A varying-coefficient method for analyzing longitudinal clinical trials data with nonignorable dropout

Contemp Clin Trials. 2012 Mar;33(2):378-85. doi: 10.1016/j.cct.2011.11.009. Epub 2011 Nov 12.


Dropout is common in longitudinal clinical trials and when the probability of dropout depends on unobserved outcomes even after conditioning on available data, it is considered missing not at random and therefore nonignorable. To address this problem, mixture models can be used to account for the relationship between a longitudinal outcome and dropout. We propose a Natural Spline Varying-coefficient mixture model (NSV), which is a straightforward extension of the parametric Conditional Linear Model (CLM). We assume that the outcome follows a varying-coefficient model conditional on a continuous dropout distribution. Natural cubic B-splines are used to allow the regression coefficients to semiparametrically depend on dropout and inference is therefore more robust. Additionally, this method is computationally stable and relatively simple to implement. We conduct simulation studies to evaluate performance and compare methodologies in settings where the longitudinal trajectories are linear and dropout time is observed for all individuals. Performance is assessed under conditions where model assumptions are both met and violated. In addition, we compare the NSV to the CLM and a standard random-effects model using an HIV/AIDS clinical trial with probable nonignorable dropout. The simulation studies suggest that the NSV is an improvement over the CLM when dropout has a nonlinear dependence on the outcome.

Publication types

  • Comparative Study
  • Randomized Controlled Trial
  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't
  • Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

  • Algorithms
  • Anti-HIV Agents / therapeutic use*
  • Biometry / methods*
  • HIV / genetics
  • HIV Infections / drug therapy*
  • Humans
  • Models, Statistical*
  • Patient Dropouts / statistics & numerical data*
  • Probability*
  • RNA, Viral / analysis


  • Anti-HIV Agents
  • RNA, Viral