In Response:

Genetic and environmental factors influence circulating insulin-like growth factor I (IGF-I) and insulin-like growth factor-binding 3 (IGFBP-3) levels. Our purpose was to estimate associations of IGF1 and IGFBP3 polymorphisms with their respective circulating protein levels in premenopausal African American and Caucasian women (1). On our a priori directed acyclic graph (2), race was the only variable affecting the polymorphisms. Given our stratification by race, it was unnecessary to adjust for covariates mentioned by Dr. Jernström (age, body mass index, parity, and oral contraceptives; ref. 3) regardless of whether they affect circulating levels. In addition, if a covariate is on the causal pathway between the polymorphisms and the circulating levels, adjustment would produce bias. For those covariates not on causal pathways, adjustment may reduce precision without improving validity. Nonetheless, we did consider age adjustment. Unsurprisingly, given the narrow age range of our study population, age adjustment had a negligible impact on the magnitude and precision of the estimated associations. Hence, we reported the unadjusted estimates.

Dr. Jernström also expressed concern about possible variability in the menstrual cycle phase at blood collection, which could have contributed to imprecision and information bias in our use of a single measurement for each individual to estimate mean circulating levels. We regret that information on the menstrual cycle phase at blood collection in our study was limited. Fortunately, previous research has indicated that absolute differences across the menstrual cycle phase were minor (4).

Dr. Jernström suggested an exploration of gene-environment interactions, which could offer a more comprehensive understanding of the biological mechanisms regulating IGF-I and IGFBP-3 expression and variability. However, these analyses were beyond the scope of our article and would be limited by the small sample sizes of both racial groups. Nevertheless, at Dr. Jernström's request, we restricted our sample to women who were not currently using oral contraceptives and produced the following results for three of our strongest IGF1 SNP findings (rs6214 in Caucasians; rs6219 and rs2946834 in African Americans). The original estimated mean differences in IGF-I levels for variants versus reference genotypes were 16 or 17 ng/mL. These estimates changed by only 1 to 2 ng/mL, with slight decreases in precision after exclusion of oral contraceptive users. Given that less than 10% of our sample were oral contraceptive users, these results were expected. Although we certainly acknowledge the probable importance of gene-environment interactions in the regulation of these proteins, this rather complex research question should be explored in larger study populations.