A Clinical Perspective of Sleep and Andrological Health: Assessment, Treatment Considerations, and Future Research

Abstract

Context

Sleep that is insufficient, misaligned, or disrupted causes hypersomnolence and neuropsychological deficits, adversely affects cardiometabolic health, and is increasingly recognized to impair other biological processes that lead to conditions important to men, such as hypogonadism, erectile dysfunction, and infertility.

Evidence Acquisition

Literature review from 1970 to December 2018.

Evidence Synthesis

High-quality and complementary epidemiological and interventional studies establish that abnormal sleep is associated with increased mortality, hypertension, and other cardiometabolic disorders (insufficient, disrupted, and misaligned sleep), as well as reduced fecundity and total sperm count (insufficient sleep), erectile dysfunction (disrupted sleep), and low testosterone (both). Circadian misalignment shifts the peak of testosterone’s diurnal rhythm to occur soon after waking up, irrespective of the biological clock time, but it does not change the mean concentration. Preliminary studies show that extending sleep in individuals who are chronically sleep deprived may become a strategy to reduce insulin resistance and hypertension. Continuous positive airway pressure therapy can improve erectile function, and possibly systemic testosterone exposure, but only when used adherently by men with obstructive sleep apnea. Both high-dose and replacement-dose testosterone therapies modestly worsen sleep-disordered breathing, but they also improve cardiometabolic function and sexual desire. Persistence of either the adverse or beneficial outcomes over the longer term requires further investigation.

Conclusions

Sleep is increasingly recognized to be essential for healthy living. Establishing the effect of abnormal sleep, and of improving sleep, on andrological issues of prime interest to men will promote prioritization of sleep, and may thereby improve overall long-term health outcomes.

Sleep is essential for healthy living. Establishing the effect of abnormal sleep on andrological issues will promote prioritization of sleep, and may thereby improve overall long-term health outcomes.

Most Americans think that sleep, diet, and exercise are essential for healthy living. Despite this, abnormal sleep is widespread in the United States: ∼20% of adults sleep insufficiently (1), 15% to 20% of the workforce are shift workers, which misaligns sleep (2), and 15% of men have moderate to severe obstructive sleep apnea (OSA), which disrupts sleep (3). Insufficient, misaligned, and/or disrupted sleep causes hypersomnolence and poor neuropsychological performance that can sometimes lead to errors and catastrophic accidents, as occurred during the Chernobyl and Challenger space shuttle disasters (4–11), as well as insulin resistance, a critical step in the pathogenesis of prediabetes, type 2 diabetes mellitus (T2DM), and other cardiometabolic disorders. As this review highlights, abnormal sleep may also induce hypogonadism, infertility, and erectile dysfunction.

Sleep Physiology Relevant to Endocrinologists

Normal sleep is divided into rapid eye movement (REM) and non-REM (NREM) sleep. NREM sleep consists of transitions between stages 1, 2, and 3, which represent progressive slowing of the electroencephalogram that occurs as sleep deepens. Stage 3 NREM sleep, also known as slow wave sleep (SWS), occurs mainly during the first half and REM sleep during the last half of the biological night. Therefore, sleep disturbances occurring at different biological times will differentially affect sleep architecture. During SWS, heart rate, blood pressure, sympathetic nervous activity, and cerebral glucose utilization are decreased, cortisol secretion is inhibited, and GH is released. Accordingly, SWS is the stage of sleep that is most metabolically and hormonally active. Alternatively, REM sleep is required for memory consolidation and vivid dreaming. Staging of sleep is therefore important because different sleep stages likely serve different biological functions.

Abnormal sleep may be insufficient (reduced in quantity), misaligned to the environment, disrupted, or a combination of these. Conceptualizing abnormal sleep in this way is useful because insufficient, misaligned, and disrupted sleep differentially impact the structural organization, that is, the architecture, of sleep. Hence, it is plausible that the effects of insufficient, misaligned, or disrupted sleep on biological systems, including those affecting cardiometabolic and reproductive health, would also differ depending on the sleep stages involved. Insufficient sleep is a reduction in total sleep duration; however, SWS is preserved relative to other stages of sleep, and the overall impact on sleep architecture may differ depending on whether sleep is truncated at the beginning or at the end of the biological night. The impact of circadian misalignment on sleep architecture also appears to depend on whether sleep is advanced (sleeping earlier than usual) or delayed (sleeping later). When sleep is advanced, there is a decrease in REM and SWS and an increase in time awake (12). In contrast, when sleep is delayed, REM sleep is increased, SWS does not change, time awake is increased, and there is a decrease in stage 2 sleep (12). Other studies confirm that advancing or delaying sleep is associated with intermittent awakenings and less sleep consolidation, although the amount of SWS can be relatively preserved under both conditions (13). Disrupted sleep completely disorganizes sleep architecture. Disrupted sleep arising from OSA is associated with frequent awakenings, fragmented sleep, and generally less REM sleep. Determining the impact of abnormal sleep on physiological processes therefore requires consideration of sleep duration, timing, and quality.

Review Strategy

Abnormal sleep that is insufficient, misaligned, and/or disrupted impairs psychomotor performance, adversely affects cardiometabolic health, and is increasingly recognized to impair other important physiological processes that are important to men. This review focuses on the effects of abnormal sleep on cardiometabolic health, the leading cause of mortality in adult men in the United States, sleep disorders such as OSA, which are male preponderant, and impacts on other andrological conditions of prime interest to men such as hypogonadism, erectile dysfunction, and infertility. The psychomotor effects of abnormal sleep are well established (4, 5), less relevant to practicing endocrinologists, and are therefore beyond the scope of this review. Common causes of abnormal sleep that are clinically relevant to adult men are highlighted instead of an exhaustive review of all possible causes. Epidemiological associations, particularly population-based studies utilizing random sampling, and longitudinal comparisons of the same subjects across time are highlighted to illustrate the potential long-term consequences of abnormal sleep. Complementary interventional studies, particularly those that are randomized and placebo controlled, are then described to provide evidence for causation; such investigations are generally of shorter duration and therefore require examination of surrogate markers that change more rapidly. The key surrogate markers of health and disease in this review are: insulin resistance measured by gold standard methods such as the hyperinsulinemic-euglycemic clamp or minimal model for metabolic health; blood pressure for vascular risk; semen analysis for male fertility; blood testosterone concentrations that are frequently assessed and/or measured by mass spectrometry for male hypogonadism; and HbA1c for glycemia. Studies utilizing the International Index of Erectile Function (IIEF) are highlighted, as it is the recognized patient-reported outcome for erectile function. Findings from both epidemiological and interventional studies are then summarized to generate practice recommendations relevant for clinical endocrinologists.

