The search strategy returned 1676 records. Of the 91 studies retained for full-text screening, we excluded 54 which did not meet the set criteria. Thirty-seven studies were therefore eligible for review.
Characteristics of Included Studies
The 37 studies were published between 2001 and 2016 and included Olympic/Commonwealth (n = 21, 57%), Paralympic (n = 4, 11%), and professional (n = 12, 32%) sports. The number of participants in the included studies ranged from 6 to 2067, with an overall age range of 18–30 years. Female athletes were under-represented generally, with only eight (21%) of the studies reporting values exclusively for women.
‘Eliteness’ of Athletes
The application of the full Swann et al.  taxonomy was limited by participant descriptions within the included studies. Only a minority of studies (n = 6, 16%) reported ‘athlete level of experience’, with only one study reporting ‘athletes’ level of success’. As a result, participants were categorised using a modified taxonomy within which only ‘semi-elite’ or ‘competitive elite’ categories could be judged (see Electronic Supplementary Material Table S2). Accordingly, 20 studies (54%) were judged to have recruited ‘competitive elite’ participants.
Evidence Quality Appraisal
The evidence quality appraisal of the 37 studies can be seen in Electronic Supplementary Material Table S1. Given the diversity of study designs, reporting standards, outcome variables, and the very limited number of control participants, a meaningful calculation of risk (as odds ratios) and a subsequent meta-analysis was not possible. However, where studies reporting sleep quality outcomes used similar instruments, pooled estimates of prevalence were calculated. Relative to the objectives of this review, studies fell into three categories: (1) studies describing sleep structure and patterns; (2) studies describing sleep quality and insomnia symptomatology; and (3) studies exploring sport-related risk for sleep disturbances.
Overall, evidence quality of the selected studies was generally ‘low’ (mean NOS score = 5, standard deviation [SD] = 2), with 23 studies (62%) scoring <5 (low quality) and only two (5%) scoring >7 (good quality). Study designs employed were generally observational (n = 34, 92%), with only three (8%) of this sample employing control-group designs. Of the observational studies, 18 (49% of all studies) were cross-sectional and 14 (38% of all studies) were longitudinal. Very few studies (14%, n = 5), adequately reported participant level of performance, age, sex, sport, and level of experience. Just under half of the studies (49%, n = 18) provided a clear description of the protocol employed to measure sleep and used validated instruments, whilst adhering to measurement standards.
Sleep Structure and Patterns
A total of 20 studies, published between 2001 and 2015, describing typical sleep profiles, assessed using wrist actigraphy (n = 11), PSG (n = 2), sleep diaries (n = 4) and questionnaires (n = 4), for mainly male elite athletes engaged in normal training are shown in Table 2. Earlier studies showed a preference for PSG, while the more recently published studies utilised wrist actigraphy and self-report inventories principally focussing on total sleep time, sleep efficiency and sleep onset latency. Only one actigraphy study compared athletes with controls. Leeder and colleagues  found no significant difference in total sleep time between 46 Summer Olympic athletes and 20 age-matched non-athletes, but did report a significantly lower sleep efficiency, and significantly higher time in bed, wake time after sleep onset, sleep onset latency and sleep fragmentation in athletes. Differences between sports were reported in three actigraphy studies [32, 50, 51]. In a comparison of team and individual sports , individual competitors showed significantly lower total sleep times and sleep efficiencies, and longer sleep latencies. On the other hand, a comparison of canoeing, diving, rowing and skating athletes reported the lowest total sleep times, the shortest sleep onset latencies, but the highest sleep efficiencies for rowers . Consistent with these findings, individual sports were reported to have shorter total sleep times than team sports and napped more frequently in the day (15% of 754 days) than team sports (11% of 613 days) . In the only study to compare sex , the time in bed of male athletes was reported to be 54 min longer than that for female athletes.
Summary of Sleep Structure and Patterns
Instrumental measurements indicate that while the typical sleep duration of elite athletes may be similar to that of non-athletes, structural differences suggest a more fragmented, lower quality sleep among athletes, with most actigraphy studies reporting athlete sleep efficiencies below 90% (see Table 2). Subjective estimates of sleep duration among athletes are broadly consistent with instrumental measures. Both the duration and structure of sleep showed between-sport and sex differences; total sleep time is shorter in individual (versus team) sports, and also shorter in women.
Sleep Quality and Insomnia Symptomatology
A total of 12 studies published between 2007 and 2016 reported data on athlete subjective sleep quality and general insomnia symptoms during normal training. While these studies utilised a range of subjective metrics, six used the Pittsburgh Sleep Quality Index (PSQI), a 19-item scale which assesses seven ‘components’ of sleep (sleep quality, sleep efficiency, sleep onset latency, sleep duration, sleep disturbance, daytime dysfunction and sleep medication use), summing the ‘component scores’ to deliver an overall ‘global’ score; global scores >5 indicate ‘poor sleepers’ . Studies reporting general characteristics of sleep quality or PSQI values are presented in Tables 3 and 4, respectively. Data in Table 3 are summarised as prevalence rates of sleep symptoms. Studies administering the PSQI adopted different reporting conventions; both mean scores and threshold (e.g. >5) prevalence rates are therefore shown in Table 4.
