The longitudinal relationship between declining levels of reproductive hormones and cognitive function remains unclear in older men.
The objective of this study was to examine the temporal relationship between changes in major reproductive hormone levels and cognitive decline over time.
Men age 70 years and older from the Concord Health and Ageing in Men Project (CHAMP) were assessed at baseline (2005–2007; n = 1705), 2-year followup (2007–2009; n = 1367), and 5-year followup (2010–2013; n = 958).
At all assessments, T, dihydrotestosterone (DHT), estradiol (E2), and estrone (E1) were measured by liquid chromatography–tandem mass spectrometry, and SHBG, LH, and FSH by immunoassay. Dementia was diagnosed at baseline by clinical assessment and review by a specialist panel. Cognitive function was measured at all three assessments by the Mini Mental State Examination.
None of the baseline reproductive hormones predicted cognitive decline in men without dementia over 5 years. However, the change in serum hormones over time was associated with cognitive decline. In univariate analysis, change in all the studied hormones, except for E2, was significantly associated with cognitive decline. However, in multivariate-adjusted models accounting for potential confounders, only change in serum T (β = 0.067), DHT (β = 0.394), calculated free T (β = 0.005), and E1 (β = 0.009) remained significantly associated (P < .05) with cognitive decline. Men who had dementia at baseline had significantly greater decline in serum T levels, but not in other studied hormones, over the 5 years.
Our findings show that decline in androgen status is associated with cognitive decline in older men, but the causality of this association requires further elucidation.
In men, the decline in reproductive hormones such as T and cognitive function coincide with aging (1). However, the longitudinal relationship between declining levels of reproductive hormones and cognitive function remains unclear.
To date there have been six longitudinal studies examining the predictive associations between baseline reproductive hormones and subsequent cognitive decline (2–8). These studies have used a range of different cognitive measures, such as Trails B, Modified Mini Mental State Examination, Selective Reminding Test, Cognitive Abilities Screening Instrument, and Mini Mental Status Examination (MMSE). The Honolulu–Asia Aging Study and the Baltimore Longitudinal Study of Aging are the only two studies to have studied new onset (incident) Alzheimer's disease and dementia. Four studies, including the Honolulu Study, found that low baseline T and its various free fraction forms was not associated with reduced cognition over time (2–5). However, the Baltimore study found that low calculated free T (cFT) was associated with a faster rate of decline in visual memory and increased risk of Alzheimer's disease, and the United Kingdom Healthy Ageing Study reported that lower baseline T was associated with a sharp cognitive decline (four points or more on MMSE) (6–8). So far no studies have examined the dynamic relationships of reproductive hormone level as concurrent predictors of cognitive decline in older men.
The objectives of our study were to 1) examine for the first time the relationship between baseline reproductive hormones and change in cognitive function by measuring serum androgen (T, DHT) and estrogen (E2, E1) levels using liquid chromatography–tandem mass spectrometry (LC-MS/MS), 2) examine the temporal longitudinal relationship between change in hormone levels and cognitive decline over time, and 3) to investigate the effect of dementia and mild cognitive impairment (MCI) on subsequent changes in serum reproductive hormones.
Materials and Methods
Study participants
The Concord Health and Ageing in Men Project (CHAMP) is a longitudinal, observational study of the epidemiology of male aging conducted among men living within the three local government areas (Burwood, Canada Bay, and Strathfield) surrounding Concord Hospital in Sydney, New South Wales, Australia (9). Men were selected from the New South Wales electoral roll; enrollment is compulsory in Australia. Potential participants were community-dwelling men age at least 70 years with no other inclusion or exclusion criteria. Letters were sent to 3627 men, of whom 2815 were successfully contacted and confirmed eligible, and 1511 participated in the study. An additional 194 eligible men living in the study area heard about the study from friends or the local media and were recruited before receiving a letter, yielding a total cohort of 1705 participants. The study design has been reported in detail elsewhere (9).
Baseline measurements were conducted between January 2005 and June 2007; data were collected using self-reported questionnaires, interviewer-administered questionnaires, and a wide range of clinical assessments. Follow-up assessments were conducted between January 2007 and October 2009 for 2-year followup, and August 2010 and July 2013 for the 5-year followup, with identical measurements as at baseline.
