Caffeine is the most widely used psychoactive drugs in the world, and estimates show that over 80% of Americans consume caffeine daily, with an average consumption of more than 200 mg/day [1]. Coffee and tea represent the main sources of caffeine (> 80% of daily intake), while soda, energy drinks, and chocolate are minor sources [2, 3]. Analysis of caffeine and its metabolites has been of great interest with respect to population-level caffeine exposure, utility in kinetic and metabolism studies of CYP450 enzymes, and for use in in vivo studies to estimate caffeine’s association with health outcomes [4]. Numerous benefits of caffeine consumption have been documented, including improved mood and wakefulness, weight loss, antioxidative properties, and potentially improved long-term memory [5,6,7]. Despite these benefits, adverse symptoms of high caffeine consumption including restlessness, insomnia, dehydration, and cardiac abnormalities are well-documented [8]. Additionally, reports of adverse effects on several organ systems including the cardiovascular, neurological, and endocrine systems have been published [8,9,10]. In the context of the endocrine system, there is preliminary evidence to suggest caffeine and its metabolites exert effects on various pathways, including those related to testosterone biosynthesis [11, 12].

Caffeine has a mean elimination half-life of 5 h, and virtually all of caffeine is metabolized in the body, with only 3% or less being excreted unchanged in urine [13]. The cytochrome P450 enzymes in the liver, mainly CYP1A2, are responsible for the metabolism of caffeine, and population level differences in P450 enzymes are known to contribute to variations in caffeine metabolism [14]. The main route of metabolism in humans (70–80%) is through N-3-demethylation to paraxanthine, also known as 1,7-dimethylxanthine or 17X. 1-N-demethylation of caffeine to theobromine accounts for approximately 7 to 8% of caffeine metabolism, and 7-N-demethylation to theophylline also around 7 to 8% [15]. The remaining 15% of caffeine undergoes C-8 hydroxylation to form 1,3,7-trimethyluric acid (Fig. 1).

Fig. 1

Showing the metabolism of caffeine into its various metabolites. Theophylline, Paraxanthine, and Theobromine represent the major metabolites of caffeine, denoted by the bold arrows (4%, 84%, and 12% respectively)

Testosterone plays an important role in men for the development of sexual features, brain function, muscle mass, and bone density [16, 17]. The most common symptoms associated with reduced serum testosterone (at or below 300 ng/dL) are decreased frequency of sexual thoughts and sexual desire, weight gain, and erectile dysfunction [18, 19]. In the U.S., the prevalence of low testosterone is as high as 40% in men over the age of 45 and is predicted to increase over the ensuing decades of life [20]. Paralleling the rise in low testosterone, there has been a decrease in fertilization rates both in the U.S. and worldwide [21]. Research shows that male factors such as hypogonadism account for over 40–50% of infertility cases [22,23,24]. A number of risk factors have been associated with low testosterone, including age, obesity, sedentary behavior, alcohol consumption, and certain medications [25, 26]. In addition to these risk factors, the potential role of the environment and dietary exposures, such as caffeine, on low testosterone etiology has come into light.

A number of population level, in vitro, and in vivo studies have investigated the effects of caffeine on testosterone levels, yet results of these studies remain conflicting, inconclusive, and lack generalizability to the U.S. adult population. In a study by Park et al., prepubertal male rats that were exposed to caffeine over four weeks showed significant reductions in serum testosterone, as well as reductions in testicular mass. In a study by Friedman et al., rats exposed to theobromine and caffeine experienced significant testicular atrophy, and plasma concentrations of testosterone were elevated in theobromine and caffeine-fed rats. In a study of human participants, Wedick et al. (2012) investigated the association between caffeinated and decaffeinated coffee on sex hormone binding globulin and endogenous sex hormone levels. Over the period of 8 weeks, no significant associations were observed, however at the 4-week interval, men showed an increase in total testosterone, while females showed a decrease in total testosterone. A previous NHANES study by Lopez et. al. found non-linear associations between recall caffeine intake and testosterone, but insignificant linear associations. Based on data from these previous studies, and the potential for caffeine and its metabolites to dysregulate testosterone pathways experimentally, the researchers hypothesized that caffeine consumption is significantly associated with testosterone in men. In a representative sample of U.S. adult men, the researchers set out to quantify the strength and direction of the association between caffeine and testosterone.

Research design and methodsNational Health and Nutrition Examination Survey (NHANES)

Data analyzed was collected from the 2013–2014 NHANES survey cycle (available from: https://wwwn.cdc.gov/Nchs/Nhanes/2013-2014/TST_H.htm). NHANES is a nationwide survey conducted annually for the purpose of collecting health and diet information from a representative, non-institutionalized U.S. population. NHANES is unique in that it combines interviews, physical examinations, and laboratory evaluations to obtain a large amount of quantitative and qualitative data. Information on NHANES survey methods are described in detail elsewhere [27]. Briefly, the survey examines about 5,000 persons each year from various counties across the U.S., which are divided into a total of 30 primary sampling units (PSUs), of which 15 are visited annually. All participants provided a written informed consent in agreement with the Public Health Service Act prior to any data collection. Household questionnaires, telephone interviews, and examinations conducted by healthcare professionals and trained personnel were utilized to collect data.