Overall, the sleep assessments shown in Table 3 show a relatively high level of sleep complaints, with reports of sleep disturbance ranging from 13 to 70%. The low prevalence of ‘abnormal sleep’ reported by Samuels et al.  involved an arbitrary cut-off applied to an as-yet unvalidated scale and may not, therefore, represent a robust estimate. In the two studies which reported the prevalence of sleep disturbance by sex [68, 69], rates were highest among women. One study  explored this further, reporting that female athletes experienced more problems both initiating and maintaining sleep when compared with their male counterparts. Sleep quality differences between sports were also identified in this study , with elite French athletes from aesthetic sports reporting a significantly higher prevalence of insomnia symptoms (33%) compared with all other sports (26% for the sample overall).
Formal (PSQI) assessments of sleep quality suggest similarly high levels of insomnia-type symptoms (Table 4), with mean values at  or above the threshold of >5 [1, 62, 63, 65, 70]. This assumption is supported by the prevalence rates reported, with levels of significant sleep disturbance ranging from 38% of multi-sport athletes  to 57% of bobsleigh competitors . The more conservative PSQI threshold of >8, indicative of highly disturbed sleep, also showed a relatively high prevalence, ranging from 22 to 26% [1, 63]. However, the possibility that the higher thresholds shown in Table 4 may mask more severe symptoms is suggested by Swinbourne et al. , who reported that 9% of elite Australian team sport athletes scored >10. Only one of the studies shown in Table 4 included control comparisons. Tsunoda et al.  compared the PSQI global scores of 14 international wheelchair basketball athletes (mean = 6; SD = 3) with 103 non-athletes (mean = 5; SD = 2), and found the difference significant (p < 0.05). In the same study, PSQI component score data also showed that athletes reported significantly lower subjective sleep quality and sleep efficiency, even though reported total sleep time showed no significant difference between the groups.
While the PSQI is not a diagnostic tool , four studies [30, 65, 68, 69] in this section used instruments validated against insomnia diagnostic criteria which allow inferences to be drawn regarding the prevalence of insomnia cases in elite athlete populations. Using the Athens Insomnia Scale (AIS—an instrument validated against the 10th revision of the International Statistical Classification of Diseases and Related Health Problems criteria for insomnia ), Dickinson and Hanrahan  reported a mean score for elite multisport athletes of 5 (range 0–16). Since scores of ≥6 indicate clinically significant insomnia symptoms, and since the reported score distribution from this study showed no significant skewness or kurtosis , then it can be assumed that, while the study does not report the prevalence of ≥6 scores, a high proportion of athletes must nevertheless have experienced serious insomnia symptoms. Consistent with this assumption, Dickinson and Hanrahan  also reported relatively high levels (for this age group) of daytime fatigue among athletes, together with consistent reports (from qualitative interviews) of nonrestorative sleep despite apparently adequate sleep durations. Using the Sleep Disorders Questionnaire (SDQ; a brief questionnaire validated against DSM-IV criteria for insomnia ), Lucidi et al.  reported that 4% of Italian Olympic athletes met diagnostic criteria for insomnia. Details of symptom chronicity, closely related to insomnia diagnosis, were provided by Schaal et al. , who reported a 6-month prevalence of insomnia symptoms of 22%, but a lifetime prevalence of insomnia symptoms of 27%, strongly indicating very high levels of sleep pathology within this nationally representative sample of elite French athletes.
Summary of Sleep Quality and Insomnia Symptomatology
The general pattern of results indicates high levels of subjective sleep disturbance and insomnia symptomatology within elite sport, with the evidence suggesting that, within athlete populations, levels of sleep disturbance are higher among women, and among aesthetic athletes. Formal measurements of subjective sleep in athletes also show findings which accord with the objective data, with similarities reported in the total sleep time of athletes and non-athletes, but significantly lower levels of sleep quality reported by athletes. Such evidence, together with that presented in the preceding section, also show that levels of daytime fatigue in athletes can be directly related to degraded night-time sleep.
Risk Factors for Sleep Disturbance
Studies reporting sleep quality, insomnia symptomatology and changes in sleep patterns broadly focussed on three challenges to athlete sleep: (1) competition (see Tables 5, 6); (2) travel (see Table 7); and (3) training (see Table 8).
Of the studies assessing sleep quality pre-competition, most employed the PSQI (n = 7), with five reporting an increased prevalence of complaints (Table 5). Silva and Paiva  found that 78% of international female gymnasts scored >5 (indicative of ‘poor’ sleep) on the PSQI prior to an international competition. Gymnasts who scored more ‘competition points’, however, reported significantly worse sleep quality (mean PSQI = 8) than those who scored less (mean PSQI = 6). Sex and inter-sport differences were considered in two studies [37, 38]. Using the Competitive Sports, Sleep and Dreams Questionnaire (CSSDQ), a metric designed to assess sleep habits and disturbances prior to competition , two studies reported high prevalence rates of pre-competition sleep disturbance (64–66%), but found no differences between male and female athletes [37, 38]. When comparing sports, however, Erlacher et al.  reported a significantly greater frequency of sleep disturbances in individuals (69%) when compared with team sport athletes (60%). Such differences were not supported by Juliff et al. , who reported similar levels of sleep disturbance between sports.