Reproductive hormone measurement
Participants had an early morning fasting blood sample taken at all three time-points, with serum stored at −80°C until assay. Measurements of serum T, DHT, E2, and E1 were by LC-MS/MS as described elsewhere (10). The steroid measurements were calibrated against certified reference materials for T and DHT (National Measurement Institute, North Ryde, Australia), E2 (European Commission's Institute for Reference Materials and Measurements) and E1 (Cerilliant Corporation). The assays had between-run coefficients of variation (CV) at three levels (low, medium, 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 184 runs including samples from this study. Overlapping QC samples were used continuously for all samples measured in this study where the steroid profiles were measured in three separate batches for the baseline, 2-year, and 5-year follow-up samples. The steroid assays had limits of quantification (defined by the U.S. Food and Drug Administration/European Medicines Agency as lowest detectable measurement with CV < 20%) of 0.025 ng/mL (T), 0.10 ng/mL (DHT), 5 pg/mL (E2), and 3 pg/mL (E1). Serum LH, FSH, and SHBG were measured by automated immunoassays (Roche Diagnostics Australia, Dee Why, Australia) subject to ongoing external QC program calibration with between-assay CV for three levels of QC specimens in each run of 2.1–2.2% for LH, 2.7–3.0 for FSH, and 2.0–2.8% (two QC levels only) SHBG. cFT levels were computed using an empirical formula validated in two large data sets consisting of more than 6000 blood samples (11, 12).
Dementia and cognition assessment
Participants were assessed for cognitive impairment at each of the three clinic assessment visits using the Mini Mental State Examination (MMSE) (13). At baseline, participants with an MMSE score less than or equal to 26 and/or an Informant Questionnaire on Cognitive Decline score (14) greater than 3.6 were invited to have a detailed clinical assessments by a study geriatrician (15). The score for MMSE ranges from 0–30 with 30 representing higher performance. The Informant Questionnaire on Cognitive Decline results in a score that ranges from 1 to 5. A score of 3 means that the men are rated on average as “no change,” score of 4 means an average of “a bit worse,” and a score of 5 means an average of “much worse.” This assessment included a review of medical comorbidities and medications, a standardized neurological assessment, a more detailed informant interview (16) and the Rowland Universal Dementia Assessment Scale (RUDAS) (17). At a weekly consensus meeting two geriatricians, a neurologist and a neuropsychologist reviewed all medical, cognitive, informant, and functional data and reached a final diagnosis of cognitive status for each participant, categorizing them into having dementia, MCI, or normal cognition. To correctly capture the change in MMSE across the three time-points over 5 years in our CHAMP men, those who were diagnosed with dementia at baseline were excluded from relevant analyses.
Participants with a MMSE decline of three or more points from baseline to 5-year followup were considered to have a clinically significant cognitive decline (18, 19). For the purpose of this analysis, men who returned to the study for the 2-year followup but not for the 5-year followup were included based on examining the change in MMSE over 2 years instead of 5 years using the same definition of a decline of three or more points.
Potential confounder measurement
Tobacco usage status (current, ex-, or never smoker) and the number of years of education were assessed through the self-reported questionnaires. Body mass index (BMI) was calculated from clinic measurements of height, using a Harpenden stadiometer, and weight. Depressive symptoms were assessed with the 15-item Geriatric Depression Scale (score ≥ 5 is indicative of depressive symptoms) (20).
Statistical analysis
In examining the relationship between reproductive hormones and cognitive decline, a total of 866 men at all three assessments were included in the analyses. Men who had dementia at baseline were excluded (n = 91) for these analyses. We also excluded men with a non-English-speaking background (n = 643) for language difficulties completing the MMSE, and 73 additional men had missing data.
In examining the relationship between dementia (n = 91) at baseline and subsequent changes in hormones, 807 of the 958 men who attended both the 2-year and 5-year followups were included. We excluded 94 men who had unknown cognitive status at baseline and 25 men with missing data. For both analyses, men receiving either androgen or anti-androgen treatment were excluded (n = 32).
Descriptive baseline characteristics were generated for men in the analytic sample. The associations between baseline reproductive hormones, fitted as continuous variables with results expressed in terms of a 1-U decline in hormone levels, and longitudinal change in cognitive function (change in MMSE) across baseline, 2-year followup, and 5-year followup were assessed by generalized estimating equations (GEE) with exchangeable working correlation and robust variance estimator. The GEE method is robust when treating missing data in longitudinal data (21). Logistic regression models were used to supplement the GEE models by evaluating the association between baseline reproductive hormones and clinically important cognitive decline (≥ 3 points on the MMSE) from baseline to 5-year followup. Finally, GEE models were used to analyze the longitudinal associations between changes in reproductive hormones and change in cognitive function (change in MMSE) across baseline, 2-year followup, and 5-year followup.
Graphs were generated to examine the mean change in serum hormone levels during followup according to baseline diagnosis of dementia, MCI, or normal cognition. Linear regression analyses and repeated measures ANOVA with confounder covariates were used to compare the mean differences in hormone levels from baseline to 5-years between men with normal cognition, with MCI or dementia. The model building for all analyses included adjustment known covariates for cognitive decline, notably, age, BMI, smoking status, years of education, and depression score (Geriatric Depression Scale). Models were fitted using SPSS software version 20 (IBM Corp., Armonk, NY) and SAS software version 9.3 (SAS Institute Inc., Cary, NC).
