Evaluating Calculated Free Testosterone as a Predictor of Morbidity and Mortality Independent of Testosterone for Cross-sectional and 5-Year Longitudinal Health Outcomes in Older Men: The Concord Health and Ageing in Men Project

Evaluating Calculated Free Testosterone as a Predictor of Morbidity and Mortality Independent of... Abstract To determine whether calculated free testosterone (cFT) provides prognostic information independent of serum T for predicting morbidity and mortality in older men in cross-sectional and 5-year longitudinal analyses. We studied men aged ≥70 years at baseline (n = 1,705), 2-year and 5-year measuring serum T (liquid chromatography-mass spectrometry), SHBG (immunoassay), cFT (an assumption-free empirical formula) together with 24 morbidity and 4 mortality outcomes. For cross-sectional and longitudinal analyses we employed a joint prediction model using generalized estimating equation models adjusted for age, smoking, comorbidities, and body mass index (BMI) with men having both normal T and normal cFT as referent group. Most morbidity and mortality outcomes were predicted by a combination of low T and cFT (LL). By contrast, only a single morbidity outcome in cross-sectional and none in longitudinal analysis was predicted by low T/normal cFT (LN) or normal T/low cFT (NL) without significant LL associations (isolated discordance). While for the few outcomes that predicted morbidity in men with discordances (LN or NL), these predictions only occurred when LL was also significant. Hence, for morbidity or mortality prediction in older men, discordance between cFT and T is unusual and isolated discordance is rare, so that cFT provides minimal independent prognostic information over serum T. Reproductive hormones, Androgen, Epidemiology, Health outcomes, Signs and symptoms In men, most circulating testosterone (T) is bound to SHBG with the remainder bound to albumin and other low-affinity binding proteins and only 1%–2% unbound to any circulating protein. The Free Hormone Hypothesis (FHH) postulates that this small unbound moiety is the most biologically active fraction of circulating serum T for its greater accessibility to tissues (1). Yet this theory cannot explain why unbound hormones are more rather than less biologically active as they are also more accessible to sites of degradation than bound hormones. Yet, despite its wide adoption, the FHH remains unproven and almost untested (2). FHH might have an empirical basis if FT provides additional independent biological or clinical information independent of serum T measurement for androgen-responsive health outcomes. There however has not been a systematic empirical evaluation of free testosterone (FT) measurement, for example, in predicting morbidity or mortality outcomes independent of accurate T measurement by mass spectrometry (MS)-based methods (3,4). As dialysis-based laboratory measurement of FT is a laborious and exacting manual method, it is rarely available so that various formulae are substituted to calculate FT (cFT) (5–7). However, comparative evaluations based on laboratory FT measurement as the gold standard show that the widely used model-based formulae (Sodergard, Vermeulen) are inaccurate due to their obligatory assumptions of plug-in estimates for the stoichiometry and affinity of testosterone binding to SHBG (5–10). We therefore developed and extensively validated an assumption-free, fully empirical formula for cFT (FTZ) that does not require plug-in estimates of binding stoichiometry and affinity of testosterone for SHBG (5–7). Therefore, the present study has primarily used this formula with comparison against a more widely used model-based (Vermeulen) formula (11). In multiple studies of older men in the Concord Health and Ageing in Men Project (CHAMP) (12–19) and Health in Men Study (HIMS) (20–25) cohorts, effect size and association of health outcomes based on accurate cFT estimates using the FTZ formula appeared not to diverge substantially from those based on serum testosterone measurement by LC-MS. Hence in this study, we aimed to determine formally whether accurately calculated FT provides additional prognostic information independent of serum T measured by LC-MS in predicting morbidity and mortality in older men in both cross-sectional and 5 year longitudinal analyses. As cFT is a deterministic function of T by its formula, we utilized a pattern of joint prediction to evaluate independent predictive contributions of health outcomes while avoiding collinearity. Methods Study Participants The CHAMP study is a longitudinal, population-based observational study of male ageing conducted among men living in the vicinity of Concord Hospital in Sydney, New South Wales, Australia as described in detail previously (26). Community dwelling men aged at least 70 years in 2005 were eligible with no other inclusion or exclusion criteria resulting in a final inception cohort of 1,705 participants. Baseline measurements were conducted between January 2005 and June 2007 using self-reported and interviewer-administered questionnaires and a wide range of clinical assessments. Follow-up assessments were conducted between January 2007 and October 2009 for 2-year follow-up, and August 2010 and July 2013 for the 5-year follow-up, with identical measurements as at baseline. All participants gave written informed consent. The study was approved by the Sydney South West Area Health Service Human Research Ethics Committee, Concord Repatriation General Hospital, Sydney, Australia. Reproductive Hormone Measurement Participants had an early morning fasting blood sample taken at baseline with serum stored at −80°C until assay. Measurement of serum T was by liquid chromatography-tandem mass spectrometry (LC-MS) as described (27) with modifications by introducing ultrapressure for high-pressure liquid chromatography with corresponding changes in extraction methodology validated according to FDA criteria (for details see Supplementary Methods in ref. (28)). The steroid measurements were calibrated against certified reference materials for T (National Measurement Institute, North Ryde, Australia). The assays had between-run coefficients of variation at three levels (low, medium, and high) of quality control (QC) specimens of 1.9%–4.5%, 3.8%–7.6%, 2.9%–13.6%, and 5.7%–8.7%, respectively, over 224 runs including all samples from this study. Overlapping QC samples were routinely run at the start, middle and end of every run with each new QC control run multiple times for calibration before use and there was no evidence of assay drift (13). Serum SHBG were measured by automated immunoassays (Roche Diagnostics Australia, Dee Why, Australia) subject to ongoing external QC program calibration with between-assay coefficients of variation for 2 levels of QC specimens in each run of 2.0%–2.8% for SHBG. The cFT levels in this study were computed using an assumption-free, empirical formula (FTZ) developed and validated against laboratory-based measurements of FT by dialysis methods which have displayed much closer conformance with laboratory-measured FT than model-based formulae (6,7). Morbidity and Mortality Outcomes Measurement Health-related quality of life and self-rated health were assessed using the 12-Item Short Form Health Survey (SF-12) (29). Functional disability was defined using the Katz activity of daily living questionnaire (30). Frailty was defined according to the criteria used in the Cardiovascular Health Study: weight loss/shrinking, weakness, exhaustion, slowness, and low activity (31). Falls were measured at the four month follow-up phone calls after their baseline assessments, participants were asked whether they had fallen in the preceding 4 months and, if so, how many times they had fallen. Participants were assessed for cognitive impairment at the clinic assessment visits using the Mini-Mental State Examination (32). Depressive symptoms were evaluated by the Geriatric Depression Scale, short form (33). The participants were asked about erectile dysfunction, sexual activity, sexual desire, and sexual satisfaction using standard, validated questionnaires (34). Metabolic syndrome was defined using the NCEP Adult Treatment Panel (ATP) III criteria (35). Physical activity was measured using the Physical Activity Scale for the Elderly (PASE) (36). Walking speed was measured at the participants’ usual pace (31). Trained staff used a stopwatch to record the time taken by the men to walk 6 m. The fastest time from two trials was used. Bone mineral density (BMD) at the total hip and femoral neck, lean mass and body fat was measured using dual x-ray absorptiometry (Discovery-W scanner Hologic, Bedford, MA). The appendicular lean mass (ALM) was calculated as the sum of lean mass of arms and legs (kg) (37). The ALM was standardized by body mass index (BMI) (ALMBMI) to take into account the body size of participants (38). Handgrip strength was measured with a Jamar dynamometer (Promedics, Blackburn, UK). Weight (by a regularly calibrated scale), height (using a Harpenden stadiometer), and waist circumference were measured by a trained professional at the clinic visit. Fasting blood samples were obtained at each visit for biochemistry tests including hemoglobin, glucose, and prostate specific antigen (PSA) performed at the accredited Clinical Pathology department of Concord Hospital. The New South Wales Registry of Births, Deaths, and Marriages were contacted to ascertain death status. The Registry also provided details recorded on the original death certificate for all participants. Based on the information provided from the death certificates, the general underlying cause of death (cancer, cardiovascular, or other) was identified independently by two medical practitioners (R.G.C., D.J.H.) (13). Potential Confounder Measurement Tobacco usage status (current, ex-, or never smoker) was by self-reported questionnaires. A comorbidity score was calculated as the sum of all conditions reported from the 19 disorders listed in the questionnaire. BMI was calculated from clinic measurements of height and weight. Statistical Analysis Descriptive characteristics of reproductive hormones at baseline and study health outcomes at baseline, 2-year and 5-year follow-up were generated for the analytic sample (Table 1). Participants were categorized into four mutually exclusive groups based on their baseline serum T and cFT defining “low” for these analyses by setting a threshold of the lowest quintile (20th centile) for serum T (10.2 nM) and cFT (156 pM). The referent groups for all cross-sectional and longitudinal analyses were men with both normal serum T and cFT (NN) with the other groups defined as men with the combinations of normal T/low cFT (NL), low T/normal cFT (LN) or low T/low cFT (LL). Of the 1,705 participants who completed the baseline assessments, a total of 1,651 were included for analyses in this article, after excluding men using androgen or antiandrogen treatments (n = 20) or with missing data on reproductive hormones (n = 34). Table 1. Characteristics of the Study Health Outcomes at Baseline, 2 Years, and 5 Years Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Note: ADL = Activities of daily living; BMD = Bone mineral density; BMI = Body mass index; MMSE = Mini-mental status examination; PASE = Physical activity scale for the elderly. †Higher values are better for MMSE (out of 30), PASE, SF-12 Physical and Mental (each out of 100). Lower values are better for comorbidity (out of 19). ‡The data for nonparticipation at 2 years and 5 years are their baseline descriptive characteristic. Death was the main reason for nonparticipation at 2 years (99 deaths) and at 5 years (382 deaths). View Large Table 1. Characteristics of the Study Health Outcomes at Baseline, 2 Years, and 5 