Randomized Discontinuation Trials: Design and Efficiency
Valerii Fedorov and Tao Liu
ABSTRACT. Randomized Discontinuation Trials (RDT) are usually two-phase designs and become more and more popular across a number of therapeutic areas (oncology is one of the most known [Rosner, Stadler, and Ratain, 2002]). In this design, a single arm trial, called open phase, is followed by a randomized blinded two-arm trial at second phase to compare two treatments (generally one needs to be placebo). Intuition and simulation exercises show that potentially RDT may increase a sensitivity of trials relatively to the more traditional patient allocations. This increase can be substantial if the the open phase provides a reliable separation of the population into two subpopulations: responders and nonresponders. We compared RDT with the traditional two-arm randomized clinical trial (RCT) when the outcomes are binary and the population of interest consists of three groups, placebo responders, treatment responders and nonresponders. All our results are derived in the “parameter estimation” setting and are based on comparison of estimator variances. Conditions under which RDT is superior to RCT and which include the response rates, misclassification rates and randomization strategy at the second stage were found. Transition to hypothesis testing is rather straightforward.
Analysis of Responses in Migraine Modelling Using Hidden Markov Models
Vladimir Anisimov, Hugo Maas1, Meindert Danhof, Oscar E. Della Pasqua
Abstract: Markov-type models have been used in the analysis of disease progression. Although standard errors of model parameters are usually estimated, available software often does not permit the construction of confidence intervals around predictions of the dependent or response variable. A method is presented to calculate means and confidence intervals of model-predicted responses in time governed by a non-homogeneous hidden Markov model in continuous time. The Kolmogorov equations serve as the basis for the calculations. The method is realized in S-Plus and is applied to the prediction of headache responses in clinical studies of anti-migraine treatment. Means and confidence intervals are calculated by numerically solving differential equations that are nonlinear in the explanatory variable. Results indicate that uncertainty on predicted drug responses is larger than that on predicted placebo responses and that pain-free responses are less precisely predicted than pain relief responses. This is due to the uncertainty in the drug-specific parameters which is not present in predicted placebo responses.
Modeling of Enrolment and Estimation of Parameters in Multicentre Trials
Vladimir Anisimov, Valerii Fedorov
ABSTRACT. We propose various stochastic enrolment models and derive probability distributions for the number of patients across the centres and the number of centres with a given number of patients. To estimate parameters of these distributions we use a few different estimators and find either analytically or using Monte Carlo simulation that these estimators produce very similar results close to the actual values of parameters.
We also analyze several GSK datasets and show that the enrolment model, where the patients arrive at centres according to Poisson processes with the recruitment rates sampled from a gamma distribution, is a parsimonious but °exible enough to describe actual data. It also helps to evaluate some features of the enrolment such as the expected number of “empty” centres or probability to have a centre with suspiciously large number of patients, etc. Finally, the model fitted to the historical data provides us the opportunity to design clinical trials for the similar drugs/treatments.