Ethics approval
The CHAMP study had ongoing approval by the Concord Hospital Human Research Ethics Committee, and study participants provided written informed consent for all procedures.
Results
The characteristics of men in the analytic sample at the three assessment time points are shown in Table 1. At baseline, they had a mean age of 76.9 ± 5.5 years (range, 70–97 y) with a mean BMI of 27.8 ± 4.1 kg/m2. The mean MMSE scores were 28.5 ± 1.3 at baseline (n = 853), 28.4 ± 1.7 at 2 years (n = 729), and 28.1 ± 2.4 at 5 years (n = 546). At baseline, 91 men had dementia, 120 men had MCI, and 1313 men had normal cognition. Overall, 11% (n = 95) of the main analytic sample excluding participants with dementia had a clinically significant cognitive decline (≥ 3 points on MMSE) from baseline to 5-year followup with7% reaching less than MMSE score of 26.
Characteristics of Men in the Analytic Sample
| Baseline, Mean (sd) or N (%) n = 853 | 2-year, Mean (sd) or N (%) n = 729 | 5-year, Mean (sd) or N (%) n = 546 | P Value | |
|---|---|---|---|---|
| Age, y | 76.9 (5.5) | 79.5 (5.3) | 81.4 (4.6) | <.001 |
| BMI, kg/m2 | 27.4 (4.1) | 27.3 (4.0) | 27.0 (3.8) | <.001 |
| Education, y | 9.5 (1.4) | 9.4 (1.4) | 9.5 (1.4) | .08 |
| Current smoker | 35 (4%) | 18 (2%) | 14 (2%) | <.001 |
| Depression | 74 (9%) | 103 (12%) | 44 (7%) | .001 |
| T, ng/mL | 4.2 (1.8) | 4.2 (1.9) | 3.4 (1.8) | <.001 |
| DHT, ng/mL | 0.4 (0.3) | 0.4 (0.2) | 0.3 (0.2) | <.001 |
| SHBG, nmol/L | 50.6 (20.6) | 52.4 (21.1) | 57.4 (24.4) | <.001 |
| E2, pg/mL | 24.9 (9.1) | 24.2 (9.8) | 36.2 (15.4) | <.001 |
| E1, pg/mL | 40.3 (15.7) | 40.1 (1.5) | 29.3 (12.1) | <.001 |
| FSH, IU/L | 14.0 (14.1) | 14.6 (14.3) | 16.6 (15.7) | <.001 |
| LH, IU/L | 9.2 (8.3) | 9.7 (8.2) | 11.0 (9.1) | <.001 |
| cFT, pg/mL | 16.9 (6.6) | 16.7 (6.7) | 13.4 (6.4) | <.001 |
| MMSE | 28.5 (1.3) | 28.4 (1.7) | 28.1 (2.4) | <.001 |
| Baseline, Mean (sd) or N (%) n = 853 | 2-year, Mean (sd) or N (%) n = 729 | 5-year, Mean (sd) or N (%) n = 546 | P Value | |
|---|---|---|---|---|
| Age, y | 76.9 (5.5) | 79.5 (5.3) | 81.4 (4.6) | <.001 |
| BMI, kg/m2 | 27.4 (4.1) | 27.3 (4.0) | 27.0 (3.8) | <.001 |
| Education, y | 9.5 (1.4) | 9.4 (1.4) | 9.5 (1.4) | .08 |
| Current smoker | 35 (4%) | 18 (2%) | 14 (2%) | <.001 |
| Depression | 74 (9%) | 103 (12%) | 44 (7%) | .001 |
| T, ng/mL | 4.2 (1.8) | 4.2 (1.9) | 3.4 (1.8) | <.001 |
| DHT, ng/mL | 0.4 (0.3) | 0.4 (0.2) | 0.3 (0.2) | <.001 |
| SHBG, nmol/L | 50.6 (20.6) | 52.4 (21.1) | 57.4 (24.4) | <.001 |
| E2, pg/mL | 24.9 (9.1) | 24.2 (9.8) | 36.2 (15.4) | <.001 |
| E1, pg/mL | 40.3 (15.7) | 40.1 (1.5) | 29.3 (12.1) | <.001 |
| FSH, IU/L | 14.0 (14.1) | 14.6 (14.3) | 16.6 (15.7) | <.001 |
| LH, IU/L | 9.2 (8.3) | 9.7 (8.2) | 11.0 (9.1) | <.001 |
| cFT, pg/mL | 16.9 (6.6) | 16.7 (6.7) | 13.4 (6.4) | <.001 |