Years Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Note: ADL = Activities of daily living; BMD = Bone mineral density; BMI = Body mass index; MMSE = Mini-mental status examination; PASE = Physical activity scale for the elderly. †Higher values are better for MMSE (out of 30), PASE, SF-12 Physical and Mental (each out of 100). Lower values are better for comorbidity (out of 19). ‡The data for nonparticipation at 2 years and 5 years are their baseline descriptive characteristic. Death was the main reason for nonparticipation at 2 years (99 deaths) and at 5 years (382 deaths). View Large Joint prediction in cross-sectional associations between the T/cFT status and health outcomes were assessed by logistic regression for categorical health outcomes, by multiple regression for continuous health outcomes and by Cox regression for mortality outcomes. Results were summarized into concordant findings (only statistically significant LL), discordant findings (either statistically significant LN and LL, or statistically significant NL and LL), and isolated discordance findings (statistically significant LN or NL without statistically significant LL). The detailed results for categorical variables are presented as odds ratios (95% confidence interval), for continuous variables are presented as β values (95% confidence interval) and for mortality are presented as hazard ratios (95% confidence interval). Similarly, longitudinal association between baseline T/cFT status and changes in health outcomes across baseline, 2 years and 5 years were assessed by generalized estimating equations (GEE) with exchangeable working correlation and robust variance estimator. The multinomial cumulative logit model was used for categorical morbidity outcomes, linear model for continuous morbidity outcomes and Poisson loglinear model for mortality outcomes. GEE method is robust and efficient when treating missing data in longitudinal studies (39). Model building for both cross-sectional and longitudinal analyses included adjustment for known major covariates, notably, age, BMI, smoking status, and comorbidity. BMI was not adjusted for in analyses for metabolic syndrome, body fat, weight, and waist circumference analysis. For post-hoc analyses, a Bonferroni adjustment to notional p values was performed to account for multiple comparisons involved in evaluating 28 outcome comparisons from a single set of data so that the conventional 0.05 level of significance was adjusted to a threshold of 0.002 (0.05/28). Models were fitted using SPSS software version 20 (IBM Corp., Armonk, NY) and SAS software 9.3 (SAS Institute Inc., Cary, NC). Results The demographic and anthropometric details of the CHAMP cohort are provided in Table 1 and the descriptive details of T, SHBG, and cFT are provided in Table 2. A total of 1,283 men (78%) were categorized into normal T/normal cFT (NN), 40 men (2%) into normal T/low cFT (NL), 38 men (2%) into low T/normal cFT (LN), and 290 men (18%) into low T/low cFT (LL). Table 2. Serum Testosterone (T), SHBG and Free Testosterone (cFT) Levels for the CHAMP Cohort at Baseline According to Different T/cFT Status Cutoff and Calculation N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 Note: cFT = calculated free testosterone; LL = Low T/Low cFT; LN = Low T/Normal cFT; NL = Normal T/Low cFT; NN = Normal T/Normal cFT. aT level below or above 10.2 nmol/L and cFT level below or above 156 pmol/L (lowest quintile). View Large Table 2. Serum Testosterone (T), SHBG and Free Testosterone (cFT) Levels for the CHAMP Cohort at Baseline According to Different T/cFT Status Cutoff and Calculation N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 Note: cFT = calculated free testosterone; LL = Low T/Low cFT; LN = Low T/Normal cFT; NL = Normal T/Low cFT; NN = Normal T/Normal cFT. aT level below or above 10.2 nmol/L and cFT level below or above 156 pmol/L (lowest quintile). View Large Cross-Sectional Morbidity Analysis The baseline cross-sectional associations between T/cFT status and morbidity outcomes are shown in detail in Supplementary Table 1 and summarized in Table 3 and Figure 1. After multivariable adjustment of the baseline cross-sectional data, low T/cFT (LL) was significantly associated with 15 of 24 outcomes (frailty, falls, sexual satisfaction, sexual desire, sexual activity, metabolic syndrome, physical activity, walking speed, physical quality of life, weight, hip BMD, body percent fat, waist circumference, glucose, hemoglobin, and PSA). Where LL was not a significantly associated with outcomes, there was only a single morbidity outcome that remained associated with either LN or NL (isolated discordance). A few morbidity outcomes displayed significant associations with discordant findings—four outcomes (metabolic syndrome, weight, fat mass and waist circumference) for LN and three outcomes (weight, fat mass, and waist circumference) for NL—but for each of these outcomes LL was also significant. After Bonferroni correction, LL remained significantly associated with 8 of 24 outcomes in cross-sectional analysis with additional discordant findings in 4 outcomes for LN and no outcomes for NL. There were no associations with isolated discordance. Table 3. Summary of the Concordant, Discordant, and Isolated Discordance for the Morbidity Outcomes Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 aPrimary; primary analysis using an empirical formula (FTZ) categorizing low T and low cFT based on lowest quintile. Sensitivity 1; sensitivity analysis using the same FTZ formula categorizing low T and low cFT based on lowest centile. Sensitivity 2; sensitivity analysis using the Vermeulen formula categorizing low T and low cFT based on lowest quintile. bLL is Low T/Low cFT; LN is Low T/Normal cFT; NL is Normal T/Low cFT; reference group is NN Normal T/Normal cFT. View Large Table 3. Summary of the Concordant, Discordant, and Isolated Discordance for the Morbidity Outcomes Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 aPrimary; primary analysis using an empirical formula (FTZ) categorizing low T and low cFT based on lowest quintile. Sensitivity 1; sensitivity analysis using the same FTZ formula categorizing low T and low cFT based on lowest centile. Sensitivity 2; sensitivity analysis using the Vermeulen formula categorizing low T and low cFT based on lowest quintile. bLL is Low T/Low cFT; LN is Low T/Normal cFT; NL is Normal T/Low cFT; reference group is NN Normal T/Normal cFT. View Large Figure 1. View largeDownload slide Summary of the cross-sectional and longitudinal findings for the morbidity outcomes. Figure 1. View largeDownload slide Summary of the cross-sectional and longitudinal findings for the morbidity outcomes. Longitudinal Morbidity Analysis The longitudinal associations over the 5-year follow-up between baseline T/cFT status and changes in morbidity outcomes are shown in detail in Supplementary Table 2 and summarized in Table 3 and Figure 1. Similar to the cross-sectional analysis, in multivariate adjusted models of the 5-year longitudinal analysis, low T/cFT (LL) was statistically significantly associated with 16 of 24 outcomes (poor self-rated health, activities of daily living disability, frailty, sexual satisfaction, sexual desire, sexual activity, metabolic syndrome, physical activity, walking speed, hip BMD, physical quality of life, weight, body percent fat, waist circumference, glucose, and hemoglobin). Significant discordant findings were present in six outcomes (metabolic syndrome, weight, fat mass, waist circumference, glucose, and hemoglobin) for LN and four (sexual activity, hip BMD, fat mass, and hemoglobin) for NL where LL was also significant. There were no significant isolated discordant findings (significant LN or NL without significant LL). After Bonferroni correction, LL remained significantly associated with 9 of 24 morbidity outcomes with additional discordant findings in 4 outcomes for LN and no outcomes for NL, all in conjunction with significant LL. Sensitivity Analysis One sensitivity analysis performed used the same empirical formula (FTZ) but lowering the threshold to define “low” from lowest quintiles (lowest 20%) to lowest centiles (lowest 10%) for the morbidity analysis produced essentially the same results. In the multivariable adjusted model, LL was significantly associated with 12 of 24 morbidity outcomes in cross-sectional and 15 of 24 outcomes in longitudinal analysis (Table 3). Additional significant discordant findings (LN or NL) in conjunction with significant LL were present for 6 outcomes in cross-sectional and 15 outcomes in longitudinal analysis. When LL was not significant (isolated discordance), only 1 outcome in cross-sectional and 1 in longitudinal analysis were significantly associated with an adverse morbidity outcome. Another sensitivity analysis was conducted using the Vermeulen cFT (lowest quintile) for the same morbidity analysis and the findings were similar to our original analysis using the more accurate FTZ formula. In multivariable model, LL was significantly associated with 12 of 24 morbidity outcomes in cross-sectional and 15 of 24 outcomes in longitudinal analysis. Where LL was not a significant predictor, isolated discordant findings (either LN or NL significant) were associated with only 1 outcome in cross-sectional and 1 in longitudinal analyses. Among men with discordant findings (significant LN or NL, with significant LL), morbidity prediction was present for 6 outcomes in cross-sectional and 15 outcomes in longitudinal analysis. Cross-Sectional and Longitudinal Mortality Analysis With multivariable adjustment, both the baseline cross-sectional as well as the longitudinal mortality analyses (Supplementary Table 3) showed significant effects for LL in all-cause, cardiovascular and other but not for cancer mortality. For no mortality outcome was there significant prediction when LL was not significant and where mortality outcomes were predicted by discordant (NL or LN), LL was also significant. Discussion Many studies have reported associations between health outcomes and low T or low cFT, considered as separate parameters, among older men (40). However, we observed consistently in a series of studies from the CHAMP (12–19) and HIMS (20–25) cohorts that cFT as a predictor rarely, if ever, provided any significantly different information on health outcomes from serum T measured by LC-MS. Furthermore, as the FHH remains largely untested, there remains minimal critical evidence to what extent, if any, FT data provides additional biological or clinical insight independent of accurate serum T measurements by LC-MS in men (41). The present findings investigating a wide range of morbidity and mortality outcomes in older men suggest that cFT rarely adds independent prognostic information to serum T measured by LC-MS in either cross-sectional or longitudinal analyses. An important caveat is that the utility of health outcome predictions by T and/or FT, depends on the accuracy of the T and FT estimates employed. Until recently, most studies relied upon T immunoassays which suffer from method-specific and other technical limitations notably if applied to reduced serum testosterone such as in older men (3,4). This became a greater problem over the decades after the 1980’s when direct, nonextraction immunoassay became almost universal in clinical practice and research. Over the last decade, more accurate measurement of serum T has become more widely feasible using modern, bench-top LC-MS to supplant direct (unextracted) T immunoassays. Currently, especially for large-scale epidemiological studies, FT is rarely measured directly by dialysis-based laboratory reference methods. These methods are laborious, exacting and require manual laboratory skills which have been largely eliminated by the deskilling