Methodological differences may have also contributed to the apparent discrepancy between analyses. We restricted our measures of complexity because of their overlapping nature (eg, number of fills and long-term medication classes) or collinearity with adherence (eg, patients with more fills are by definition more adherent). Geers et al entered many variables into 1 model; this may have spread the predictive ability among multiple measures of the same construct. We evaluated a linear outcome in contrast to the binary measures used by Geers et al. While these models produce different outputs, the nonsignificant 23% increase in the odds of nonpersistence from poor refill consolidation in the much smaller study by Geers et al generates inferences very similar to those from ours. We assessed complexity during the 90-day period after the first statin fill, whereas Geers et al evaluated complexity during the year prior to the first fill. The burden of medication access and filling may be particularly high after, rather than before, starting therapy, and thus the evaluation by Geers et al may have measured exposure variables that would be expected to have less relationship to adherence.