Answers are grounded in this paper and cite their sources.
This editorial highlights methodological concerns, including collinearity and selection bias, in a study assessing baPWV as a predictor of diastolic function.
We have read the article by Park and colleagues1 that was recently published in the Journal of Clinical Hypertension. The authors assessed the effect of low brachial-ankle pulse wave velocity (baPWV) on normal diastolic function in Korean adults without documented cardiovascular disease older than 50 years. The results provide evidence on the role of baPWV in the prediction of normal left ventricular diastolic function.1 Although the results were interesting, we think some methodological and statistical issues should be noted. It appears that the estimated confidence interval for effect of estimated glomerular filtration rate ≥60 mL/min per 1.73 m2 on normal diastolic function is excessively wide, eg, 0.54 to 43.02.1 Here, it seems that collinearity among the determinants of normal diastolic function reduce the precision of the estimated effect for some of the predictors. When studying the relatively correlated predictors, it may be wiser to use efficient statistical testing such as variance inflation factor to check multicollinearity before performing regression analysis.2, 3 The authors stated that the study was a consecutive sample of an elderly population without overt cardiovascular disease and cardiac structural changes1; however, they did not perform a sensitivity analysis to test how much the results were influenced by this nonrandom selection (selection bias). Several techniques are currently available to adjust for selection bias.4 In addition, although the authors expressed that the model that included baPWV compared with the model without baPWV had good performance for identifying normal diastolic function (P=.001),1 it is unclear how much the discrimination improvement index changed by adding baPWV to the model. For application in clinical practice, we suggest that the authors report the difference in discrimination power of the two models with and without baPWV. None.
No takes yet. Share an insight, caveat, or question.
Safiri et al. (2017) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: