Abstract: SA-PO770
Methods to Identify Novel, Repeatedly Measured Biomarkers of CKD Outcomes
Session Information
- CKD: Epidemiology, Risk Factors, Prevention - III
October 27, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 1901 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Liu, Qian, Arbor Research Collaborative for Health, Ann Arbor, Michigan, United States
- Smith, Abigail R., Arbor Research Collaborative for Health, Ann Arbor, Michigan, United States
- Mariani, Laura H., University of Michigan, Ann Arbor, Michigan, United States
- Nair, Viji, University of Michigan, Ann Arbor, Michigan, United States
- Zee, Jarcy, Arbor Research Collaborative for Health, Ann Arbor, Michigan, United States
Background
Identifying novel biomarkers is critical to advancing diagnosis and treatment of CKD, but relies heavily on the statistical methods used. Inappropriate methods can lead to both false positive and false negative associations between biomarkers and outcomes. This study assessed accuracy of methods using computer simulations and compared biomarker effect estimates in NEPTUNE, a prospective cohort study of patients with glomerular disease.
Methods
We compared four methods for analyzing repeatedly measured biomarkers in Cox models: 1) baseline, that only uses the single baseline biomarker value, 2) time-dependent with last observation carried forward (LOCF), that assumes the biomarker is unchanged until the next value, 3) time-averaged, that averages the biomarker values over all follow up and uses that average as a baseline covariate, and 4) time-dependent cumulative average, that updates the average using biomarker values at or before each time point.
Results
Simulation results showed the time-averaged method often gave extremely biased, inaccurate results. In NEPTUNE, all assessed urinary biomarkers were positively associated with 40% reduction in eGFR. Compared to the LOCF method, the baseline method had 6-21% lower hazard ratios (HRs), and the time-averaged method had 3-32% higher HRs [figure]. For complete remission, Eotaxin and TIMP1 were negatively associated with the outcome, and PDGF-BB had little association. HRs from the time-averaged method were always higher (1-13%) than the LOCF method.
Conclusion
Different analytic methods resulted in markedly different results. Using inappropriate methods such as time-averaging can bias effect estimates, while other methods provide prognostic (baseline), additive (cumulative average), or instantaneous (LOCF) effects depending on the hypothesized underlying mechanism.
Funding
- NIDDK Support