Insufficient Sleep

Causes of insufficient sleep and its cardiometabolic consequences

An expert panel of the National Sleep Foundation determined, after evaluating a systematic review of the literature, that adults aged 18 to 64 years should sleep for 7 to 9 hours per night, and adults >65 years of age should sleep for 7 to 8 hours per night (1). Accordingly, at least 18% of adults in the United States do not sleep enough. Sleep loss typically arises from environmental factors, including work schedules and noise and light pollution, but it may also accompany sleep disorders, such as insomnia and OSA (4). Large-scale population-based longitudinal studies generally show that insufficient sleep is associated with increased risk of mortality, incident myocardial infarction, hypertension, obesity, and T2DM (4, 14–19). One influential meta-analysis showed that short sleep duration increased the risk of mortality, and that this risk was numerically similar using self-reported usual sleep duration cutoffs of either 5 or 7 hours per night (14). However, a qualitative reexamination of many of the studies included in this meta-analysis revealed inconsistencies in the reported association between short sleep and mortality, identified the lack of objectively measured sleep duration in almost all studies, and recognized that overarching health and sociodemographic factors, rather than sleep duration, could explain the association (20). For example, low socioeconomic status could cause shorter sleep when sleeping in more affordable locations resulted in exposure to greater environmental noise at night, such as residing in apartments that are closer to freeways or railways. Likewise, sleep opportunity could be curtailed by low socioeconomic status when it was necessary to work longer hours to generate more income. Low socioeconomic status could also increase mortality due to reduced access to health care. Hence, socioeconomic status could entirely explain the association between short sleep and mortality, without short sleep actually causing increased mortality. Similarly, the association between long sleep and mortality identified in some studies could be explained by other overarching factors such as cancer-related fatigue or depression. For this reason, interventional studies manipulating sleep to examine appropriate surrogate endpoints that track mortality or predict diseases are important to verify the putative causal mechanisms suggested by epidemiological associations.

Both epidemiological and interventional studies show that insufficient sleep induces insulin resistance. In fact only two population-representative cross-sectional studies examining this relationship are available (21, 22). In the more comprehensive study, 788 clinically healthy men and women (with approximately half being men) aged 30 to 60 years from 14 European countries were recruited and sleep duration was objectively measured by single-axis actigraphy, and insulin resistance was determined by hyperinsulinemic-euglycemic clamp (21). A U-shaped relationship between sleep duration and insulin resistance was reported only in men, indicating that short, or long, sleep was associated with insulin resistance (21). This study using gold standard methods to assess both insulin resistance and sleep duration is supported by another study where insulin resistance was assessed by minimal model of a 3-hour intravenous glucose tolerance test, but sleep was self-reported (22). In this smaller study of 224 healthy men and women aged 30 to 54 years from Allegheny County, Pennsylvania, recruited by mass mail solicitation, short sleep was associated with insulin resistance, but again only in men (22). Interventional studies where sleep is restricted under controlled laboratory conditions also show the induction of insulin resistance, an important factor in the development of T2DM in healthy young adults with normal sleeping patterns (19). Table 1 (23–30) shows the seven in-laboratory studies that use gold standard methods to measure insulin resistance (23–29). The five randomized-order studies are shown above the solid line, and the two fixed-order studies are shown below the solid line. All studies, except one nonrandomized study that was not fully conducted in the laboratory (27), show a statistically significant increase in insulin resistance with sleep restriction. Only 15 women, and 84 men, were studied in total (Table 1), so there was insufficient power to determine whether sex is an important biological variable in modulating this relationship. Nevertheless, among the five randomized-order studies, sleep restriction for 1, 4, 5, and 14 days consistently increased insulin resistance by 15% to 25%. The timing of sleep loss does not influence this effect (30). Furthermore these in-laboratory studies have been verified by an in-the-field study where 19 healthy lean [body mass index (BMI) of 19 to 26 kg/m²] young (20 to 30 years) men were randomized to either 3 weeks of usual sleep (n = 9) or 3 weeks where sleep was curtailed by waking 1.5 hours earlier (n = 10) (31). Sleep was reduced by an average of 1.5 hours, as confirmed by at-home actigraphy. Insulin resistance assessed by a hyperinsulinemic-euglycemic clamp was worsened after 1 week, but not 3 weeks, of sleep restriction, compared with the group that underwent usual sleep.

Table 1.

In-Laboratory Studies Showing the Effect of Sleep Restriction on Insulin Resistance

Study	Subjects (n)	Age (y) (Mean ± SD)	BMI (kg/m2) (Mean ± SD)	Sleep Restriction (d)	Sleep Opportunity	Control Sleep	Sleep Opportunity	Random Order	Method to Measure IR	Effect of SR on IR
Donga (23)	5 M, 4 F	45 ± 15	24 ± 2	1 × 4 h	0100–0500	1 × 8.5 h	2300–0730	Yes	EC	25% ↑a
Broussard (24)	19 M	24 ± 3	23 ± 1	4 × 4.5 h	0100–0530	4 × 8.5 h	2300–0730	Yes	MM	23% ↑a
Rao (25)	8 M, 6 F	27 ± 5	24 ± 4	5 × 4 h	0100–0500	5 × 8 h	2300–0700	Yes	EC	25% ↑a
Nedeltcheva (26)	6 M, 5 F	39 ± 5	27 ± 2	14 × 5.5 h	Individualized centered	14 × 8.5 h	Individualized centered	Yes	MM	20% ↑b
Wilms (30)	15 M	25 ± 3	23 ± 2	1 × 4 h	2230–0300 or 0215–0645	1 × 8 h	2215–0645	Yes	EC	16% ↑b
Spiegel (27, 28)c	11 M	22 ± 3	23 ± 2	6 × 4 h	0100–0500	3 × 8 h	2300–0700	No	MM	None
Buxton (29)	20 M	27 ± 5	23 ± 3	7 × 5 h	0030–0530	3 × 10 h	2200–0800	No	EC	11% ↑b
									MM	20% ↑a

Abbreviations: ↑, increased; BMI, body mass index; EC, euglycemic-hyperinsulinemic clamp; F, female; IR, insulin resistance; M, male; MM, minimal model; SR, sleep restriction.

^a P < 0.001.

^b P < 0.05.

^cSubjects were allowed outside of the laboratory to follow their usual daytime activities.

Effects of extending sleep on cardiometabolic markers in individuals who are chronically sleep deprived

The effect of extending sleep on cardiometabolic markers in individuals who are chronically sleep deprived has also been examined, but only in two studies (32, 33). The first study was a carefully controlled in-laboratory study of 19 men who underwent three weekend nights of 10 hours per night “catch up” sleep and three weekend nights of continued sleep restriction of 6 hours per night, in random order (32). Insulin resistance measured by minimal model improved with weekend catch up sleep. The second was a nonrandomized in-the-field study of 16 adults (13 women, 3 men) that did not include a control group, and it determined insulin resistance with fasting indices and not by a gold standard method (33). Sleep opportunity was extended by 1 hour per night for 6 weeks, fasting insulin and fasting glucose did not change, and the effect on insulin resistance was not reported. Nevertheless, significant correlations between change in total sleep time and change in fasting indices of insulin resistance were observed. One other study of sleep extension used blood pressure as the surrogate marker of cardiometabolic health. This in-the-field study examined 22 adults (13 women, 9 men) with prehypertension or hypertension who also self-reported chronic sleep restriction. Adults were randomized to extend the opportunity to sleep by 1 hour per night (n = 13), or to maintain habitual sleep patterns (n = 9), for 6 weeks. Blood pressure fell by a greater numerical amount in the sleep extension group, but there were no significant differences between the sleep extension and sleep maintenance (control) groups (34). Conversely, carefully conducted in-laboratory studies of sleep restriction have not consistently reported worsening of blood pressure, particularly under conditions of partial sleep restriction (35–37). Nevertheless, these limited studies show that extending sleep may be a promising approach to reduce cardiometabolic risk, but large-scale studies examining cardiometabolic events are yet to be performed.