automation of chemical pathology laboratories. Furthermore FT measurements lack quality control programs or validated reference ranges. Instead, FT is usually calculated by a variety of formulae which fall into two classes, model-based equilibrium binding and fully empirical equations. These differ in their assumptions and in conformance in accuracy to dialysis-based laboratory gold standard reference methods. The accuracy of cFT is crucial because any formula produces a deterministic (inverse) function of age as it compounds two age-dependent variables—testosterone and SHBG. Unless the formula accurately represents the authentic laboratory-based FT measurement it intends to represent, it will display a spurious correlation with any age-dependent variable regardless of whether that variable has any genuine biological relationship to testosterone. The FTZ equation was originally developed from a large data set of 3975 serum samples by identifying the optimal regression formula of laboratory dialysis-based reference FT measurements on serum testosterone and SHBG measured in the same samples. This formula was cross-validated against a separate set of 124 serum samples (6) and then subsequently confirmed as highly accurate when tested in a different large data set of 2,159 samples from another laboratory using different methods to measure FT, testosterone and SHBG (7). A key finding from the extensive validation involving over 6,000 serum samples was that the widely used, model-based equilibrium binding equation-based formulae by Vermeulen (11) and Sodergard (42) display marked bias deviating from the laboratory-measured FT. These deviations were due to both wrong stoichiometry as well as arbitrary plug-in binding affinity coefficients for T binding to SHBG (6,7), the latter varying fivefold among the various implementations of model-based equilibrium binding formulae (43). The present study uses this FTZ formula to evaluate the impact of accurately estimated cFT, corresponding most closely to laboratory-based FT measurements, on morbidity and mortality outcome predictions. The novelty of the current longitudinal study is that it investigates both serum T and cFT levels concurrently as joint predictors of a wide range of health outcomes over time. Our analysis revealed that both cross-sectionally and longitudinally over 5 years, men with concordant low serum T and cFT as well as those with concordant normal T and cFT were more likely to die or experience adverse health outcomes. On the contrary, only a minority of men had variables displaying discrepancies between T and cFT values and where there was an isolated discordance—that is discordance between T and cFT but not accompanied significant association or prediction by LL—was rare. Hence, not only are discrepancies unusual but cFT alone predicts almost no health outcomes among older men independent of an accurately measured serum T. Altogether, the present analysis provides a comprehensive analysis of a wide range of health outcomes including nonspecific symptoms resembling those of androgen deficiency or many other chronic diseases. A recent study from the EMAS cohort evaluated the FHH among older men by analyzing cross-sectionally the joint association of cFT and T with health outcomes. They reported that low Vermeulen cFT, even in the presence of normal T, but not the combination of normal cFT and low T, was associated with a range of nonspecific symptoms including sexual and physical symptoms (44). In a previous study, they reported that low T and cFT were associated with sexual but not physical or psychological symptoms (45) although the sexual symptoms had high rates of false positive and false negatives reflecting their nonspecificity and the direction of causality could not be determined. This reflects the fact that genuine androgen deficiency symptoms are, for any individual, are highly reproducible at consistent blood testosterone concentrations (46); however, as the actual symptoms differ widely between individuals, grouping individuals according to symptoms erodes the relationship of symptoms to blood testosterone concentrations (47). Furthermore, as the Vermeulen model-based cFT formula systematically deviates from laboratory measured FT values as reported by several independent groups (5–10). Yet, as any cFT remains a deterministic (inverse) function of age, failure to correspond accurately to laboratory-measured FT makes it likely that any relationship to age-related symptoms may reflect residual confounding due to the age-mismatch of the subgroups (persisting after linear age adjustment) rather than any authentic relationship with serum T. The present analysis, using a more accurate and extensively validated cFT formula so that it corresponds more closely to laboratory-measured FT, showed that cFT and T were usually concordant and, in the unusual instances where there was a discordance, that almost always occurred only when the concordant low T/low cFT was also significant. In our analysis as well as that using the Vermeulen formula, significant isolated discordant association or prediction by a low cFT was rare and had little impact on prediction of mortality or morbidity over the next 5 years. Instead it was the combination of both a low T and low cFT that was significantly associated with most outcomes although the direction of causality remains undetermined. Our study shows that men with low T were most likely to have low cFT while men with normal T were most likely to have normal cFT. Only a very small proportion had discordance with either normal T and low cFT, or low T and normal cFT and when this occurred it was almost invariably in the setting where the concordant combination of low T/low CFT was also significant. The major finding in this study is consistent with previous studies showing low T and low cFT, as separate parameters, are associated with these many health outcomes such as general health status, functional ability, metabolic syndrome, bone health, cognition, sexual function, etc. These findings confirm our impression from previous studies in the CHAMP (12–19) and HIMS (20–25) cohorts that show very similar effect size and associations in either low T or low cFT with a wide variety of health outcomes (12–25). This suggest that cFT provides minimal independent predictive information for health outcomes independent of accurately measured serum T and questions whether cFT estimates, even when accurately calculated, provide any useful information for clinical practice. The strengths of this study include the use of longitudinal data to investigate a comprehensive profile of T/cFT status in conjunction with a wide array of key morbidity and mortality outcomes over three follow-up time-points spanning 5 years. Another is the use of the LC-MS, the current gold standard for steroid assays, providing multianalyte steroid profiling. This improves upon direct immunoassay methods which, lacking extraction and chromatography, feature poor accuracy at low levels of circulating sex steroids, which is particularly problematic for measuring circulating T in older men (3,4). Furthermore, we used an extensively validated, assumption-free formula for cFT which corresponds more accurately to laboratory-measured FT than previous model-based equilibrium binding formula that rely on arbitrary plug-in coefficients. A further strength of CHAMP is that it includes a large and representative group of older Australian men, as demonstrated by similar sociodemographic and health characteristics compared to older men in the nationally representative MATeS study (48). A significant limitation of our study is the impact of survivor bias. This applies to the survivorship in the cohort with most losses due to mortality which accounted for nearly 35% of loss to follow-up in our cohort. On the other hand, mortality was evaluated as an outcome so that this cohort provides a more complete view of the causes and determinant of mortality among living older men. To avoid the impact of potential diurnal variation in hormone concentrations, a rhythm that is mostly lost in ageing men (49), we obtained fasting morning blood samples in this study and evaluated joint prediction to avoid collinearity between cFT and T. In conclusion, concordant low serum T and cFT levels were strongly associated with many health outcomes in older men whereas among the minority of men with discrepancies between T and cFT, such discordance was associated with or predicted few health outcomes and only then when for the same outcome, there was also a significant association or prediction by the combination of both low T and cFT. Hence, in addition to the ambiguous theoretical basis of the FHH, the present findings suggest that even accurately cFT estimates provide minimal additional clinical or biological information independent of accurate measurement of serum T concentrations for mortality or morbidity outcomes in older men. These findings provide little support for the application of the FHH to studies of testosterone and clinical outcomes in older men. Supplementary Material Supplementary data is available at The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences online. Funding The CHAMP study is funded by the NHMRC Project Grant (No. 301916), Sydney Medical School Foundation, and Ageing and Alzheimer’s Institute. Conflict of Interest D.G.L.C. is a co-deputy editor for the Journal of Gerontology: Biological Sciences. All the other authors have nothing to declare. Acknowledgments R.G.C., D.J.H., M.J.S., L.M.W., V.N., D.G.L.C., and F.M.B. contributed to the formulation of the study concept, design, methods, subject recruitment, and data collection. B.H. wrote the manuscript and performed the analyses. D.J.H. wrote portions of the manuscript. R.G.C., F.M.B., V.N., D.G.L.C., M.J.S., and L.M.W. reviewed the manuscript and contributed to discussion. References 1. Goldman AL , Bhasin S , Wu FCW , Krishna M , Matsumoto AM , Jasuja R . A reappraisal of testosterone’s binding in circulation: physiological and clinical implications . Endocr Rev . 2017 ;doi: 10.1210/er.2017-00025 2. Handelsman D . Free testosterone: pumping up the tires or ending the free ride ? Endocr Rev . 2017 ; 38 : 297 – 301 .doi: 10.1210/er.2017-00171 Google Scholar CrossRef Search ADS PubMed 3. Rosner W , Auchus RJ , Azziz R , Sluss PM , Raff H . Position statement: utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement . J Clin Endocrinol Metab . 2007 ; 92 : 405 – 413 . doi: 10.1210/jc.2006-1864 Google Scholar CrossRef Search ADS PubMed 4. Handelsman DJ , Wartofsky L . Requirement for mass spectrometry sex steroid assays in the Journal of Clinical Endocrinology and Metabolism . J Clin Endocrinol Metab . 2013 ; 98 : 3971 – 3973 . doi: 10.1210/jc.2013-3375 Google Scholar CrossRef Search ADS PubMed 5. Ly LP , Handelsman DJ . Empirical estimation of free testosterone from testosterone and sex hormone-binding globulin immunoassays . Eur J Endocrinol . 2005 ; 152 : 471 – 478 . doi: 10.1530/eje.1.01844 Google Scholar CrossRef Search ADS PubMed 6. Sartorius G , Ly LP , Sikaris K , McLachlan R , Handelsman DJ . Predictive accuracy and sources of variability in calculated free testosterone estimates . Ann Clin Biochem . 2009 ; 46 ( Pt 2 ): 137 – 143 . doi: 10.1258/acb.2008.008171 Google Scholar CrossRef Search ADS PubMed 7. Ly LP , Sartorius G , Hull L et al. Accuracy of calculated free testosterone formulae in men . Clin Endocrinol (Oxf) . 2010 ; 73 : 382 – 388 . doi: 10.1111/j.1365-2265.2010.03804.x Google Scholar CrossRef Search ADS PubMed 8. Hackbarth JS , Hoyne JB , Grebe SK , Singh RJ . Accuracy of calculated free testosterone differs between equations and depends on gender and SHBG concentration . Steroids . 