Association tests of interleukin-6 (IL-6) and type II tumor necrosis factor receptor (TNFR2) genes with bone mineral density in Caucasians using a re-sampling approach.

Interleukin 6 (IL-6) and tumor necrosis factor (TNF) are important cytokines for bone turnover. In this study, a promoter C-174G single-nucleotide polymorphism (SNP) within the IL-6 gene affecting the transcription rate of IL-6 and an exon 6 T676G SNP of the TNF receptor 2 (TNFR2) gene causing an M196R amino-acid change were examined for their relationship with bone mineral density (BMD). Four hundred and five multi-offspring Caucasian families, including 389 male children and 744 female children, were used. One thousand re-samplings were conducted and in each data set, one child was randomly chosen from each family. For each data set, one-way analysis of variance (ANOVA) test was independently implemented using age, age2, sex, height and weight as covariates. There were 523, 288, 204 and 369 significant results out of 1,000-replicate re-samplings of the data of the IL-6 SNP (P<0.05) for one-third, mid-distal, ultradistal radius BMD, and the first principal component (PC1) extracted from the three radial BMDs, respectively, which means that the confidences for associations of the C-174G SNP in the IL-6 gene with one-third, mid-distal, ultradistal radius (totally called distal forearm) BMDs, and PC1, were 52.3, 28.8, 20.4 and 36.9%, respectively. For this SNP with BMD at other skeletal sites and the TNFR2 T676G SNP with BMD at any site, significant results were far less than 200 times out of 1,000 re-sampling replicates. The exceedingly consistent permutation results further improved the confidence of the associations. It may imply that the IL-6 C-174G SNP is associated with distal forearm BMD, but there is no evidence that the TNFR2 T676G SNP is related with BMD in US Caucasians. This is the first attempt to conduct association test utilizing a re-sampling approach. Our results may be more informative than other association analyses that were only based on one sampling result. The results also suggest that different samplings could produce significantly diverse results even for the same population and the results from one sampling are unlikely to be conclusive. Our results have significant implications for association studies and interpretation of non-reproducible association findings.