Abstract
Purpose
The role of testosterone (T) deficiency (T ≤ 300 ng/dL) and hypercholesterolemia (total cholesterol ≥ 240mg/dL) in the risk of all-cause cardiovascular diseases (CVD) and cancer mortality among a nationally representative sample of non-Hispanic White (NHW), non-Hispanic Black (NHB) and Hispanic men remains poorly understood.
Methods
Data included a full sample (NHANES 1988–1991, 1999–2004, 2011–2014) and subset sample (excluding 2011–2012, no estradiol and SHBG levels available) of 5379 and 3740 men, respectively. Participants were aged ≥ 20 y with serum T and cholesterol data (median follow-up 7.6 years). Weighted multivariable-adjusted Cox proportional hazards models were used in this study.
Results
In the overall population of full and subset samples, hypercholesterolemia was inversely associated with all-cause (HR = 0.76, 95% CI, 0.63–0.91) and cancer mortality (HR = 0.56, 95% CI, 0.34–0.90). Similar findings were observed among NHW men, but higher T levels increased the risk of CVD mortality in the subset sample (T3 vs T1, Ptrend = 0.02). Among NHB men in the full and subset samples, T deficiency increased the risk of CVD mortality, but T3 vs. T1 decreased it (Ptrend = 0.03), and hypercholesterolemia decreased cancer mortality. Among Hispanic men in the full and subset samples, T deficiency increased, and hypercholesterolemia decreased the risk of CVD mortality.
Conclusion
Hypercholesterolemia was inversely associated with cancer mortality. However, higher levels of T were positively associated with CVD mortality among NHW and were inversely associated with CVD mortality among NHB and Hispanic men. Larger prospective studies are warranted to clarify the underlying relationship between T and cholesterol with mortality among racial and ethnic groups.
Keywords: Serum testosterone, Cardiovascular, Cancer, Mortality
Introduction
Testosterone (T) is linked to numerous positive physiologic effects, including improved sexual function, physical performance, muscle strength, lean body mass, and cognitive function [1, 2]. Low levels of T or T deficiency (total T ≤ 300ng/dL) [3] are associated with cardiovascular diseases (CVD), [4] prostate cancer (PCa), [5, 6] and all-cause mortality. [7, 8] Over the last decade, studies have shown that as many as 38.7% of men over 45 years old demonstrate low T or T deficiency. [9, 10] In the USA, it has been reported that there are approximately 2.4 million men in the age groups 40–69 years old with T deficiency. [11] However, an even greater concern is the projection that by 2025, approximately 6.5 million American men aged 30 to 80 will suffer from T deficiency, partly due to the increasing older population. [10] Although the above studies have provided valuable insight, they often had small samples of understudied and underserved populations with limited generalizability to non-Hispanic Black (NHB) and Hispanic men.
In parallel to the trends of T deficiency, there are approximately 12.1 million adult men ≥ 20 years of age with hypercholesterolemia (total cholesterol ≥ 240ng/dL) in the USA. [12] Total cholesterol has previously been linked with PCa and CVD. [13, 14] These earlier studies have also included small numbers of NHB and Hispanic men. It is moreover suggested that the obesity epidemic and aging of the US population have played a significant role in increasing prevalence of T deficiency and hypercholesterolemia.
Due to the rising prevalence of T deficiency and hypercholesterolemia, it is possible that their treatment with testosterone therapy (TTh) and statins, respectively, have also increased. [15, 16] A previous US commercial claims database of 189,491 men aged 40–69 years stated that 21% of TTh users have also reported statin use. [17] Furthermore, we previously reported an inverse independent and joint effect of TTh [17] and statin use on PCa in a case-only analysis [18] and prospective-analysis. [19] These two latter studies showed racial and ethnic differences in the combination of TTh and statins in relation to PCa. There is a potential biological plausibility in the interaction between total T and cholesterol, and their treatments, TTh, and statins given that cholesterol is a required intermediate precursor in steroidogenesis. [20]
Therefore, the objectives of this study are to investigate the association of T deficiency and hypercholesterolemia with all-cause, CVD, and cancer mortality and to examine whether these associations vary among a representative sample of US non-Hispanic White (NHW), NHB, and Hispanic men.
Methods
Study population
The National Health and Nutrition Examination Survey (NHANES) is a program of studies undertaken by the National Center for Health Statistics (NCHS) of the US Centers for Disease Control and Prevention (CDC) to assess the health and nutritional status of adults and children in the US. [21] NHANES is a cross-sectional study that uses a multistage, stratified, and clustered probability sampling strategy in which Hispanic-Americans, non-Hispanic Blacks, and the elderly are oversampled to ensure adequate sample size and to represent the total US civilian, non-institutionalized population. [22] Details of the survey design, methods, and data collections are available on the NHANES website (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx. Accessed Jan. 2020).
This study included data from male participants in the 1988–1991 (NHANES III-Phase 1), 1999–2004, and 2011–2014 NHANES cycles. Endogenous sex hormones were measured from stored surplus serum samples by the study investigators in 5379 males aged ≥ 20y, who were stratified in a random sample of participants in the morning examination sessions of each cycle. Morning sample participants were chosen to reduce extraneous variation due to diurnal production of hormones. Serum total cholesterol and triglycerides were measured enzymatically; LDL-C was calculated using the Friedewald equation if the triglyceride level was ≤ 4.5mmol/L(400mg/dL). [23]
Male participants younger than 20 y were not included in this study. We excluded participants with a self-reported history of cardiovascular disease and prostate cancer because certain treatments may affect hormone levels. We also excluded men with prevalent type 2 diabetes (defined by having fasting plasma glucose of ≥ 126mg/dl, [24] diabetes medication, or being “told by a doctor you have diabetes or sugar diabetes”), missing cholesterol data, missing information on covariates, missing sex hormone measurements, and having extreme hormone measurements based on identification of outliers from a graphical representation of the distribution of each hormone. The following threshold points were used to determine extreme measurements based on a visual inspection of the distribution: testosterone = 5000ng/dL (n = 3), estradiol = 360 pg/mL (n = 4), and sex hormone binding globulin (SHBG) > 170 nmol/L (n = 6), leaving a final full sample of 5379 men. The NHANES 2011–2012 cycle did not include estradiol and SHBG, and a subset sample of 1988–1991, 1999–2004, and 2013–2014 cycles was created to be able to adjust for estradiol and SHBG (n = 3740).
