Extending the scope of pooled analyses of individual patient biomarker data from heterogeneous laboratory platforms and cohorts using merging algorithms

Abstract

Background: A common challenge in medicine, exemplified in the analysis of biomarker data, is that large studies are needed for sufficient statistical power. Often, this may only be achievable by aggregating multiple cohorts. However, different studies may use disparate platforms for laboratory analysis, which can hinder merging. Methods: Using circulating placental growth factor (PIGF), a potential biomarker for hypertensive disorders of pregnancy (HDP) such as preeclampsia, as an example, we investigated how such issues can be overcome by inter-platform standardization and merging algorithms. We studied 16,462 pregnancies from 22 study cohorts. PIGF measurements (gestational age >= 20 weeks) analyzed on one of four platforms: R & Systems, Alere (R) Triage, Roche (R) Elecsys or Abbott (R) Architect, were available for 13,429 women. Two merging algorithms, using Z-Score and Multiple of Median transformations, were applied. Results: Best reference curves (BRC), based on merged, transformed PIGF measurements in uncomplicated pregnancy across six gestational age groups, were estimated. Identification of HDP by these PIGF-BRCS was compared to that of platform-specific curves. Conclusions: We demonstrate the feasibility of merging PIGF concentrations from different analytical platforms. Overall BRC identification of HDP performed at least as well as platform-specific curves. Our method can be extended to any set of biomarkers obtained from different laboratory platforms in any field. Merged biomarker data from multiple studies will improve statistical power and enlarge our understanding of the pathophysiology and management of medical syndromes. (C) 2015 International Society for the Study of Hypertension in Pregnancy. Published by Elsevier B.V. All rights reserved.