| MMSE | 28.5 (1.3) | 28.4 (1.7) | 28.1 (2.4) | <.001 |
Characteristics of Men in the Analytic Sample
| Baseline, Mean (sd) or N (%) n = 853 | 2-year, Mean (sd) or N (%) n = 729 | 5-year, Mean (sd) or N (%) n = 546 | P Value | |
|---|---|---|---|---|
| Age, y | 76.9 (5.5) | 79.5 (5.3) | 81.4 (4.6) | <.001 |
| BMI, kg/m2 | 27.4 (4.1) | 27.3 (4.0) | 27.0 (3.8) | <.001 |
| Education, y | 9.5 (1.4) | 9.4 (1.4) | 9.5 (1.4) | .08 |
| Current smoker | 35 (4%) | 18 (2%) | 14 (2%) | <.001 |
| Depression | 74 (9%) | 103 (12%) | 44 (7%) | .001 |
| T, ng/mL | 4.2 (1.8) | 4.2 (1.9) | 3.4 (1.8) | <.001 |
| DHT, ng/mL | 0.4 (0.3) | 0.4 (0.2) | 0.3 (0.2) | <.001 |
| SHBG, nmol/L | 50.6 (20.6) | 52.4 (21.1) | 57.4 (24.4) | <.001 |
| E2, pg/mL | 24.9 (9.1) | 24.2 (9.8) | 36.2 (15.4) | <.001 |
| E1, pg/mL | 40.3 (15.7) | 40.1 (1.5) | 29.3 (12.1) | <.001 |
| FSH, IU/L | 14.0 (14.1) | 14.6 (14.3) | 16.6 (15.7) | <.001 |
| LH, IU/L | 9.2 (8.3) | 9.7 (8.2) | 11.0 (9.1) | <.001 |
| cFT, pg/mL | 16.9 (6.6) | 16.7 (6.7) | 13.4 (6.4) | <.001 |
| MMSE | 28.5 (1.3) | 28.4 (1.7) | 28.1 (2.4) | <.001 |
| Baseline, Mean (sd) or N (%) n = 853 | 2-year, Mean (sd) or N (%) n = 729 | 5-year, Mean (sd) or N (%) n = 546 | P Value | |
|---|---|---|---|---|
| Age, y | 76.9 (5.5) | 79.5 (5.3) | 81.4 (4.6) | <.001 |
| BMI, kg/m2 | 27.4 (4.1) | 27.3 (4.0) | 27.0 (3.8) | <.001 |
| Education, y | 9.5 (1.4) | 9.4 (1.4) | 9.5 (1.4) | .08 |
| Current smoker | 35 (4%) | 18 (2%) | 14 (2%) | <.001 |
| Depression | 74 (9%) | 103 (12%) | 44 (7%) | .001 |
| T, ng/mL | 4.2 (1.8) | 4.2 (1.9) | 3.4 (1.8) | <.001 |
| DHT, ng/mL | 0.4 (0.3) | 0.4 (0.2) | 0.3 (0.2) | <.001 |
| SHBG, nmol/L | 50.6 (20.6) | 52.4 (21.1) | 57.4 (24.4) | <.001 |
| E2, pg/mL | 24.9 (9.1) | 24.2 (9.8) | 36.2 (15.4) | <.001 |
| E1, pg/mL | 40.3 (15.7) | 40.1 (1.5) | 29.3 (12.1) | <.001 |
| FSH, IU/L | 14.0 (14.1) | 14.6 (14.3) | 16.6 (15.7) | <.001 |
| LH, IU/L | 9.2 (8.3) | 9.7 (8.2) | 11.0 (9.1) | <.001 |
| cFT, pg/mL | 16.9 (6.6) | 16.7 (6.7) | 13.4 (6.4) | <.001 |
| MMSE | 28.5 (1.3) | 28.4 (1.7) | 28.1 (2.4) | <.001 |
When comparing men at baseline to men lost to followup at 2-year, there were no difference in the mean levels of all the studied reproductive hormones (results not shown). Men lost to followup at 2 years were significantly older (79 vs 78 y), but there were no differences in the number of comorbidities, BMI status, and MMSE scores. However, when comparing men at baseline to men lost to followup at either 2-year or 5-year, the lost-to-followup men were significantly older (79 vs 75 y), lower BMI (27.5 vs 28.0 kg/m2), more comorbidities (3 vs 2), and had lower MMSE (26 vs 28). Death was the main reason for nonparticipation at 2 years (99 deaths) and at 5 years (382 deaths). The other main reason for failure to attend the followup clinic visits was illness (n = 110 at 2 y and n = 174 at 5 y).