2011 ; 76 : 48 – 55 . doi: 10.1016/j.steroids.2010.08.008 Google Scholar CrossRef Search ADS PubMed 9. Salameh WA , Redor-Goldman MM , Clarke NJ , Reitz RE , Caulfield MP . Validation of a total testosterone assay using high-turbulence liquid chromatography tandem mass spectrometry: total and free testosterone reference ranges . Steroids . 2010 ; 75 : 169 – 175 . doi: 10.1016/j.steroids.2009.11.004 Google Scholar CrossRef Search ADS PubMed 10. Zakharov MN , Bhasin S , Travison TG et al. A multi-step, dynamic allosteric model of testosterone’s binding to sex hormone binding globulin . Mol Cell Endocrinol . 2015 ; 399 : 190 – 200 . doi: 10.1016/j.mce.2014.09.001 Google Scholar CrossRef Search ADS PubMed 11. Vermeulen A , Verdonck L , Kaufman JM . A critical evaluation of simple methods for the estimation of free testosterone in serum . J Clin Endocrinol Metab . 1999 ; 84 : 3666 – 3672 . doi: 10.1210/jcem.84.10.6079 Google Scholar CrossRef Search ADS PubMed 12. Hsu B , Hirani V , Naganathan V et al. Sexual function and mortality in older men: the Concord Health and Ageing in Men Project . J Gerontol A Biol Sci Med Sci . 2017 ; 72 : 520 – 527 . doi: 10.1093/gerona/glw101 Google Scholar PubMed 13. Hsu B , Cumming RG , Naganathan V et al. Temporal changes in androgens and estrogens are associated with all-cause and cause-specific mortality in older men . J Clin Endocrinol Metab . 2016 ; 101 : 2201 – 2210 . doi: 10.1210/jc.2016-1025 . Google Scholar CrossRef Search ADS PubMed 14. Hsu B , Cumming RG , Blyth FM , Naganathan V , Le Couteur DG , Seibel MJ et al. Longitudinal and cross-sectional relationships of circulating reproductive hormone levels to self-rated health and health-related quality of life in community-dwelling older men . J Clin Endocrinol Metab . 2014 ; 99 : 1638 – 1647 . doi: 10.1210/jc.2013–3984 Google Scholar CrossRef Search ADS PubMed 15. Hsu B , Cumming RG , Naganathan V et al. Associations between circulating reproductive hormones and SHBG and prevalent and incident metabolic syndrome in community-dwelling older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2014 ; 99 : E2686 – E2691 . doi: 10.1210/jc.2014-2464 Google Scholar CrossRef Search ADS PubMed 16. Hsu B , Cumming RG , Naganathan V et al. Associations between circulating reproductive hormones and SHBG and prevalent and incident metabolic syndrome in community-dwelling older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2014 ; 99 : E2686 – E2691 . doi: 10.1210/jc.2014-2464 Google Scholar CrossRef Search ADS PubMed 17. Hsu B , Cumming RG , Waite LM et al. Longitudinal relationships between reproductive hormones and cognitive decline in older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2015 ; 100 : 2223 – 2230 . doi: 10.1210/jc.2015-1016 Google Scholar CrossRef Search ADS PubMed 18. Hsu B , Cumming RG , Seibel MJ et al. Reproductive hormones and longitudinal change in bone mineral density and incident fracture risk in older men: the Concord Health and Aging in Men Project . J Bone Miner Res . 2015 ; 30 : 1701 – 1708 . doi: 10.1002/jbmr.2493 Google Scholar CrossRef Search ADS PubMed 19. Hsu B , Cumming RG , Waite LM et al. Longitudinal relationships between reproductive hormones and cognitive decline in older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2015 ; 100 : 2223 – 2230 . doi: 10.1210/jc.2015-1016 Google Scholar CrossRef Search ADS PubMed 20. Yeap BB , Alfonso H , Chubb SA et al. In older men, higher plasma testosterone or dihydrotestosterone is an independent predictor for reduced incidence of stroke but not myocardial infarction . J Clin Endocrinol Metab . 2014 ; 99 : 4565 – 4573 . doi: 10.1210/jc.2014-2664 Google Scholar CrossRef Search ADS PubMed 21. Yeap BB , Alfonso H , Chubb SA et al. Higher serum undercarboxylated osteocalcin and other bone turnover markers are associated with reduced diabetes risk and lower estradiol concentrations in older men . J Clin Endocrinol Metab . 2015 ; 100 : 63 – 71 .doi: 10.1210/jc.2014–3019 Google Scholar CrossRef Search ADS PubMed 22. Yeap BB , Knuiman MW , Divitini ML et al. Epidemiological and Mendelian randomization studies of dihydrotestosterone and estradiol and leukocyte telomere length in men . J Clin Endocrinol Metab . 2016 ; 101 : 1299 – 1306 . doi: 10.1210/jc.2015-4139 Google Scholar CrossRef Search ADS PubMed 23. Yeap BB , Alfonso H , Chubb SA et al. In older men, higher plasma testosterone or dihydrotestosterone is an independent predictor for reduced incidence of stroke but not myocardial infarction . J Clin Endocrinol Metab . 2014 ; 99 : 4565 – 4573 . doi: 10.1210/jc.2014-2664 Google Scholar CrossRef Search ADS PubMed 24. Yeap BB , Alfonso H , Paul Chubb SA et al. In older men, higher plasma testosterone or dihydrotestosterone are independent predictors for reduced incidence of stroke but not myocardial infarction . J Clin Endocrinol Metab . 2014 Dec;99:4565–73 . doi: 10.1210/jc.2014–2664 25. Yeap BB , Alfonso H , Chubb SA et al. Reference ranges and determinants of testosterone, dihydrotestosterone, and estradiol levels measured using liquid chromatography-tandem mass spectrometry in a population-based cohort of older men . J Clin Endocrinol Metab . 2012 ; 97 : 4030 – 4039 . doi: 10.1210/jc.2012-2265 Google Scholar CrossRef Search ADS PubMed 26. Cumming RG , Handelsman D , Seibel MJ et al. Cohort Profile: the Concord Health and Ageing in Men Project (CHAMP) . Int J Epidemiol . 2009 ; 38 : 374 – 378 . doi: 10.1093/ije/dyn071 Google Scholar CrossRef Search ADS PubMed 27. Harwood DT , Handelsman DJ . Development and validation of a sensitive liquid chromatography-tandem mass spectrometry assay to simultaneously measure androgens and estrogens in serum without derivatization . Clin Chim Acta . 2009 ; 409 : 78 – 84 . doi: 10.1016/j.cca.2009.09.003 Google Scholar CrossRef Search ADS PubMed 28. Keski-Rahkonen P , Desai R , Jimenez M , Harwood DT , Handelsman DJ . Measurement of estradiol in human serum by LC-MS/MS using a novel estrogen-specific derivatization reagent . Anal Chem . 2015 ; 87 : 7180 – 7186 . doi: 10.1021/acs.analchem.5b01042 Google Scholar CrossRef Search ADS PubMed 29. Ware J Jr , Kosinski M , Keller SD . A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity . Med Care . 1996 ; 34 : 220 – 233 . Google Scholar CrossRef Search ADS 30. Smith LA , Branch LG , Scherr PA et al. Short-term variability of measures of physical function in older people . J Am Geriatr Soc . 1990 ; 38 : 993 – 998 . Google Scholar CrossRef Search ADS PubMed 31. Fried LP , Tangen CM , Walston J et al. ; Cardiovascular Health Study Collaborative Research Group . Frailty in older adults: evidence for a phenotype . J Gerontol A Biol Sci Med Sci . 2001 ; 56 : M146 – M156 . Google Scholar CrossRef Search ADS PubMed 32. Folstein MF , Folstein SE , McHugh PR . “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res . 1975 ; 12 : 189 – 198 . Google Scholar CrossRef Search ADS PubMed 33. Shiekh J , Yesavage JA . Geriatric depression scale: recent findings and developments of a short version . QualityMetric Inc ; 2009 . 34. O’Donnell AB , Araujo AB , Goldstein I , McKinlay JB . The validity of a single-question self-report of erectile dysfunction. Results from the Massachusetts Male Aging Study . J Gen Intern Med . 2005 ; 20 : 515 – 519 . doi: 10.1111/j.1525-1497.2005.0076.x Google Scholar CrossRef Search ADS PubMed 35. Grundy SM , Brewer HB , Jr. , Cleeman JI , Smith SC , Jr. , Lenfant C . Definition of metabolic syndrome: report of the National Heart, Lung, and Blood Institute/American Heart Association conference on scientific issues related to definition . Circulation . 2004 ; 109 : 433 – 438 . doi: 10.1161/01.cir.0000111245.75752.c6 Google Scholar CrossRef Search ADS PubMed 36. Washburn RA , McAuley E , Katula J , Mihalko SL , Boileau RA . The physical activity scale for the elderly (PASE): evidence for validity . J Clin Epidemiol . 1999 ; 52 : 643 – 651 . Google Scholar CrossRef Search ADS PubMed 37. Heymsfield SB , Smith R , Aulet M et al. Appendicular skeletal muscle mass: measurement by dual-photon absorptiometry . Am J Clin Nutr . 1990 ; 52 : 214 – 218 . Google Scholar CrossRef Search ADS PubMed 38. Studenski SA , Peters KW , Alley DE , Cawthon PM , McLean RR , Harris TB et al. The FNIH sarcopenia project: rationale, study description, conference recommendations, and final estimates . J Gerontol A Biol Sci Med Sci . 2014 ; 69 : 547 – 558 . doi: 10.1093/gerona/glu010 Google Scholar CrossRef Search ADS PubMed 39. Zeger SL , Liang KY . Longitudinal data analysis for discrete and continuous outcomes . Biometrics . 1986 ; 42 : 121 – 130 . Google Scholar CrossRef Search ADS PubMed 40. Kaufman JM , Vermeulen A . The decline of androgen levels in elderly men and its clinical and therapeutic implications . Endocr Rev . 2005 ; 26 : 833 – 876 . doi: 10.1210/er.2004-0013 Google Scholar CrossRef Search ADS PubMed 41. Bhasin S , Cunningham GR , Hayes FJ et al. ; Task Force, Endocrine Society . Testosterone therapy in men with androgen deficiency syndromes: an Endocrine Society clinical practice guideline . J Clin Endocrinol Metab . 2010 ; 95 : 2536 – 2559 . doi: 10.1210/jc.2009-2354 Google Scholar CrossRef Search ADS PubMed 42. Södergård R , Bäckström T , Shanbhag V , Carstensen H . Calculation of free and bound fractions of testosterone and estradiol-17 beta to human plasma proteins at body temperature . J Steroid Biochem . 1982 ; 16 : 801 – 810 . Google Scholar CrossRef Search ADS PubMed 43. Mazer NA . A novel spreadsheet method for calculating the free serum concentrations of testosterone, dihydrotestosterone, estradiol, estrone and cortisol: with illustrative examples from male and female populations . Steroids . 2009 ; 74 : 512 – 519 . doi: 10.1016/j.steroids.2009.01.008 Google Scholar CrossRef Search ADS PubMed 44. Antonio L , Wu FC , O’Neill TW et al. ; European Male Ageing Study Study Group . Low free testosterone is associated with hypogonadal signs and symptoms in men with normal total testosterone . J Clin Endocrinol Metab . 2016 ; 101 : 2647 – 2657 . doi: 10.1210/jc.2015-4106 Google Scholar CrossRef Search ADS PubMed 45. Wu FC , Tajar A , Beynon JM et al. ; EMAS Group . Identification of late-onset hypogonadism in middle-aged and elderly men . N Engl J Med . 2010 ; 363 : 123 – 135 . doi: 10.1056/NEJMoa0911101 Google Scholar CrossRef Search ADS PubMed 46. Kelleher S , Conway AJ , Handelsman DJ . Blood testosterone threshold for androgen deficiency symptoms . J Clin Endocrinol Metab . 2004 ; 89 : 3813 – 3817 . doi: 10.1210/jc.2004-0143 Google Scholar CrossRef Search ADS PubMed 47. Zitzmann M , Faber S , Nieschlag E . Association of specific symptoms and metabolic risks with serum testosterone in older men . J Clin Endocrinol Metab . 2006 ; 91 : 4335 – 4343 . doi: 10.1210/jc.2006-0401 Google Scholar CrossRef Search ADS PubMed 48. Holden CA , McLachlan RI , Cumming R et al. Sexual activity, fertility and contraceptive use in middle-aged and older men: men in Australia, Telephone Survey (MATeS) . Hum Reprod . 2005 ; 20 : 3429 – 3434 . doi: 10.1093/humrep/dei307 Google Scholar CrossRef Search ADS PubMed 49. Bremner WJ , Vitiello MV , Prinz PN . Loss of circadian rhythmicity in blood testosterone levels with aging in normal men . J Clin Endocrinol Metab . 1983 ; 56 : 1278 – 1281 . doi: 10.1210/jcem-56-6-1278 Google Scholar CrossRef Search ADS PubMed © The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences Oxford University Press

Evaluating Calculated Free Testosterone as a Predictor of Morbidity and Mortality Independent of Testosterone for Cross-sectional and 5-Year Longitudinal Health Outcomes in Older Men: The Concord Health and Ageing in Men Project

Loading next page...
 
/lp/ou_press/evaluating-calculated-free-testosterone-as-a-predictor-of-morbidity-tXX93hx8aH
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1079-5006
eISSN
1758-535X
D.O.I.