Sleep duration and reproductive health

The epidemiological relationship between male reproductive health and sleep duration has also been studied in large cohorts. In a preconception cohort of 1176 young couples in a stable relationship, reduced fecundability was observed in couples where the male partner self-reported sleeping <6 hours per night (38). These couples were currently planning a pregnancy, having regular unprotected intercourse, and not undergoing infertility treatment. However, bed partners often share sleeping patterns, so it can be difficult to determine which partner’s sleeping pattern is responsible for the couple’s reduced fertility, even after statistical adjustment. Assessing effects on direct measures of male fertility, such as semen analysis, is required to resolve this uncertainty. Three studies of varying applicability have examined the cross-sectional relationship between sleep and semen parameters that are widely used in clinical practice to assess male fertility. The first study is not applicable to the general population because it was conducted in 382 men seeking fertility treatment, and an additional problem was that it evaluated sleep disturbances, not sleep duration (39). The second study was of 953 Danish military recruits, and it reported that higher and lower sleep disturbances were both associated with lower sperm concentrations and total sperm counts (40). However, sleep disturbances do not directly correspond with sleep duration, and it is difficult to conceptualize why both high and low sleep disturbances would be associated with reduced sperm concentration. The final cohort was of 796 military cadets in China, and the study reported associations between short (<6.5 hours per night) and long (>9 hours per night) sleep with reduced semen volume, reduced total sperm count (15), and high sperm DNA stainability, but not with other semen analysis parameters or with DNA fragmentation (41). A subgroup of 592 of these men provided a second semen sample after 2 more years. In a post hoc longitudinal analysis, men who happened to change their sleep to become closer to an arbitrary ideal sleep duration of 7.0 to 7.5 hours per night had a significantly higher total sperm number compared with those who did not (15). Sleep duration was not experimentally altered in this observational study, and the reasons why sleep duration changed were not assessed. Other studies have examined relationships with testis volume, which is another recognized marker of male reproductive potential. Preliminary cross-sectional analyses show that higher sleep disturbances (40) and lower sleep duration (42) are each associated with lower testis volumes. Interventional studies prospectively altering sleep patterns to assess effects on fecundability, testis volume, or semen parameters in humans have not been performed. However, Wistar Hannover rats (10 per group) randomized to stay awake through constant movement on multiple platforms for 18 hours per day for 21 days have lower viable sperm count and higher testicular nitric oxide synthase expression compared with littermates in the control group that were kept in the home cage (43).

Sleep duration and testosterone

Large epidemiological cohort studies in young (15), older (44, 45), and young and older (46) adult men have not consistently found a cross-sectional association between sleep duration and testosterone concentrations. Longitudinal observations have not been reported. Nevertheless, experimentally restricting sleep in men in controlled laboratory conditions generally reduces blood testosterone concentrations (47–49). Recently, it has been proposed that sleep loss must occur during the second half of the biological night to reduce testosterone (49), and this hypothesis is consistent with the available interventional studies of partial (47–49) and total (50–53) sleep restriction. This hypothesis may explain why the observational cross-sectional studies have not consistently reported an association between sleep duration and testosterone, because sleep loss in relationship to timing of the biological night was not considered.

Misaligned Sleep

Circadian physiology relevant to endocrinologists

The central circadian pacemaker (CCP) is located in the suprachiasmatic nucleus of the hypothalamus and entrains all other peripheral clocks throughout the body to coordinate bodily processes with the environment (54, 55). The periodicity of the CCP is nearly 24 hours (56) and must therefore be repeatedly reset to match the environment through exogenous cues called “zeitgebers,” such as light in the blue spectrum, feeding, and physical activity (57). Endogenous cues such as the timing of sleep and systemic melatonin exposure also entrain the CCP (58).

Once entrained, the CCP synchronizes all peripheral clocks through different hormonal and neural signaling pathways. Melatonin is the key hormonal signal for many physiological processes, but relevant for men’s health, experimental evidence implicates cortisol as the key metabolic central synchronizing signal (59, 60). Glucocorticoid administration shifts the timing of peripheral clocks in organs relevant to metabolism, that is, liver, muscle, and adipose tissue, and putatively also in pancreas and gut (59, 61). Recently, it was shown that the circadian rhythm of glucocorticoid administration in those with adrenal insufficiency entrains clock genes in human peripheral blood mononuclear cells (62). This synchronization by glucocorticoids is likely direct because silencing of clock genes Bmal1, Cry1, Per1, and Per2 impairs this action, glucocorticoid-response elements are located in the regulatory regions of the core clock genes, and both peripheral clocks and glucocorticoid receptors are present in these tissues (59, 60, 63). In addition to this central synchronizing function, cortisol is the main hormone controlling catabolism, acting with testosterone, the quintessential male anabolic hormone, to maintain catabolic/anabolic homeostasis in men.

Coupling between the hypothalamic–pituitary–adrenal and hypothalamic–pituitary–testicular axes occurs at the hypothalamus, pituitary, adrenal, and testis (64–66). At a molecular level, direct interactions can occur, because glucocorticoid and androgen receptors can physically interact to regulate transcription at common DNA binding sites (64, 65). Glucocorticoids inhibit the release of GnRH in rodents and primates (67–69). Glucocorticoids also suppress enzymes critical for testosterone synthesis, but likely through indirect mechanisms, as classical glucocorticoid response elements are not present in the promoter of many of these enzymes (65). Conversely, testosterone suppresses CRH-stimulated cortisol production, but it increases CRH-stimulated ACTH in men (70).

Blood testosterone concentrations vary diurnally (71, 72), and normative reference ranges applied to confirm the diagnosis of hypogonadism have been established for the early morning when concentrations are maximal in non–shift workers (73, 74). If the diurnal change in testosterone is intrinsically driven by the CCP, then measurement of testosterone to confirm hypogonadism should be timed to the biological clock. However, experimental evidence shows that the diurnal variation in testosterone, and in particular the peak in testosterone, is related to sleep and not to intrinsic circadian rhythms. In a carefully controlled in-laboratory study of seven young men, the timing of sleep was manipulated to day (0700 to 1500 hours) or night (2300 to 0700 hours) in a balanced order crossover fashion. The diurnal rhythm of testosterone, determined by measurements of blood sampled every hour, corresponded to the timing of sleep, and the mean 24-hour testosterone concentration did not change (75). In this study, only a minor change in testosterone due to the intrinsic circadian rhythms was observed. These data linking the diurnal rhythm of testosterone with sleep are consistent with two other studies that have additionally recognized the importance of REM sleep. The first reported a positive relationship between the number of REM sleep episodes and testosterone concentrations measured in blood sampled every 20 minutes during sleep from 67 middle-aged men (76). The second study fragmented night time sleep (7 minutes of sleep followed by 13 minutes awake, every 20 minutes) in 10 young men, and it showed that the rise in testosterone did not occur in those in whom REM sleep did not occur (77).

