Single-Nucleotide Polymorphism Data Support the General Unrelatedness of the Males in the Agricultural Health Study

The Agricultural Health Study (AHS; ref. 1) is a cohort study that collected longitudinal data on farmers and their spouses to understand the health risks of exposure to pesticides and other agricultural agents. From 1993 to 1997, a total of 57,310 licensed pesticide applicators enrolled in the study, representing 82% of the licensed private pesticide applicators in Iowa and North Carolina. The AHS has evaluated a number of health outcomes, including cancer, respiratory disease, neurologic outcomes, and diabetes. In addition to the questionnaire data collected for the main study, a nested case–control genetic association study was conducted to identify genetic factors that modified pesticide risks for prostate cancer. In this study, men with or without prostate cancer were selected for genotyping. Results of the gene–environment analyses have previously been reported (2–4).

These gene–environment interaction studies suggested that specific single-nucleotide polymorphisms (SNP) may interact with use of specific pesticides by these farmers to contribute to prostate cancer risk. As farming is often a family-based business that is passed from father to son, we hypothesized that some of the individuals in the case–control study could potentially be related. As relatedness among the farmers could contribute bias in evaluating gene–environment interactions, we used the SNP data to estimate the overall genomic sharing to test to see whether cryptic relatedness was present.

The SNP data were previously described in ref. (2). Briefly, genotyping was performed with the Custom InifiniumR BeadChip Assays from Illumina (iSelect) as part of an array of 26,512 SNPs. Quality control procedures, including STRUCTURE, linkage disequilibrium, and principal components analysis (PCA), identified 25,009 SNPs across 2,220 individuals for further analysis. For this analysis, we performed additional quality control steps. Markers and individuals were removed if they had low genotyping efficiency (missing data >1%). SNPs showing significant deviation from Hardy–Weinberg proportions (P < 0.00001 after a χ² test of associations) were also removed. A total of 22,706 SNPs and 1,837 total individuals passed the entire quality control scheme.

Next, we used all of the control data (1196 individuals), along with a random sampling of 10% of the cases (64 of 641 individuals), to ensure sufficient representation of the male AHS population in general, and to avoid over-representing patients with prostate cancer (5) during the kinship analysis. Hence, 1,260 total individuals were used in this analysis.

The genetic data allowed us to directly test for potential relatedness in the cohort by assessing the overall genomic sharing across the genome. We used PLINK (6), to estimate inbreeding coefficients for each individual and genome-wide identity-by-descent (IBD) estimates for all pairs of individuals. In addition, PCA via EIGENSOFT (7) was performed to visualize the overall patterns of sharing for a high number of principal components.

The results of our PLINK analyses for potential cryptic relatedness are shown in Fig. 1: The inbreeding coefficients for each individual and the mean inbreeding coefficient (solid red line) are shown in Fig. 1A. The maximum inbreeding coefficient for all of the individuals in the study was 0.1433 whereas the majority (99.52%) of individuals had an inbreeding coefficient less than 0.0625, which indicates that the individuals in this study are overwhelmingly unrelated.

Using IBD estimates, we generally saw no evidence for potential cryptic relatedness. In Fig. 1B, the probabilities of IBD for all possible pairs of individuals are shown—P(IBD = 0) and P(IBD = 1) on the y- and x-axis, respectively. The vast majority of individual pairs have P(IBD = 0) close to 1.0, which indicates an unrelated population. In fact, only six individual–individual pairs scored P(IBD = 1) > 0.38 of 793,170 total pairwise combinations.

In Fig. 2, we show the first three as well as the ninth principal components to visualize the variance across the data. The first three principal components show no significant subpopulations. The ninth component is the first display of separation for the individuals with relatively higher P(IBD = 1). Cryptic relatedness is often not revealed in lower components, but is evident in higher components (8).

Overall, the results of our analysis indicate that there is generally no cryptic relatedness among this subset of male applicators in the AHS. The inbreeding and IBD results indicate that levels of genetic sharing are not higher than expected for an unrelated cohort in an unstratified population. This is an important null result, as it justifies treating individuals in the AHS as independent observations for studies of gene–environment interactions.

No potential conflicts of interest were disclosed.

Conception and design: A.A. Motsinger-Reif, M.C. Alavanja, J.A. Hoppin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.C. Alavanja, J.A. Hoppin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.R. Jack, A.A. Motsinger-Reif, S. Koutros, J.A. Hoppin

Writing, review, and/or revision of the manuscript: J.R. Jack, A.A. Motsinger-Reif, S. Koutros, M.C. Alavanja, L.E. Beane Freeman, J.A. Hoppin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Koutros, L.E. Beane Freeman, J.A. Hoppin

Study supervision: A.A. Motsinger-Reif

This work was supported in part by the intramural research program of the NIH, National Cancer Institute (Z01-ES049030; PIs, M.C. Alavanja and L.E. Beane Freeman), National Institute of Environmental Health Sciences (Z01-CP010119; PI, D.P. Sandler), and Pilot Project Program (PI, Alison Motsinger-Reif) of the North Carolina State University's Center for Human Health and the Environment.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.