The associations between baseline reproductive hormones and longitudinal change in MMSE across baseline, 2-year followup, and 5-year followup are shown in Table 2. Univariate analyses revealed a positive association for DHT (β = 0.292) and a negative association for FSH (β = −0.008) and LH (β = −0.019), and no association of T, SHBG, E2, E1, and cFT with longitudinal change in MMSE. This means that lower DHT, or higher FSH and LH levels at baseline were associated with greater decline in cognitive function over time. However, in multivariate-adjusted models that account for potential confounders, none of these hormones remained statistically significantly associated with change in MMSE.
Associations between Baseline Reproductive Hormones and Longitudinal Change in MMSE across Baseline, 2-Year, and 5-Year Followup
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.018 | .5 | −0.003 | .9 | 0.012 | .7 |
| DHT | 0.292 | .05 | 0.203 | .2 | 0.236 | .2 |
| SHBG | −0.004 | .08 | 0.002 | .3 | 0.003 | .2 |
| E2 | 0.005 | .4 | 0.004 | .5 | 0.006 | .3 |
| E1 | 0.006 | .07 | 0.004 | .2 | 0.005 | .08 |
| FSH | −0.008 | .04 | −0.001 | .9 | −0.001 | .9 |
| LH | −0.019 | .002 | −0.004 | .5 | −0.003 | .6 |
| cFT | 0.002 | .3 | −0.001 | .7 | 0.0004 | .8 |
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.018 | .5 | −0.003 | .9 | 0.012 | .7 |
| DHT | 0.292 | .05 | 0.203 | .2 | 0.236 | .2 |
| SHBG | −0.004 | .08 | 0.002 | .3 | 0.003 | .2 |
| E2 | 0.005 | .4 | 0.004 | .5 | 0.006 | .3 |
| E1 | 0.006 | .07 | 0.004 | .2 | 0.005 | .08 |
| FSH | −0.008 | .04 | −0.001 | .9 | −0.001 | .9 |
| LH | −0.019 | .002 | −0.004 | .5 | −0.003 | .6 |
| cFT | 0.002 | .3 | −0.001 | .7 | 0.0004 | .8 |
Associations between Baseline Reproductive Hormones and Longitudinal Change in MMSE across Baseline, 2-Year, and 5-Year Followup
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.018 | .5 | −0.003 | .9 | 0.012 | .7 |
| DHT | 0.292 | .05 | 0.203 | .2 | 0.236 | .2 |
| SHBG | −0.004 | .08 | 0.002 | .3 | 0.003 | .2 |
| E2 | 0.005 | .4 | 0.004 | .5 | 0.006 | .3 |
| E1 | 0.006 | .07 | 0.004 | .2 | 0.005 | .08 |
| FSH | −0.008 | .04 | −0.001 | .9 | −0.001 | .9 |
| LH | −0.019 | .002 | −0.004 | .5 | −0.003 | .6 |
| cFT | 0.002 | .3 | −0.001 | .7 | 0.0004 | .8 |
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.018 | .5 | −0.003 | .9 | 0.012 | .7 |
| DHT | 0.292 | .05 | 0.203 | .2 | 0.236 | .2 |
| SHBG | −0.004 | .08 | 0.002 | .3 | 0.003 | .2 |
| E2 | 0.005 | .4 | 0.004 | .5 | 0.006 | .3 |
| E1 | 0.006 | .07 | 0.004 | .2 | 0.005 | .08 |
| FSH | −0.008 | .04 | −0.001 | .9 | −0.001 | .9 |
| LH | −0.019 | .002 | −0.004 | .5 | −0.003 | .6 |
| cFT | 0.002 | .3 | −0.001 | .7 | 0.0004 | .8 |
The associations between baseline reproductive hormones and clinically significant cognitive decline over 5-years followup (data not shown) were consistent with the GEE model (see Table 2). The univariate logistic regression model (data not shown) revealed that men with high LH were more likely to have a clinically significant cognitive decline (odds ratio [OR], 0.76; 95% confidence interval [CI], 0.64–0.91). However, this statistically significant association between LH and cognitive decline was nullified by adjustment for potential confounders. No associations were observed between baseline androgen (T, DHT, cFT), estrogen (E2, E1), FSH, or SHBG levels and cognitive decline over time in either univariate or multivariate-adjusted models.
The associations between longitudinal changes in reproductive hormones and in MMSE across baseline, 2-year, and 5-year followups are shown in Table 3. Univariate analyses revealed positive associations for T (β = 0.077; P = .01), DHT (β = 0.426; P = .02), E1 (β = 0.009; P = .004), and cFT (β = 0.007; P = .002), and a negative association for SHBG (β = −0.005; P = .02) with decline in MMSE across the three time points. In the multivariate model adjusted for confounders (age, BMI, smoking status, years of education, and depression score), only the androgens, T (β = 0.067; P = .03), DHT (β = 0.394; P = .04), cFT (β = 0.005; P = .02), and E1 (β = 0.009; P = .002) remained statistically significantly predictors. Therefore, even after multivariate adjustment for confounders, men with declining serum T, DHT, cFT, and E1 (but not other hormones) over the follow-up period were more likely to have cognitive decline.