Causes of circadian misalignment and its cardiometabolic consequences

The most common causes of circadian misalignment are shift work and jet lag. In fact, 15% to 20% of the US workforce are shift workers (2). Cross-sectional analyses of population-representative cohorts show that those who experience shift work or social jet lag are more likely to have metabolic diseases such as obesity, prediabetes, and T2DM (78–81). A meta-analysis of 34 studies (and >2 million individuals), which included 12 cohorts followed longitudinally, shows that shift work is associated with T2DM and vascular events, including acute myocardial infarction and ischemic stroke (82). Additionally, three large cross-sectional studies, which collectively involve >100,000 participants, show that shift workers are more likely to have hypertension (78). Despite these associations, whether shift work increases mortality specifically in men remains controversial. An earlier meta-analysis that included men did not report an increase in all-cause or cardiovascular mortality (82). Since then, two large population-based longitudinal cohorts of 28,731 Danish and 74,862 US nurses who were exclusively female show that shift work is associated with increases in the incidence of all-cause and cardiovascular mortality (83, 84). Although not yet definitively demonstrated, shift work could plausibly increase mortality in men as well.

Interventional studies in humans conducted under controlled in-laboratory conditions show that circadian misalignment induced by forced desynchrony (13, 85) or simulated night shift work (86–92) impairs glucose tolerance (13, 85–87, 92) and induces insulin resistance (88–91). As little as 1 day of circadian misalignment is sufficient to induce insulin resistance. One study was performed in actual shift workers (92), and two studies also restricted sleep in all individuals, that is, in those whose sleep was or was not aligned (85, 91), and so were not designed to assess the effect of circadian misalignment alone. Of the four studies examining insulin resistance (88–91), all used a simulated night shift routine in individuals who were not shift workers, and only two were randomized (Table 2). The two randomized-order studies are shown first, followed by the two fixed-order studies. Circadian misalignment for 1, 2, 3, and 4 days increased insulin resistance by 14% to 26%, and the combination of circadian misalignment with sleep restriction increased insulin resistance by 55% (Table 2. Blood pressure was also assessed in some of these studies. Circadian misalignment from experimental shift work in actual shift workers (93), and from forced desynchrony (13) or experimental shift work (94) in non–shift workers, increases blood pressure.

Table 2.

In-Laboratory Studies Showing the Effect of Simulated Shift Work on Insulin Resistance

Study	Day Shift (n)	Night Shift (n)	Age (y) (Mean ± SD or IQR)	BMI (kg/m2) (Mean ± SD)	Shift Work Exposure (d)	Random Order	Method to Measure IR	Effect of CM on IR
Wefers (88)	14 M	14 M	22 ± 3	22 ± 2	2	Yes	EC	14% ↑a
Qian (89)	8 M, 6 F	8 M, 6 F	28 ± 9	25 ± 3	1 and 3	Yes	MM	17% ↑b
Bescos (90)	4 M, 4 F	4 M, 5 F	26 ± 5	22 ± 3	4	No	EC	26% ↑c
Leproult (91)	10 M, 3 F	9 M, 4 F	22-26	23 ± 3	2-4	No	MM	55% ↑d

Abbreviations: ↑, increased; CM, circadian misalignment; EC, euglycemic-hyperinsulinemic clamp; F, females; IQR, interquartile range; IR, insulin resistance; M, males; MM, minimal model.

^a P = 0.029.

^b P = 0.0007 (IR was induced after the first day of night shift work and remained unchanged after the third day).

^c P = 0.03.

^d P = 0.011 (only in men).

Circadian misalignment and reproductive health

Two population-based studies of couples randomly selected from discrete geographical areas suggest that shift work in men does not impair fertility, and it may not induce hypogonadism in the absence of actual sleep loss. The first study was a cross-sectional analysis from five European countries of 6630 couples planning pregnancy and another 4035 couples where the female partner was already pregnant. Shift work by the male partner was not related to fecundity, which is the probability of having a live birth for each menstrual cycle (95). The second study involved 501 couples from four counties in Michigan and 12 counties in Texas discontinuing contraception who were followed for 12 months (96). Semen parameters, including DNA fragmentation, did not differ between men who were, or were not, shift workers (96).

The remaining studies performed in cohorts that are not population representative are conflicting and inconclusive. Retrospective studies assessing fecundity in 202 men obtained from an infertility clinic (97) and 1201 men from a hospital (98) did and did not, respectively, show a relationship with shift work. However, neither finding is likely to be applicable to the normal population. A case control study of 267 fertile men compared with 255 infertile men showed a larger number of shift workers in the infertile group (99). However, the fertile men refused to provide semen samples, suggesting that the groups were fundamentally unmatched. Another study of 365 infertile couples reported no significant differences in semen parameters between shift workers and non–shift workers (100). Randomized controlled interventional studies of circadian misalignment induced by experimental shift work or by forced desynchrony that assess the effects on semen parameters or other measures of male fertility are not available.

Circadian misalignment and testosterone

Complaints of fatigue, lack of energy, and reduced sexual drive are common in shift workers and overlap that of hypogonadism. Studies assessing the relationship between shift work and hypogonadism should therefore be interpreted cautiously because symptoms of hypogonadism are not specific. Nevertheless, in some of the few data available, symptoms of hypogonadism are associated with sleep quality in shift workers, but not with shift work per se (101). Most other studies are naturalistic observations in actual shift workers that have examined testosterone concentrations as a surrogate for the clinical diagnosis of hypogonadism, which requires symptoms (102). Shift work did not change testosterone concentrations, but only a single blood sample was used to estimate the entire 24-hour testosterone rhythm in three of these studies (101, 103, 104). The fourth study assessed blood testosterone concentrations on multiple occasions but only at night when shift workers were awake, and control non–shift workers were asleep (105). Testosterone was reduced at night while awake, but this finding is expected because carefully conducted in-laboratory experiments show that the nocturnal increase in testosterone requires sleep (75). The effect across the entire 24-hour period was not assessed.

