Confidence intervals and adjusted odds ratios: a commend on interpreting the results of Killip classification and glucose levels.

摘要

I have read with great pleasure the letters by Lina et al [1] and Cheng [2] about the article investigating the relationship between Killip classification and glucose levels in patients with acute myocardial infarction by Cheng and Yen [3]. Lina et al [1] claimed that the width confidence intervals (CIs) of glucose levels of Killip classes make the difference insignificant despite significant result by 1-way analysis of variance, and Cheng [2] accepted this claim and stated that it could be due to the limited sample size. However, the statistical significance by CIs is calculated whether the difference between 2 means includes zero; and when we look at the glucose levels of Killip class 1 (186.8 +- 82.5) and Killip class 4 (236.2 +- 115.5), we get a difference of 49.4 with a 95% CI of 19.4 to 79.4, which has a P value of .0014. This P value is significant even with the Bonferroni correction. So, the statistical significance of differences between 2 means calculated by CIs is not related to the width of the Clbut the CI of differences including zero or not.