Causal Link Between Vitamin D and Total Testosterone in Men

A Mendelian Randomization Analysis

Chi Chen; Hualing Zhai; Jing Cheng; Pan Weng; Yi Chen; Qin Li; Chiyu Wang; Fangzhen Xia; Ningjian Wang; Yingli Lu


J Clin Endocrinol Metab. 2019;104(8):3148-3156. 

In This Article

Abstract and Introduction


Context: Low circulating vitamin D levels have been associated with lower total testosterone levels. These epidemiological associations, if true, would have public health importance because vitamin D deficiency is common and correctable. We tested whether genetically lowered vitamin D levels were associated with lower total testosterone (T), using Mendelian randomization (MR) methodology.

Design and Setting: A total of 4254 men were enrolled from the Survey on Prevalence in East China for Metabolic Diseases and Risk Factors (SPECT-China) study, which was performed at 23 sites in eastern China during 2014 to 2016. Using four single-nucleotide polymorphisms strongly associated with 25-hydroxyvitamin D [25(OH)D] levels, we created a genetic risk score (GRS) as an instrumental variable to estimate the effect of genetically lowered 25(OH)D on total T.

Main Outcome Measure: Total T was detected by chemiluminescence assay.

Results: Lower 25(OH)D levels were associated with lower total T (β, 0.40; 95% CI, 0.23 to 0.58) after multivariable adjustment. Per-SD increase in the vitamin D GRS (VD_GRS) was significantly associated with 25(OH)D (β, −1.64; 95% CI, −2.04 to −1.24) and with total T (β, −0.19; 95% CI, −0.37 to −0.02). Using VD_GRS as the instrumental variable in the MR analysis, the causal regression coefficient of genetically determined per-SD increase for 25(OH)D on total T was 0.12 (95% CI, 0.02 to 0.22).

Conclusion: We provide evidence for the biologically plausible causal effects of 25(OH)D on total T using MR analysis. Whether vitamin D supplementation can raise androgen levels merits further investigation in long-term, randomized controlled trials.


Vitamin D deficiency has been becoming a pandemic public health problem in developing and developed countries and in all age groups.[1] Beyond the well-known relationship between vitamin D deficiency and musculoskeletal diseases, accumulating evidence has indicated the effect of vitamin D deficiency on extra-skeletal disorders such as certain cancers,[2] autoimmune diseases,[3] infectious diseases,[4] diabetes mellitus,[5] and neurocognitive disorders.[6] It is noteworthy that vitamin D is a relatively cheap and safe supplement for most individuals, thus there may be potentially a great upside to increasing the vitamin D status in the population worldwide to improve health problems, if the casual links are determined.[7]

Vitamin D deficiency might lower testosterone (T) levels. Observational evidence supports this speculation: Our previous study found that lower 25-hydroxyvitamin D [25(OH)D] level was associated with lower T level and increased OR of hypogonadism,[8] which was in accordance with findings of other cross-sectional studies,[9,10] and the association revealed a concordant seasonal variation in one study.[11] Furthermore, one trial indicated that vitamin D supplementation increased T levels in overweight subjects,[12] and another study in men with normal baseline T concentrations showed that vitamin D supplementation did not increase T concentrations.[13] Because of the nature of cross-sectional studies and inconsistent results of small intervention studies, the causality between vitamin D and total T should be further confirmed.

An alternative to classical observational study is Mendelian randomization (MR) study. MR uses an instrumental variable built from genetic variants in nonexperimental data to make a causal inference about the effect of an exposure on an outcome.[14] In this study, if low 25(OH)D levels causally lead to lower total T levels, genetic variants associated with lower25(OH)D status should be associated with lower total T levels. Because allelic variants are inherited randomly from parent to offspring during meiosis and do not change over time, they are not prone to potential confounding factors.[15] Thus, analogous to natural randomized controlled trials (RCT), MR could avoid problems in traditional epidemiological studies such as inadequate control of confounders and reverse causation.[14]

On the basis of the large, community-based sample from the Survey on Prevalence in East China for Metabolic Diseases and Risk Factors (SPECT-China) study, we adopted an MR approach to test the hypotheses that low plasma 25(OH)D concentrations caused by genetic variants were associated with low androgen levels. A vitamin D genetic risk score (VD_GRS) was constructed to represent the genetic susceptibility.