Serum levels of insulin-like growth factor-I and insulin-like growth factor-I binding protein-3: Quality control for studies of stored serum

The insulin-like growth factor (IGF) axis, particularly IGF-I and IGF binding protein-3 (IGFBP-3), has been the subject of much attention because of its role in juvenile growth and their association with cancers at several sites. However, epidemiologic studies of IGF-I and IGFBP-3 have had mixed results and several authors have speculated that quality control (QC), sample storage history, and other methodologic concerns could play a role in this heterogeneity. This article documents the results of storage history and QC efforts for a study of IGF-I and IGFBP-3 in 6,226 serum samples from the National Health and Nutrition Examination Survey III (NHANES III). The study was carried out on site at Diagnostic Systems Laboratories in Webster, Texas, using the IGF-I ELISA (DSL 10-5600) and the IGFBP-3 immunoradiometric assay (DSL 6600). A run-in study of assay performance suggested that plates, days, and weeks significantly affected the variance of both assays. Analysis of samples with different storage histories also indicated strong effects of storage history. Serum samples disbursed to laboratories for measurement of diverse analytes and then returned for storage showed reductions in serum IGF-I level averaging 43% and reductions in IGFBP-3 of 25% compared with samples shipped immediately to the repository for long-term storage at -80 degrees C. Therefore, the main study was carried out using samples that had been shipped directly to the National Center for Health Statistics/NHANES collection center for storage. Laboratory analyses of NHANES III and QC samples were carried out over similar to 10 months. QC was monitored through repeated testing of blood samples from six individuals, with two individuals tested twice on each plate. Assay performance was stable over the entire study and coefficients of variation averaged 2% to 3% within plates and similar to 14% for IGF-I and similar to 11.5% for IGFBP-3 over the entire study. Coefficients of variation varied significantly among individual QC subjects, ranging from 12.3% to 17.6% for IGF-I and 8.9% to 12.8% for IGFBP-3. Based on Levy-Jennings plots, similar to 5% of the plates used for IGF-I in the main study were out of compliance. Finally, location on a plate had small but significant effects on IGF-I level. Together, these results highlight the need for care in large studies of putative biomarkers for cancer risk and illustrate some probable sources of heterogeneity in past epidemiologic studies of the IGF axis and cancer.