Reply to McDonald.

TO THE EDITOR—We thank Dr Mc-Donald for his interest in our work and concur that the role of particular ribo-types in the clinical outcome of Clostrid-ium difficile infection (CDI) is complex [1]. Our study identified several factors that predicted severe CDI in unadjusted analysis, including ribotype 027/078 (Table 2 in our article) [2]. We agree that including intermediate white blood cell (WBC) counts in our final model (Table 4 in our article) may have potentially attenuated direct effects of ribotype on outcome. In Table 3 of our article, we show, however, that adjustment for any of the significant predictors rendered ribotype 027/078 effect on CDI severity nonsignificant [2], indicating that WBC count was not required for the conclusions drawn in our study. To directly address the possibility that WBC count was influenced by ribotype-specific host responses requires temporal patient data, including treatment, which we did not collect. To indirectly address this issue, we performed additional analyses for our combined dataset (N = 412 cases with complete data) in the following 4 contexts: 1. We added abnormal WBC count (<4000 or >12 000 cells/mL) to the definition of severe CDI. 2. We determined whether ribotype 027/078 predicted an abnormal WBC count. 3. We excluded both WBC count and albumin levels from the analysis, as is appropriate if they are potentially confounded. 4. We again added abnormal WBC to the severe CDI definition and excluded albumin level from the analysis. Ribotype 027/078 was not significant in any of these contexts (adjusted analyses). Although these results do not exclude the possibility raised by McDonald, they do add support to our original conclusion that predictors other than ribotype may be more important in explaining the variability of CDI severity. We did not explore the influence of treatment on outcome, so the influence of ribotype was either not strong enough to overcome these effects or they were relatively minor. Why the incidence and severity of CDI increased in the early 2000s remains unclear. However, epidemic spread of clonal pathogens need not result from advantageous phenotypes (toxin production , sporulation, antibiotic resistance, etc). Such population expansions may result simply by chance, as transmission rates overcome clonal interference. Temporal fluctuations in ribotype prevalence support a " drift " scenario [3]. Such fluctuations suggest that many ribotypes reach a high local prevalence, which would increase transmission and potentiate a high global