Probiotics and synbiotics may affect sleep in people who exercise: a review of controlled trials

Inclusion criteria were: (a) randomised controlled design (RCT); (b) populations comprising athletes or individuals engaged in exercise; (c) original studies examining short- or long-term effects of probiotic and synbiotic supplementation; and (d) sufficient reporting of sleep-related outcomes. Exclusion criteria comprised: (a) in vitro, in silico, or in vivo animal experiments; and (b) ecological, cross-sectional, or case-control designs; non-randomised trials; and secondary research such as systematic reviews or meta-analyses. Records identified in the initial search were screened using EndNote X9.3.3. Study titles and abstracts were independently screened against the inclusion criteria by two reviewers (EM and FT). Records meeting eligibility at the title and abstract stage were advanced to full-text assessment. Discrepancies during screening were resolved through discussion; when consensus was not reached, a third senior reviewer (ED) adjudicated the final decision.

In this study, the outcomes were categorised into primary and secondary outcomes (Table 1). The primary outcomes consist of sleep quality, sleep efficiency, sleep latency, sleep quantity, theta wave, and delta brain waves. In addition, the secondary outcomes across the included studies were fatigue, anxiety, depression, heart rate, electrodermal response, DVT accuracy, motivation, muscle soreness, inflammatory markers such as CRP, STAI-state and STAI-trait anxiety, HADS anxiety and depression, Chalder Fatigue Scale scores (general, physical, and mental fatigue), salivary chromogranin A, cortisol, perceived stress, dopamine, sleepiness, tenseness, and general mood.

Two separate researchers extracted the data from each selected article (AD and KD). The following information was extracted from each study: first author's name, publication year, country, study design, duration of intervention, subjects' age and sex, training status, type of intervention (probiotics and synbiotics), number of participants, comparison group, and findings of the included studies. In addition, Discrepancies in data extraction were resolved through discussion between the two extractors; if consensus could not be reached, a third senior reviewer (ED) adjudicated.

Risk of bias in the included studies was independently assessed by two authors (FA and ME) using RoB 2 (the revised Cochrane risk-of-bias tool for randomised trials). RoB 2 evaluates five domains: the randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Judgements within these domains inform an overall rating of “low risk of bias,” “some concerns,” or “high risk of bias” [21]. Additionally, the results of these assessments were visualised using the online robvis tool.

According to the SWiM criteria, the findings of the included articles were narratively synthesised and reported [22]. SWiM was developed to enhance transparency in reporting narrative synthesis, based on systematic reviews that do not include meta-analysis. In this review, meta-analysis was not feasible, even under random-effects assumptions, owing to substantial heterogeneity in supplement interventions, population characteristics, outcome measures, and study designs. Following SWiM guidance, studies were grouped by intervention type (probiotic vs synbiotic) and by sleep outcome domain, including sleep quality, sleep latency, sleep efficiency, sleep quantity, and EEG-derived outcomes. Vote-counting was used based on the direction of effect for each prespecified sleep outcome, using between-group comparisons at the end of the intervention. Outcomes were classified as favouring supplementation if the intervention group improved more than the comparison group, as no effect if there was no between-group difference, and as negative if results favoured control or placebo. Quantitative and qualitative findings were summarised using the direction of effects, the p-value for each outcome, and combined P-values via Fisher’s method, as implemented in the Corbi packages within R, for each intervention; these results are presented in Table 2. Furthermore, Outcome data were extracted from per-protocol analyses as reported in the included trials. The combined P-value was calculated only for studies that reported a significance level. Key study characteristics were qualitatively summarised in Table 1, and quantitative results were tabulated separately. To facilitate transparent presentation of the direction-of-effect synthesis, the results were summarised visually using a harvest-style plot.

Following the removal of duplicates, 184 records were screened, and 172 were excluded based on titles and abstracts. Twelve full-text articles were assessed, of which 6 randomised controlled trials met the eligibility criteria and were included in this review (Figure 1) [13-18]. Additionally, the reasons for excluding the remaining six full-text articles are reported in Figure 1. The included trials comprised four probiotic [13,14,16,18] and two synbiotic [15,17] interventions in exercised populations; intervention duration ranged from 4 to 17 weeks, and publications spanned from 2019 to 2024. All included trials were randomised and blinded. The included articles were five double-blind RCTs [13,14,16-18] and one triple-blinded pilot RCT [15]. Also, trials were conducted in Malaysia [13], Australia [14], Japan [16], Spain [15], Brazil [17], and Poland [18], which they are covering Asia, Europe, South America, and Oceania. In addition, all studies enroled trained populations, including football players [13], rugby players [14], runners [16], female dancers [18], soccer players [15], and military personnel [17]. Four trials were male-only [13-16], one was female-only [18], and one included both sexes [17]. Across studies, sample sizes ranged from n = 13 (15) to n = 65 [17], with a total of 180 participants across the included RCTs. The characteristics of the included studies are summarised in Table 1.

The primary outcomes consist of sleep-related outcomes such as sleep quality (n = 5) [14-18], sleep efficiency (n = 1)[15], sleep latency (n = 2) [15,18], sleep quantity (n = 1) [14], theta wave (n = 1) [13], and delta brain waves (n = 1) [13].

Significant between-group improvements were observed for theta and delta brain waves at week 4 (p < 0.05 for both of them) [13]. Significant improvements were also reported for sleep quality and sleep quantity (both p < 0.05) [14]. A positive, non-significant trend was noted for sleep quality (p ≥ 0.05) [18], whereas sleep latency (p ≥ 0.05) [18] and sleep quality (p < 0.05) [16] showed no between-group effect. As mentioned in Table 2, the combined p-value for probiotic trials (Fisher’s method) was p < 0.01 [13,14,16,18]. Figure 2 presents a harvest-style plot summarising the direction of effects across sleep outcome domains and intervention types.

