Howard Barkan, Kaiser Permanente Medical Center, Oakland, Calif, USA. howardbarkan@cs.com
21 April 2015
This is apparently a surgical series treated as a pair of cohort studies. The grounds for the surgeons' using or not using statins are never specified. Multiple risk factors for the target outcomes are described. There are statistically significant inter-cohort differences in the frequencies of many of these potential confounding variables. Nonetheless, these are not considered in the bivariate analyses of associations with the target outcomes (via multivariate analyses or other methods). The authors examine the association of statin use with several outcomes. There is no adjustment for multiple comparisons. Although key analyses as presented do not have statistically significant results, no statistical power analyses are presented in the paper. As such, one is forced to see this as a partial analysis of a convenience sample. This investigation of an interesting question deserves a more rigorous examination.
Methodological questions
21 April 2015
This is apparently a surgical series treated as a pair of cohort studies. The grounds for the surgeons' using or not using statins are never specified. Multiple risk factors for the target outcomes are described. There are statistically significant inter-cohort differences in the frequencies of many of these potential confounding variables. Nonetheless, these are not considered in the bivariate analyses of associations with the target outcomes (via multivariate analyses or other methods). The authors examine the association of statin use with several outcomes. There is no adjustment for multiple comparisons. Although key analyses as presented do not have statistically significant results, no statistical power analyses are presented in the paper. As such, one is forced to see this as a partial analysis of a convenience sample. This investigation of an interesting question deserves a more rigorous examination.
Competing interests
None declared