Biological variation estimates of thyroid related measurands – meta-analysis of BIVAC compliant studies
Pilar Fernández–Calle, Jorge Díaz–Garzón, William A. Bartlett, Sverre Sandberg, Federica Braga, Beatriz Boned, Anna Carobene, Abdurrahman Coşkun, Elisabet González-Lao, Fernándo Marqués-García, Carmen Perich, Margarida Simón, Aasne K. Aarsand
Abstract
Abstract Objectives Testing for thyroid disease constitutes a high proportion of the workloads of clinical laboratories worldwide. The setting of analytical performance specifications (APS) for testing methods and aiding clinical interpretation of test results requires biological variation (BV) data. A critical review of published BV studies of thyroid disease related measurands has therefore been undertaken and meta-analysis applied to deliver robust BV estimates. Methods A systematic literature search was conducted for BV studies of thyroid related analytes. BV data from studies compliant with the Biological Variation Data Critical Appraisal Checklist (BIVAC) were subjected to meta-analysis. Global estimates of within subject variation (CV I ) enabled determination of APS (imprecision and bias), indices of individuality, and indicative estimates of reference change values. Results The systematic review identified 17 relevant BV studies. Only one study (EuBIVAS) achieved a BIVAC grade of A. Methodological and statistical issues were the reason for B and C scores. The meta-analysis derived CV I generally delivered lower APS for imprecision than the mean CV A of the studies included in this systematic review. Conclusions Systematic review and meta-analysis of studies of BV of thyroid disease biomarkers have enabled delivery of well characterized estimates of BV for some, but not all measurands. The newly derived APS for imprecision for both free thyroxine and triiodothyronine may be considered challenging. The high degree of individuality identified for thyroid related measurands reinforces the importance of RCVs. Generation of BV data applicable to multiple scenarios may require definition using “big data” instead of the demanding experimental approach.