SUN-LB46 Differences in IGF-I Concentrations Between European and US Populations - Consequences for Reference Intervals

Background: IGF-I is the most widely used biomarker for management of GH related diseases. Reproducible assays and method-specific reference intervals (RIs) are crucial determinants of its clinical utility. Assay validation and RIs based on >15,000 subjects were published for the IDS iSYS IGF-I assay (J Clin Endocrinol Metab 2014). We now analyzed distribution of IGF-I results obtained in routine samples analyzed by accredited laboratories in the US and Europe, all using the IDS iSYS assay. Methods: All results from routine IGF-I measurements during the past 5 years in 4 laboratories were included (US lab n=778,173 males/710,752 females; European labs (Germany/Belgium, n=23,220 males/40,183 females). Assay performance across laboratories was confirmed through proficiency testing schemes and exchange of patient samples. We constructed RIs adjusted for age/sex from European and US cohorts separately using a modified Hoffmann approach (Am J Clin Pathol 2015), and compared to the originally published RIs (n=6697 males/8317 females, adults from Europe). A subset of US samples was used to compare IGF-I between regions with lower (Colorado) and higher (Alabama) mean body mass index (BMI). Results: Lower limits (LLs) of RIs calculated from routine results are superimposable to LLs from the original publication for all ages and sexes, regardless whether IGF-I results were from Europe or the US. For groups with sufficient n, upper limits (ULs) of RIs calculated from European routine data were also not statistically different from the originally published central 95%. However, a striking difference exists in calculated ULs from data of European and US origin: For ages 10-18 years, calculated UL on average was 149.3 ng/mL (34.6%) higher in boys and 94.9 ng/mL (19.8%) in girls from the US. In adults (19-95 years), calculated UL on average was 45 ng/mL (20.3%) higher in males and 29.7 ng/mL (13.8%) in females from the US. Within the US, mean IGF-I was significantly higher in samples from Colorado (lower mean BMI) than in Alabama (p<0.0001) across age- and sex groups, although the difference between the two states was smaller than between each of them and Europe. Conclusion: This study provides evidence that in sufficiently large datasets, both, direct sampling (as in the original publication) and the indirect Hoffmann algorithms provide statistically comparable RI limits and may be considered as accurate representation of results distribution in the disease-free populations. More importantly, we demonstrate that even with tight cross-correlation and continuous monitoring of IGF-I assay performance RIs generated in different populations can be different. Notably, in our extremely large study, the difference between Europe and the US was clinically relevant only at the UL. Although our study cannot reveal the cause of the difference, we suggest using adapted RIs for the US.