Letters Section Editor: Jody W. Zylke, MD, Senior Editor.
Author Affiliations: Department of Medicine, University of Texas Health Science Center, San Antonio (Mr Williams; firstname.lastname@example.org); and Department of Mathematics, University of the Incarnate Word, San Antonio, Texas (Dr Monroe).
To the Editor: The recent ERFC meta-analysis1 indicated that the standard hazard ratio (HR) for cardiovascular disease using baseline levels of non-HDL-C was 1.27 (95% CI, 1.22 to 1.33) vs 1.24 (1.19 to 1.29) using apolipoprotein B. In a head-to-head comparison, the HR for apolipoprotein B relative to the HR for non-HDL-C was lower: −2.4% (95% CI, −4.1% to −0.6%), assuming a correlation (r = 0.91) between the 2 markers.2 This contrasts with the within-study comparisons in our meta-analysis,2 which found the standard HR for coronary heart disease of apolipoprotein B averaged 5.7% (95% CI, 2.4% to 9.1%) higher than that of non-HDL-C. Our meta-analysis focused on HRs rather than the C statistic or the Cohen d statistic to aid assessment of the relative utility of each marker to target treatment rather than to just assess risk. Thus each meta-analysis estimated statistically significant associations in opposite directions with different but related outcomes. We wonder how there could be such a systematic difference in findings.
Williams K, Monroe HM. Cardiovascular Disease Risk Prediction Factors. JAMA. 2012;308(19):1969. doi:10.1001/jama.2012.14042