocks and a sequential method.

Stratified permuted blocks randomization is commonly applied in clinical trials, but other randomization methods that attempt to balance treatment counts marginally for the stratification variables are able to accommodate more stratification variables. When the analysis stratifies on the cells formed by crossing the stratification variables, these other randomization methods yield treatment effect estimates with larger variance than does stratified permuted blocks. When it is truly necessary to balance the randomization on many stratification variables, it is shown how this inefficiency can be improved by using a sequential randomization method where the first level balances on the crossing of the strata used in the analysis and further stratification variables fall lower in the sequential hierarchy.

Analysis of tumor burden versus progression-free survival for Phase II decision making.

PURPOSE: There have been recent recommendations to use percentage change in tumor burden (dTB) as a primary endpoint in randomized Phase II trials. We assessed whether dTB is better for the decision to start a Phase III trial than is progression-free survival (PFS). METHODS: We repeatedly sampled patients from six large randomized trials to obtain simulated Phase II trials. We derived PFS and dTB endpoints on the trial patients and determined the fraction of simulated trials with positive results for each endpoint. We supplemented these analyses with regression analyses to assess the ability of PFS and dTB to predict overall survival (OS). RESULTS: The best PFS endpoint included tumor assessments through 6 months after the last patient enrolled. With 70 patients in each simulated Phase II trial, the estimated rate of a correct 'Phase III go' decision ranged from 0.74 to 0.91 across the six parent studies. The best dTB endpoint was the last dTB through 6 months after the last patient enrolled, with corresponding rates of 0.54 to 0.81. The PFS rate was better than the dTB rate in five studies. PFS and dTB are individually statistically significant predictors of OS (p < 0.05). In all six studies PFS added significantly to the regression models with dTB included, while in only two studies did dTB add significantly to the regression model with PFS included. CONCLUSION: Analysis of PFS in randomized Phase II trials generally leads to better 'Phase III go' decisions than does analysis of dTB. Tumor burden analyses should be used in supportive analyses to a primary PFS analysis.

Optimizing collection of adverse event data in cancer clinical trials supporting supplemental indications.

PURPOSE: Although much is known about the safety of an anticancer agent at the time of initial marketing approval, sponsors customarily collect comprehensive safety data for studies that support supplemental indications. This adds significant cost and complexity to the study but may not provide useful new information. The main purpose of this analysis was to assess the amount of safety and concomitant medication data collected to determine a more optimal approach in the collection of these data when used in support of supplemental applications. METHODS: Following a prospectively developed statistical analysis plan, we reanalyzed safety data from eight previously completed prospective randomized trials. RESULTS: A total of 107,884 adverse events and 136,608 concomitant medication records were reviewed for the analysis. Of these, four grade 1 to 2 and nine grade 3 and higher events were identified as drug effects that were not included in the previously established safety profiles and could potentially have been missed using subsampling. These events were frequently detected in subsamples of 400 patients or larger. Furthermore, none of the concomitant medication records contributed to labeling changes for the supplemental indications. CONCLUSION: Our study found that applying the optimized methodologic approach, described herein, has a high probability of detecting new drug safety signals. Focusing data collection on signals that cause physicians to modify or discontinue treatment ensures that safety issues of the highest concern for patients and regulators are captured and has significant potential to relieve strain on the clinical trials system.

Effectiveness of ipsapirone, a 5-HT-1A partial agonist, in major depressive disorder: support for the role of 5-HT-1A receptors in the mechanism of action of serotonergic antidepressants.

Desensitisation of serotonin 1A (5-HT-1A) receptors is a leading hypothesis for the mechanism of action of antidepressants which block serotonin reuptake. This hypothesis predicts that direct-acting 5-HT-1A agonists should also exhibit anti-depressant properties. Here we report the results of the first large-scale controlled study of the efficacy and tolerability of a 5-HT-1A agonist in outpatients with major depressive disorder (MDD). Three hundred and seventy-three subjects meeting DSM-III-R criteria for MDD participated in this randomised, double-blind comparison of the 5-HT-1A partial agonist ipsapirone (5 mg, 7.5 mg and 10 mg t.i.d.) to placebo t.i.d. Improvement in depressive symptoms relative to placebo, as measured by the Hamilton Depression Rating Scale, occurred in the ipsapirone (7.5 mg t.i.d.) group with a magnitude of effect (D=-2.53 points) that was statistically significant (p=0.010). Adverse events occurred in 76% of the placebo patients and 92% of the ipsapirone patients. A dose-related increase in the incidence of adverse events led to discontinuation of treatment with the 10 mg t.i.d. Results of this study demonstrate that ipsapirone, at a dose of 7.5 mg t.i.d., is an effective antidepressant agent in the treatment of MDD, supporting the hypothesised role of 5-HT-1A receptors in the mechanism of action of serotonin reuptake inhibitors. However, as a potential therapeutic agent for depression, ipsapirone shows only a modest magnitude of drug-placebo differences as well as a side-effect profile less favorable than many of the newer antidepressants.