Associations between Longitudinal Change in Reproductive Hormones and MMSE across Baseline, 2-Year, and 5-Year Followup
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.077 | .01 | 0.059 | .04 | 0.067 | .03 |
| DHT | 0.426 | .02 | 0.366 | .04 | 0.394 | .04 |
| SHBG | −0.005 | .02 | −0.0004 | .9 | −0.00007 | .9 |
| E2 | −0.005 | .2 | −0.007 | .1 | −0.006 | .2 |
| E1 | 0.009 | .004 | 0.008 | .005 | 0.009 | .002 |
| FSH | −0.008 | .01 | −0.002 | .5 | −0.002 | .4 |
| LH | −0.018 | .0003 | −0.005 | .3 | −0.005 | .3 |
| cFT | 0.007 | .002 | 0.005 | .03 | 0.005 | .02 |
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.077 | .01 | 0.059 | .04 | 0.067 | .03 |
| DHT | 0.426 | .02 | 0.366 | .04 | 0.394 | .04 |
| SHBG | −0.005 | .02 | −0.0004 | .9 | −0.00007 | .9 |
| E2 | −0.005 | .2 | −0.007 | .1 | −0.006 | .2 |
| E1 | 0.009 | .004 | 0.008 | .005 | 0.009 | .002 |
| FSH | −0.008 | .01 | −0.002 | .5 | −0.002 | .4 |
| LH | −0.018 | .0003 | −0.005 | .3 | −0.005 | .3 |
| cFT | 0.007 | .002 | 0.005 | .03 | 0.005 | .02 |
Associations between Longitudinal Change in Reproductive Hormones and MMSE across Baseline, 2-Year, and 5-Year Followup
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.077 | .01 | 0.059 | .04 | 0.067 | .03 |
| DHT | 0.426 | .02 | 0.366 | .04 | 0.394 | .04 |
| SHBG | −0.005 | .02 | −0.0004 | .9 | −0.00007 | .9 |
| E2 | −0.005 | .2 | −0.007 | .1 | −0.006 | .2 |
| E1 | 0.009 | .004 | 0.008 | .005 | 0.009 | .002 |
| FSH | −0.008 | .01 | −0.002 | .5 | −0.002 | .4 |
| LH | −0.018 | .0003 | −0.005 | .3 | −0.005 | .3 |
| cFT | 0.007 | .002 | 0.005 | .03 | 0.005 | .02 |
| Hormone | Unadjusted | Plus Age | Plus Education, BMI, Smoking, Depression | |||
|---|---|---|---|---|---|---|
| β | P | β | P | β | P | |
| T | 0.077 | .01 | 0.059 | .04 | 0.067 | .03 |
| DHT | 0.426 | .02 | 0.366 | .04 | 0.394 | .04 |
| SHBG | −0.005 | .02 | −0.0004 | .9 | −0.00007 | .9 |
| E2 | −0.005 | .2 | −0.007 | .1 | −0.006 | .2 |
| E1 | 0.009 | .004 | 0.008 | .005 | 0.009 | .002 |
| FSH | −0.008 | .01 | −0.002 | .5 | −0.002 | .4 |
| LH | −0.018 | .0003 | −0.005 | .3 | −0.005 | .3 |
| cFT | 0.007 | .002 | 0.005 | .03 | 0.005 | .02 |
Although there were no differences in either serum T or cFT levels between men with dementia and men with normal cognition at baseline (data not shown), men who had dementia at baseline experienced significantly greater decline in serum T (P = .04, repeated measures ANOVA), but not in cFT (P = .29, repeated measures ANOVA) over the follow-up period (see Figure 1). The declines in serum T (P = .08, repeated measures ANOVA) and cFT (P = .38, repeated measures ANOVA) were no different in men with MCI than in men with normal cognition. Likewise, linear regression analyses revealed similar findings in both univariate- and multivariate-adjusted models in which men with dementia were more likely to have decline in serum T (OR, 1.08; 95% CI, 1.00–1.16), but not for cFT (OR, 1.07; 95% CI, 0.99–1.10) (data not shown). There was no difference in the change over time for the other studied hormones (DHT, E2, E1, FSH, LH) and SHBG between dementia, MCI, and normal cognition.
Multivariate-adjusted change in T, DHT, E2, and E1 over the follow-up period plotted against dementia, mild cognitive impairment (MCI), and normal cognition at baseline.