In athletes, sleep efficiency and sleep latency improved significantly (both p < 0.05) [15]. In military personnel, sleep quality improved (p < 0.05), and sleepiness decreased (p < 0.05) [17]. The combined p-value for synbiotic trials was p < 0.001 [15,17] (Table 2) (Figure 2).

Pooling all twelve sleep outcomes from six RCTs, the direction of effect favoured supplementation (positive n = 9; no-effect n = 3; negative n = 0). Where exact p-values were reported in the source trials, they ranged from 0.009 to 0.418. Fisher’s combined test across all sleep outcomes indicated a significant overall effect, p < 0.0001 [13-18] (Table 2).

Regarding secondary outcomes, Wiącek et al. found no significant effects on fatigue, anxiety, or depression [18], while Adikari et al. observed no changes in heart rate, electrodermal response, or DVT accuracy [13]. In one study, better sleep was associated with higher motivation, longer sleep, lower muscle soreness, and reduced inflammation [14]. Also, significant reductions in anxiety, depression, fatigue, and salivary chromogranin A were demonstrated, though cortisol remained unchanged [16]. A decrease in stress, anxiety, and depression, with increased dopamine in athletes, was highlighted [15], and significant improvements in sleepiness and tenseness, alongside transient mood changes, were shown [17].

Risk of bias was appraised using the RoB 2 tool across five domains (Figure 3). Randomisation and adherence to intended interventions were consistently judged low risk across all trials. Also, Randomisation procedures were not consistently described in the included RCTs. For missing outcome data, judgements were low risk in five trials, with some concerns in one trial [18]. For the measurement of the outcome, all studies were rated low risk. In contrast, selection of the reported result was the principal concern: some concerns were assigned in five trials [13-17], and high risk in one trial (Wiącek et al.) [18]. This was primarily because outcome prespecification was unclear in the five trials rated as ‘some concerns’, while the trial rated as ‘high risk’ showed indications consistent with selective outcome reporting. Overall risk of bias was therefore some concerns in five trials [13-17] and high risk in one trial [18].

Strengths of the evidence include randomised, blinded designs in genuinely trained cohorts spanning multiple sports and stress profiles, which improve ecological validity for performance settings.

Across included trials, the total sample size was small (n = 180), limiting precision and generalisability. However, samples are small and predominantly male. Four of the six trials were conducted exclusively in men, with only one female-only trial and one mixed-sex trial; therefore, the generalisability of these findings to female athletes remains uncertain, and interventions vary widely in strain composition and matrix. Interpretation across trials is further limited by differences in product characteristics; notably, one included study used a heat-inactivated non-viable preparation (Sawada et al.), which may not be directly comparable to live probiotic formulations [16]. In addition, intervention durations were relatively short, 4–17 weeks, which may limit the ability to detect durable changes in sleep. Sleep outcomes are heterogeneous, including PSQI subscales, latency/efficiency, spectral EEG features, athlete-specific tools, and device metrics, which constrain comparability and preclude a single pooled effect size. Moreover, outcomes relied largely on subjective sleep measures, while objective sleep assessment, such as actigraphy, was limited, and polysomnography was not used. Risk of bias assessments indicate “some concerns” for selective reporting in several trials and “high risk” in one, and meta-analytic work in the general population highlights potential publication bias [12,14]. These concerns were driven primarily by the RoB 2 domain in the selection of the reported result. In several trials, outcome prespecification was unclear, and one trial was rated high risk due to indications consistent with selective outcome reporting. Together, these issues argue for cautious interpretation of effect size while acknowledging a consistent direction of benefit.

For practice, probiotic or synbiotic supplementation can be considered as an adjunct to established sleep hygiene and scheduling strategies during periods of predictable sleep burden, such as congested fixtures, intensified training blocks, or travel. Expectations should be calibrated: improvements are most likely to manifest in perceived sleep quality and, in some cases, latency, rather than immediate, robust changes in actigraphy-derived duration or efficiency over short cycles. Given strain specificity and the current predominance of multi-species blends, formulations with prior evidence in stress-exposed or athletic contexts are a reasonable starting point, while monitoring both subjective and objective metrics to track response [14,15,17,18]. Also, regarding the limitations of the current evidence base, probiotics and synbiotics should be considered an adjunct rather than a replacement for established sleep strategies.

Future research in exercised populations should pre-register primary sleep endpoints and adopt a core outcome set that pairs validated questionnaires, such as PSQI or insomnia scales, with objective measures like actigraphy and, where feasible by using polysomnography. To improve consistency across trials, objective outcomes should include actigraphy-derived total sleep time, sleep efficiency, sleep latency, and wake after sleep onset, and assessment windows should be aligned with predictable high-stress periods to reduce noise and enhance sensitivity. Mechanistic markers such as salivary melatonin should be assessed at baseline and at prespecified timepoints during the intervention, with follow-up at approximately four to six weeks and again at the end of the intervention, which is commonly eight to sixteen weeks. Trials should be adequately powered, extend beyond eight to twelve weeks to assess durability, and move toward strain-resolved, dose–response designs, including head-to-head comparisons of probiotic versus synbiotic approaches. Mechanistic integration—microbiome profiling, metabolomics, and endocrine markers such as melatonin and cortisol—will be essential to map causal pathways from microbial modulation to sleep outcomes [7–9,12,14]. Finally, broader sampling across sexes, sport types, and competitive calendars is needed, with reporting that is SWiM-ready (complete statistics, exact P values, and transparent outcome hierarchies) to facilitate future synthesis and, ultimately, meta-analysis [21,22].

The data extracted and analysed during this study are available from the corresponding author upon reasonable request.