Uncovering the complex relationship between balding, testosterone and skin cancers in men

Skin cancers comprise keratinocyte cancers (basal cell carcinoma, BCC; squamous cell carcinoma, SCC) and melanoma. Ultraviolet radiation (UVR) exposure is a key risk factor, and sex differences in incidence are observed in several Western countries with higher rates in men. Male-pattern baldness (MPB), associated with androgenic pathways (notably dihydrotestosterone/testosterone), has been linked to increased scalp melanoma risk. The research question is whether the observed associations between MPB and skin cancers are due to (a) increased UV exposure of the scalp from hair loss, (b) a direct causal role of androgens/testosterone, or (c) other androgen-regulated pathways including immune modulation. The study aims to disentangle these pathways using Mendelian randomization (MR) leveraging large-scale genetic datasets, including site-specific analyses, to clarify causal relationships and explain sex and body-site differences in skin cancer risk.

Observational studies have reported higher incidence of melanoma and KCs in men and site-specific differences with more head and neck cases in men. Prior work has linked MPB with increased risk of scalp melanoma and suggested an androgen-related mechanism. Some studies reported higher endogenous testosterone in MPB and a recent observational report suggested a positive association between free testosterone and melanoma risk. However, behavioral factors (sun exposure patterns, healthcare engagement) and biological differences (immune responses, pigmentation) also differ by sex and may confound associations. Large-scale causal evaluations, particularly for KCs and by primary tumor site, have been limited, motivating a genetics-based MR approach to distinguish UV exposure effects from androgenic or other biological pathways.

The study implemented a two-sample Mendelian randomization (MR) framework to evaluate causal relationships between MPB, endogenous sex hormones, and risk of skin cancers in men. Exposure instruments: 444 independent SNPs for MPB explaining ~13.95% of variance; 118 SNPs for total testosterone (totalT; ~8.13% variance), 83 for estimated free testosterone (freeT; ~4.94%), and 204 for sex-hormone-binding globulin (SHBG; ~14.5%). Instruments were primarily derived from UK Biobank (UKB) GWAS, with re-estimation in non-overlapping subsets to minimize sample overlap with outcome GWAS. SNPs were LD-clumped (10 Mb, r² ≤ 0.001) and harmonized with outcomes; palindromic SNPs with high MAF and variants with problematic standard errors were excluded. Outcomes: male-only GWAS/meta-analyses for melanoma (12,232 cases; 20,566 controls) and KCs (all KC, BCC, SCC) from UKB and QSkin; additional site-specific melanoma analyses (head/neck, scalp, other sites) using UKB and Melanoma Institute Australia (MIA) datasets. For KCs, sex-stratified GWAS in UKB were conducted using SAIGE; QSkin KC results were combined with UKB via fixed-effect inverse-variance meta-analysis. Primary MR estimator was inverse-variance weighted (IVW) to estimate OR per 1 SD increase in genetically predicted exposure. Sensitivity analyses included MR-PRESSO to detect/remove outlier instruments, MR-Egger, weighted median, and robust MR estimators to account for horizontal pleiotropy. To probe pleiotropy, PheWAS of MR-PRESSO-identified outlier SNPs was performed against pigmentation and related traits (e.g., skin color, tanning, hair color). Multivariable MR (MVMR) models were fitted conditioning MPB effects on candidate traits (including pigmentation and hormones) to estimate marginal effects and assess mediation/confounding. Instrument strength was quantified via conditional F-statistics; power was assessed from explained variance and sample sizes. Site-stratified MR assessed heterogeneity of MPB associations by primary tumor location (head/neck versus other sites; scalp-specific where possible) and by Breslow thickness in MIA.

