In Reply: In response to Drs van de Poll-Franse and Coebergh, we agree that the weight assigned to the risk estimate from the study by Kouvaris et al1 seems surprisingly large given its modest sample size. However, the weight reflects the narrow CI the article reported for diabetes and mortality in the unadjusted Cox model (HR, 1.038; 95% CI, 1.006-1.099). Such narrow CIs reflect the kind of precision typically attained in much larger studies. We did not request the data required to double-check precision estimates as part of our protocol. However, we are certain that the study by Kouvaris et al did not unduly influence our overall pooled results, since exclusion of that study in our influence analysis produced a robust estimate of long-term mortality risk related to preexisting diabetes (HR, 1.43; 95% CI, 1.32-1.55).
Barone BB, Yeh H, Brancati FL. Long-term Mortality in Cancer Patients With Preexisting Diabetes—Reply. JAMA. 2009;301(15):1541. doi:10.1001/jama.2009.489