Pattern-mixture models provide a general and flexible framework for sensitivity analyses of nonignorable missing data in longitudinal studies. The placebo-based pattern-mixture model handles missing data in a clinically interpretable manner and has been used as sensitivity analysis addressing the de facto (effectiveness) hypothesis, while a likelihood-based approach that assumes missing at random is often used as the primary analysis addressing the de jure (efficacy) hypothesis. We describe a sample size calculation method that takes into account both types of treatment effect estimates. This method can be used to power the study to control the probability of reaching conflicting conclusions from the primary and sensitivity analyses. To apply the method, we only need to specify the pattern probabilities at postbaseline time points, the expected treatment differences at postbaseline time points, and the conditional covariance matrix of postbaseline measurements given the baseline measurement. Simulation studies are conducted to assess the performance of the proposed method.