Association Between Testosterone Replacement Therapy and Prostatic Disorders in Elderly Hypogonadal Men

A Medicare study of 547K men found TRT linked to 16% lower prostate cancer hazard but 13% higher BPH risk

Journal: The Journal of Clinical Endocrinology & Metabolism | Published: 2026-02-20 | Type: Retrospective cohort study | PMID: 40922669 Authors: Hyon Baik Seo, Fitsum Baye, Kin Wah Fung, Haotian Xian, Clement J McDonald (Division of Intramural Research, National Library of Medicine, NIH) Funding/COI: NIH/NLM Division of Intramural Research; COI not listed

Summary

This retrospective cohort study used Medicare claims data from 2007–2020 to analyze prostate outcomes in over 546,000 hypogonadal men aged 65 and older. Using propensity score matching to balance baseline characteristics, TRT users showed a 16% lower hazard of prostate cancer compared to untreated hypogonadal men — a finding that cuts against decades of clinical anxiety about testosterone and prostate cancer. The trade-off: a 13% increased hazard of benign prostatic hyperplasia, with parenteral administration showing larger increases than topical.

Claims

• TRT was associated with a 16% reduction in prostate cancer hazard (HR = 0.84; 95% CI: 0.82–0.86) in 546,964 propensity-matched men
• Reductions were consistent across long-term use (15%), short-term use (16%), parenteral (12%), and topical (13%) administration
• A sensitivity analysis excluding men with elevated PSA or family history of PCa still showed an 18% reduction in PCa hazard among TRT users
• TRT was associated with a 13% increase in BPH hazard (HR = 1.13; 95% CI: 1.11–1.14) in 412,782 matched men
• Parenteral TRT showed a larger BPH hazard increase (19%) than topical TRT (12%)
• Long-term topical use showed the smallest BPH hazard increase at 9%
• Aromatase inhibitor co-administration altered the association in both outcomes

Study Quality

This is a large-scale retrospective cohort study — not a randomized trial. At nearly 547,000 men in the PCa arm, it is likely one of the largest analyses of TRT and prostate outcomes to date. The use of 1:1 propensity score matching stratified by age group substantially reduces, but does not eliminate, confounding. Medicare Parts A/B/C/D data capture a fairly comprehensive picture of diagnoses and prescriptions, and the 13-year follow-up window provides meaningful long-term exposure data.

The core limitation is residual confounding inherent to any claims-based observational design. Men selected for TRT differ systematically from those who are not — they tend to be healthier, more engaged with the healthcare system, and more frequently screened. This "healthy user bias" could artificially suppress observed PCa incidence in the TRT group, and the sensitivity analyses, while helpful, cannot fully correct for unmeasured confounders like PSA screening frequency, digital rectal exam rates, or lifestyle factors.

Red Flags

• Healthy user bias: Men who receive TRT tend to be healthier overall than untreated hypogonadal men — this alone could explain the lower observed PCa hazard, independent of any TRT effect
• Surveillance paradox: TRT users undergo more frequent PSA monitoring; more clinical contact increases the chance of BPH being coded, which could inflate the BPH hazard without reflecting a true biological increase
• Claims data limitations: ICD codes for hypogonadism and prostate disorders are subject to miscoding; actual serum testosterone levels are not captured in claims data
• Elderly cohort only: Men ≥65 have a different baseline prostate cancer risk profile; results may not generalize to younger hypogonadal men
• No causal mechanism: Hazard ratios from observational data cannot explain why the association exists — the paper does not resolve the underlying biology
• COI disclosure absent: NIH intramural funding is low-risk, but the absence of an explicit COI statement rather than a confirmed negative is a minor gap

Strengths

• Massive sample size (~547K for PCa, ~413K for BPH) yields narrow confidence intervals and ample statistical power
• 1:1 propensity score matching stratified by age group substantially controls for measured confounders at baseline
• Long observational window (2007–2020) captures long-term exposure effects
• Subgroup analyses by treatment duration, administration route, and AI co-administration add granularity beyond most prior studies
• Government (NIH intramural) funding with no apparent industry affiliation
• Sensitivity analyses excluding high-risk individuals (elevated PSA, family history) reinforce the primary PCa finding

Verdict

Methodologically solid observational work from a government-funded NIH team with no apparent industry conflict — the kind of large, carefully matched retrospective study that can meaningfully shift a long-standing clinical debate. The 16% PCa hazard reduction will generate headlines, and it deserves serious engagement rather than reflexive dismissal. But "associated with lower hazard" is not "protective": healthy user bias is a real and substantial problem in this population, and no retrospective design fully neutralizes it. The BPH finding — a consistent 13% hazard increase, larger with parenteral dosing — is arguably the more credible signal, being mechanistically coherent and less vulnerable to detection bias artifacts. Read the full paper; don't treat it as the last word.