Assessment of testosterone, estradiol, and SHBG
Details on the blood draw, processing, storage, and shipping methods are provided elsewhere. [22] In NHANES 1988–1991 and 1999–2004, testosterone, estradiol, and SHBG were measured using the electrochemiluminescence immunoassays on the 2010 Elecsys system (Roche Diagnostics, Laval, QC, Canada; and Roche Diagnostics, Indianapolis, IN, USA). The laboratory technicians were blinded to the participant characteristics of the samples. The lower limits of detection of the assays were 2 ng/dL for testosterone, 5 pg/mL for estradiol, and 3 nmol/L for SHBG. One sample had a concentration below the limit of detection for testosterone and 10 samples for estradiol, which were assigned to half the limit of detection. Twenty-one samples were assayed in duplicate for quality control purposes, and the coefficients of variation were 4.8% for testosterone, 21.4% for estradiol, and 5.6% for SHBG. In NHANES 2011–2014, testosterone was measured with LC-MS/MS and was isolated from 100 μL serum by 2 serial liquid-liquid extraction steps and quantified with [13C] stable isotope-labeled testosterone as the internal standard. The lower limit of detection was 0.3ng/dL. SHBG was quantified based on the reaction of SHBG with immunoantibodies and chemoluminescence measurements of the reaction products that occur after two incubation periods and subjected to a magnetic field. Total T below or equal to 300 ng/dL was considered T deficiency. [3] In this study, the term T deficiency does not imply that a deficit needs to be replaced. Total T was also categorized in tertiles to compare risk of mortality between the highest to lowest tertile category of total T. Calculated free testosterone was derived using published formulas with information for total testosterone, total estradiol, SHBG, and serum albumin (measured in NHANES). [25, 26]
Assessment of serum cholesterol and cholesterol-lowering medications
Serum total cholesterol and triglycerides were measured enzymatically; LDL-C was calculated using the Friedewald equation if the triglyceride level was ≤ 4.5mmol/L (400 mg/dL). [23] Standardization of serum lipid measurement was performed according to the criteria of CDC’s Lipid Standardization Program. [27] Hypercholesterolemia was defined with total cholesterol ≥ 240 ng/dL. [12] In the in-person interviews, participants were asked “Have you ever been told by a doctor or other health professional that your blood cholesterol level was high?” If the answer was “yes,” then they were asked “Because of your high cholesterol, have you ever been told by a doctor or other health professional to take prescribed medicine?” If the answer was “yes,” they were then asked “To lower your blood cholesterol, are you now following this advice to take prescribed medicine?” From the positive answers (“yes”) of these questions, we categorized participants as self-reported cholesterol lipid-lowering drugs and prescribed cholesterol lipid-lowering drugs in this analysis. [28]
Outcome ascertainment (mortality follow-up)
NHANES 1988–1991, 1999–2004, and 2011–2014 is a cross-sectional study; however, the NCHS created the Linked Mortality File that provides mortality follow-up data that matched records from NHANES waves with data in the National Death Index (NDI) as of December 31, 2015; the mortality status of > 98% of subjects was correctly classified. [29] The method of probabilistic matching was used to link these files to ascertain vital status and cause of death. [29] The NDI is a NCHS-centralized database of all US deaths beginning in 1979. For some participants, other sources of mortality information included the Social Security Administration, the Centers for Medicare and Medicaid Services, and death certificates. The Linked Mortality File included information on mortality from all causes, nine leading causes, and a residual category of mortality in the US population that was based on International Classification of Diseases (ICD)-9th and 10th cause-of-death codes. Years 1988–1991 focused on ICD-9 codes and years 1999–2004 and 2011–2014 on ICD-10 codes. Cause-specific mortality was ascertained for CVD (ICD-9th codes 390–434 and 436–459; ICD-10th codes I00-I99) and cancer (ICD-9th codes 140–208; ICD-10th codes C00-C97). The numbers of deaths due to all-cause, CVD, and cancer were 665, 160, and 156, respectively.
Statistical analyses
Sampling weights were applied to account for selection probabilities, over-sampling, non-response, and differences between the sample and the total US population. The person-time of follow-up was computed from the date of the examination in the Mobile Examination Center (1988–1991, 1999–2004, and 2011–2014) to the last date that the person was known to be alive or to December 31, 2015. The assessment of covariates has been previously described. [11] In brief, for descriptive purposes, we compared the distribution of sociodemographic and lifestyle factors by categories of serum testosterone and total cholesterol using Student’s t-statistic for means from continuous variables and Chi-square for categorical factors. We used Cox proportional hazards regression to estimate the age-race/ethnicity and multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for all-cause, CVD and cancer mortality associated independently with total testosterone and cholesterol and their categories. Multivariable analyses were conducted in two models, namely, full sample (1988–1991, 1999–2004, and 2011–2014) and subset sample (1988–1991, 1999–2004, and 2013–2014). Full sample models were adjusted for age, race and ethnicity, smoking status, history of hypertension, physical activity, alcohol consumption, BMI, and mutual adjustment between T deficiency and hypercholesterolemia. Similarly, subset sample models were adjusted for the aforementioned confounders plus estradiol and SHBG (not included in the NHANES 2011–2012 wave). Because total testosterone and estradiol have previously been correlated [30] and SHBG is a major carrier of circulating testosterone, they were included in the multivariable models. To test for a linear trend across categories of total testosterone and calculated free testosterone, we modeled categories of total and calculated free testosterone as continuous variables using the median for each category.
Stratified and multivariable analyses were conducted by race and ethnicity (NHW, NHB, and Hispanics), and hypercholesterolemia because these variables are known to modify testosterone levels. All p-values were two-sided; alpha = 0.05 was considered the cut-off for statistical significance. All statistical analyses were performed using STATA version 12.0 (College Station, TX, USA).
Results
In the full and subset samples, we identified 5379 and 3740 men, respectively. A total of 2516 men were NHWs, 1370 NHBs, and 1493 Hispanics. Mean age was 46.64 and the median follow-up time from baseline to mortality or end of study (12/31/2015) was 7.6 years. The distribution of baseline characteristics by categories of total testosterone and cholesterol in both populations, full and subset samples, after applying sample weights, is reported in Tables 1 and 2. Supplemental Table 1 displays the distribution of baseline characteristics by all-cause, CVD, and cancer mortality. Approximately 33.30% of men were included in the low testosterone category (tertile 1), 33.6% in tertile 2, and 33.10% in tertile 3, while 34.20% had hypercholesterolemia (Table 1). Similar results were observed in the subset sample (Table 2). Frequencies of mortality included 12.31% all-cause, 3.30% CVD, and 3.20% cancer mortality (Supplemental Table 1). Compared to men in testosterone tertile 1, men in tertiles 2 and 3 were less likely to have hypertension and be obese but more likely to be younger, to be NHW, to smoke, to be physically active, and to consume alcohol (Table 1). Compared to men with no hypercholesterolemia, men with hypercholesterolemia were less likely to smoke but they were more likely to be older, to be NHW, to have hypertension, to be physically active, and to have higher BMI. In general, similar results were observed in the subset sample analysis.
Table 1.
Selected characteristics of the US population of adult men 20 y and older by categories of levels of total testosterone and total cholesterol in NHANES 1988–1991, 1999–2004, and 2011–2014
| NHANES 1988–1991, 1999–2004, and 2011–2014 | Testosterone T1 (N = 1791) |
Testosterone T2 (N = 1808) |
Testosterone T3 (N = 1780) |
Low cholesterol (N = 3542) |
High cholesterol (N = 1837) |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| n | % | n | % | n | % | p-value | n | % | n | % | P-value | |
|
| ||||||||||||
| Age, mean (SE) | 49.82 | 0.38 | 46.73 | 0.39 | 43.37 | 0.39 | <0.01 | 43.39 | 0.28 | 52.92 | 0.34 | <0.01 |
| Race/Ethnicity | <0.01 | |||||||||||
| NH White | 850 | 47.46 | 889 | 49.17 | 777 | 43.65 | - | 1581 | 72.20 | 935 | 79.36 | |
| NH Black | 444 | 24.79 | 414 | 22.90 | 512 | 28.76 | 0.16 | 976 | 12.04 | 394 | 7.76 | |
| Hispanic | 497 | 27.75 | 505 | 27.93 | 491 | 27.58 | 0.16 | 985 | 15.76 | 588 | 12.87 | |
| Smoking status | <0.01 | |||||||||||
| Never smoked | 875 | 48.94 | 851 | 47.09 | 685 | 38.53 | - | 1653 | 47.05 | 758 | 46.13 | |
| Former smoker | 571 | 31.94 | 530 | 29.33 | 398 | 22.38 | 0.77 | 840 | 24.79 | 659 | 27.58 | |
| Current smoker | 342 | 19.13 | 426 | 23.57 | 695 | 39.09 | <0.01 | 1044 | 28.16 | 419 | 26.30 | |
| Hypertension | <0.0 | |||||||||||
| No | 923 | 53.02 | 1098 | 62.42 | 1251 | 72.10 | - | 2411 | 75.12 | 861 | 52.05 | |
| Yes | 818 | 46.98 | 661 | 37.58 | 484 | 27.90 | <0.01 | 1030 | 24.88 | 933 | 47.95 | |
| Physically active | 0.06 | |||||||||||
| No | 612 | 34.17 | 582 | 32.19 | 626 | 35.17 | - | 1239 | 32.15 | 581 | 28.33 | |
| Yes | 1179 | 65.83 | 1226 | 67.81 | 1154 | 64.83 | 0.06 | 2303 | 67.85 | 1256 | 71.67 | |
| Alcohol, mean (SE) | 13.67 | 0.79 | 15.7 | 0.86 | 18.61 | 1.09 | 0.01 | 16.9 | 0.7 | 14.32 | 0.78 | 0.48 |
| BMI, mean (SE) | 31.09 | 0.16 | 28.09 | 0.11 | 25.32 | 0.10 | <0.01 | 27.67 | 0.09 | 29.1 | 0.12 | <0.01 |
Table 2.