Discussion
This study provides the first comprehensive examination of the longitudinal associations between serum reproductive hormone levels, cognitive function, and dementia in older men. By using LC-MS steroid analysis, this study is also able to provide a full appraisal of androgen status in men; this appraisal requires concurrent evaluation of circulating levels of both potent androgens T and DHT as well as E2 (22) together with serum LH, FSH, and SHBG. Androgens (T, DHT, cFT), estrogens (E2, E1), FSH, LH, and SHBG at baseline were not predictive of cognitive decline over the 5-year follow-up period. However, declines in serum androgen levels and E1 levels over time were significantly associated with declines in cognitive function. Furthermore, men with dementia at baseline had a greater decline in serum T and cFT but not in other reproductive measures over the follow-up period. Given that decline in both cognition and reproductive hormones are correlated and both associated with aging and that cognitive decline is not a recognized clinical feature of life-long hypogonadism (23) nor is cognitive function significantly improved by T treatment in placebo-controlled clinical trials (24, 25), these findings suggest that low T, DHT, cFT, and E1 may be more likely a consequence of cognitive impairment rather than its cause. The strong correlation makes unlikely the possibility of both hormones and cognitive function being independent consequences of aging or other third factors.
Most previous longitudinal studies have focused only on serum T, which provides an incomplete analysis of male reproductive endocrine status. Those studies reported no association between T and cognitive decline over time. The Osteoporotic Fractures in Men Study (MrOS) (n = 1364), the Health ABC Study (n = 439), the Honolulu-Asia Aging Study (n = 2974), the NHLBI Twin Study (n = 348), and the Baltimore Longitudinal Study of Aging (n = 901) all reported no relationship between low baseline T and cognitive decline over time (2–6) but did not study other related hormonal variables. The Baltimore study did find that men with low cFT had significantly faster rates of decline in visual memory (but not verbal memory) and increased risk of Alzheimer's disease (5, 8). However, the cFT in the Baltimore study was calculated by an equilibrium-binding formula (26). The present study used a more accurate empirical method validated by reference to direct laboratory measurements that avoids the errors of equilibrium binding formulae, which incorporate inaccurate stoichiometry and arbitrary plug-in constants for binding affinities (11, 12, 27, 28). The Healthy Ageing Study in the United Kingdom (n = 257) reported that men with low T had a sharp drop in cognitive function (≥ 4 points on MMSE) (7). Compared with CHAMP, the United Kingdom study had a short follow-up period (2 y) and a different cutoff for clinically significant cognitive decline. None of these studies, except MrOS, have measured serum T by mass spectrometry–based methods. Most studies have used direct sex steroid immunoassays, which has poorer accuracy and method-specific bias, especially at low circulating T levels such as those prevailing among older men as in the CHAMP study (29, 30).
Although we found that baseline reproductive hormone levels did not predict subsequent cognitive decline, we did find a longitudinal relationship between decline in cognitive function and decline in serum T, DHT, and cFT over 5 years. We also found that men with dementia at baseline had greater declines in T over the next 5 years than men with normal cognition or MCI.
The exact biological mechanism for a causal relationship between dementia and changes in androgen status remains unclear due to limited analysis in this field. One possibility is that reduced androgen status in older men may cause detrimental effects on cerebral neuronal function. The wide distribution of androgen receptors throughout the brain is consistent with such effects (31, 32). For example, men with dementia may have impaired hypothalamic function thereby reducing secretion of GnRH and eventually lowering T, DHT, and E2 (33). However this hypothesis is not consistent with the raised LH and FSH we observed, nor with lack of wider decline in hypothalamic function involving other hypothalamic-pituitary hormonal axes.
Alternatively, cognitive decline and/or dementia could also indirectly lead to changes in androgen status. For example, older men and women with dementia are more likely to have reduced physical function (34, 35), which is associated with a lowering of androgen status as partially reflected by serum T in older men (36). Similarly, the cognitive decline may also invoke other nonspecific adaptive responses that impair hypothalamic regulation of reproductive function and androgen status. Therefore, there is the possibility of an indirect effect of poor cognitive function on reduced androgen status over time due to poor physical function and underlying comorbidities.
Placebo-controlled and nonplacebo randomized clinical trials examining T supplementation and cognitive function in older men have revealed inconsistent findings. Recent reviews have concluded that there is no evidence for an improvement in attention or executive function after T treatment in either eugonadal or hypogonadal men (24, 25). However, these reviews have shown that some but not all, studies of T in eugonadal men have found positive effects on memory and visuospatial functions (24, 25), whereas in hypogonadal men, those studies of T supplementation have found no significant effect on memory and visuospatial functions (24, 25).
