Age-Dependent Directional Reversal of Total Testosterone Associations across the Phenome in NHANES 2011–2023

Background: Observational cohorts have persistently produced contradictory findings regarding disease associations of sex hormones in various diseases. This inconsistency is commonly attributed to cohort selection or methodological heterogeneity, but a structural explanation rooted in age-dependent biology has not been systematically tested. Methods: We analysed 21,847 adults from NHANES 2011-2023 with total testosterone (TT) measured by isotope-dilution liquid chromatography tandem mass spectrometry. Across 39 disease-sex pairs meeting a pre-specified case floor, we scanned age boundaries between 40 and 60 years for a change in the direction of the TT disease association and gated candidate reversals through a two-stage within-age-bin permutation null, an initial 500 draws promoted to 1,000 for pairs near the significance threshold. Each surviving reversal was characterised across five analytical layers, namely categorical onset, continuous laboratory severity markers, dose- response curvature, adiposity-adjusted robustness, and reproductive-stage stratification. Results: Eleven age-dependent directional reversals (ADRs) survived the permutation gate, spanning cardiometabolic (coronary heart disease, obesity, abdominal obesity, insulin resistance, low HDL cholesterol), musculoskeletal (fracture), sensory (hearing loss), neuropsychiatric (insomnia), and gastrointestinal (irritable bowel syndrome) systems. Six were sign flips where the slope changed direction across the boundary, five were magnitude-change reversals with same-signed but stratum-specific effects. The same age-cancellation appeared in continuous severity markers and in dose-response curvature. Adiposity adjustment mapped the ADRs onto four distinct patterns. Menarche stratification concentrated female signals by pubertal timing, and menopause stratification localised the protective testosterone-diabetes association to premenopausal women. Cardiovascular mortality in older men decreased with higher testosterone. All eleven ADRs held under exclusion of direct testosterone-modifier drug users. Conclusion: TT reverses direction across a definable midlife boundary, so a pooled null may be two opposing age-specific effects cancelling one another, making age-stratified reporting a necessary standard for future hormone-disease research.