Lipid trait-associated genetic variation is associated with gallstone disease in the diverse Third National Health and Nutrition Examination Survey (NHANES III)

The study population described here includes phase 2 participants ascertained between 1991 and 1994 as part of the Third National Health and Nutrition Examination Survey (NHANES III). NHANES is now conducted yearly by the National Center on Health Statistics (NCHS) at the Center for Disease Control and Prevention (CDC) with the major goal of assessing the health and nutrition status of Americans regardless of health status. Beginning with NHANES III phase 2, CDC began collecting bio-specimens from consenting participants for genetic studies. NHANES is a complex survey design that oversamples specific age groups (such as the elderly) and racial/ethnic groups (such as non-Hispanic blacks).

All eligible participants were given an approximately hour long interview. Participants also underwent a health examination by the Mobile Exam Centers (MEC), and blood and urine samples were also obtained [25]. Participants were eligible for NHANES at two months of age or older, while participants consenting for bio-specimen collection for DNA extraction had to be at least twelve years of age. Non-eligible individuals included military personnel and institutionalized civilians.

Genetic NHANES III includes 7,159 participants, of which 2,631 are self-described non-Hispanic white, 2,108 are self-described non-Hispanic black, and 2,073 are self-described Mexican American. All procedures were approved by the CDC Ethics Review Board and written informed consent was obtained from all participants. Because no identifying information was accessed by the investigators, Vanderbilt University’s Institutional Review Board determined that this study met the criteria of “non-human subjects.”

Gallstones were defined by a positive “yes” response to an administered questionnaire exam asking, “Has the doctor ever told you that you had gallstones?” or a positive ultrasound reading (presence of scar tissue (surgical removal of gallbladder or cholecystectomy) or demonstration of gallstones present in the gallbladder) from trained technicians. Controls were defined as gallstone-free participants who answered “no” response to the administered questionnaire or had no history of gallstone removal surgeries. To assess the potential for misclassification, we examined participants with both ultrasound and questionnaire data to validate asymptomatic cases and to identify misclassified controls.

A total of 49 GWAS-identified SNPs were selected for study based on evidence of being previously associated with at least one lipid profile (HDL-C, LDL-C, TG) in published candidate gene and genome-wide association studies [10, 12, 18, 20, 22]. The SNP selection criteria meet the genome-wide significance of p < 10^-8 in previous published studies [22]. SNP data was generated by genotyping using Sequenom and Illumina BeadXpress platforms. Thorough genotyping and SNP details (gene region, physical location, coding type, etc.) have been previously published in Text S1 of [22]. CDC quality control metrics were implemented on the 49 SNPs and included tests of Hardy-Weinberg Equilibrium (threshold of p > 0.0001 in at least two of the three race/ethnicities) as well as concordance with blinded duplicates supplied by CDC (at least 95%). Also, as part of the PAGE study [23], we genotyped these lipid-associated SNPs in 360 HapMap samples for network-wide quality control assessment.

Participants less than 18 years of age were excluded from this study. Single SNP tests of association were performed using logistic regression. Gallstone status (yes/no) was the dependent variable and each SNP assuming an additive genetic model was the independent variable. All models were adjusted for age, sex, and body mass index. Logistic regressions were performed stratified by self-reported race/ethnicity, and genetic effect estimates were expressed as odds ratios (ORs) with 95% confidence intervals (95% CIs). All analyses were conducted in SAS v9.2 (SAS Institute, Cary, NC) using the Analytic Data Research by Email (ANDRE) portal of the CDC Research Data Center in Hyattsville, MD. Results were plotted using Synthesis-View [26]. Linkage disequilibrium was assessed using SNP Annotation and Proxy Search (SNAP) [27].

Study population characteristics by self-described race/ethnicity and gallstone disease status are given in Table 1. Overall, we identified 446, 179, and 227 cases of gallstone disease among non-Hispanic whites, non-Hispanic blacks, and Mexican Americans, respectively. As expected, based on the known epidemiology of gallstone disease [3, 13], cases tended to be female, older, and have a higher BMI compared with controls (Table 1). Also, overall, there were proportionally fewer cases among non-Hispanic blacks compared with non-Hispanic whites and Mexican Americans, which is consistent with the lower prevalence of this disease among this group [3].

