Search

This paper introduces a test of superiority of new anti-infective drug B over comparator drug A based on a randomized clinical trial. This test can be used to demonstrate assay (trial) sensitivity for noninferiority trials and rigorously tailor drug choice for individual patients. Our approach uses specialized baseline covariates X A ,X B , which should predict the benefits of drug A and drug B, respectively. Using a response surface model for the treatment effect, we test for superiority at the (X A ,X B ) point that is most likely to show superiority. We identify this point based on estimates from a novel half-blind pseudo likelihood, where we augment a blinded likelihood (mixed over the treatment indicator) with likelihoods for the overall success rates for drug A and drug B (mixed over X A ,X B ). The augmentation results in much better estimates than those based on the mixed blinded likelihood alone but, interestingly, the estimates almost behave as if they were based on fully blinded data. We also develop an analogous univariate method using X A for settings where X B has little variation. Permutation methods are used for testing. If the "half-blind" test rejects, pointwise confidence interval can be used to identify patients who would benefit from drug B. We compare the new tests to other methods with an example and via simulations., (Published 2018. This article is a U.S. Government work and is in the public domain in the USA.)

We propose an approach to analyze survival time in a crossover clinical trial by performing a primary ranking on whether events occur and a secondary ranking on event times. This hierarchical ranking method is meant to reflect the idea that the goal of therapy is to prevent a clinical event and, failing that, to delay the occurrence of the event, hopefully for a substantial amount of time. We compare our approach with other methods including one method proposed by Feingold and Gillespie, a recommended procedure. The power is similar in many settings, but the hierarchical ranking can have substantially greater power under certain censoring patterns and also under a cure model, or models where treatment induces a substantial delay in some fraction of patients. We additionally feel that the hierarchical ranking method should be more clinically relevant in many settings. The method can also be applied to continuous outcomes censored by a limit of detection, such as HIV viremia., (Copyright © 2011 John Wiley & Sons, Ltd.)

While the intent-to-treat (ITT) analysis is widely accepted for superiority trials, there remains debate about its role in non-inferiority trials. It is often said that the ITT tends to be anti-conservative in the demonstration of non-inferiority. This concern has led to some reliance on per-protocol (PP) analyses that exclude patients on the basis of post-baseline events, despite the inherent bias of such analyses. We compare ITT and PP results from antibiotic trials presented to the public at the FDA's Anti-infective Drug Advisory Committee from 1999 to 2003. While the number of available trials is too small to produce clear conclusions, these data did not support the assumption that the ITT would lead to smaller treatment difference than the PP, in the setting of antibiotic trials. Possible explanations are discussed., (2004 by John Wiley & Sons, Ltd.)

Investigators often design clinical trials without knowing precisely the values of such necessary parameters as the variances or the event rates in the control group. In order to determine reasonable values for such parameters, they may design a small pilot study external to the main trial. In this paper we propose designs, which we term internal pilot studies, that designate a portion of the main trial as a pilot phase. At the end of the internal pilot study, the investigators recompute preselected parameters and recalculate required sample size. The study then proceeds with the modifications dictated by the internal pilot. Final analyses of the results incorporate all data, disregarding the fact that part of the data came from a pilot phase. As one example of this type of design, we consider a study to compare two normally distributed means. By simulation, we show a numerical example for which the effect of the procedure on the alpha-level is negligible, but the potential gain in power considerable. We urge considering a similar approach for a number of types of endpoints.

Factorial designs in clinical trials allow for the study of several medical treatments simultaneously. This paper distinguishes among different types of settings in which factorial designs are useful. For the experiment that involves investigation of several new or untested therapies, we introduce a model that incorporates rates of non-compliance to therapy as well as various degrees of subadditivity of treatment effects. We compare the operating characteristics of the factorial under this model with those of competing designs and show that a modest negative interaction can considerably diminish the power to detect treatment effects in the factorial even in cases that have little power to detect this interaction. We urge, therefore, that designers of clinical trials with factorial layouts posit realistic estimates of interactions among treatments in order to assure adequate power to detect beneficial effects of treatment.