10.1093/gerona/glx170
Publisher site
See Article on Publisher Site

Abstract

Abstract To determine whether calculated free testosterone (cFT) provides prognostic information independent of serum T for predicting morbidity and mortality in older men in cross-sectional and 5-year longitudinal analyses. We studied men aged ≥70 years at baseline (n = 1,705), 2-year and 5-year measuring serum T (liquid chromatography-mass spectrometry), SHBG (immunoassay), cFT (an assumption-free empirical formula) together with 24 morbidity and 4 mortality outcomes. For cross-sectional and longitudinal analyses we employed a joint prediction model using generalized estimating equation models adjusted for age, smoking, comorbidities, and body mass index (BMI) with men having both normal T and normal cFT as referent group. Most morbidity and mortality outcomes were predicted by a combination of low T and cFT (LL). By contrast, only a single morbidity outcome in cross-sectional and none in longitudinal analysis was predicted by low T/normal cFT (LN) or normal T/low cFT (NL) without significant LL associations (isolated discordance). While for the few outcomes that predicted morbidity in men with discordances (LN or NL), these predictions only occurred when LL was also significant. Hence, for morbidity or mortality prediction in older men, discordance between cFT and T is unusual and isolated discordance is rare, so that cFT provides minimal independent prognostic information over serum T. Reproductive hormones, Androgen, Epidemiology, Health outcomes, Signs and symptoms In men, most circulating testosterone (T) is bound to SHBG with the remainder bound to albumin and other low-affinity binding proteins and only 1%–2% unbound to any circulating protein. The Free Hormone Hypothesis (FHH) postulates that this small unbound moiety is the most biologically active fraction of circulating serum T for its greater accessibility to tissues (1). Yet this theory cannot explain why unbound hormones are more rather than less biologically active as they are also more accessible to sites of degradation than bound hormones. Yet, despite its wide adoption, the FHH remains unproven and almost untested (2). FHH might have an empirical basis if FT provides additional independent biological or clinical information independent of serum T measurement for androgen-responsive health outcomes. There however has not been a systematic empirical evaluation of free testosterone (FT) measurement, for example, in predicting morbidity or mortality outcomes independent of accurate T measurement by mass spectrometry (MS)-based methods (3,4). As dialysis-based laboratory measurement of FT is a laborious and exacting manual method, it is rarely available so that various formulae are substituted to calculate FT (cFT) (5–7). However, comparative evaluations based on laboratory FT measurement as the gold standard show that the widely used model-based formulae (Sodergard, Vermeulen) are inaccurate due to their obligatory assumptions of plug-in estimates for the stoichiometry and affinity of testosterone binding to SHBG (5–10). We therefore developed and extensively validated an assumption-free, fully empirical formula for cFT (FTZ) that does not require plug-in estimates of binding stoichiometry and affinity of testosterone for SHBG (5–7). Therefore, the present study has primarily used this formula with comparison against a more widely used model-based (Vermeulen) formula (11). In multiple studies of older men in the Concord Health and Ageing in Men Project (CHAMP) (12–19) and Health in Men Study (HIMS) (20–25) cohorts, effect size and association of health outcomes based on accurate cFT estimates using the FTZ formula appeared not to diverge substantially from those based on serum testosterone measurement by LC-MS. Hence in this study, we aimed to determine formally whether accurately calculated FT provides additional prognostic information independent of serum T measured by LC-MS in predicting morbidity and mortality in older men in both cross-sectional and 5 year longitudinal analyses. As cFT is a deterministic function of T by its formula, we utilized a pattern of joint prediction to evaluate independent predictive contributions of health outcomes while avoiding collinearity. Methods Study Participants The CHAMP study is a longitudinal, population-based observational study of male ageing conducted among men living in the vicinity of Concord Hospital in Sydney, New South Wales, Australia as described in detail previously (26). Community dwelling men aged at least 70 years in 2005 were eligible with no other inclusion or exclusion criteria resulting in a final inception cohort of 1,705 participants. Baseline measurements were conducted between January 2005 and June 2007 using self-reported and interviewer-administered questionnaires and a wide range of clinical assessments. Follow-up assessments were conducted between January 2007 and October 2009 for 2-year follow-up, and August 2010 and July 2013 for the 5-year follow-up, with identical measurements as at baseline. All participants gave written informed consent. The study was approved by the Sydney South West Area Health Service Human Research Ethics Committee, Concord Repatriation General Hospital, Sydney, Australia. Reproductive Hormone Measurement Participants had an early morning fasting blood sample taken at baseline with serum stored at −80°C until assay. Measurement of serum T was by liquid chromatography-tandem mass spectrometry (LC-MS) as described (27) with modifications by introducing ultrapressure for high-pressure liquid chromatography with corresponding changes in extraction methodology validated according to FDA criteria (for details see Supplementary Methods in ref. (28)). The steroid measurements were calibrated against certified reference materials for T (National Measurement Institute, North Ryde, Australia). The assays had between-run coefficients of variation at three levels (low, medium, and high) of quality control (QC) specimens of 1.9%–4.5%, 3.8%–7.6%, 2.9%–13.6%, and 5.7%–8.7%, respectively, over 224 runs including all samples from this study. Overlapping QC samples were routinely run at the start, middle and end of every run with each new QC control run multiple times for calibration before use and there was no evidence of assay drift (13). Serum SHBG were measured by automated immunoassays (Roche Diagnostics Australia, Dee Why, Australia) subject to ongoing external QC program calibration with between-assay coefficients of variation for 2 levels of QC specimens in each run of 2.0%–2.8% for SHBG. The cFT levels in this study were computed using an assumption-free, empirical formula (FTZ) developed and validated against laboratory-based measurements of FT by dialysis methods which have displayed much closer conformance with laboratory-measured FT than model-based formulae (6,7). Morbidity and Mortality Outcomes Measurement Health-related quality of life and self-rated health were assessed using the 12-Item Short Form Health Survey (SF-12) (29). Functional disability was defined using the Katz activity of daily living questionnaire (30). Frailty was defined according to the criteria used in the Cardiovascular Health Study: weight loss/shrinking, weakness, exhaustion, slowness, and low activity (31). Falls were measured at the four month follow-up phone calls after their baseline assessments, participants were asked whether they had fallen in the preceding 4 months and, if so, how many times they had fallen. Participants were assessed for cognitive impairment at the clinic assessment visits using the Mini-Mental State Examination (32). Depressive symptoms were evaluated by the Geriatric Depression Scale, short form (33). The participants were asked about erectile dysfunction, sexual activity, sexual desire, and sexual satisfaction using standard, validated questionnaires (34). Metabolic syndrome was defined using the NCEP Adult Treatment Panel (ATP) III criteria (35). Physical activity was measured using the Physical Activity Scale for the Elderly (PASE) (36). Walking speed was measured at the participants’ usual pace (31). Trained staff used a stopwatch to record the time taken by the men to walk 6 m. The fastest time from two trials was used. Bone mineral density (BMD) at the total hip and femoral neck, lean mass and body fat was measured using dual x-ray absorptiometry (Discovery-W scanner Hologic, Bedford, MA). The appendicular lean mass (ALM) was calculated as the sum of lean mass of arms and legs (kg) (37). The ALM was standardized by body mass index (BMI) (ALMBMI) to take into account the body size of participants (38). Handgrip strength was measured with a Jamar dynamometer (Promedics, Blackburn, UK). Weight (by a regularly calibrated scale), height (using a Harpenden stadiometer), and waist circumference were measured by a trained professional at the clinic visit. Fasting blood samples were obtained at each visit for biochemistry tests including hemoglobin, glucose, and prostate specific antigen (PSA) performed at the accredited Clinical Pathology department of Concord Hospital. The New South Wales Registry of Births, Deaths, and Marriages were contacted to ascertain death status. The Registry also provided details recorded on the original death certificate for all participants. Based on the information provided from the death certificates, the general underlying cause of death (cancer, cardiovascular, or other) was identified independently by two medical practitioners (R.G.C., D.J.H.) (13). Potential Confounder Measurement Tobacco usage status (current, ex-, or never smoker) was by self-reported questionnaires. A comorbidity score was calculated as the sum of all conditions reported from the 19 disorders listed in the questionnaire. BMI was calculated from clinic measurements of height and weight. Statistical Analysis Descriptive characteristics of reproductive hormones at baseline and study health outcomes at baseline, 2-year and 5-year follow-up were generated for the analytic sample (Table 1). Participants were categorized into four mutually exclusive groups based on their baseline serum T and cFT defining “low” for these analyses by setting a threshold of the lowest quintile (20th centile) for serum T (10.2 nM) and cFT (156 pM). The referent groups for all cross-sectional and longitudinal analyses were men with both normal serum T and cFT (NN) with the other groups defined as men with the combinations of normal T/low cFT (NL), low T/normal cFT (LN) or low T/low cFT (LL). Of the 1,705 participants who completed the baseline assessments, a total of 1,651 were included for analyses in this article, after excluding men using androgen or antiandrogen treatments (n = 20) or with missing data on reproductive hormones (n = 34). Table 1. Characteristics of the Study Health Outcomes at Baseline, 2 Years, and 5 Years Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Note: ADL = Activities of daily living; BMD = Bone mineral density; BMI = Body mass index; MMSE = Mini-mental status examination; PASE = Physical activity scale for the elderly. †Higher values are better for MMSE (out of 30), PASE, SF-12 Physical and Mental (each out of 100). Lower values are better for comorbidity (out of 19). ‡The data for nonparticipation at 2 years and 5 years are their baseline descriptive characteristic. Death was the main reason for nonparticipation at 2 years (99 deaths) and at 5 years (382 deaths). View Large Table 1. Characteristics of the Study Health Outcomes at Baseline, 2 Years, and 5 Years Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Baseline (n = 1,651) Mean (SD) or N (%) 2 y (n = 1,291) Mean (SD) or N (%) 5 y (n = 910) Mean (SD) or N (%) Nonparticipation at 2 y‡ (n = 345) Nonparticipation at 5 y‡ (n = 747) Age (years) 76.9 ± 5.5 79.0 ± 25.8 81.4 ± 4.6 79.1 ± 6.1 78.8 ± 6.0 Comorbidity† 2.6 ± 1.8 2.5 ± 1.7 2.5 ± 1.6 2.9 ± 1.9 2.9 ± 1.9 BMI (kg/m2) 27.8 ± 1.8 27.8 ± 4.0 27.6 ± 4.0 27.5 ± 4.2 27.6 ± 4.3 MMSE† 27.1 ± 3.05 27.4 ± 2.8 27.2 ± 3.1 26.1 ± 3.7 26.4 ± 3.4 PASE† 124.4 ± 62.1 119.8 ± 59.7 117.4 ± 63.2 100.6 ± 62.8 107.9 ± 61.2 Walking speed (m/s) 0.9 ± 0.2 0.9 ± 0.2 0.9 ± 0.2 0.8 ± 0.2 0.8 ± 0.2 Hip BMD (g/cm2) 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.2 0.9 ± 0.2 SF-12 Physical† 48.6 ± 10.5 48.6 ± 10.5 47.6 ± 10.6 45.7 ± 11.7 46.1 ± 11.1 SF-12 Mental† 49.1 ± 6.4 49.3 ± 6.3 49.6 ± 6.5 48.3 ± 7.0 48.6 ± 7.0 Weight (kg) 79.4 ± 13.0 79.2 ± 12.8 78.1 ± 12.7 