Study participants and exclusion criteria

The 2013–2014 NHANES cycle collected data on 10,175 individuals. In the analysis, the researchers excluded a total 7,217 women and all children under the age of 18, leaving 2,958 men. Boys under 18 were excluded since low testosterone is a rare outcome in this age group and wouldn’t provide a sufficient sample size for robust analyses (n < 10). Those individuals presenting with low testosterone so early is likely attributable to genetic conditions or unusually high exposure to medications or toxicants, levels exceeding everyday exposure amounts [28, 29]. From these remaining individuals, analysis was restricted to men with valid serum testosterone concentrations, as well as complete information on demographic, anthropometric, questionnaire, and laboratory variables including BMI, alcohol use, diabetes status, creatinine and albumin concentration, ethnicity, smoking status, and sex hormone binding globulin concentrations, resulting in a final analysis sample size of 372.

Assessment of serum testosterone

Following an overnight fast, serum samples were first taken between 8:30 a.m. and 11:30 a.m. and then testosterone concentrations were determined using a competitive electrochemiluminescence immunoassay on the 2010 Elecsys autoanalyzer (Roche Diagnostics, Indianapolis, IN, USA) with the lowest detection limit of the assay being 0.02 ng/mL. All sex steroid hormones from the present NHANES cycle were assayed at Boston Children’s Hospital (Boston, MA, USA) by laboratory technicians blinded to participant characteristics. The details for the NHANES laboratory methodology for testosterone determination are available from: https://wwwn.cdc.gov/Nchs/Nhanes/2013-2014/TST_H.htm.

Quantification of caffeine metabolites

Caffeine and 14 of its metabolites were quantified in urine by use of high-performance liquid chromatography-electrospray ionization-tandem quadrupole mass spectrometry (HPLC–ESI–MS/MS), and with stable isotope labeled internal standards. With the exception of paraxanthine, which is readily obtained from plasma and minimally excreted in the urine, these metabolites are readily obtained from urine and serve as good proxies for caffeine exposure [30]. In addition to serving as proxies for caffeine, some of these metabolites are biologically active (e.g. methylxanthines, theophylline, theobromine), and may exert their own effects on endocrine pathways related to testosterone production. To begin sample preparation, 50-µL aliquots of urine were diluted with 450 µL of water. Then, 100 µL of the diluted urine was combined with 120 µL of a 0.2 N NaOH solution containing stable isotope labeled internal standards. The mixture was incubated for 30 min at room temperature. Samples were then acidified with 30 µL of 2.0 N HCl and 250 µL of a 1:9 methanol/water solution containing 0.1% formic acid. Quantitation by HPLC–ESI–MS/MS was based on peak area ratios interpolated against an 11-point calibration curve derived from calibrators in synthetic urine. A further detailed description on laboratory procedures can be found elsewhere [31].

Defining demographic variables

Methods for questionnaire data collection are described in the NHANES procedures guide [32]. Covariates related to low testosterone, as well as potential confounders were included and based on results from literature searches. Participants were classified according to highest level of education attainment, insurance coverage status, smoking status, alcohol use, diabetes status, and cholesterol status. Levels of education were based on responses by participants during the home interview. Insurance coverage status and smoking status were recorded as a yes or no response from the home interview. Alcohol use was divided into three categories of “non-drinker”, “moderate drinker”, and “heavy drinker.” Non-drinkers were defined by individuals stating they drank less than 1 alcoholic beverage a week. Moderate drinkers reported drinking between 2–8 drinks a week. Heavy drinkers were defined as drinking over 10 alcoholic drinks a week. Diabetes status was defined as a fasting serum glucose greater than 126 mg/dL, having answered yes to taking diabetic medications, or having been diagnosed by a physician with diabetes. Cholesterol status was defined by whether or not a person was told he/she has high cholesterol by a physician, if the serum total cholesterol was greater than 240 mg/dL, and/or if that person is currently taking hypercholesterolemia medications.

Statistical analyses

Continuous variables were compared using one-way ANOVA, while categorical variables were compared using the Chi-squared test. Multivariable, ordinary least squares regression models were used to measure the association between caffeine and its urinary metabolites, and serum testosterone concentrations. Additionally, theobromine and theophylline are biologically active metabolites with known involvement in pathways that may be related to testosterone production and maintenance. Therefore, theobromine and theophylline were included in the logistic regression models, to model the odds of low testosterone based on quartile of metabolite concentration. The lowest quartile was used as the reference in each case. The complex survey design assigns a weight to each individual as a function of their probability of being randomly selected and this was considered when building the regression models. The final models were adjusted for potential confounders including age, BMI, smoking, drinking, and creatinine to control for urinary dilution.

All statistical analyses were performed using SAS 9.4 and SUDAAN software packages accounting for the complex survey design of NHANES [33]. A p-value < 0.05 was used as the criterion for significance.