- Power and instruments: MPB (444 SNPs; 13.95% variance), totalT (118; 8.13%), freeT (83; 4.94%), SHBG (204; 14.5%). Strong instrument strength and ≥90% power to detect modest effects. - Melanoma (overall): No evidence for association for MPB or for endogenous sex hormones; OR point estimates ~1.0 (e.g., IVW IRD 0.96–0.99). Genetic evidence did not support a causal role for totalT, freeT, or SHBG in melanoma susceptibility. - Keratinocyte cancers (KCs): Genetically predicted MPB was associated with higher risk of KCs. Per 1 SD increase in MPB genetic liability, ORs were: all KC 1.17 (95% CI 1.08–1.27), BCC 1.15 (95% CI 1.05–1.26), SCC 1.31 (95% CI 1.17–1.46). No associations detected for totalT, freeT, or SHBG with any KC outcome. - Pleiotropy and pigmentation: MR-PRESSO identified outlier MPB instruments (e.g., variants near KRT73, CUX1, IRF4). PheWAS showed these variants associated with pigmentation traits (skin color, tanning, hair color). Including pigmentation-related traits in MVMR attenuated the MPB–KC association; excluding IRF4 also reduced effects, indicating confounding via pigmentation pathways and pleiotropy. - Site-stratified melanoma: MPB genetic liability was associated with head/neck melanoma: pooled OR ~1.31 (95% CI 1.07–1.61) in main model; OR ~1.23 (95% CI 1.00–1.50) in outlier-robust model. Scalp-specific melanoma OR ~1.66 (95% CI 0.99–2.80) with attenuation in outlier-robust analyses. No association for melanoma at other body sites. No evidence that associations differed by Breslow thickness (thick OR 1.12 [0.89–1.39]; thin OR 1.04 [0.82–1.33]). - Site-stratified KCs: Limited evidence for associations of MPB with head/neck BCC (OR 1.08 [0.99–1.17]) or SCC (OR 0.92 [0.80–1.05]) in pleiotropy-robust models; scalp-site estimates were consistent with null. - Overall interpretation: Findings support scalp UV exposure due to hair loss as the main driver of the MPB–skin cancer relationship rather than androgen levels; genetic confounding via pigmentation loci contributes to observed associations, especially for KCs.

The study addressed whether MPB increases skin cancer risk through androgenic pathways or via increased UV exposure of the scalp. Comprehensive MR analyses showed no causal role for endogenous testosterone (total or free) or SHBG in melanoma or KCs, contradicting prior observational associations that may be confounded. MPB was positively associated with overall KC risk and with head/neck melanoma specifically, consistent with increased scalp UV exposure as a key mechanism. Pleiotropic effects from pigmentation-related loci (notably IRF4) partly explained the MPB–KC association, underscoring shared genetic architecture between balding, pigmentation, and skin cancer susceptibility. Site-specific analyses indicated that MPB-related increased risk was localized to head/neck (particularly scalp) for melanoma with no evidence of effect heterogeneity by tumor thickness. Together, results suggest that male-specific higher incidence of head/neck melanoma may be partly explained by hair loss exposing the scalp rather than androgen-driven carcinogenesis. Possible detection/ascertainment biases for scalp tumors in balding men are acknowledged, but the genetic causality framework and sensitivity analyses strengthen the inference that UV exposure and pigmentation pathways, not androgens, are the principal drivers.

This genetic study finds minimal support for a causal role of endogenous sex hormones in melanoma and KC susceptibility. MPB increases risk of KCs and is associated with head/neck melanoma, likely via increased UV exposure of the scalp, with pigmentation-related pleiotropy contributing to associations. These findings help explain higher male incidence of head/neck melanoma and emphasize UV protection of the scalp. Future research should: (1) replicate site-specific MR findings in additional populations with moderate ambient UV and in non-European ancestries; (2) refine instruments and MVMR models to better separate pigmentation, immune, and balding pathways (e.g., include immunological markers); (3) obtain richer phenotyping (objective MPB measures, detailed tumor characteristics) and larger scalp-specific case numbers; and (4) assess potential detection biases and behavioral mediators related to sun exposure and hair coverage.

- MPB measured via self-report in UKB (4-point scale), introducing potential misclassification that reduces instrument precision. - Some overlap between exposure and outcome samples (melanoma meta-analysis includes a UKB subset ~25% of male cases), though overlap <30% and re-estimations in non-overlapping subsets were performed. - MR assumes linear, no-threshold effects; true dose-response may be non-linear. - Limited availability of Breslow thickness and scalp-specific data in UKB; thickness-stratified analyses only in MIA. - Site-specific analyses had limited power, especially for scalp-specific outcomes; estimates less precise. - Pleiotropy from pigmentation and potentially immune-related loci complicates causal attribution; conditioning on multiple traits in MVMR constrained by instrument strength (conditional F-statistics). - Generalizability limited to predominantly European ancestry and two settings with very different ambient UV (UK, Australia); external validation in other ancestries and UV environments is needed.