On Random Center Grouping in Multicenter Clinical Trials |Biomedgrid

With the attention of achieving desired power within a pre-specified timeframe in clinical development, a multicenter trial is often conducted to expedite the patient recruitment. However, a multicenter trial with plethoric number of centers is very likely to result in numerous small centers, which will cause the problem of treatment imbalance (unequal number of patients per arm within a center) and/or center imbalance (a few large centers with a number of small centers). In practice, treatment imbalance within and between centers inevitably occurs regardless the randomization models and/or methods used. Treatment imbalance within center and/or center imbalance will not only (i) increase the probability of observing treatment-by-center interaction, but also (ii) decrease the power for detecting clinically meaningful difference (or treatment effect) of the test treatment under investigation. In this paper, we propose a method of determining whether an observed significant interaction is a false alarm and a method of random grouping for an unbiased and reliable assessment of treatment effect when significant treatment and center imbalance are observed. The impact of treatment and center imbalance will also be examined in terms of the potential loss of power.