In:
Circulation, Ovid Technologies (Wolters Kluwer Health), Vol. 125, No. suppl_10 ( 2012-03-13)
Kurzfassung:
Purpose: Risk prediction models are assessed by calibration, discrimination, and reclassification. Net reclassification improvement (NRI) is a popular reclassification statistic in clinical epidemiology. However, methods to assess NRIs in a meta-analysis are not well described. Using an example of the CKD-EPI and the MDRD Study equations, we demonstrate a method to meta-analyze NRIs for eGFR category based on different equations. Methods: In a sample of 29 of 46 cohorts joining the CKD Prognosis Consortium (CKD-PC), NRI was calculated for reclassification across the following six eGFR categories (≥90, 60–89, 45–59, 30–44, 15–29, 〈 15 ml/min/1.73m 2 ) for CKD-EPI vs. MDRD Study equations. Standard errors of NRIs were calculated in analogy to McNemar’s test for paired proportions as √{(proportion of reclassification to high risk [low eGFR] category in those who developed events + proportion of reclassification to low risk [high eGFR] category in those who developed)/(# of events) + (reclassification to low risk category in those who did not develop events + reclassification to high risk category in those who did not develop)/(# of no events)}. NRIs were meta-analyzed in a random effects model applying the metan command in STATA. Results: There were a total of 830,132 participants and 22,647 deaths due to cardiovascular disease (coronary heart disease, stroke, or heart failure). NRIs for cardiovascular mortality based on eGFR category varied from −0.07 to 0.27 (p-values from 〈 0.001 to 0.921; 22 of 29 studies with p 〈 0.05) among cohorts. Standard errors also varied from 0.005 to 0.049. Meta-analyzed NRI was 0.09 (95% CI: 0.07–0.12, p 〈 0.001), suggesting that the CKD-EPI equation classifies individuals more correctly than the MDRD Study equation. The I-squared was 96.0% (p 〈 0.001), suggesting high heterogeneity across studies. Conclusion: We demonstrated how to meta-analyze NRIs, which estimate the clinical risk implication of new eGFR equations on cardiovascular disease, a leading cause of death among individuals with CKD.
Materialart:
Online-Ressource
ISSN:
0009-7322
,
1524-4539
DOI:
10.1161/circ.125.suppl_10.AP095
Sprache:
Englisch
Verlag:
Ovid Technologies (Wolters Kluwer Health)
Publikationsdatum:
2012
ZDB Id:
1466401-X
Permalink