Among the 49 lipid trait-associated SNPs tested for association with gallstone disease, five, four, and four SNPs were associated with disease risk at a liberal significance threshold of 0.05 among non-Hispanic whites, non-Hispanic blacks, and Mexican Americans, respectively (Table 2; Additional file 1: Table S1). Among all groups, the most significant finding was observed in non-Hispanic whites (OR = 1.89; 95% CI = 1.44-2.49; p = 0.0001) for non-synonymous ABCG5 rs6756629, a variant whose minor allele (A) was previously associated with decreased LDL-C and total cholesterol and increased triglycerides among European-descent populations [28]. This associated SNP is in strong linkage disequilibrium with ABCG5 rs11887534 identified in a genome-wide association within a European-descent population for gallstone disease [20, 24]. The second association identified for gallstone disease in non-Hispanic whites involved ABCG8 rs6544713 (OR = 1.35; 95% CI = 1.14-1.61; p = 0.0007), a variant previously associated with LDL-C levels in European-descent populations [29]. ABCG8 rs6544713 is not in linkage disequilibrium with rs6756629 or rs11887534 (both r² = 0.047 in CEU 1000 Genomes Project) and therefore most likely represents an independent association with gallstone disease risk.

Interestingly, despite the smaller sample size, four associations at p < 0.05 were observed among non-Hispanic blacks. The most significant finding among non-Hispanic blacks was BUD13 rs28927680 (OR = 1.40; 95% CI = 1.06-1.86, p = 0.019) previously associated with HDL-C and TG in European-descent populations [29]. Other nominal associations observed in non-Hispanic blacks included genetic variants previously associated with HDL-C (CETP rs3764261 (OR = 0.71; 95% CI 0.56-0.9)) and LDL-C (HNF1A/LEF1 rs2650000 (OR = 1.50; 95% CI 1.02-2.21) and APOB rs754523 (OR = 1.36; 95% CI 1.04-1.78)).

Among Mexican Americans, four associations were nominally associated with gallstone disease risk: HDL-C (ABCA1 rs1883025 (OR = 0.78; 95% CI 0.62-0.98; p = 0.033), ABCA1 rs4149268 (OR = 0.75; 95% CI 0.60-0.95; p = 0.015), and PCIF1 rs7679 (OR = 1.51; 95% CI 1.13-2.01; p = 0.048)) and TG (DOCK7 rs10889353 (OR = 1.23; 95% CI 1.05-1.67; p = 0.017). No association was associated at p < 0.05 across all three racial/ethnicity groups (Table 2).

In addition to examining genetic associations for gallstone disease risk by race/ethnicity, we also examined the associations by associated lipid trait (Figures 1, 2 and 3; Additional file 2: Figure S1). That is, among the 49 SNPs tested here, 24, 14, and 17 SNPs were reported to be associated with HDL-C, LDL-C, and TG levels, respectively. As previously mentioned, the most significant finding involved an LDL-C associated variant, and, overall, approximately one-third (4 out of 14 or 29%) of LDL-C associated SNPs were nominally associated with gallstone disease risk in any one population. The other nominal findings reported at p < 0.05 were associated with HDL-C (6 out of 24 or 25%) or triglycerides (4 out of 17 or 24%) associated SNPs.

ABCG5 rs6756629 (R50C) is in strong linkage disequilibrium with rs11887534 (D19H), a variant previously associated with gallstone disease risk in populations of European-descent [20]. Pairwise linkage disequilibrium (r²) was 1.00 in CEU from the 1000 Genomes Project Pilot Study [30] and 0.95 in a German study population [24]. Thus, the association observed here with rs6756629 likely represents a replication of the association identified in the original GWAS [20]. The magnitude of effect (OR = 1.89) is somewhat smaller compared with the original (OR = 2.2) [20] and subsequent reports [18, 21, 24, 31], but the confidence intervals overlap suggesting similar overall genetic effect sizes for this population.