77.5 ± 13.5 77.9 ± 13.5 Grip strength (kg) 34.5 ± 7.5 34.7 ± 8.0 32.7 ± 8.3 32.0 ± 7.6 32.5 ± 7.0 Lean mass (kg) 7.2 ± 1.2 7.2 ± 1.2 7.2 ± 1.2 7.0 ± 1.3 7.1 ± 1.2 Fat percentage (%) 28.9 ± 6.0 29.2 ± 6.0 29.7 ± 6.1 28.9 ± 6.4 29.0 ± 6.2 Glucose (mmol/L) 5.6 ± 1.4 5.7 ± 1.5 5.7 ± 1.5 5.6 ± 1.4 5.6 ± 1.5 Hemoglobin (g/dL) 142.9 ± 14.0 142.2 ± 13.5 141.0 ± 14.4 138.6 ± 17.0 140.2 ± 15.8 Waist (cm) 103.4 ± 19.9 102.1 ± 11.2 101.1 ± 11.0 104.3 ± 39.1 103.6 ± 27.7 Current Smoker 101 (6%) 52 (4%) 35 (4%) 32 (9%) 49 (7%) Poor Self-rated Health 500 (30%) 389 (29%) 251 (26%) 132 (41%) 272 (38%) ADL disability 138 (8%) 141 (10%) 119 (13%) 69 (20%) 104 (14%) Frail 158 (10%) 129 (10%) 93 (10%) 74 (23%) 130 (18%) Previous Falls 138 (8%) 125 (15%) 115 (12%) 47 (14%) 84 (11%) Depression 242 (15%) 206 (15%) 122 (13%) 86 (26%) 159 (22%) Erectile dysfunction 441 (36%) 306 (33%) 121 (20%) 62 (31%) 153 (32%) No sexual activity 532 (44%) 379 (41%) 241 (41%) 65 (33%) 143 (31%) Low sexual satisfaction 1031 (89%) 783 (90%) 509 (90%) 161 (85%) 368 (86%) Low sexual desire 371 (30%) 276 (30%) 179 (30%) 65 (33%) 133 (28%) Metabolic syndrome 481 (37%) 472 (40%) 474 (53%) 90 (36%) 202 (36%) Note: ADL = Activities of daily living; BMD = Bone mineral density; BMI = Body mass index; MMSE = Mini-mental status examination; PASE = Physical activity scale for the elderly. †Higher values are better for MMSE (out of 30), PASE, SF-12 Physical and Mental (each out of 100). Lower values are better for comorbidity (out of 19). ‡The data for nonparticipation at 2 years and 5 years are their baseline descriptive characteristic. Death was the main reason for nonparticipation at 2 years (99 deaths) and at 5 years (382 deaths). View Large Joint prediction in cross-sectional associations between the T/cFT status and health outcomes were assessed by logistic regression for categorical health outcomes, by multiple regression for continuous health outcomes and by Cox regression for mortality outcomes. Results were summarized into concordant findings (only statistically significant LL), discordant findings (either statistically significant LN and LL, or statistically significant NL and LL), and isolated discordance findings (statistically significant LN or NL without statistically significant LL). The detailed results for categorical variables are presented as odds ratios (95% confidence interval), for continuous variables are presented as β values (95% confidence interval) and for mortality are presented as hazard ratios (95% confidence interval). Similarly, longitudinal association between baseline T/cFT status and changes in health outcomes across baseline, 2 years and 5 years were assessed by generalized estimating equations (GEE) with exchangeable working correlation and robust variance estimator. The multinomial cumulative logit model was used for categorical morbidity outcomes, linear model for continuous morbidity outcomes and Poisson loglinear model for mortality outcomes. GEE method is robust and efficient when treating missing data in longitudinal studies (39). Model building for both cross-sectional and longitudinal analyses included adjustment for known major covariates, notably, age, BMI, smoking status, and comorbidity. BMI was not adjusted for in analyses for metabolic syndrome, body fat, weight, and waist circumference analysis. For post-hoc analyses, a Bonferroni adjustment to notional p values was performed to account for multiple comparisons involved in evaluating 28 outcome comparisons from a single set of data so that the conventional 0.05 level of significance was adjusted to a threshold of 0.002 (0.05/28). Models were fitted using SPSS software version 20 (IBM Corp., Armonk, NY) and SAS software 9.3 (SAS Institute Inc., Cary, NC). Results The demographic and anthropometric details of the CHAMP cohort are provided in Table 1 and the descriptive details of T, SHBG, and cFT are provided in Table 2. A total of 1,283 men (78%) were categorized into normal T/normal cFT (NN), 40 men (2%) into normal T/low cFT (NL), 38 men (2%) into low T/normal cFT (LN), and 290 men (18%) into low T/low cFT (LL). Table 2. Serum Testosterone (T), SHBG and Free Testosterone (cFT) Levels for the CHAMP Cohort at Baseline According to Different T/cFT Status Cutoff and Calculation N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 Note: cFT = calculated free testosterone; LL = Low T/Low cFT; LN = Low T/Normal cFT; NL = Normal T/Low cFT; NN = Normal T/Normal cFT. aT level below or above 10.2 nmol/L and cFT level below or above 156 pmol/L (lowest quintile). View Large Table 2. Serum Testosterone (T), SHBG and Free Testosterone (cFT) Levels for the CHAMP Cohort at Baseline According to Different T/cFT Status Cutoff and Calculation N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 N (%) T ((nmol/L) Mean (SD) SHBG (nmol/L) Mean (SD) cFT (pmol/L) Mean (SD) CHAMP cutoff and calculationa  All 1,651, 100% 14.7 ± 6.4 50.1 ± 20.7 206.6 ± 78.0  Normal T/Normal cFT (NN) 1,283, 78% 17.0 ± 5.1 52.3 ± 20.2 235.5 ± 56.8  Normal T/Low cFT (NL) 40, 2% 11.2 ± 0.7 61.9 ± 17.2 149.2 ± 5.0  Low T/Normal cFT (LN) 38, 2% 9.4 ± 0.8 24.3 ± 5.9 164.1 ± 8.2  Low T/Low cFT (LL) 290, 18% 5.9 ± 3.4 42.4 ± 20.6 91.8 ± 52.6 Note: cFT = calculated free testosterone; LL = Low T/Low cFT; LN = Low T/Normal cFT; NL = Normal T/Low cFT; NN = Normal T/Normal cFT. aT level below or above 10.2 nmol/L and cFT level below or above 156 pmol/L (lowest quintile). View Large Cross-Sectional Morbidity Analysis The baseline cross-sectional associations between T/cFT status and morbidity outcomes are shown in detail in Supplementary Table 1 and summarized in Table 3 and Figure 1. After multivariable adjustment of the baseline cross-sectional data, low T/cFT (LL) was significantly associated with 15 of 24 outcomes (frailty, falls, sexual satisfaction, sexual desire, sexual activity, metabolic syndrome, physical activity, walking speed, physical quality of life, weight, hip BMD, body percent fat, waist circumference, glucose, hemoglobin, and PSA). Where LL was not a significantly associated with outcomes, there was only a single morbidity outcome that remained associated with either LN or NL (isolated discordance). A few morbidity outcomes displayed significant associations with discordant findings—four outcomes (metabolic syndrome, weight, fat mass and waist circumference) for LN and three outcomes (weight, fat mass, and waist circumference) for NL—but for each of these outcomes LL was also significant. After Bonferroni correction, LL remained significantly associated with 8 of 24 outcomes in cross-sectional analysis with additional discordant findings in 4 outcomes for LN and no outcomes for NL. There were no associations with isolated discordance. Table 3. Summary of the Concordant, Discordant, and Isolated Discordance for the Morbidity Outcomes Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 aPrimary; primary analysis using an empirical formula (FTZ) categorizing low T and low cFT based on lowest quintile. Sensitivity 1; sensitivity analysis using the same FTZ formula categorizing low T and low cFT based on lowest centile. Sensitivity 2; sensitivity analysis using the Vermeulen formula categorizing low T and low cFT based on lowest quintile. bLL is Low T/Low cFT; LN is Low T/Normal cFT; NL is Normal T/Low cFT; reference group is NN Normal T/Normal cFT. View Large Table 3. Summary of the Concordant, Discordant, and Isolated Discordance for the Morbidity Outcomes Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 Primarya Sensitivity 1a Sensitivity 2a Cross-sectional Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL)b 19 15 17 12 17 12 Discordant (both significant LN and significant LL) 8 4 0 3 0 3 Discordant (both significant NL and significant LL) 8 3 7 3 7 3 Isolated Discordance (significant LN or NL without significant LL) 1 1 1 1 1 1 Longitudinal Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Concordant (only significant LL) 18 16 17 15 17 15 Discordant (both significant LN and significant LL) 7 6 13 8 13 8 Discordant (both significant NL and significant LL) 10 4 10 7 10 7 Isolated Discordance (significant LN or NL without significant LL) 2 0 1 1 1 1 aPrimary; primary analysis using an empirical formula (FTZ) categorizing low T and low cFT based on lowest quintile. Sensitivity 1; sensitivity analysis using the same FTZ formula categorizing low T and low cFT based on lowest centile. Sensitivity 2; sensitivity analysis using the Vermeulen formula categorizing low T and low cFT based on lowest quintile. bLL is Low T/Low cFT; LN is Low T/Normal cFT; NL is Normal T/Low cFT; reference group is NN Normal T/Normal cFT. View Large Figure 1. View largeDownload slide Summary of the cross-sectional and longitudinal findings for the morbidity outcomes. Figure 1. View largeDownload slide Summary of the cross-sectional and longitudinal findings for the morbidity outcomes. Longitudinal Morbidity Analysis The longitudinal associations over the 5-year follow-up between baseline T/cFT status and changes in morbidity outcomes are shown in detail in Supplementary Table 2 and summarized in Table 3 and Figure 1. Similar to the cross-sectional analysis, in multivariate adjusted models of the 5-year longitudinal analysis, low T/cFT (LL) was statistically significantly associated with 16 of 24 outcomes (poor self-rated health, activities of daily living disability, frailty, sexual satisfaction, sexual desire, sexual activity, metabolic syndrome, physical activity, walking speed, hip BMD, physical quality of life, weight, body percent fat, waist circumference, glucose, and hemoglobin). Significant discordant findings were present in six outcomes (metabolic syndrome, weight, fat mass, waist circumference, glucose, and hemoglobin) for LN and four (sexual activity, hip BMD, fat mass, and hemoglobin) for NL where LL was also significant. There were no significant isolated discordant findings (significant LN or NL without significant LL). After Bonferroni correction, LL remained significantly associated with 9 of 24 morbidity outcomes with additional discordant findings in 4 outcomes for LN and no outcomes for NL, all in conjunction with significant LL. Sensitivity Analysis One sensitivity analysis performed used the same empirical formula (FTZ) but lowering the threshold to define “low” from lowest quintiles (lowest 20%) to lowest centiles (lowest 10%) for the morbidity analysis produced essentially the same results. In the multivariable adjusted model, LL was significantly associated with 12 of 24 morbidity outcomes in cross-sectional and 15 of 24 outcomes in longitudinal analysis (Table 3). Additional significant discordant findings (LN or NL) in conjunction with significant LL were present for 6 outcomes in cross-sectional and 15 outcomes in longitudinal analysis. When LL was not significant (isolated discordance), only 1 outcome in cross-sectional and 1 in longitudinal analysis were significantly associated with an adverse morbidity outcome. Another sensitivity analysis was conducted using the Vermeulen cFT (lowest quintile) for the same morbidity analysis and the findings were similar to our original analysis using the more accurate FTZ formula. In multivariable model, LL was significantly associated with 12 of 24 morbidity outcomes in cross-sectional and 15 of 24 outcomes in longitudinal analysis. Where LL was not a significant predictor, isolated discordant findings (either LN or NL significant) were associated with only 1 outcome in cross-sectional and 1 in longitudinal analyses. Among men with discordant findings (significant LN or NL, with significant LL), morbidity prediction was present for 6 outcomes in cross-sectional and 15 outcomes in longitudinal analysis. Cross-Sectional and Longitudinal Mortality Analysis With multivariable adjustment, both the baseline cross-sectional as well as the longitudinal mortality analyses (Supplementary Table 3) showed significant effects for LL in all-cause, cardiovascular and other but not for cancer mortality. For no mortality outcome was there significant prediction when LL was not significant and where mortality outcomes were predicted by discordant (NL or LN), LL was also significant. Discussion Many studies have reported associations between health outcomes and low T or low cFT, considered as separate parameters, among older men (40). However, we observed consistently in a series of studies from the CHAMP (12–19) and HIMS (20–25) cohorts that cFT as