Associations of Serum Testosterone and SHBG With Incident Fractures in Middle-aged to Older Men

Abstract

Context: As men age, circulating testosterone (T) decreases, circulating SHBG increases, and the risk of fracture increases. It is unclear if circulating T, independently of comorbidities, is associated with fracture risk in men.

Objectives: To determine associations for T and SHBG with incident fractures in men.

Methods: We utilized the large (n = 205 973 participants, 11 088 any fracture cases, 1680 hip fracture cases, 1366 forearm fracture cases) and well-characterized UK Biobank cohort. Associations were modeled using Cox regressions, adjusting for multiple comorbidities/covariates, imputing for missing information, and assessing nonlinearity using cubic splines.

Results: For T, not considering SHBG, there was a nonlinear association with hip but not forearm fractures, with the lowest risk in the second quintile. However, in models adjusted for SHBG or using calculated free T, lower T was associated with a higher risk for fractures at all evaluated bone sites. Lower SHBG was strongly associated with a lower risk of hip and forearm fractures (Q1 vs Q5, hip 0.55, 0.47-0.65; forearm 0.62, 0.52-0.74).

Conclusion: Low circulating SHBG is strongly associated with a low risk of fracture at all evaluated bone sites, while the associations of circulating T with fracture risk are of lesser magnitude, nonlinear, inconsistent among fracture site, and affected by adjustment for SHBG. These findings demonstrate that circulating SHBG, rather than T, is a major independent biomarker of fracture risk in men. Consequently, both total T and SHBG should be assessed when examining the relationship of endogenous T concentrations with fractures in middle-aged to older men.

Keywords: SHBG; incident fractures; men; testosterone.

Figure 1.

Estimated association of baseline serum testosterone concentration with risk of fracture in models with or without consideration of SHBG. (A–C) Estimates for testosterone models not considering SHBG, model 3, adjusted for time of blood sampling; geographic region; thyroid disease; renal impairment; ethnicity (White vs not White); participant age; other FRAX-related clinical risk factors: body mass index, fracture in past 5 years, smoking status, glucocorticoid use, secondary osteoporosis, and alcohol consumption; living with partner status; educational attainment; diet (red meat: high vs low vs none); physical activity; waist circumference; chronic obstructive pulmonary disease; serum vitamin D concentration; opioids; anticonvulsants; vitamin D supplementation; and total number of medications (proxy for overall comorbidity status). (D–I) Estimates for testosterone models considering SHBG either by (D–F) adjusting for SHBG and model 3 covariates (ie, model 4) or (G–I) using calculated free testosterone as exposure and model 3 covariates. Shaded areas are 95% confidence intervals and the locations of hazard ratios (medians of sample quintiles, as presented in Table 1 for testosterone and Table 2 for calculated free testosterone) are indicated. Horizontal axes are truncated to exclude values outside of boundary knots, where data are sparsely distributed and trends are constrained to linearity. Abbreviation: FRAX, Fracture Risk Assessment Tool.

Figure 2.

Estimated association of baseline serum SHBG concentration with risk of fracture. (A–C) Estimates for SHBG using model 3, adjusted for time of blood sampling; geographic region; thyroid disease; renal impairment; ethnicity (White vs not White); participant age; other FRAX-related clinical risk factors: body mass index, fracture in past five years, smoking status, glucocorticoid use, secondary osteoporosis, and alcohol consumption; living with partner status; educational attainment; diet (red meat: high vs low vs none); physical activity; waist circumference; chronic obstructive pulmonary disease; serum vitamin D concentration; opioids; anticonvulsants; vitamin D supplementation; and total number of medications (proxy for overall comorbidity status). (D–F) Estimates for SHBG models adjusting for testosterone and model 3 covariates (ie, model 4). Shaded areas are 95% confidence intervals and the locations of hazard ratios (medians of sample quintiles, as presented in Table 3) are indicated. Horizontal axes are truncated to exclude values outside of boundary knots, where data are sparsely distributed and trends are constrained to linearity. Abbreviation: FRAX, Fracture Risk Assessment Tool.