Two naturalistic studies have examined the effect of circadian misalignment on diurnal testosterone concentrations measured in urine or saliva frequently collected across a 24-hour day by men while awake (106, 107). Although the duration and frequency of sampling were more extensive than in the aforementioned studies, the lack of sampling during sleep is an important limitation. Nevertheless, such studies do allow estimation of how the timing of the diurnal rhythm in blood testosterone concentration is altered in a naturalistic setting. In both studies, subjects were actual shift workers who altered sleep timing as part of their usual shift work schedules, with night shift concluding at 7:00 am, and testosterone was measured by gold standard mass spectrometry–based methods. Despite larger claims, neither study showed major statistically significant effects of night shift work on testosterone (106, 107). In the first study of car manufacturing or rail transport workers, 60 male permanent day (n = 21) and permanent night (n = 39) shift workers provided multiple urine samples while awake during a 24-hour day (106). The 39 male night shift workers had worked on average for 3 consecutive nights immediately prior to assessment, and on average for 7 nights during the previous 14. The peak in testosterone was shifted from around 8:30 am to 12:30 pm, that is, delayed by 4 hours, in these night shift workers. However, when circadian time was adjusted for, the delay with night shift work decreased to ∼1.5 hours and was not statistically different from no delay (106). In the other study, saliva was collected every 4 hours from 0700 hours while awake, on awakening, and at bedtime in 73 male police officers undergoing a rotating day and night shift schedule that consisted of a consecutive number of day shifts followed or proceeded by an equal consecutive number of recovery days (which were day shifts or days off) (107). In this nonrandomized crossover study, the pattern of testosterone compared from time awake (not according to clock time) and the mean of testosterone did not differ between 2, 4, and 7 days of night shift work, nor after 2, 4, or 7 days of recovery (107). These data are consistent with the previously mentioned in-laboratory study that showed that one night of experimental night shift work altered the diurnal variation in testosterone, that this variation was educed by sleep, not clock time, and that the mean concentrations are not changed (75). Collectively, all three studies (75, 106, 107) suggest that after 1 day of night shift, the diurnal pattern of testosterone exposure is shifted to track sleep, that this testosterone pattern remains stable after 2, 3, 4, or 7 days of night shift. Mean testosterone did not change. However, studies of more than one night of shift work that include testosterone measurements across both wake and sleep for an entire 24-hour day are lacking.

Disrupted Sleep

Causes of disrupted sleep and its cardiometabolic consequences

The most important clinical cause of disrupted sleep is OSA. Other medical causes include nocturnal asthma and nocturia, as well as nonmedical causes such as environmental noise. OSA is particularly relevant to men’s health because it is a male-preponderant disease (108); is associated with many negative health consequences, including hypertension, increased cardiovascular events such as stroke and myocardial infarction, cardiac arrhythmia, T2DM, and metabolic dysfunction, hypogonadism and erectile dysfunction, hypersomnolence, accidents, memory and cognitive impairment, and mood disorders; and can be ameliorated by effective therapies such as continuous positive airway pressure (CPAP), mandibular advancement, or uvulopalatopharyngoplasty in selected situations. Currently, 13% of men and 6% of women have moderate to severe sleep-disordered breathing, an increase in recent times that is due to rising obesity rates (3).

Longitudinal studies of population-based cohorts reveal that OSA independently increases the risk of all-cause mortality (109, 110), T2DM (111, 112), hypertension (113), and cardiovascular events (113). Meta-analyses of randomized controlled trials show that CPAP decreases hypertension (114) but increases weight (115). However, large-scale randomized controlled trials have not shown that CPAP reduces cardiovascular events (116–118) unless there is adherent CPAP use, where the mask is on the face for at least 4 hours per night (119). Table 3 (120–138) summarizes the studies utilizing gold standard methods to assess insulin resistance and/or HbA1c, a gold standard method to assess longer-term trends in glycemia. The rows above the solid line show the three (120–123) randomized sham-controlled studies (upper), the two (124, 125) randomized non–sham-controlled studies (middle), and the four (120, 126–128) nonrandomized studies (lower) examining the effect on gold standard measures of insulin resistance. The rows below the solid line show the randomized sham-controlled (129–135) studies (upper) and randomized non–sham-controlled (136–138) studies (lower) on the homeostatic model assessment of insulin resistance. This table illustrates the importance of using precise gold standard methods, as eight of the nine studies utilizing homeostatic model assessment of insulin resistance did not show any effect of CPAP on insulin resistance. In contrast, seven of the nine studies utilizing gold standard methods have shown that CPAP improves insulin resistance by a median of 40% (range, 25% to 90%) in many different individuals, including those who are healthy, at risk for T2DM, or have actual T2D. However, reduced insulin resistance did not translate to greater euglycemia. In fact, HbA1c was only significantly reduced in one study (135) and the prevalence of impaired glucose tolerance was significantly reduced in only one other (138).

Table 3.

Studies Showing the Effect of CPAP on Insulin Resistance or HbA1c

Study	Subjects (n)	Age (y) (Mean ± SD)	BMI (kg/m2) (Mean ± SD)	AHI/ODI (Events per Hour) (Mean ± SD)	Population	CPAP Exposure (mo)	CPAP Adherence (Hours per Night) (Mean ± SD)	Random	Sham	Crossover	Method to Measure IR	Effect of CPAP on IR (%)	Effect of CPAP on HbA1c
Hoyos (120)	65 M	49 ± 12	31 ± 5	40 ± 18	Healthy	3	4 ± NR	Yes	Yes	No	MM	None	NR
					No T2DM
Lam (121)	61 M	46 ± 10	28 ± 4	32 ± NR	Healthy	0.25	5 ± 2	Yes	Yes	No	MMa	90%	NR
					No T2DM							P = 0.02
West (122, 123)	42 M	56 ± 10	37 ± 5	36 ± 23	T2DM	3	3 ± 3	Yes	Yes	No	EC	None	None
Pamidi (124)	26 M, 13 F	54 ± 7	35 ± 7	57 ± NR	Prediabetes	0.5	8 ± 0	Yes	No	No	MM	30%	NR
												P = 0.04
Duarte (125)	7 M, 5 F	50 ± 9	33 ± 6	41 ± 13	Acromegaly	3	6 ± NR	Yes	No	Yes	EC	40%	None
												P = 0.03
Hoyos (120)	65 M	49 ± 12	31 ± 5	40 ± 18	Healthy	6	4 ± NR	No	No	No	MM	50%	NR
					No T2DM							P = 0.009
Harsh (126)	34 M	54 ± 12	33 ± 7	43 ± 11	Healthy	3	5 ± 1	No	No	No	EC	25%	NR
	6 F				No T2DM							P < 0.001
Brooks (127)	7 M, 3 F	51 ± 10	43 ± 4	47 ± 32	T2DM	4	NR	No	No	No	EC	30%	None
												P < 0.05b
Harsh (128)	7 M, 2 F	56 ± 8	37 ± 6	43 ± 21	T2DM	3	6 ± 1	No	No	No	EC	60%	None
												P = 0.02
Kritikou (129, 130)	20 M	56 ± 6	29 ± 3	37 ± 20	Healthy	2	6 ± 1	Yes	Yes	Yes	HOMA	None	NR
	18 F				No T2DM
Coughlin (131)	34 M	49 ± 8	36 ± 8	40 ± 14	Healthy	1.5	3 ± NR	Yes	Yes	Yes	HOMA	None	NR
					No T2DM
Kohler (132)	40 M	63 ± 6	33 ± 6	41 ± 20	Healthy	0.5c	6 ± 2	Yes	Yes	No	HOMA	None	NR
	1 F
Weinstock (133)	21 M, 29 F	54 ± 10	39 ± 8	44 ± 27	Prediabetes	2	4 ± 2	Yes	Yes	Yes	HOMA	None	NR
Comondore (136)	9 M, 4 F	56 ± 7	31 ± 6	28 ± NR	Medically stable	1	6±NR	Yes	No	Yes	HOMA	None	NR
Chirakalwasan (137)	32 F	32 ± 6	31 ± 4	10 ± NR	Prediabetes (Gestational diabetes)	0.5	4 ± 2	Yes	No	No	HOMA	None	NR
Salord (138)	22 M, 58 F	47 ± 9	48 ± 6	60 ± NR	Prediabetes (Obese)	3	5 ± 2	Yes	No	No	HOMA	None	Noned
Barcelo (134)	22 M, 58 F	47 ± 9	48 ± 6	60 ± NR	Prediabetes	3	Only 6 subjects used <4 h	Yes	No	No	HOMA	None	None
Martínez-Cerón (135)	30 M, 20F	61 ± 9	33 ± 5	32 ± 21	T2DM	6	5 ± 2	Yes	No	No	HOMA	25%	0.4
												P = 0.03	P = 0.03

Abbreviations: AHI, apnea–hypopnea index; EC, euglycemic-hyperinsulinemic clamp; F, females; HOMA, homeostatic model assessment; IR, insulin resistance; M, males; MM, minimal model; NR, not reported; ODI, oxygen desaturation index.