Selected characteristics of the US population of adult men 20 y and older by categories of levels of total testosterone and total cholesterol in NHANES 1988–1991, 1999–2004, and 2013–2014
| NHANES 1988–1991, 1999–2004, 2013–2014 | Testosterone T1 (N = 1243) |
Testosterone T2 (N = 1241) |
Testosterone T3 (N = 1255) |
Low cholesterol (N = 2506) |
High cholesterol (N = 1234) |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| n | % | n | % | n | % | p-value | n | % | n | % | p-value | |
|
| ||||||||||||
| Age, mean (SE) | 59.93 | 0.50 | 47.11 | 0.47 | 42.66 | 0.44 | <0.01 | 43.71 | 0.34 | 52.4 | 0.42 | <0.01 |
| Race/Ethnicity | <0.01 | |||||||||||
| NH White | 546 | 50.00 | 646 | 51.23 | 597 | 43.04 | - | 1139 | 73.34 | 650 | 80.12 | |
| NH Black | 236 | 21.61 | 257 | 20.38 | 390 | 28.12 | 0.04 | 643 | 11.89 | 240 | 7.78 | |
| Hispanic | 310 | 28.39 | 358 | 28.39 | 400 | 28.84 | 0.49 | 724 | 14.78 | 344 | 12.11 | |
| Smoking status | <0.01 | |||||||||||
| Never smoked | 517 | 47.39 | 591 | 46.87 | 527 | 38.02 | - | 1124 | 44.88 | 511 | 43.75 | |
| Former smoker | 387 | 35.47 | 381 | 30.21 | 298 | 21.5 | 0.29 | 621 | 25.61 | 445 | 32.75 | |
| Current smoker | 187 | 17.14 | 289 | 22.92 | 561 | 40.48 | <0.01 | 759 | 29.50 | 278 | 23.5 | |
| Hypertension | <0.01 | |||||||||||
| No | 540 | 50.99 | 779 | 63.49 | 996 | 73.67 | - | 1708 | 75.16 | 607 | 52.83 | |
| Yes | 519 | 49.01 | 448 | 36.51 | 356 | 26.33 | <0.01 | 724 | 24.84 | 599 | 47.17 | |
| Physically active | 0.08 | |||||||||||
| No | 394 | 36.08 | 446 | 35.37 | 537 | 38.72 | - | 964 | 36.04 | 413 | 31.21 | |
| Yes | 698 | 63.92 | 815 | 64.63 | 850 | 61.58 | 0.08 | 1542 | 63.96 | 821 | 68.79 | |
| Alcohol, mean (SE) | 13.66 | 0.94 | 14.95 | 0.97 | 17.76 | 1.17 | 0.02 | 16.44 | 0.81 | 13.99 | 0.85 | 0.5 |
| BMI, mean (SE) | 30.86 | 0.20 | 28.16 | 0.13 | 25.24 | 0.11 | <0.01 | 27.35 | 0.11 | 28.88 | 0.15 | <0.01 |
Table 3 illustrates the multivariable-adjusted associations of total (T), T deficiency, calculated free-testosterone, hypercholesterolemia, and lipid-lowering medications with all-cause, CVD, and cancer mortality in the full sample and subset sample. No significant associations were observed between categories of total T (T3 vs. T1), T deficiency, and calculated-free T (T3 vs. T1) with all-cause, CVD, and cancer mortality. In the full sample, hypercholesterolemia was inversely associated with all-cause (HR = 0.76, 95% CI, 0.63–0.91) and cancer mortality (HR = 0.56,95%CI, 0.34–0.90). Similar inverse associations were observed in the subset sample for all-cause (HR = 0.78, 95% CI, 0.64–0.95) and cancer mortality (HR = 0.60, 95% CI, 0.36–0.98) after adjusting for estradiol, SHBG, and other confounders. No association was identified among men who took lipid-lowering medications.
Table 3.
Multivariable associations of total testosterone, hypercholesterolemia, and self-reported and prescribed lipid-lowering medication with all-cause, cardiovascular (CVD) and cancer mortality: NHANES 1988–2014 linked with 2015 mortality data
| Mortality full sample†a
N = 5,379 |
Mortality subset sample†b (Adjusted for estradiol and SHBG) N = 3,740 |
||||||
|---|---|---|---|---|---|---|---|
| All-cause | CVD | Cancer | All-cause | CVD | Cancer | ||
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | ||
|
| |||||||
| T deficiency (≤ 300 ng/dL), N = 1137 |
T deficiency (≤ 300 ng/dL), N = 659 |
||||||
| No | 1.0 | 1.0 | 1.0 | No | 1.0 | 1.0 | 1.0 |
| Yes | 0.90 (0.64, 1.26) |
1.11 (0.62, 2.00) |
0.92 (0.39, 2.14) |
Yes | 0.88 (0.61, 1.26) |
0.93 (0.50, 1.73) |
1.22 (0.51, 2.90) |
| Total T (ng/dL) | Total T (ng/dL) | ||||||
| T1 (100–354) (N = 1791) | 1.00 | 1.00 | 1.00 | T1 (100–374) (N = 1243) | 1.00 | 1.00 | 1.00 |
| T2 (355–519) (N = 1808) | 0.91 (0.68, 1.23) |
1.23 (0.70, 2.17) |
0.92 (0.44, 1.90) |
T2 (375–539) (N = 1242) | 0.91 (0.67, 1.25) |
1.58 (0.85, 2.94) |
0.74 (0.36, 1.51) |
| T3 (≥ 520) (N = 1780) | 1.11 (0.81, 1.53) |
1.19 (0.59, 2.39) |
1.74 (0.86, 3.50) |
T3 (≥ 540) (N = 1255) | 1.06 (0.75, 1.50) |
2.28 (0.86, 6.04) |
1.11 (0.54, 2.32) |
| P trend | 0.44 | 0.63 | 0.07 | P trend | 0.71 | 0.10 | 0.62 |
| Calculated free-T (pg/dL) | Calculated free-T (pg/dL) | ||||||
| T1 (≤ 15.5) (N = 1162) | 1.00 | 1.00 | 1.00 | T1 (≤ 15.5) (N = 1162) | 1.00 | 1.00 | 1.00 |
| T2 (15.6–22.3) (N = 1165) | 1.09 (0.81, 1.48) |
1.25 (0.78, 2.01) |
1.10 (0.47, 2.56) |
T2 (15.6–22.3) (N = 1,165) | 1.04 (0.76, 1.43) |
1.51 (0.84, 2.68) |
0.96 (0.42, 2.22) |
| T3 (22.3–24.70) (N = 1164) | 1.31 (0.99, 1.74) |
1.06 (0.46, 2.42) |
1.74 (0.82, 3.70) |
T3 (22.3–24.70) (N = 1164) | 1.19 (0.87, 1.63) |
1.76 (0.57, 5.48) |
1.29 (0.64, 2.61) |
| p trend | 0.06 | 0.86 | 0.12 | p trend | 0.27 | 0.29 | 0.43 |
| Hypercholesterolemia, ng/mL (N = 1837) | Hypercholesterolemia, ng/mL (N = 1234) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.76 (0.63, 0.91) |
1.05 (0.58, 1.92) |
0.56 (0.34,0.90) |
Yes | 0.78 (0.64, 0.95) |
1.08 (0.60, 1.92) |
0.60 (0.36, 0.98) |
| Self-reported lipid-lowering (N = 557) | 0.66 (0.42, 1.05) |
0.82 (0.34, 1.98) |
0.57 (0.19, 1.74) |
Self-reported lipid-lowering (N = 342) | 0.76 (0.47, 1.23) |
0.81 (0.34, 1.88) |
0.86 (0.25, 2.95) |
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.66 (0.42, 1.05) |
0.82 (0.34, 1.98) |
0.57 (0.19, 1.74) |
Yes | 0.76 (0.47, 1.23) |
0.81 (0.34, 1.88) |
0.86 (0.25, 2.95) |
| Prescribed lipid-lowering (N = 617) | Prescribed lipid-lowering (N = 373) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.74 (0.45, 1.21) |
0.99 (0.41, 2.38) |
0.52 (0.18, 1.52) |
Yes | 0.70 (0.40, 1.22) |
0.82 (0.34, 2.01) |
0.71 (0.22, 2.29) |
Multivariable model adjusted for age, race and ethnicity, smoking status, history of hypertension, physical activity, alcohol consumption, body mass index, and mutual adjustment between sex steroid hormones and hypercholesterolemia
NHANES waves 1988–1991, 1999–2004, 2011–2014
NHANES waves 1988–1991, 1999–2004, 2013–2014
Racial and ethnic stratified multivariable-adjusted associations of total (T), T deficiency, calculated free-testosterone, hypercholesterolemia, and lipid-lowering medications with all-cause, CVD, and cancer mortality are shown in Tables 4, 5, and 6. Among NHW men (Table 4), only in the subset sample did the second and third tertiles of testosterone (T2 and T3 vs T1) increase the risk of CVD-mortality (Ptrend = 0.02). As in Table 3, similar findings were observed in the significant inverse associations between hypercholesterolemia and all-cause and cancer mortality in the full sample (HR = 0.73 and 0.52, respectively) and subset sample (HR = 0.77 and 0.58 [borderline significant]), respectively. No association was identified among men who took lipid-lowering medications.