a predictor rarely, if ever, provided any significantly different information on health outcomes from serum T measured by LC-MS. Furthermore, as the FHH remains largely untested, there remains minimal critical evidence to what extent, if any, FT data provides additional biological or clinical insight independent of accurate serum T measurements by LC-MS in men (41). The present findings investigating a wide range of morbidity and mortality outcomes in older men suggest that cFT rarely adds independent prognostic information to serum T measured by LC-MS in either cross-sectional or longitudinal analyses. An important caveat is that the utility of health outcome predictions by T and/or FT, depends on the accuracy of the T and FT estimates employed. Until recently, most studies relied upon T immunoassays which suffer from method-specific and other technical limitations notably if applied to reduced serum testosterone such as in older men (3,4). This became a greater problem over the decades after the 1980’s when direct, nonextraction immunoassay became almost universal in clinical practice and research. Over the last decade, more accurate measurement of serum T has become more widely feasible using modern, bench-top LC-MS to supplant direct (unextracted) T immunoassays. Currently, especially for large-scale epidemiological studies, FT is rarely measured directly by dialysis-based laboratory reference methods. These methods are laborious, exacting and require manual laboratory skills which have been largely eliminated by the deskilling automation of chemical pathology laboratories. Furthermore FT measurements lack quality control programs or validated reference ranges. Instead, FT is usually calculated by a variety of formulae which fall into two classes, model-based equilibrium binding and fully empirical equations. These differ in their assumptions and in conformance in accuracy to dialysis-based laboratory gold standard reference methods. The accuracy of cFT is crucial because any formula produces a deterministic (inverse) function of age as it compounds two age-dependent variables—testosterone and SHBG. Unless the formula accurately represents the authentic laboratory-based FT measurement it intends to represent, it will display a spurious correlation with any age-dependent variable regardless of whether that variable has any genuine biological relationship to testosterone. The FTZ equation was originally developed from a large data set of 3975 serum samples by identifying the optimal regression formula of laboratory dialysis-based reference FT measurements on serum testosterone and SHBG measured in the same samples. This formula was cross-validated against a separate set of 124 serum samples (6) and then subsequently confirmed as highly accurate when tested in a different large data set of 2,159 samples from another laboratory using different methods to measure FT, testosterone and SHBG (7). A key finding from the extensive validation involving over 6,000 serum samples was that the widely used, model-based equilibrium binding equation-based formulae by Vermeulen (11) and Sodergard (42) display marked bias deviating from the laboratory-measured FT. These deviations were due to both wrong stoichiometry as well as arbitrary plug-in binding affinity coefficients for T binding to SHBG (6,7), the latter varying fivefold among the various implementations of model-based equilibrium binding formulae (43). The present study uses this FTZ formula to evaluate the impact of accurately estimated cFT, corresponding most closely to laboratory-based FT measurements, on morbidity and mortality outcome predictions. The novelty of the current longitudinal study is that it investigates both serum T and cFT levels concurrently as joint predictors of a wide range of health outcomes over time. Our analysis revealed that both cross-sectionally and longitudinally over 5 years, men with concordant low serum T and cFT as well as those with concordant normal T and cFT were more likely to die or experience adverse health outcomes. On the contrary, only a minority of men had variables displaying discrepancies between T and cFT values and where there was an isolated discordance—that is discordance between T and cFT but not accompanied significant association or prediction by LL—was rare. Hence, not only are discrepancies unusual but cFT alone predicts almost no health outcomes among older men independent of an accurately measured serum T. Altogether, the present analysis provides a comprehensive analysis of a wide range of health outcomes including nonspecific symptoms resembling those of androgen deficiency or many other chronic diseases. A recent study from the EMAS cohort evaluated the FHH among older men by analyzing cross-sectionally the joint association of cFT and T with health outcomes. They reported that low Vermeulen cFT, even in the presence of normal T, but not the combination of normal cFT and low T, was associated with a range of nonspecific symptoms including sexual and physical symptoms (44). In a previous study, they reported that low T and cFT were associated with sexual but not physical or psychological symptoms (45) although the sexual symptoms had high rates of false positive and false negatives reflecting their nonspecificity and the direction of causality could not be determined. This reflects the fact that genuine androgen deficiency symptoms are, for any individual, are highly reproducible at consistent blood testosterone concentrations (46); however, as the actual symptoms differ widely between individuals, grouping individuals according to symptoms erodes the relationship of symptoms to blood testosterone concentrations (47). Furthermore, as the Vermeulen model-based cFT formula systematically deviates from laboratory measured FT values as reported by several independent groups (5–10). Yet, as any cFT remains a deterministic (inverse) function of age, failure to correspond accurately to laboratory-measured FT makes it likely that any relationship to age-related symptoms may reflect residual confounding due to the age-mismatch of the subgroups (persisting after linear age adjustment) rather than any authentic relationship with serum T. The present analysis, using a more accurate and extensively validated cFT formula so that it corresponds more closely to laboratory-measured FT, showed that cFT and T were usually concordant and, in the unusual instances where there was a discordance, that almost always occurred only when the concordant low T/low cFT was also significant. In our analysis as well as that using the Vermeulen formula, significant isolated discordant association or prediction by a low cFT was rare and had little impact on prediction of mortality or morbidity over the next 5 years. Instead it was the combination of both a low T and low cFT that was significantly associated with most outcomes although the direction of causality remains undetermined. Our study shows that men with low T were most likely to have low cFT while men with normal T were most likely to have normal cFT. Only a very small proportion had discordance with either normal T and low cFT, or low T and normal cFT and when this occurred it was almost invariably in the setting where the concordant combination of low T/low CFT was also significant. The major finding in this study is consistent with previous studies showing low T and low cFT, as separate parameters, are associated with these many health outcomes such as general health status, functional ability, metabolic syndrome, bone health, cognition, sexual function, etc. These findings confirm our impression from previous studies in the CHAMP (12–19) and HIMS (20–25) cohorts that show very similar effect size and associations in either low T or low cFT with a wide variety of health outcomes (12–25). This suggest that cFT provides minimal independent predictive information for health outcomes independent of accurately measured serum T and questions whether cFT estimates, even when accurately calculated, provide any useful information for clinical practice. The strengths of this study include the use of longitudinal data to investigate a comprehensive profile of T/cFT status in conjunction with a wide array of key morbidity and mortality outcomes over three follow-up time-points spanning 5 years. Another is the use of the LC-MS, the current gold standard for steroid assays, providing multianalyte steroid profiling. This improves upon direct immunoassay methods which, lacking extraction and chromatography, feature poor accuracy at low levels of circulating sex steroids, which is particularly problematic for measuring circulating T in older men (3,4). Furthermore, we used an extensively validated, assumption-free formula for cFT which corresponds more accurately to laboratory-measured FT than previous model-based equilibrium binding formula that rely on arbitrary plug-in coefficients. A further strength of CHAMP is that it includes a large and representative group of older Australian men, as demonstrated by similar sociodemographic and health characteristics compared to older men in the nationally representative MATeS study (48). A significant limitation of our study is the impact of survivor bias. This applies to the survivorship in the cohort with most losses due to mortality which accounted for nearly 35% of loss to follow-up in our cohort. On the other hand, mortality was evaluated as an outcome so that this cohort provides a more complete view of the causes and determinant of mortality among living older men. To avoid the impact of potential diurnal variation in hormone concentrations, a rhythm that is mostly lost in ageing men (49), we obtained fasting morning blood samples in this study and evaluated joint prediction to avoid collinearity between cFT and T. In conclusion, concordant low serum T and cFT levels were strongly associated with many health outcomes in older men whereas among the minority of men with discrepancies between T and cFT, such discordance was associated with or predicted few health outcomes and only then when for the same outcome, there was also a significant association or prediction by the combination of both low T and cFT. Hence, in addition to the ambiguous theoretical basis of the FHH, the present findings suggest that even accurately cFT estimates provide minimal additional clinical or biological information independent of accurate measurement of serum T concentrations for mortality or morbidity outcomes in older men. These findings provide little support for the application of the FHH to studies of testosterone and clinical outcomes in older men. Supplementary Material Supplementary data is available at The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences online. Funding The CHAMP study is funded by the NHMRC Project Grant (No. 301916), Sydney Medical School Foundation, and Ageing and Alzheimer’s Institute. Conflict of Interest D.G.L.C. is a co-deputy editor for the Journal of Gerontology: Biological Sciences. All the other authors have nothing to declare. Acknowledgments R.G.C., D.J.H., M.J.S., L.M.W., V.N., D.G.L.C., and F.M.B. contributed to the formulation of the study concept, design, methods, subject recruitment, and data collection. B.H. wrote the manuscript and performed the analyses. D.J.H. wrote portions of the manuscript. R.G.C., F.M.B., V.N., D.G.L.C., M.J.S., and L.M.W. reviewed the manuscript and contributed to discussion. References 1. Goldman AL , Bhasin S , Wu FCW , Krishna M , Matsumoto AM , Jasuja R . A reappraisal of testosterone’s binding in circulation: physiological and clinical implications . Endocr Rev . 2017 ;doi: 10.1210/er.2017-00025 2. Handelsman D . Free testosterone: pumping up the tires or ending the free ride ? Endocr Rev . 2017 ; 38 : 297 – 301 .doi: 10.1210/er.2017-00171 Google Scholar CrossRef Search ADS PubMed 3. Rosner W , Auchus RJ , Azziz R , Sluss PM , Raff H . Position statement: utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement . J Clin Endocrinol Metab . 2007 ; 92 : 405 – 413 . doi: 10.1210/jc.2006-1864 Google Scholar CrossRef Search ADS PubMed 4. Handelsman DJ , Wartofsky L . Requirement for mass spectrometry sex steroid assays in the Journal of Clinical Endocrinology and Metabolism . J Clin Endocrinol Metab . 2013 ; 98 : 3971 – 3973 . doi: 10.1210/jc.2013-3375 Google Scholar CrossRef Search ADS PubMed 5. Ly LP , Handelsman DJ . Empirical estimation of free testosterone from testosterone and sex hormone-binding globulin immunoassays . Eur J Endocrinol . 