^aUsed truncated version of minimal model KITT. Calculated from a short insulin tolerance test.

^bAfter one participant with extremely infrequent reported CPAP use was excluded from the analysis. Otherwise, the analysis was not statistically significant.

^cAt least 7 d of therapeutic CPAP before randomized to subtherapeutic CPAP or therapeutic CPAP (i.e., withdrawal design). The study states that the patient was blinded.

^dCPAP decreased prevalence of impaired glucose tolerance.

Disrupted sleep and reproductive health

The relationship between OSA and fecundity or semen parameters has not yet been studied in large epidemiological cohorts. Such an undertaking may first require the development of a simple and accurate biomarker of OSA. Nevertheless, andrological issues are highly prevalent in men with OSA (139). For example, half of all men with OSA have erectile dysfunction; and reduced libido is a common complaint among obese men with OSA (140). Table 4 (141–155) summarizes all studies examining the effect of CPAP therapy on the erectile function domain of the full 15-item IIEF (141–147) (above the solid line) or the abridged 5-item version (148–155) (below the solid line). Although there are three randomized controlled trials, two utilizing the 15-item IIEF and one using the 5-item IIEF, only one study was sham controlled, and therefore double-blinded. Double blinding is essential for a patient-reported outcome such as erectile dysfunction. In this randomized sham-controlled parallel group study, 3 months of adherent CPAP use improved erectile function by 6 U in 61 men with moderate to severe erectile dysfunction (95) (see Table 4). This effect is clinically relevant because the minimal clinically important difference in the erectile function domain of the 15-item IIEF is 4 U (156). Most studies show a clinically relevant improvement in erectile function with CPAP, and there is a suggestion across all studies, and in individual studies (149), that this effect is greater in those who already have more severe erectile dysfunction (see Table 4). If true, this would explain the very small effect of CPAP of 2.5 U compared with a no-treatment control group (P = 0.06) in the other randomized study that was conducted in men with less severe erectile dysfunction (142). With the caveat that only three studies are randomized, and only one of these is sham controlled, the vast majority of studies utilizing the gold standard IIEF show that CPAP therapy improves erectile function to a clinically meaningful degree (Table 4). Additionally, many other studies have also reported that CPAP improves erectile function assessed by methods other than the standardized IIEF (157–163).

Table 4.

Studies Showing the Effect of CPAP on Erectile Dysfunction

Study	Subjects (n)	Age (y) (Mean ± SD)	BMI (kg/m2) (Mean ± SD)	AHI/ODI (Events per Hour) (Mean ± SD)	CPAP Exposure (mo)	CPAP Adherence (Hours per Night) (Mean ± SD)	Random Group Allocation	Sham	IIEF Version	Baseline IIEFa	Change in IIEFa
Melehan (141)	61	54 ± 9	33 ± 5	46 ± 26	3	3 ± 2	Yes	Yes	15	13 ± 7	6.0b
											P = 0.04
Pascual (142)	75	55 ± 1	33 ± 1	52 ± 21	3	5 ± NR	Yes	No	15	16 ± 9	None
Irer (143)	54	42 ± 7	31 ± 5	38 ± 18	3	5 ± 1	No	No	15	17 ± 5	7.2
											P < 0.001
Pastore (144)	41	49 ± 9	26 ± 4	47 ± 15	3	NR	No	No	15	7 ± 1	4.3
											P < 0.001
Acar (145)	21	47 ± 9	31 ± 4	55 ± 21	3	NR	No	No	15	18 ± 7	6.0
											P < 0.01
Karkoulias (146, 147)	15	56 ± 4	NR	7 ± 1	3	NR	No	No	15	7 ± 2	2.3
											P = 0.018
Li (148)	27	NR	NR	43 ± 13	1	NR	Yes	No	5	11 ± 3	7.0
											P < 0.05
Schulz (149)	94	52 ± 1	34 ± 1	56 ± 2	7	6 ± 0.2	No	No	5	18 ± 1	3.5c
											7.0d
											P < 0.05
Shin (150)	16	53 ± 8	28 ± 4	52 ± 17	7	NR	No	No	5	15 ± 7	None
Husnu (151)	28	49 ± 11e	32 ± 4e	59 ± 19e	3	NR	No	No	5	17 ± 6	4.3
											P = 0.001
Khafagy (152)	57	42 ± 9e	31 ± 3e	NR	3	>4 ± NR	No	No	5	16 ± 5	4.6
											P < 0.001
Zhang (153)	53	44 ± 9	29 ± 3	63 ± 12	3	7 ± NR	No	No	5	18 ± 4	1.0
											P = 0.001
Li (154)	32	55 ± 7e	29 ± 3e	52 ± 16	1	6 ± NR	No	No	5	14 ± 3	5.0
											P = 0.04
Taskin (155)	20	51 ± 7	NR	35 ± 19	1	7 ± NR	No	No	5	16 ± 5	3.4
											P < 0.001

Abbreviations: AHI, apnea–hyponea index; F, femal; M, male; NR, not reported; ODI, oxygen desaturation index.

^aIncrease in erectile function domain of the 15-item IIEF and of total score of the 5-item IIEF.

^bPer protocol analysis.

^cIn men with moderate erectile dysfunction.

^dIn men with severe erectile dysfunction.

^eApproximate estimates because the actual values are not reported.

Disrupted sleep and testosterone

The concurrence of low blood testosterone concentrations, obesity, and OSA has been recognized for decades, but how these observations are linked mechanistically remains controversial (139). Cross-sectional studies consistently show that more severe OSA (i.e., worse hypoxemia) is associated with lower blood testosterone concentrations (44, 164–166), but differ in whether adiposity does (44) or does not (164–166) explain this relationship. Three studies have used repetitive blood sampling and concluded that OSA is associated with either low testosterone or a change in the diurnal pattern of testosterone, and that these associations are not explained by adiposity when adiposity is controlled for (167–169). However, the small numbers of men and lack of close matching for obesity limits these conclusions. One study sampled blood during an entire 24-hour day, but relatively infrequently every 4 hours (167), whereas the other two studies sampled blood frequently every 20 minutes but mainly during sleep from 10:00 pm to 7:00 am (168) or from 7:00 pm to 7:00 am (169). In the first study, a post hoc analysis in 24 men found a significant correlation between the duration of hypoxemia with an arbitrary measurement that corresponded to the reduction in the diurnal fall in testosterone that occurs at the time of waking (167). Obesity was not adjusted or controlled for. In the second study, LH and testosterone were lower in the 10 men with OSA compared with the 5 controls, even after adjustment for age and adiposity (168). The third study reported similar findings in five men with OSA and five age- and BMI-matched controls; however, the mean BMI was 30.9 and 26.3 kg/m² in those with and without OSA, respectively (169). Despite matching, this degree of discrepancy was typical of all of these studies that repetitively assessed testosterone (166, 168, 169), and it may not have been adequately adjusted for statistically owing to small sample sizes.