Table 4.
NHW men: Multivariable associations of total testosterone, hypercholesterolemia, and self-reported and prescribed lipid-lowering medication with all-cause, cardiovascular (CVD) and cancer mortality: NHANES 1988–2014 linked with 2015 mortality data
| Mortality Full-Sample†a (N = 2,516) |
Mortality Subset-Sample†b (Adjusted for estradiol and SHBG) (N = 1789) |
||||||
|---|---|---|---|---|---|---|---|
| All-cause (n = 336) | CVD (n = 83) | Cancer (n = 79) | All-cause (n = 307) | CVD (n = 79) | Cancer (n = 75) | ||
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | ||
|
| |||||||
| T deficiency (≤300 ng/dL), (N = 549) | T deficiency (≤300 ng/dL), (N = 331) | ||||||
| No | 1.0 | 1.0 | 1.0 | No | 1.0 | 1.0 | 1.0 |
| Yes | 0.79 (0.52, 1.20) |
0.88 (0.44, 1.78) |
0.84 (0.29, 2.43) |
Yes | 0.74 (0.46, 1.17) |
0.65 (0.31, 1.35) |
1.12 (0.37, 3.38) |
| Total T (ng/dL) | Total T (ng/dL) | ||||||
| T1 (100–354 ng/dL) (N = 850) | 1.00 | 1.00 | 1.00 | T1 (100–374 ng/dL) (N = 546) | 1.00 | 1.00 | 1.00 |
| T2 (355–519) (N = 889) | 0.89 (0.62, 1.27) |
1.43 (0.72, 2.83) |
0.84 (0.33, 2.13) |
T2(375–539) (N = 646) | 0.93 (0.65, 1.34) |
2.22 (1.03, 4.79) |
0.70 (0.29, 1.68) |
| T3 (≥ 520) (N = 777) | 1.26 (0.87, 1.82) |
1.65 (0.70, 3.90) |
1.78 (0.74, 4.27) |
T3 (≥ 540) (N = 597) | 1.30 (0.88, 1.91) |
4.73 (1.37, 16.29) |
1.03 (0.41, 2.58) |
| p trend | 0.17 | 0.26 | 0.12 | p trend | 0.18 | 0.02 | 0.82 |
| Calculated free-T (pg/dL) | Calculated free-T (pg/dL) | ||||||
| T1 (15.5) (N = 596) | 1.00 | 1.00 | 1.00 | T1 (15.5) (N = 596) | 1.00 | 1.00 | 1.00 |
| T2 (15.6–22.3) (N = 580) | 1.02 (0.72, 1.43) |
1.30 (0.72, 2.32) |
1.13 (0.41, 3.08) |
T2 (15.6–22.3) (N = 580) | 0.97 (0.69, 1.37) |
1.76 (0.81, 3.81) |
0.97 (0.37, 2.57) |
| T3 (22.3–24.70) (N = 483) | 1.35 (0.95, 1.91) |
1.12 (0.39, 3.26) |
1.69 (0.62, 4.64) |
T3 (22.3–24.70) (N = 483) | 1.23 (0.84, 1.80) |
2.47 (0.53, 11.37) |
1.15 (0.44, 3.04) |
| P trend | 0.10 | 0.79 | 0.28 | P trend | 0.31 | 0.22 | 0.76 |
| Hypercholesterolemia, ng/mL (N = 935) | Hypercholesterolemia, ng/mL (N = 650) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.73 (0.57, 0.92) |
1.14 (0.58, 2.24) |
0.52 (0.29, 0.92) |
Yes | 0.77 (0.60, 0.98) |
1.17 (0.63, 2.19) |
0.58 (0.32, 1.06) |
| Self-reported lipid-lowering (N = 300) | Self-reported lipid-lowering (N = 205) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.66 (0.38, 1.12) |
0.85 (0.32, 2.25) |
0.51 (0.13, 2.07) |
Yes | 0.79 (0.45, 1.38) |
0.84 (0.35, 2.00) |
0.81 (0.18, 3.57) |
| Prescribed lipid-lowering (N = 332) | Prescribed lipid-lowering (N = 227) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.72 (0.40, 1.30) |
1.07 (0.40, 2.85) |
0.47 (0.13, 1.71) |
Yes | 0.73 (0.39, 1.38) |
0.87 (0.34, 2.21) |
0.76 (0.22, 2.64) |
Multivariable model adjusted for age, race and ethnicity, smoking status, history of hypertension, physical activity, alcohol consumption, body mass index, and mutual adjustment between sex steroid hormones and hypercholesterolemia
NHANES waves 1988–1991, 1999–2004, 2011–2014
NHANES waves 1988–1991, 1999–2004, 2013–2014
ND = not determined due to small sample size
Table 5.