Of six studies which used wrist actigraphy to assess sleep patterns prior to competition, most reported no significant changes in sleep efficiency and sleep onset latency when compared with normal training. Two studies, however, reported a significant increase in pre-competition total sleep time [39, 40], while one study  reported a significant increase in sleep efficiency. Again, however, there was evidence of sleep–performance relationships. Chennaoui et al.  reported that elite swimmers who finished above fourth position at the French national championships exhibited more consistent total sleep times across the competition when compared with swimmers who finished fourth or below, with this latter group reporting significantly longer total sleep times the night before the final race.
The five actigraphy and one sleep diary study shown in Table 6 all reported a significant decrease in total sleep time, and a significantly delayed bedtime, following night competitions, with no study reporting significant changes in sleep efficiency or sleep onset latency. Actigraphy studies assessing post-competition sleep quality more generally were equivocal, with three studies showing a significant decrease [39, 41, 78] and one study showing no change . Using sleep diary assessments, Fullagar et al.  also reported a decrease in sleep ‘restfulness’ in elite football players following a night competition, compared with day matches and training days. In the only study assessing PSG-measured sleep structure, Netzer et al.  reported a significant increase in stage 3 sleep following competition (compared with rest days) and a significantly increased REM sleep onset latency. However, no changes in sleep onset latency or sleep efficiency were reported.
Studies investigating the impact of both long- (n = 4) and short-haul (n = 3) travel on athlete sleep are shown in Table 7. Overall, studies used a range of designs and methodological approaches. Three of the studies investigating long-haul travel reported a rating of jet lag in addition to sleep outcomes [41, 42, 58]. Time zone changes ranged from 1 to 11 h, with the maximum time zone change when travelling east being 8 h and the maximum when travelling west being 11 h. Across all studies, however, no change in sleep onset latency or sleep efficiency was reported following travel compared with pre-travel assessments. Changes in total sleep time and sleep quality reported were mixed, with one study showing a significant decrease in total sleep time following long-haul eastward travel  and another showing an increase following long-haul westward travel . The majority of studies reported no change in sleep quality; however, Richmond et al.  reported a significant decrease in sleep quality prior to away matches following a 2-h eastward time-zone change (mean score = 3.4), when compared with home matches (mean score = 3.8) in elite Australian Rules Football players. Ratings of jet lag following westward travel (no assessments of jet lag were reported following eastward travel) showed a positive trend with increasing time-zone change, with a 1, 4 (both p < 0.05) and 11 h time zone (p < 0.01) change showing a significant increase from pre-travel assessments.
Studies investigating the impact of training on sleep also showed methodological differences, either comparing training days or rest days, or comparing intensified training with normal training. Most studies reported instrumental measures to assess changes in sleep patterns with only one study using a questionnaire. In studies comparing training and rest days, all studies reported significantly earlier rise times and decreased total sleep time on training days (p < 0.01). Sargent et al.  reported a significant total sleep time gradient relative to training start times across different sports, with earlier start times associated with lower total sleep times and greater pre-training levels of fatigue. Only two studies assessed sleep quality, with one of these reporting a reduction in ‘sleep restfulness’ on a training day compared with a rest day. Three studies quantified levels of daytime sleep [32, 33, 60]. Sargent et al.  found elite athletes to nap at similar frequencies on training (15% of 14 days) and rest days (16% of 14 days). However, Kölling et al.  reported that the proportion of elite rowers who napped on training days (43%, n = 24) was greater than that for rest days (16%, n = 9).
In the studies reporting comparisons between intensified and normal training, significant decreases in total sleep time were observed in all actigraphy studies [53, 60]. However, changes in rise time, sleep onset latency, sleep efficiency and sleep quality were equivocal. Schaal et al.  reported a significant decrease in sleep efficiency and increase in sleep onset latency, but no change in sleep quality during two weeks of intensified training when compared with baseline values in elite synchronised swimmers. Consistent with this, Juliff et al. , using questionnaire assessments, reported that 28% of elite athletes experienced sleep disturbances during periods of heavy training.
Summary of Risk Factors
Among elite athletes, predictable events in the training/competition cycle are associated with an increased risk of insomnia symptomatology and disturbed sleep patterns: competition, long- and short-haul travel, and training. Typically, sleep quality significantly declines prior to competition for both men and women. Following competitions, the impact on sleep is related to the timing of events, with late-evening competitions delaying bed times and reducing total sleep time. While the circadian de-synchrony (jet lag) associated with long-haul travel significantly affects sleep patterns, it appears that sleep quality, and instrumental indices of sleep quality such as sleep onset latency and sleep efficiency, are more resilient. Nevertheless, few studies of jet lag or travel fatigue in elite athletes have used formal assessments of insomnia symptoms. Finally, training days can require earlier rise times, with consequent reductions in total sleep time, increased daytime fatigue, and an increased likelihood of daytime napping in some sports.