Very few genetic association studies and no GWA studies of gallstone disease risk have been performed in populations of non-European descent. A few studies have examined the association between ABCG5 rs11887534 and gallstone disease in populations from China [24], Taiwan [32], and India [24]. To our knowledge, no study has examined ABCG5 rs11887534 or tagged rs6756629 in African Americans or Mexican Americans for gallstone disease risk. As expected based on the known epidemiology of gallstone disease, the sample sizes available for non-Hispanic blacks (n = 179 cases) and Mexican Americans (n = 227 cases) were smaller compared with non-Hispanic whites (n = 446 cases). ABCG5 rs6756629 was not associated with gallstone disease risk in non-Hispanic blacks (Table 2). The ABCG5 variants rs11887534 and rs6756629 have lower pair-wise linkage disequilibrium in African-descent populations (YRI r² = 0.613) compared with European-descent populations (CEU r² = 1.000) based on data from the 1000 Genomes Project Pilot Study [30]. Recent statistical and functional data suggest that ABCG5 rs11887534 (D19H) is likely the functional variant responsible for the association with gallstone disease risk while rs6756629 is likely a tagSNP [24]. The smaller sample sizes available for non-European descent populations coupled with differences in linkage disequilibrium between the genotyped rs6756629 and putative functional rs11887534 likely resulted in low statistical power to detect this association in non-Hispanic blacks and Mexican Americans. Further ABCG5 fine-mapping and functional studies are needed to fully catalogue gallstone disease risk variants in diverse populations.

This study has several limitations and strengths. A major limitation is sample size and power. Although genetic NHANES III is large overall (n = 7,159), the number of cases and controls available for this gallstone disease study is small when stratified by race/ethnicity. More recent NHANES did not collect survey or ultrasound data; therefore, large sample sizes or independent replication datasets for genetic associations for gallstone disease risk are not available in NHANES. To maximize the power of the present study within NHANES III, we used both questionnaire and ultrasound data (where ultrasound data represented the “truth” or gold standard when both were available per participant) to define cases and controls. Indeed, limiting the dataset to participants with both data types would have resulted in a 14% reduction in sample size. A trade-off of this approach to maximize sample size, however, is that we may have misclassified participants who only had questionnaire data but no corroborating ultrasound data. To examine the possible extent of misclassification, we compared our case/control definition (questionnaire or ultrasound data) with case/control status among participants with both data types (excluding those participants with questionnaire data only). Overall, we found little evidence for misclassification using questionnaire data only (positive predictive value = 99%, negative predictive value = 100%, sensitivity = 100%, and specificity = 99%). We also compared results of tests of association between the two definitions and found little difference (Additional file 3: Figure S2).

Another limitation to the study is that we did not adjust our significance level for multiple statistical tests. Our liberal significance threshold coupled with the lack of an independent dataset can be cause for concerns for false positive associations. However, the most significant association reported here replicated an association previously reported in a GWAS in a population of similar race/ethnicity. The association between rs6756629 and gallstone disease risk in non-Hispanic whites, along with the association involving ABCG5 rs6544713, would be considered significant with a conservative Bonferroni correction. The facts that the associations represent a replication of a previous report [20], that rs6756629 tags a likely functional SNP associated with gallstone disease risk [24], and that both survive correction for multiple testing support the conclusion that these are not a false positive findings.

A major strength of the current study is its diversity. Little to no data exist for the genetics of gallstone disease risk in African Americans and Mexican Americans. Further discovery studies are needed to identify the full genetic architecture of gallstone disease risk in diverse populations [33]. Another major strength of this current study is the combination of questionnaire and ultrasound data available in NHANES III. For participants with overlapping questionnaire and ultrasound data, we were able to identify asymptomatic cases of gallstones that would have otherwise been misclassified using questionnaire data only. Misclassification of case/control status lowers power to detect genetic associations; therefore, careful phenotyping is one of many essential components in the conduction of genetic association studies [34]. Combination of these data types maximized sample size for this study whereas restricting eligible cases and controls to participants with ultrasound data only would have reduced the sample size by 14% and therefore reduced power.