NHW men: Multivariable associations of total testosterone, hypercholesterolemia, and self-reported and prescribed lipid-lowering medication with all-cause, cardiovascular (CVD) and cancer mortality: NHANES 1988–2014 linked with 2015 mortality data
| Mortality full sample†a (N = 1370) |
Mortality subset sample†b (Adjusted for estradiol and SHBG) (N = 833) |
||||||
|---|---|---|---|---|---|---|---|
| All-cause (n = 167) | CVD (n = 47) | Cancer (n = 43) | All-cause (n= 152) | CVD (n = 42) |
Cancer (n = 39) | ||
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | ||
|
| |||||||
| T deficiency (≤300 ng/dL), (N = 275) | T deficiency (≤300 ng/dL), (N = 137) | ||||||
| No | 1.0 | 1.0 | 1.0 | No | 1.0 | 1.0 | 1.0 |
| Yes | 1.45 (0.86, 2.45) |
2.89 (1.03, 8.15) |
1.64 (0.43, 6.19) |
Yes | 1.66 (1.00, 2.75) |
3.22 (1.10, 9.39) |
2.51 (0.58, 10.82) |
| Total T (ng/dL) | Total T (ng/dL) | ||||||
| T1 (100–354 ng/dL) (N = 444) | 1.00 | 1.00 | 1.00 | T1 (100–374 ng/dL) (N = 236) | 1.00 | 1.00 | 1.00 |
| T2 (355–519) (N = 414) | 1.15 (0.71, 1.84) |
0.84 (0.28, 2.57) |
1.10 (0.28, 4.36) |
T2 (375–539) (N = 257) | 1.00 (0.62, 1.64) |
0.64 (0.14, 2.96) |
0.88 (0.18, 4.21) |
| T3 (≥ 520) (N = 512) | 0.78 (0.49, 1.23) |
0.35 (0.13, 0.99) |
1.23 (0.42, 3.58) |
T3 (≥ 540) (N = 390) | 0.59 (0.37, 0.94) |
0.23 (0.05, 1.02) |
1.03 (0.29, 3.68) |
| p trend | 0.13 | 0.03 | 0.63 | p trend | 0.01 | 0.03 | 0.87 |
| Calculated free-T (pg/dL) | Calculated free-T (pg/dL) | ||||||
| T1 (15.5) (N = 226) | 1.00 | 1.00 | 1.00 | T1 (15.5) (N = 226) | 1.00 | 1.00 | 1.00 |
| T2 (15.6–22.3) (N = 238) | 2.08 (1.23, 3.52) |
1.12 (0.38, 3.32) |
1.14 (0.31, 4.17) |
T2 (15.6–22.3) (N = 238) | 1.75 (1.05, 2.91) |
1.42 (0.29, 6.88) |
0.69 (0.18, 2.59) |
| T3 (22.3–24.70) (N = 361) | 1.45 (0.93, 2.25) |
1.12 (0.38, 3.32) |
1.87 (0.76, 4.60) |
T3 (22.3–24.70) (N = 361) | 1.13 (0.79, 1.60) |
0.99 (0.35, 2.85) |
1.92 (0.64, 5.73) |
| P trend | 0.40 | 0.99 | 0.10 | P trend | 0.66 | 0.80 | 0.13 |
| Hypercholesterolemia, ng/mL (N = 964) | Hypercholesterolemia, ng/mL (N = 240) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 1.07 (0.68, 1.69) |
1.15 (0.49, 2.69) |
0.42 (0.21, 0.83) |
Yes | 1.03 (0.63, 1.66) |
1.08 (0.43, 2.71) |
0.46 (0.22, 0.96) |
| Self-reported lipid-lowering (N = 149) | Self-reported lipid-lowering (N = 75) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 1.03 (0.52, 2.04) |
0.92 (0.25, 3.44) |
1.08 (0.31, 3.78) |
Yes | 1.43 (0.68, 2.98) |
0.57 (0.07, 4.43) |
2.60 (0.61, 11.17) |
| Prescribed lipid-lowering (N = 168) | Prescribed lipid-lowering (N = 82) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.96 (0.46, 2.00) |
0.53 (0.16, 1.75) |
1.01 (0.29, 3.50) |
Yes | 0.70 (0.23, 2.16) |
ND | 0.72 (0.06, 8.15) |
Multivariable model adjusted for age, race and ethnicity, smoking status, history of hypertension, physical activity, alcohol consumption, body mass index, and mutual adjustment between sex steroid hormones and hypercholesterolemia
NHANES waves 1988–1991, 1999–2004, 2011–2014
NHANES waves 1988–1991, 1999–2004, 2013–2014
ND = not determined due to small sample size
Table 6.
Hispanic men: Multivariable associations of total testosterone, hypercholesterolemia, and self-reported and prescribed lipid-lowering medication with all-cause, cardiovascular, and cancer mortality: NHANES 1988–2014 linked with 2015 mortality data
| Mortality full sample†a (N = 1493) |
Mortality subset sample†b (Adjusted for estradiol and SHBG) (N = 1068) |
||||||
|---|---|---|---|---|---|---|---|
| All-cause (n = 156) | CVD (n = 30) | Cancer (n = 34) | All-cause (n = 146) | CVD (n = 30) | Cancer (n = 31) | ||
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | ||
|
| |||||||
| T deficiency (≤300 ng/dL), (N = 313) | T deficiency (≤300 ng/dL), (N = 191) | ||||||
| No | 1.0 | 1.0 | 1.0 | No | 1.0 | 1.0 | 1.0 |
| Yes | 1.43 (0.87, 2.34) |
5.99 (3.01, 11.88) |
0.79 (0.28, 2.27) |
Yes | 1.59 (0.91, 2.79) |
5.80 (2.63, 12.81) |
1.40 (0.48, 4.09) |
| Total T (ng/dL) | Total T (ng/dL) | ||||||
| T1 (100–354 ng/dL) (N = 497) | 1.00 | 1.00 | 1.00 | T1 (100–374 ng/dL) (N = 310) | 1.00 | 1.00 | 1.00 |
| T2 (355–519) (N = 505) | 0.86 (0.57, 1.30) |
0.26 (0.12, 0.55) |
2.23 (0.79, 6.33) |
T2 (375–539) (N = 358) | 0.77 (0.47, 1.24) |
0.35 (0.16, 0.75) |
1.30 (0.41, 4.10) |
| T3 (≥ 520) (N = 491) | 0.75 (0.39, 1.46) |
0.56 (0.24, 1.31) |
2.70 (0.86, 8.49) |
T3 (≥ 540) (N = 400) | 0.67 (0.40, 1.12) |
1.46 (0.48, 4.39) |
3.13 (0.98, 9.99) |
| p trend | 0.38 | 0.15 | 0.10 | P trend | 0.10 | 0.91 | 0.05 |
| Calculated free-T (pg/dL) | Calculated free-T(pg/dL) | ||||||
| T1 (15.5) (N = 340) | 1.00 | 1.00 | 1.00 | T1 (15.5) (N = 340) | 1.00 | 1.00 | 1.00 |
| T2 (15.6–22.3) (N = 347) | 1.01 (0.57, 1.82) |
0.41 (0.21, 0.78) |
0.67 (0.17, 2.60) |
T2 (15.6–22.3) (N = 347) | 1.06 (0.59, 1.90) |
0.55 (0.24, 1.24) |
0.69 (0.18, 2.59) |
| T3 (22.3–24.70) (N = 320) | 0.99 (0.49, 2.01) |
0.67 (0.30, 1.51) |
2.11 (0.74, 6.02) |
T3 (22.3–24.70) (N = 320) | 1.11 (0.58, 2.15) |
1.75 (0.77, 3.98) |
2.30 (0.90, 5.91) |
| P trend | 0.98 | 0.30 | 0.22 | P trend | 0.74 | 0.66 | 0.18 |
| Hypercholesterolemia, ng/mL (N = 508) | Hypercholesterolemia, ng/mL (N = 344) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.94 (0.56, 1.60) |
0.28 (0.11, 0.74) |
2.08 (0.55, 7.89 |
Yes | 0.94 (0.56, 1.60) |
0.29 (0.11, 0.75) |
2.57 (0.50, 13.09) |
| Self-reported lipid-lowering (N = 108) | Self-reported lipid-lowering (N = 62) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.47 (0.14, 1.53) |
0.45 (0.05, 4.43) |
0.68 (0.08, 5.60) |
Yes | 0.14 (0.02, 0.82) |
0.52 (0.05, 5.58) |
ND |
| Prescribed lipid-lowering (N = 117) | Prescribed lipid-lowering (N = 64) | ||||||
| No | 1.00 | 1.00 | 1.00 | No | 1.00 | 1.00 | 1.00 |
| Yes | 0.77 (0.31, 1.95) |
0.54 (0.05, 5.41 |
0.70 (0.12, 4.30) |
Yes | 0.25 (0.05, 1.22) |
0.59 (0.05, 6.66) |
0.26 (0.03, 2.32) |
Multivariable model adjusted for age, race and ethnicity, smoking status, history of hypertension, physical activity, alcohol consumption, body mass index, and mutual adjustment between sex steroid hormones and hypercholesterolemia
NHANES waves 1988–1991, 1999–2004, 2011–2014
NHANES waves 1988–1991, 1999–2004, 2013–2014
ND = not determined due to small sample size
Among NHB men (Table 5), T deficiency increased the risk of CVD mortality in the full sample and subset sample. The highest tertile of total T (T3 vs T1) was inversely associated with CVD mortality (HR = 0.35, 95% CI, 0.13–0.99, Ptrend = 0.03) in the full sample, and with all-cause (HR = 0.59, 95% CI, 0.37–0.94, Ptrend = 0.01) and CVD mortality (HR = 0.23, 95% CI, 0.05–1.02, Ptrend = 0.03) in the subset sample. Hypercholesterolemia was inversely associated with cancer mortality in the full sample (HR = 0.42, 95% CI, 0.21–0.83) and subset sample (HR = 0.46, 95% CI, 0.22–0.96). No association was identified among men who took lipid-lowering medications.