The effect of reversing OSA on testosterone concentrations is also controversial. Only seven studies that include 232 men in total have examined the effect of CPAP, and a recent meta-analysis of these studies concluded that CPAP does not alter testosterone (170). However, this conclusion may not be warranted when studies of higher quality and/or that more effectively reversed OSA are considered. For example, the three most thoroughly conducted studies have all individually concluded that reversing OSA increases testosterone. The first study is the only randomized sham-controlled trial (171) and accounts for almost half of all participants included in the aforementioned meta-analysis (170). The change in morning blood testosterone was significantly higher in the 52 men treated with CPAP compared with the 49 men treated with sham CPAP after 1 month; however, this effect was due to a reduction in testosterone in the sham group. The second study is the only study that frequently sampled blood (every 20 minutes from 7:00 pm to 7:00 am) and treated men with CPAP compliantly (average time the mask was on the face was 5.2 hours per night) for the longest duration (at least 9 months) (172). CPAP significantly increased mean, incremental, and area under the curve testosterone concentrations in the five men studied (172). A third study was of 12 men in whom sleep-disordered breathing was almost completely reversed for the entire duration of sleep through surgical uvulopalatopharyngoplasty (173). Blood testosterone increased after 3 months, whereas the number of apneas fell from an average of almost 40 events per hour to ∼5 events per hour. Studies in rodents confirm and extend these findings, but they are not exactly comparable with sleep-disordered breathing. Nevertheless, chronic intermittent hypoxia during sleep (which occurs during the light cycle in rodents) decreases blood testosterone (174), sperm motility (175), and fertility (174, 175), and it increases circulating (174) and testicular (175) oxidative stress. Comparable studies in humans are not available; however, 5 or more days of high-altitude hypobaric hypoxia in male mountaineers decreases sperm motility (176) and concentration (177). However, these changes in sperm parameters could conceivably be due to intense physical activity, which was also experienced by these mountaineers.

Effect of testosterone therapy on sleep and sleep-disordered breathing

The effect of short-term testosterone therapy on sleep and breathing has been examined in two randomized controlled trials (178, 179). In both studies, many men had some degree of mild to moderate OSA at study entry, although this was not a specific inclusion criterion. In the first study (178), 11 men with hypogonadism aged 19 to 72 years were already being treated with intramuscular testosterone 200 or 400 mg of enanthate every 2 weeks and were assessed in random order off treatment of at least 30 (mean, 53) days after the last injection, and on treatment ∼3 to 7 (mean, 3.5) days after the last injection. The number of apneas and hypopneas were higher by an average of nine events per hour shortly after a testosterone injection, and there were no differences in sleep architecture on and off treatment. In the other study (179), 17 eugonadal men >60 years of age were randomized to receive either three injections of intramuscular mixed short-chain testosterone esters (500 mg, 250 mg, and then 250 mg) every week or matching oil-based placebo, and then crossed over to the other treatment after 8 weeks’ washout. High-dose testosterone exposure for 2 to 3 weeks increased sleep-disordered breathing by seven events per hour, lengthened the duration of hypoxemia by 2%, and reduced total sleep time by 1 hour, compared with placebos. Upper airway dimensions assessed by CT in the first study (178) or by acoustic reflectometry in the other study (179) did not differ, and they were not anticipated to differ because of the short-term nature of either study, with testosterone therapy. Nevertheless, this suggests that testosterone is acting through mechanisms not related to upper airway dimensions, such as previously recognized testosterone-induced changes to ventilatory responses to hypoxemia (180) or hypercapnia (181) that can reduce the apneic threshold and thereby induce sleep-disordered breathing. Furthermore, both studies used testosterone doses that were often (178) or always (179) supraphysiological, as well as testosterone formulations that were likely to result in large excursions in blood concentrations.

Two other randomized placebo-controlled parallel group studies have examined the effect of longer-term replacement dose and more steady-state testosterone therapy on sleep-disordered breathing (182, 183). In the first study, 108 men >65 years of age were randomized to receive a dose-titrated testosterone scrotal patch (∼6 mg/d on average) or matching placebo for 3 years (182). Many, but not all, men had mild sleep apnea before starting therapy, and testosterone therapy did not worsen sleep breathing at 6, 12, 24, or 36 months, compared with placebo treatment. However, the portable at-home device was relatively insensitive, and it may have not be able to detect a small increase in the number of hypopneas and apneas. The other study recruited 67 middle-aged men with obesity, and unique to this study, the degree of sleep-disordered breathing prior to treatment was moderate to severe, and all men were required to have some degree of polysomnographically proven OSA (183). In addition to being placed on a hypocaloric diet, men were randomized to also receive three intramuscular injections of either testosterone undecanoate (1000 mg) or matching placebo every 6 weeks. Weight loss occurred in both groups, and the degree of weight loss was similar between groups overall and at every time point. The overall changes in the oxygen desaturation index and the duration of hypoxemia were significantly higher in the 33 men treated with testosterone compared with the 34 men treated with placebo. Further analyses showed that this was due to differences at 7 weeks where testosterone therapy worsened the oxygen desaturation index by 10 events per hour and the duration of hypoxemia by 6%, compared with placebo. These between-group changes from baseline were no longer statistically different by 18 weeks (183). These data show that even replacement-dose more steady-state testosterone therapy can worsen sleep-disordered breathing, at least in the short term. The improvement in sleep-disordered breathing with time requires further investigation,

These four studies in total show that testosterone therapy only modestly worsens sleep-disordered breathing by on average 7 to 10 events per hour, and it does so in a wide range of adult men aged from 20 to 80 years with mild, moderate, and severe OSA. Treatment with either therapeutic or supratherapeutic testosterone doses, including preparations that result in more steady-state blood testosterone concentrations, can worsen OSA. The available studies also suggest the possibility that the worsening in OSA dissipates with longer-term therapy. If true, preliminary observations suggest that this could be due to time-dependent changes in the relationship between hyperoxic ventilatory recruitment threshold to carbon dioxide and blood testosterone concentrations (184), because changes in ventilatory responses can promote apneas (180, 181). Further studies confirming this time dependency are required. Another possibility is that the induction of OSA is dependent on the mode of testosterone delivery. All three randomized controlled trials showing induction of OSA used intramuscular testosterone esters (178, 179, 183), whereas the single study that used transdermal testosterone did not (182). If true, transdermal conversion of testosterone to dihydrotestosterone, which does not occur with intramuscular injections, or some other mechanism related to daily vs long-acting testosterone administration would need to be identified to explain why mode of delivery should influence the effect of testosterone on OSA.