2005 ; 152 : 471 – 478 . doi: 10.1530/eje.1.01844 Google Scholar CrossRef Search ADS PubMed 6. Sartorius G , Ly LP , Sikaris K , McLachlan R , Handelsman DJ . Predictive accuracy and sources of variability in calculated free testosterone estimates . Ann Clin Biochem . 2009 ; 46 ( Pt 2 ): 137 – 143 . doi: 10.1258/acb.2008.008171 Google Scholar CrossRef Search ADS PubMed 7. Ly LP , Sartorius G , Hull L et al. Accuracy of calculated free testosterone formulae in men . Clin Endocrinol (Oxf) . 2010 ; 73 : 382 – 388 . doi: 10.1111/j.1365-2265.2010.03804.x Google Scholar CrossRef Search ADS PubMed 8. Hackbarth JS , Hoyne JB , Grebe SK , Singh RJ . Accuracy of calculated free testosterone differs between equations and depends on gender and SHBG concentration . Steroids . 2011 ; 76 : 48 – 55 . doi: 10.1016/j.steroids.2010.08.008 Google Scholar CrossRef Search ADS PubMed 9. Salameh WA , Redor-Goldman MM , Clarke NJ , Reitz RE , Caulfield MP . Validation of a total testosterone assay using high-turbulence liquid chromatography tandem mass spectrometry: total and free testosterone reference ranges . Steroids . 2010 ; 75 : 169 – 175 . doi: 10.1016/j.steroids.2009.11.004 Google Scholar CrossRef Search ADS PubMed 10. Zakharov MN , Bhasin S , Travison TG et al. A multi-step, dynamic allosteric model of testosterone’s binding to sex hormone binding globulin . Mol Cell Endocrinol . 2015 ; 399 : 190 – 200 . doi: 10.1016/j.mce.2014.09.001 Google Scholar CrossRef Search ADS PubMed 11. Vermeulen A , Verdonck L , Kaufman JM . A critical evaluation of simple methods for the estimation of free testosterone in serum . J Clin Endocrinol Metab . 1999 ; 84 : 3666 – 3672 . doi: 10.1210/jcem.84.10.6079 Google Scholar CrossRef Search ADS PubMed 12. Hsu B , Hirani V , Naganathan V et al. Sexual function and mortality in older men: the Concord Health and Ageing in Men Project . J Gerontol A Biol Sci Med Sci . 2017 ; 72 : 520 – 527 . doi: 10.1093/gerona/glw101 Google Scholar PubMed 13. Hsu B , Cumming RG , Naganathan V et al. Temporal changes in androgens and estrogens are associated with all-cause and cause-specific mortality in older men . J Clin Endocrinol Metab . 2016 ; 101 : 2201 – 2210 . doi: 10.1210/jc.2016-1025 . Google Scholar CrossRef Search ADS PubMed 14. Hsu B , Cumming RG , Blyth FM , Naganathan V , Le Couteur DG , Seibel MJ et al. Longitudinal and cross-sectional relationships of circulating reproductive hormone levels to self-rated health and health-related quality of life in community-dwelling older men . J Clin Endocrinol Metab . 2014 ; 99 : 1638 – 1647 . doi: 10.1210/jc.2013–3984 Google Scholar CrossRef Search ADS PubMed 15. Hsu B , Cumming RG , Naganathan V et al. Associations between circulating reproductive hormones and SHBG and prevalent and incident metabolic syndrome in community-dwelling older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2014 ; 99 : E2686 – E2691 . doi: 10.1210/jc.2014-2464 Google Scholar CrossRef Search ADS PubMed 16. Hsu B , Cumming RG , Naganathan V et al. Associations between circulating reproductive hormones and SHBG and prevalent and incident metabolic syndrome in community-dwelling older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2014 ; 99 : E2686 – E2691 . doi: 10.1210/jc.2014-2464 Google Scholar CrossRef Search ADS PubMed 17. Hsu B , Cumming RG , Waite LM et al. Longitudinal relationships between reproductive hormones and cognitive decline in older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2015 ; 100 : 2223 – 2230 . doi: 10.1210/jc.2015-1016 Google Scholar CrossRef Search ADS PubMed 18. Hsu B , Cumming RG , Seibel MJ et al. Reproductive hormones and longitudinal change in bone mineral density and incident fracture risk in older men: the Concord Health and Aging in Men Project . J Bone Miner Res . 2015 ; 30 : 1701 – 1708 . doi: 10.1002/jbmr.2493 Google Scholar CrossRef Search ADS PubMed 19. Hsu B , Cumming RG , Waite LM et al. Longitudinal relationships between reproductive hormones and cognitive decline in older men: the Concord Health and Ageing in Men Project . J Clin Endocrinol Metab . 2015 ; 100 : 2223 – 2230 . doi: 10.1210/jc.2015-1016 Google Scholar CrossRef Search ADS PubMed 20. Yeap BB , Alfonso H , Chubb SA et al. In older men, higher plasma testosterone or dihydrotestosterone is an independent predictor for reduced incidence of stroke but not myocardial infarction . J Clin Endocrinol Metab . 2014 ; 99 : 4565 – 4573 . doi: 10.1210/jc.2014-2664 Google Scholar CrossRef Search ADS PubMed 21. Yeap BB , Alfonso H , Chubb SA et al. Higher serum undercarboxylated osteocalcin and other bone turnover markers are associated with reduced diabetes risk and lower estradiol concentrations in older men . J Clin Endocrinol Metab . 2015 ; 100 : 63 – 71 .doi: 10.1210/jc.2014–3019 Google Scholar CrossRef Search ADS PubMed 22. Yeap BB , Knuiman MW , Divitini ML et al. Epidemiological and Mendelian randomization studies of dihydrotestosterone and estradiol and leukocyte telomere length in men . J Clin Endocrinol Metab . 2016 ; 101 : 1299 – 1306 . doi: 10.1210/jc.2015-4139 Google Scholar CrossRef Search ADS PubMed 23. Yeap BB , Alfonso H , Chubb SA et al. In older men, higher plasma testosterone or dihydrotestosterone is an independent predictor for reduced incidence of stroke but not myocardial infarction . J Clin Endocrinol Metab . 2014 ; 99 : 4565 – 4573 . doi: 10.1210/jc.2014-2664 Google Scholar CrossRef Search ADS PubMed 24. Yeap BB , Alfonso H , Paul Chubb SA et al. In older men, higher plasma testosterone or dihydrotestosterone are independent predictors for reduced incidence of stroke but not myocardial infarction . J Clin Endocrinol Metab . 2014 Dec;99:4565–73 . doi: 10.1210/jc.2014–2664 25. Yeap BB , Alfonso H , Chubb SA et al. Reference ranges and determinants of testosterone, dihydrotestosterone, and estradiol levels measured using liquid chromatography-tandem mass spectrometry in a population-based cohort of older men . J Clin Endocrinol Metab . 2012 ; 97 : 4030 – 4039 . doi: 10.1210/jc.2012-2265 Google Scholar CrossRef Search ADS PubMed 26. Cumming RG , Handelsman D , Seibel MJ et al. Cohort Profile: the Concord Health and Ageing in Men Project (CHAMP) . Int J Epidemiol . 2009 ; 38 : 374 – 378 . doi: 10.1093/ije/dyn071 Google Scholar CrossRef Search ADS PubMed 27. Harwood DT , Handelsman DJ . Development and validation of a sensitive liquid chromatography-tandem mass spectrometry assay to simultaneously measure androgens and estrogens in serum without derivatization . Clin Chim Acta . 2009 ; 409 : 78 – 84 . doi: 10.1016/j.cca.2009.09.003 Google Scholar CrossRef Search ADS PubMed 28. Keski-Rahkonen P , Desai R , Jimenez M , Harwood DT , Handelsman DJ . Measurement of estradiol in human serum by LC-MS/MS using a novel estrogen-specific derivatization reagent . Anal Chem . 2015 ; 87 : 7180 – 7186 . doi: 10.1021/acs.analchem.5b01042 Google Scholar CrossRef Search ADS PubMed 29. Ware J Jr , Kosinski M , Keller SD . A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity . Med Care . 1996 ; 34 : 220 – 233 . Google Scholar CrossRef Search ADS 30. Smith LA , Branch LG , Scherr PA et al. Short-term variability of measures of physical function in older people . J Am Geriatr Soc . 1990 ; 38 : 993 – 998 . Google Scholar CrossRef Search ADS PubMed 31. Fried LP , Tangen CM , Walston J et al. ; Cardiovascular Health Study Collaborative Research Group . Frailty in older adults: evidence for a phenotype . J Gerontol A Biol Sci Med Sci . 2001 ; 56 : M146 – M156 . Google Scholar CrossRef Search ADS PubMed 32. Folstein MF , Folstein SE , McHugh PR . “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res . 1975 ; 12 : 189 – 198 . Google Scholar CrossRef Search ADS PubMed 33. Shiekh J , Yesavage JA . Geriatric depression scale: recent findings and developments of a short version . QualityMetric Inc ; 2009 . 34. O’Donnell AB , Araujo AB , Goldstein I , McKinlay JB . The validity of a single-question self-report of erectile dysfunction. Results from the Massachusetts Male Aging Study . J Gen Intern Med . 2005 ; 20 : 515 – 519 . doi: 10.1111/j.1525-1497.2005.0076.x Google Scholar CrossRef Search ADS PubMed 35. Grundy SM , Brewer HB , Jr. , Cleeman JI , Smith SC , Jr. , Lenfant C . Definition of metabolic syndrome: report of the National Heart, Lung, and Blood Institute/American Heart Association conference on scientific issues related to definition . Circulation . 2004 ; 109 : 433 – 438 . doi: 10.1161/01.cir.0000111245.75752.c6 Google Scholar CrossRef Search ADS PubMed 36. Washburn RA , McAuley E , Katula J , Mihalko SL , Boileau RA . The physical activity scale for the elderly (PASE): evidence for validity . J Clin Epidemiol . 1999 ; 52 : 643 – 651 . Google Scholar CrossRef Search ADS PubMed 37. Heymsfield SB , Smith R , Aulet M et al. Appendicular skeletal muscle mass: measurement by dual-photon absorptiometry . Am J Clin Nutr . 1990 ; 52 : 214 – 218 . Google Scholar CrossRef Search ADS PubMed 38. Studenski SA , Peters KW , Alley DE , Cawthon PM , McLean RR , Harris TB et al. The FNIH sarcopenia project: rationale, study description, conference recommendations, and final estimates . J Gerontol A Biol Sci Med Sci . 2014 ; 69 : 547 – 558 . doi: 10.1093/gerona/glu010 Google Scholar CrossRef Search ADS PubMed 39. Zeger SL , Liang KY . Longitudinal data analysis for discrete and continuous outcomes . Biometrics . 1986 ; 42 : 121 – 130 . Google Scholar CrossRef Search ADS PubMed 40. Kaufman JM , Vermeulen A . The decline of androgen levels in elderly men and its clinical and therapeutic implications . Endocr Rev . 2005 ; 26 : 833 – 876 . doi: 10.1210/er.2004-0013 Google Scholar CrossRef Search ADS PubMed 41. Bhasin S , Cunningham GR , Hayes FJ et al. ; Task Force, Endocrine Society . Testosterone therapy in men with androgen deficiency syndromes: an Endocrine Society clinical practice guideline . J Clin Endocrinol Metab . 2010 ; 95 : 2536 – 2559 . doi: 10.1210/jc.2009-2354 Google Scholar CrossRef Search ADS PubMed 42. Södergård R , Bäckström T , Shanbhag V , Carstensen H . Calculation of free and bound fractions of testosterone and estradiol-17 beta to human plasma proteins at body temperature . J Steroid Biochem . 1982 ; 16 : 801 – 810 . Google Scholar CrossRef Search ADS PubMed 43. Mazer NA . A novel spreadsheet method for calculating the free serum concentrations of testosterone, dihydrotestosterone, estradiol, estrone and cortisol: with illustrative examples from male and female populations . Steroids . 2009 ; 74 : 512 – 519 . doi: 10.1016/j.steroids.2009.01.008 Google Scholar CrossRef Search ADS PubMed 44. Antonio L , Wu FC , O’Neill TW et al. ; European Male Ageing Study Study Group . Low free testosterone is associated with hypogonadal signs and symptoms in men with normal total testosterone . J Clin Endocrinol Metab . 2016 ; 101 : 2647 – 2657 . doi: 10.1210/jc.2015-4106 Google Scholar CrossRef Search ADS PubMed 45. Wu FC , Tajar A , Beynon JM et al. ; EMAS Group . Identification of late-onset hypogonadism in middle-aged and elderly men . N Engl J Med . 2010 ; 363 : 123 – 135 . doi: 10.1056/NEJMoa0911101 Google Scholar CrossRef Search ADS PubMed 46. Kelleher S , Conway AJ , Handelsman DJ . Blood testosterone threshold for androgen deficiency symptoms . J Clin Endocrinol Metab . 2004 ; 89 : 3813 – 3817 . doi: 10.1210/jc.2004-0143 Google Scholar CrossRef Search ADS PubMed 47. Zitzmann M , Faber S , Nieschlag E . Association of specific symptoms and metabolic risks with serum testosterone in older men . J Clin Endocrinol Metab . 2006 ; 91 : 4335 – 4343 . doi: 10.1210/jc.2006-0401 Google Scholar CrossRef Search ADS PubMed 48. Holden CA , McLachlan RI , Cumming R et al. Sexual activity, fertility and contraceptive use in middle-aged and older men: men in Australia, Telephone Survey (MATeS) . Hum Reprod . 2005 ; 20 : 3429 – 3434 . doi: 10.1093/humrep/dei307 Google Scholar CrossRef Search ADS PubMed 49. Bremner WJ , Vitiello MV , Prinz PN . Loss of circadian rhythmicity in blood testosterone levels with aging in normal men . J Clin Endocrinol Metab . 1983 ; 56 : 1278 – 1281 . doi: 10.1210/jcem-56-6-1278 Google Scholar CrossRef Search ADS PubMed © The Author(s) 2017. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

The Journals of Gerontology Series A: Biomedical Sciences and Medical SciencesOxford University Press

Published: Sep 15, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off