Among Hispanic men (Table 6), T deficiency increased the risk of CVD mortality in the full sample (HR = 5.99, 95% CI, 3.01–11.88) and subset sample (HR = 5.80, 95% CI, 2.63–12.81). However, hypercholesterolemia was inversely associated with CVD mortality in the full sample (HR = 0.28, 95% CI, 0.11–0.74) and subset sample (HR = 0.29, 95% CI, 0.11–0.75). No consistent association was identified among men who took lipid-lowering medications.
We also analyzed the association of total T and calculated free-T with all-cause, CVD, and cancer mortality stratified by hypercholesterolemia (Table 7). In general, we did not find significant associations in the full sample and subset sample stratified by hypercholesterolemia (Yes/No). There were some suggestive positive associations in the full sample between total T (T3 vs T1) and cancer mortality (Ptrend = 0.03) and calculated free-T (T3 vs T1) and all-cause mortality (Ptrend = 0.01) among men without hypercholesterolemia. In the sensitivity analysis, we also investigated the association of high levels of LDL-cholesterol (160–189 mg/dL) with all-cause, CVD, and cancer mortality in the total, NHW, NHB, and Hispanic population (Supplemental Table 2). In the total population and NHW men, there were suggestive associations between high LDL-cholesterol and cancer mortality, but they did not reach statistical significance.
Table 7.
Hypercholesterolemia status: Multivariable associations of total testosterone and calculated free testosterone with all-cause, cardiovascular (CVD), and cancer mortality: NHANES 1988–2014 linked with 2015 mortality data
| Mortality full sample†a (N = 5,379) |
Mortality subset sample†b (Adjusted for estradiol and SHBG) N = 3,740 |
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Hypercholesterolemia (N = 1,837) |
No-hypercholesterolemia (N =3,542) |
Hypercholesterolemia (N = 1,234) |
No-hypercholesterolemia (N = 2,506 |
||||||||||
| All-cause (n = 214) | CVD (n = 57) | Cancer (n = 49) | All-cause (n = 445) | CVD (n=103) | Cancer (n = 107) | All-cause (n = 191) | CVD (n=52) | Cancer (n=45) | All-cause (n = 414) | CVD (n = 99) | Cancer (n = 100) | ||
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | ||
|
| |||||||||||||
| T deficiency (≤300 ng/dL), (N = 1137) | T deficiency (≤ 300 ng/dL), (N = 659) | ||||||||||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | No | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| Yes | 1.09 (0.71, 1.68) |
1.34 (0.67, 2.69) |
0.87 (0.31, 2.46) |
0.78 (0.50, 1.22) |
1.02 (0.49, 2.10) |
0.96 (0.35, 2.66) |
Yes | 0.91 (0.57, 1.46) |
1.11 (0.45, 2.74) |
1.21 (0.29, 5.02) |
0.81 (0.49, 1.33) |
0.81 (0.38, 1.74) |
1.20 (0.40, 3.60) |
| Total T (ng/dL) | Total T (ng/dL) | ||||||||||||
| T1 (100–354 ng/dL) (N = 1791) | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | T1 (100–374 ng/dL) (N = 1243) | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| T2 (355–519) (N = 1808) | 0.82 (0.48, 1.42) |
1.21 (0.61, 2.39) |
0.56 (0.21, 1.46) |
0.98 (0.67, 1.42) |
1.18 (0.58, 2.37) |
1.16 (0.46, 2.95) |
T2 (375–539) (N = 1242) | 0.80 (0.41, 1.54) |
1.62 (0.73, 3.62) |
0.36 (0.12, 1.09) |
0.99 (0.69, 1.44) |
1.57 (0.71, 3.49) |
1.05 (0.41, 2.69) |
| T3 (≥520) (N = 1780) | 0.74 (0.41, 1.33) |
1.06 (0.53, 2.12) |
0.76 (0.29, 1.96) |
1.32 (0.90, 1.95) |
1.13 (0.49, 2.59) |
2.51 (0.96, 6.60) |
T3 (≥ 540) (N = 1255) | 0.70 (0.33, 1.49) |
1.97 (0.65, 6.00) |
0.37 (0.09, 1.60) |
1.31 (0.86, 1.99) |
2.39 (0.77, 7.38) |
1.86 (0.62, 5.55) |
| p trend | 0.31 | 0.92 | 0.60 | 0.12 | 0.77 | 0.03 | p trend | 0.35 | 0.24 | 0.19 | 0.19 | 0.13 | 0.21 |
| Calculated free-T (pg/dL) | Calculated free-T (pg/dL) | ||||||||||||
| T1 (15.5) (N = 1162) | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | T1 (15.5) (N = 1,162) | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| T2 (15.6–22.3) (N = 1165) | 0.89 (0.50, 1.58) |
1.19 (0.53, 2.67) |
0.75 (0.27, 2.10) |
1.25 (0.84, 1.87) |
1.36 (0.75, 2.49) |
1.35 (0.43, 4.27) |
T2 (15.6–22.3) (N = 1165) | 0.86 (0.47, 1.56) |
1.55 (0.59, 4.09) |
0.65 (0.24, 1.80) |
1.19 (0.78, 1.80) |
1.61 (0.79, 3.25) |
1.21 (0.37, 3.94) |
| T3 (22.3–24.70) (N = 1164) | 0.82 (0.47, 1.43) |
0.97 (0.50, 1.87) |
0.83 (0.31, 2.19) |
1.69 (1.14, 2.52) |
1.03 (0.35, 3.03) |
2.61 (0.88, 7.75) |
T3 (22.3–24.70) (N = 1164) | 0.79 (0.43, 1.43) |
1.93 (0.50, 7.46) |
0.60 (0.22, 1.64) |
1.52 (0.96, 2.40) |
1.61 (0.42, 6.12) |
2.02 (0.66, 6.21) |
| P trend | 0.48 | 0.95 | 0.69 | 0.01 | 0.92 | 0.05 | P trend | 0.44 | 0.33 | 0.32 | 0.07 | 0.42 | 0.16 |
Multivariable model adjusted for age, race and ethnicity, smoking status, history of hypertension, physical activity, alcohol consumption, and body mass index
NHANES waves 1988–1991, 1999–2004, 2011–2014
NHANES waves 1988–1991, 1999–2004, 2013–2014
ND = not determined due to small sample size
Discussion
In this prospective analysis of US men from NHANES 1988–2014 linked with 2015 mortality data, there were no significant associations between total T, T deficiency, and calculated free-T and mortality, whereas hypercholesterolemia was inversely associated with all-cause and cancer mortality in the full sample and subset samples. This latter observation was also made among NHW men, but high levels of T increased the risk of CVD-mortality. In NHB men, T deficiency increased the risk of CVD mortality, but high levels of T decreased CVD and all-cause mortality. Hypercholesterolemia was inversely associated with cancer mortality in the full sample and subset samples, as exhibited in the overall population. In Hispanic men, T deficiency increased the risk of CVD mortality, but hypercholesterolemia decreased it. To our knowledge, this is the first epidemiological study that capitalized on all the available total T and mortality data from NHANES stratified by race and ethnicity.
Total T and mortality
Previous studies have suggested a link between low serum levels of T and risk of mortality. [8,31] In the overall population, our findings related to the association between T deficiency, total T, and calculated free T, did not reach statistical significance in the full and subset samples. These differences may be attributed, in part, to sample size, years of follow-up, and study design differences. Furthermore, our study did not find significant associations between T levels and CVD and cancer mortality, these findings being similar to those of previous studies capitalizing on NHANES and UK Biobank databases. [30, 31] However, these associations seemed to vary by race and ethnicity in our analysis. Previous studies have reported an association between T and mortality among NHW men, but this body of literature has been inconsistent. [4, 32] Our findings among NHW men concurs with those of previous studies that have demonstrated an increased risk of CVD mortality in men with high levels of T compared with low T levels. [33, 34] However, we found opposite associations among NHB and Hispanic men showing that T deficiency (total testosterone ≤ 300ng/dL) increased the risk of CVD mortality. Our findings in NHB men do not concur with those of a previous small study of 622 NHB men that showed no association between T levels and mortality. [35] To our knowledge, no other study has reported that T deficiency (total testosterone ≤ 300ng/dL) [3] increased the risk of CVD mortality among NHB and Hispanic men.