We found that changes over time in serum E1, a circulating precursor of E2, was associated with subsequent cognitive decline. When compared with E2, serum E1 is usually considered to have no clinical significance in men due to its minimal intrinsic estrogenic bioactivity (37, 38). However, it could function as a biological buffer as an immediate precursor to E2. Our results suggest that more studies should be conducted to better define the mechanistic and prognostic significance of change in serum E1 as a novel biomarker of cognitive health. In previous analyses based on the same CHAMP cohort, we have reported that men with low serum E1 were more likely to have a subsequent deterioration in their general health status and decline in functional ability (39, 40) and the Framingham Heart Study revealed a significant independent relationship between E1 and diabetes and cardiovascular disease (41).
We found no relationships between E2 and cognition, consistent with all previous longitudinal studies except the Honolulu–Asia Aging Study, which reported that older men with high calculated “bio-available” E2 levels were at increased risk for cognitive decline and Alzheimer's disease (4). The use of calculated “bio-available” E2 in the Honolulu study is questionable both for its uncertain biological interpretation for estrogen action as well as limited validation as an accurate estimate of laboratory measured “bio-available” E2. Calculation of derived fractions of E2 are modeled on analogous calculation for derived fractions of T but have even less validation against equilibrium dialysis (for “free”) (42) or ammonium sulfate precipitation (for “bio-available”) methods (43, 44).
A major strength of our study was that we were able to use longitudinal data, which enabled us to examine the longitudinal relationships between change in reproductive hormones and change in cognitive function over several follow-up time points. Another strength was the use of the LC-MS/MS, the current gold standard for steroid assays providing multianalyte steroid profiling as well as a more accurate formula for cFT (45, 46). Direct immunoassay methods that do not include extraction and chromatography have poor accuracy in measuring low levels of sex steroids, which is particularly problematic for measuring circulating T, DHT, E2, and E1 in older men (10, 26, 47, 48). Our study also included detailed assessment of cognitive status at baseline, which allowed us to make clinical diagnoses of dementia and MCI. Unfortunately, we did not have clinical diagnoses of dementia at follow-up assessments. A further strength of CHAMP is that it includes a large and representative group of older Australian men, as demonstrated by similar socio-demographic and health characteristics in CHAMP men compared with older men in the nationally representative Men in Australia Telephone Survey (MATeS) study (49).
A significant limitation of our study is the 20% loss to followup from baseline to 2-year and a further 30% loss to followup from 2 to 5 years. However, loss to follow up in cohort studies of older people is inevitable because of the high mortality rate, which accounted for nearly 35% of the loss in our cohort. Furthermore, the GEE analysis methodology is robust with regard to data missing at random in longitudinal analyses (50). Our use of single morning fasting blood samples served to minimize any possible variations due to diurnal variation (51) while seasonal variation of male reproductive hormones is negligible (52). An unavoidable limitation is that, although samples from the different study waves were assayed together at similar times in storage, all the study samples from individuals could not be either stored for same length of time or assayed together. Nevertheless, steroids are highly stable in long-term frozen storage and to repeated freeze/thaw cycles (53–57) so differences in when study wave batches were assayed is unlikely to create systemic errors.
This study has revealed longitudinal temporal relationships between androgen status reflected in serum T, DHT, cFT, and E1, and cognitive decline over multiple time points in older men. Our findings raise the possibility that decline in androgen status may be the consequence rather than the cause of cognitive decline consistent with the absence of significant cognitive decline in life-long hypogonadism as well as randomized placebo-controlled trials that find little or no benefits of T treatment on cognitive function in older men. However, this study is observational, it could not fully rule out the possibility that cognitive decline and reproductive hormone decline are both independent result of aging. Nevertheless, our study suggests that reduced cognitive function may be an important but under recognized contributor to the lowering of androgen status in older men.
Acknowledgments
The Concord Health and Ageing in Men Project study is funded by the National Health and Medical Research Council Project Grant No. 301916, Sydney Medical School Foundation and Ageing, and Alzheimer's Institute. B.H is funded by the Sydney Medical School Foundation.
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. performed the analyses and wrote the manuscript; R.G.C. and D.J.H. wrote portions of the manuscript; F.M.B., V.N., D.G.L.C., M.J.S., and L.M.W. reviewed the manuscript and contributed to discussion.
Disclosure Summary: R.G.C. received an honorarium from Eli Lilly Australia for an education event. B.H., L.M.W., F.M.B., V.N., D.G.L.C., M.J.S., and D.J.H. have nothing to disclose.
Abbreviations
- BMI
body mass index
- cFT
calculated free T
- CHAMP
Concord Health and Ageing in Men Project
- CI
confidence interval
- CV
coefficient of variation
- E1
estrone
- E2
estradiol
- GEE
generalized estimating equations
- LC-MS/MS
liquid chromatography–tandem mass spectrometry
- MCI
mild cognitive impairment
- MMSE
Mini Mental Status Examination
- MrOS
Osteoporotic Fractures in Men Study
- OR
odds ratio
- QC
quality control
- RUDAS
Rowland Universal Dementia Assessment Scale.