Despite these adverse effects on breathing during sleep, 18 weeks of testosterone therapy in men with OSA increased muscle mass, reduced liver fat, and improved insulin sensitivity and sexual desire compared with placebo therapy (185, 186). Studies advancing our knowledge regarding the relative risks and benefits of testosterone therapy in older men have been recently summarized (187). Further studies examining the risks and benefits of testosterone therapy in men with OSA over the longer term are required. Until such data are available, expert societal guidelines will continue to caution against testosterone use in men with untreated severe OSA (74), despite a paucity of high-quality data examining the effect of testosterone therapy exclusively in men with OSA, and specifically in men with clinically significant OSA.

Summary and Conclusions

Epidemiological data firmly establish that insufficient, misaligned, and disrupted sleep are each associated with increased mortality, hypertension, cardiovascular events, T2DM, insulin resistance, and other metabolic disorders (Table 5). Reproductive disorders, including reduced fecundity and total sperm count (insufficient sleep), erectile dysfunction (disrupted sleep due to OSA), and low testosterone (both) are also features of insufficient sleep and disrupted sleep due to OSA. Alternatively, misaligned sleep changes the diurnal pattern of testosterone secretion, but it does not alter mean testosterone levels and has not yet been conclusively shown to be associated with reproductive disorders such as reduced fecundity in men. However, a potential interaction between the circadian timing of sleep loss and its effect on reproductive health, particularly on testosterone secretion, has been postulated and warrants further investigation.

Table 5.

Summary of Epidemiological Studies

	Sleep Restriction	Circadian Misalignment	Sleep Disordered Breathing
Higher mortality	Yes	Unknown in men	Yes
Hypertension and/or cardiovascular events	Yes	Yes	Yes
T2DM and/or obesity	Yes	Yes	Yes
Lower fecundity or sperm or testicular parameters	Yes	No	Unknown
Lower testosterone	Yes	No	Yesa

^aMight be explained by adiposity.

Carefully conducted experiments verify and extend these observational findings (Table 6). Preliminary data now show that extending sleep in those who are chronically sleep deprived may possibly improve insulin resistance, hypertension, and potentially sperm output, but larger studies in more diverse populations are required. Studies of short-term experimental shift work conditions reproduced in the laboratory now show that misaligned sleep, as occurs in night shift work, induces insulin resistance and hypertension and changes the diurnal rhythm of testosterone secretion without altering the mean concentrations in the short term. Randomized controlled trials now show that adherent CPAP reduces hypertension and cardiovascular events, improves insulin resistance, ameliorates erectile dysfunction, and may increase testosterone, although data for the latter are preliminary and inconsistent. Improved erectile function may motivate the men who highly value this to adherently use CPAP and thus obtain additional cardiometabolic benefits.

Table 6.

Summary of Interventional Studies

	Normal Men		Chronically Sleep Deprived, or Chronically Misaligned, or Men With OSA
	Sleep Restriction	Circadian Misalignment	Sleep Extension	Circadian Realignment	OSA Reversala
Insulin resistance	Induced	Induced	Reversed	Unknown	Reversed
Hypertension	Possibly induced	Induced	Possibly reversed	Unknown	Reversed
Sperm count	Unknownb	Unknown	Increasedc	Unknown	Unknown
Testosterone	Reducedd	Diurnal shift to waking time	Unknown	Diurnal shift to waking timec	Possibly increased

^aReversal with adherent CPAP, mandibular advancement, or uvulopalatopharyngoplasty.

^bReduced viable sperm count in rodents.

^cNaturalistic study.

^dWhen appropriately timed.

Fifteen percent of the US workforce are shift workers in whom complaints of fatigue, lack of energy, and reduced sexual drive are common and overlap those of hypogonadism. The correct assessment of blood testosterone levels to confirm the diagnosis is therefore critical, but societal guidelines make no specific recommendations for shift workers and instead recommend two separate measurements in the early morning drawn when the patient is fasting with free testosterone calculated or directly measured in selected situations for the assessment of hypogonadism (74). This review shows that testosterone should be measured soon after waking, not at a specific clock time, to obtain levels that are most comparable with normative ranges established in healthy non–shift workers (73). This is because the diurnal rhythm of testosterone is educed mainly by sleep, with only a minor endogenous circadian effect. Therefore, blue light exposure (e.g., from late night use of cell phones or computer monitors) should not meaningfully alter testosterone levels, assuming no change in the timing of sleep and no direct effect of blue light on testicular testosterone secretion. Free testosterone may still be calculated, but it should incorporate recent advances in the understanding of the relevant stoichiometry (188). Additionally, this review shows that shift work itself does not decrease (mean) testosterone levels, in the absence of actual sleep loss, but are limited to naturalistic studies of up to 7 days in duration. This is important because if decreases in (mean) testosterone were observed in shift workers, then changing work schedules rather than prescribing testosterone would be the initial treatment of choice.

The effect of circadian misalignment on insulin resistance is increasingly becoming recognized to be a causal mechanism that explains the relationship between shift work and T2DM and other cardiometabolic disorders. Whether such associations lead to increased mortality has not yet been conclusively shown in men, but it is plausible. Available studies examining the effects of shift work on male fertility are probably underpowered to detect an effect of modest size. Although mean testosterone exposure does not change, mistimed testosterone exposure could possibly lead to androgen-sensitive diseases that are more prevalent in shift workers, such as prostate cancer (189), and deserves further investigation. Until such information becomes available, modulation of the circadian timing of testosterone replacement, as might occur with daily testosterone gel application, in men with hypogonadism who happen to be shift workers cannot be definitively recommended. Future research should account for differences in shift work schedules, between permanent night shift vs rotating night shift workers, and include in-laboratory interventional studies. Longer-term longitudinal studies are needed to establish whether mean 24-hour testosterone levels are reduced by prolonged night shift work.

Although more research is needed, the overwhelming evidence of the effects of sleep on numerous biological systems highlights the importance of regular sleep on human health. Accordingly, society needs to prioritize and value sufficient and appropriately timed sleep. In parallel, studies to develop countermeasures to reduce the impact of insufficient sleep or circadian misalignment on cardiometabolic health are also required because insufficient or mistimed sleep may sometimes be unavoidable. Methods to promote rapid resynchronization would seem to be appropriate, as might shift work schedules tailored to promote cardiometabolic health or neurobiological performance. New approaches to promote adherence to CPAP, or novel therapies that are more effective or more easily adhered to, are needed to reduce the impact of disrupted sleep due to OSA. Until that occurs, physicians, including endocrinologists, need to be aware of the importance of sleep on health to properly advise their patients.

Glossary

Abbreviations:

BMI

body mass index

CCP

central circadian pacemaker

CPAP

continuous positive airway pressure

IIEF

International Index of Erectile Function

NREM

non–rapid eye movement

OSA

obstructive sleep apnea

REM

rapid eye movement

SWS

slow wave sleep

T2DM

type 2 diabetes mellitus