Hypercholesterolemia and mortality
In parallel, a number of studies have reported that high levels of serum total cholesterol are inversely associated with all-cause and cancer mortality, but not CVD mortality. [36, 37] Our findings seem to be in agreement with a few studies that have reported an inverse relationship between hypercholesterolemia (total cholesterol ≥ 240mg/dL) and reduced all-cause (total cholesterol levels 210–249mg/dL in men ≥ 35 years) [37] and cancer mortality [36] among men, but further investigations have reported a U-curve association in both men and women. [37, 38] These differences compared with those of our study may be due, in part, to the larger sample sizes in the aforementioned previous studies which included men and women, and also to the fact that our study further adjusted for serum testosterone, estradiol, and SHBG levels in the multivariable analysis models. Nevertheless, it is still unclear whether and to what extent the associations of cholesterol with mortality vary among NHW, NHB, and Hispanic men. Wang et al. and Yi et al. in significant large sample sizes of European White and Asian men, respectively, reported that high levels of cholesterol seemed to decrease the risk of mortality. [32, 37] Our findings among NHW and Black men (only cancer mortality) appear to be in agreement with these results. However, among Hispanics, hypercholesterolemia was inversely associated with CVD mortality, which may be due to the small sample size in this group. On the other hand, a previous study reported a similar inconsistency, coining the term HDL-cholesterol paradox among Hispanics due to a link between HDL-cholesterol and mortality. [27, 39]
Combined serum total T and serum total cholesterol
Since it has been suggested that cholesterol may be a precursor for the biosynthesis of T, [20] previous epidemiological studies have explored the interplay between them, including lipid-cholesterol treatment. The latter epidemiological findings have, however, been inconsistent. [20, 28, 40] In a prospective cohort study, Oluleye et al. reported no significant association between total T and statin use. [20] However, in a meta-analysis of five homogenous randomized controlled trials in men, Schooling et al. concluded that statins may partially operate by lowering T. [40] To our knowledge, no other study has investigated the association of T deficiency with all-cause, CVD, and cancer mortality stratified by hypercholesterolemia status. Our findings demonstrated a potential association between total T and calculated free T with cancer mortality and all-cause mortality, respectively, among men without hypercholesterolemia in the full sample model. Nevertheless, it is possible that these observations may be driven by the large sample size in the full sample model, as they were attenuated after adjusting for estradiol and SHBG in the subset sample. Furthermore, it is also possible that potential residual confounding remains in those significant associations for unidentified confounders.
Strengths and limitations
Our study has a number of strengths. NHANES is a program of studies that is representative of the civilian non-institutionalized US population, which factor aids in the generalizability of these results. In addition, NHANES follows a rigorous protocol with extensive quality control procedures for the collection of the exposures, outcome of interest, and potential confounding factors analyzed and adjusted in this study. We also mutually adjusted total T, SHBG, and estradiol. Furthermore, this study capitalized on all the sex steroid hormone data readily available in NHANES (1988–2014) linked with the mortality-NDI file (2015).
Despite these strengths, the current study has several limitations that may influence interpretation of the results. First, we relied on a single measurement of sex steroid hormones. Estradiol and SHBG were not available in all the NHANES cycles; therefore, two different datasets were created, namely, full and subset samples. The subset sample is adjusted for estradiol and SHBG level, while the full sample is not. This will allow comparison with future studies where estradiol and SHBG may or may not be available. Other covariates may have complete data and there might be significant differences between the comparison groups, but these differences could be driven by the large sample size of the study rather than constituting a clinically relevant difference (e.g., points difference between mean BMI). Second, measurement of testosterone was obtained with electrochemiluminescence immunoassays and not HPLC tandem mass spectrophotometry, which has been recommended by some investigators. Third, cut-off points to define T deficiency and hypercholesterolemia are surrogate markers and may not fully or accurately capture their biological effect and sequelae on the risk of mortality among men. Fourth, there were small sample size categories among NHB and Hispanic men related to the interplay between T deficiency, hypercholesterolemia, and CVD and cancer mortality. We had no information on the treatment for T deficiency with testosterone replacement therapy, which can help normalize the levels of T. Fifth, we had no information on mortality from specific cancers, especially hormone-related cancers (e.g., prostate and testicular); therefore, we could not conduct these specific analyses. Although we adjusted for several potential confounders, there is still the possibility of residual confounding from strong risk factors (e.g., age) and unmeasured confounding from additional unmeasured factors. However, due to our detailed adjustment for strong confounders (e.g., BMI, hypertension, age, and smoking) and exclusion of major risk factors for mortality (e.g., diabetes and cardiovascular diseases), it is unlikely these would fully account for the observed findings. Finally, one important component that our study could not address, but that future studies with available data should explore in depth, is the role by population genomics and how genomic variation can influence the interplay between sex steroid hormones, total cholesterol, HDL- and LDL-cholesterol, and mortality in different racial and ethnic groups. [41, 42]
Conclusion
In summary, the results of this study confirm and elaborate on the observation that men with hypercholesterolemia have a reduced risk for cancer mortality, including NHW and NHB men. Only among NHW men did higher levels of T increase the risk of CVD mortality. However, among NHB and Hispanic men, T deficiency increased the risk of CVD mortality. Future studies with larger sample sizes of NHW, NHB, and Hispanic men are required to further confirm these findings.
Supplementary Material
Funding
David S. Lopez was supported by the National Institutes of Health (NIH) and National Institute on Aging, Grant #: P30 AG059301.
Footnotes
Declarations
Conflict of interest The authors declare that they have no conflict of interest.
Ethical approval The protocols for the conduct of the NHANES were approved by the Institutional Review Board. Informed consent was obtained from all participants.
Informed consent We conducted a secondary data analysis using data from NHANES. NHANES obtained informed consents from participants that also included consent for publication. The University of Texas Medical Branch (UTMB)-Institutional Review Board (IRB) approved the study.
Submission declaration and verification This work has not been published previously and is not under consideration for publication elsewhere. The publication is approved by all authors and, if accepted, it will not be published elsewhere in the same form, in English or in any other language.
Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s42000-022-00360-3.
Data availability
All data generated or analyzed in this study was provided by the National Center for Health Statistics (NCHS) of the US Centers for Disease Control and Prevention (CDC)-https://www.cdc.gov/nchs/nhanes/index.htm. For further data inquiries, please contact the corresponding author of this manuscript (David S. Lopez).
