Pooling biospecimens within individuals, a solution to highly variable expensive biomarkers: practical or idealistic?

ABSTRACT Background: Existing methodology to analyze pooled biospecimens, a cost-effective way for measuring expensive biomarkers, consider study designs where biospecimens are pooled across individuals, but not within individuals. Pooling biospecimens across individuals is not effective when the target biomarker is highly variable. Furthermore, ignoring the variability leads to grossly biased effect estimates (s). Repeated samples within subject are recommended but not always practical. A novel study design is needed to bridge this limitation. Methods: We propose a study design where the target biomarker is measured from pooled biospecimens collected over time from every study participant and derived a bias-correction factor corresponding to this pooling strategy. We simulated 1000 different samples of size 1088 with different pooling strategies and 3 different scenarios of biomarker variability, namely 1) pooling error > measurement error, 2) pooling error < measurement error, and 3) pooling error = measurement error. With this simulation study we assessed the robustness of our proposed design by comparing the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) for uncorrected and bias-corrected s for a binary outcome. Results: The MAE and RMSE for bias-corrected s were lower compared to uncorrected s, across all scenarios of biomarker variability. The MAE and RMSE are minimum when pooling error < measurement error. Conclusion: Combined with an appropriate regression calibration method, pooling biospecimens within individuals is a statistically robust and cost-effective solution to measuring highly variable expensive biomarkers.