References
- 1.Khera M, Adaikan G, Buvat J et al. (2016) Diagnosis and treatment of testosterone deficiency: Recommendations from the fourth international consultation for sexual medicine (ICSM 2015). J Sex Med 13:1787–1804 [DOI] [PubMed] [Google Scholar]
- 2.Buvat J, Maggi M, Guay A, Torres LO (2013) Testosterone deficiency in men: Systematic review and standard operating procedures for diagnosis and treatment. J Sex Med 10:245–284 [DOI] [PubMed] [Google Scholar]
- 3.Mulhall JP, Trost LW, Brannigan RE et al. (2018) Evaluation and management of testosterone deficiency: AUA guideline. J Urol 200:423–432 [DOI] [PubMed] [Google Scholar]
- 4.Lopez DS, Canfield S, Wang R (2016) Testosterone replacement therapy and the heart: Friend, foe or bystander? Transl Androl Urol 5:898–908 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Lopez DS, Advani S, Tsilidis KK et al. (2017) Endogenous and exogenous testosterone and prostate cancer: Decreased-, increased- or null-risk? Transl Androl Urol 6:566–579 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Morgentaler A, Conners WP (2015) Testosterone therapy in men with prostate cancer: Literature review, clinical experience, and recommendations. Asian J Androl 17:206–211 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Platz EA (2008) Low testosterone and risk of premature death in older men: Analytical and preanalytical issues in measuring circulating testosterone. Clin Chem 54:1110–1112 [DOI] [PubMed] [Google Scholar]
- 8.Lopez DS, Qiu X, Advani S et al. (2017) Double trouble: Co-occurrence of testosterone deficiency and body fatness associated with all-cause mortality in US men. Clin Endocrinol (Oxf) 88:58–65 [DOI] [PubMed] [Google Scholar]
- 9.Travison TG, Araujo AB, O’Donnell AB et al. (2007) A population-level decline in serum testosterone levels in american men. J Clin Endocrinol Metab 92:196–202 [DOI] [PubMed] [Google Scholar]
- 10.Araujo AB, Esche GR, Kupelian V et al. (2007) Prevalence of symptomatic androgen deficiency in men. J Clin Endocrinol Metab 92:4241–4247 [DOI] [PubMed] [Google Scholar]
- 11.Malik RD, Lapin B, Wang CE et al. (2015) Are we testing appropriately for low testosterone?: Characterization of tested men and compliance with current guidelines. J Sex Med 12:66–75 [DOI] [PubMed] [Google Scholar]
- 12.Carroll MD, Fryar CD, Nguyen DT (2017) Total and high-density lipoprotein cholesterol in adults: United States, 2015–2016. NCHS Data Brief 290:1–8 [PubMed] [Google Scholar]
- 13.Platz EA, Till C, Goodman PJ et al. (2009) Men with low serum cholesterol have a lower risk of high-grade prostate cancer in the placebo arm of the prostate cancer prevention trial. Cancer Epidemiol Biomarkers Prev 18:2807–2813 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Chou R, Dana T, Blazina I, Daeges M, Jeanne TL (2016) Statins for prevention of cardiovascular disease in adults: Evidence report and systematic review for the US preventive services task force. JAMA 316:2008–2024 [DOI] [PubMed] [Google Scholar]
- 15.Baillargeon J, Kuo YF, Westra JR, Urban RJ et al. (2018) Testosterone prescribing in the United States, 2002–2016. JAMA 320:200–202 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Mefford MT, Rosenson RS, Deng L et al. (2019) Trends in statin use among US adults with chronic kidney disease 1999–2014. J Am Heart Assoc 8:e010640. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lopez DS, Huang D, Tsilidis KK et al. (2019) Association of the extent of therapy with prostate cancer in those receiving testosterone therapy in a US commercial insurance claims database. Clin Endocrinol (Oxf) 91:885–891 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Lopez DS, Huang D, Tsilidis KK et al. (2021) The role of testosterone replacement therapy and statin use, and their combination, in prostate cancer. Cancer Causes Control 32:965–976 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lopez DS, Polychronopoulou E, Tsilidis KK et al. (2021) Independent and joint effects of testosterone replacement therapy and statins use on the risk of prostate cancer among white, black and Hispanic men. Cancer Prev Res (Phila) 14:719–728 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Oluleye OW, Kronmal RA, Folsom AR et al. (2019) Association between statin use and sex hormone in the multi-ethnic study of atherosclerosis (MESA) cohort. J Clin Endocrinol Metab 104:4600–4606 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Centers for Disease Control and Prevention (CDC) National Center for Health Statistics. National Health and Nutrition Examination Survey: Overview. Accessed January 2020. https://www.cdc.gov/nchs/data/nhanes/nhanes_13_14/NHANES_Overview_Brochure.pdf.
- 22.National Center for Health Statistics (1994) Plan and operation of the third national health and nutrition examination survey 1988–94. Series 1 Programs and collection procedures vital health stat 1:1–407.1994 [PubMed] [Google Scholar]
- 23.Friedewald WT, Levy RI, Fredrickson DS (1972) Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 18:499–502 [PubMed] [Google Scholar]
- 24.Ford ES, Giles WH (2013) A comparison of the prevalence of the metabolic syndrome using two proposed definitions. Diabetes Care 26:575–581 [DOI] [PubMed] [Google Scholar]
- 25.Sodergard R, Backstrom T, Shanbhag V, Carstensen H (1982) Calculation of free and bound fractions of testosterone and estradiol-17 beta to human plasma proteins at body temperature. J Steroid Biochem 16:801–810 [DOI] [PubMed] [Google Scholar]
- 26.Vermeulen A, Verdonck L, Kaufman JM (1999) A critical evaluation of simple methods for the estimation of free testosterone in serum. J Clin Endocrinol Metab 84:3666–3672 [DOI] [PubMed] [Google Scholar]
- 27.Mazidi M, Mikhailidis DP, Banach M (2019) Associations between risk of overall mortality, cause-specific mortality and level of inflammatory factors with extremely low and high high-density lipoprotein cholesterol levels among american adults. Int J Cardiol 276:242–247 [DOI] [PubMed] [Google Scholar]
- 28.Mondul AM, Selvin E, Rohrmann S et al. (2010) Association of serum cholesterol and cholesterol-lowering drug use with serum sex steroid hormones in men in NHANES III. Cancer Causes Control 21:1575–1583 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.National Center for Health Statistics, Office of Analysis and Epidemiology. Public-use linked mortality file [internet]. hyattsville (MD): National center for health statistics; 2015. [cited 2015 jun 15]. available from: http://Www.cdc.gov/nchs/data_access/data_linkage/mortality.htm. [Google Scholar]
- 30.Menke A, Guallar E, Rohrmann S et al. (2010) Sex steroid hormone concentrations and risk of death in US men. Am J Epidemiol 171:583–592 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Yeap BB, Marriott RJ, Antonio L et al. (2021) Serum testosterone is inversely, and sex hormone-binding globulin is directly associated with all-cause mortality in men. J Clin Endocrinol Metab 106:e625–e637 [DOI] [PubMed] [Google Scholar]
- 32.Wang J, Fan X, Yang M et al. (2021) Sex-specific associations of circulating testosterone levels with all-cause and cause-specific mortality. Eur J Endocrinol 184:723–732 [DOI] [PubMed] [Google Scholar]
- 33.Soisson V, Brailly-Tabard S, Helmer C et al. (2013) A J-shaped association between plasma testosterone and risk of ischemic arterial event in elderly men: The french 3C cohort study. Maturitas 75:282–288 [DOI] [PubMed] [Google Scholar]
- 34.Vikan T, Schirmer H, Njolstad I, Svartberg J (2009) Endogenous sex hormones and the prospective association with cardiovascular disease and mortality in men: The tromso study. Eur J Endocrinol 161:435–442 [DOI] [PubMed] [Google Scholar]
- 35.Cummings-Vaughn LA, Malmstrom TK, Morley JE, Miller DK (2011) Testosterone is not associated with mortality in older African-American males. Aging Male 14:132–140 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Schuit AJ, Van Dijk CE, Dekker JM, Schouten EG et al. (1993) Inverse association between serum total cholesterol and cancer mortality in Dutch civil servants. Am J Epidemiol 137:966–976 [DOI] [PubMed] [Google Scholar]
- 37.Yi SW, Yi JJ, Ohrr H (2019) Total cholesterol and all-cause mortality by sex and age: A prospective cohort study among 128 million adults. Sci Rep 9:1596–018-38461 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.He GD, Liu XC, Liu L et al. (2021) A nonlinear association of total cholesterol with all-cause and cause-specific mortality. Nutr Metab (Lond) 18:25–021-00548–1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Cybulska B, Klosiewicz-Latoszek L (2014) The HDL paradox: What does it mean and how to manage low serum HDL cholesterol level? Kardiol Pol 72:681–686 [DOI] [PubMed] [Google Scholar]
- 40.Schooling CM, Au Yeung SL, Freeman G, Cowling BJ (2013) The effect of statins on testosterone in men and women, a systematic review and meta-analysis of randomized controlled trials. BMC Med 11:5770151157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Duello TM, Rivedal S, Wickland C, Weller A (2021) Race and genetics versus “race” in genetics: A systematic review of the use of African ancestry in genetic studies. Evol Med Public Health 9:232–245 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Garcia-Ortiz H, Barajas-Olmos F, Contreras-Cubas C et al. (2021) The genomic landscape of mexican indigenous populations brings insights into the peopling of the americas. Nat Commun 12:594202126188. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data generated or analyzed in this study was provided by the National Center for Health Statistics (NCHS) of the US Centers for Disease Control and Prevention (CDC)-https://www.cdc.gov/nchs/nhanes/index.htm. For further data inquiries, please contact the corresponding author of this manuscript (David S. Lopez).