Effects of Caffeine Intake on Endurance Running Performance and Time to Exhaustion: A Systematic Review and Meta-Analysis

This systematic review was developed following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2020 statement [27]. The search terms included a mix of Medical Subject Headings (MeSH) and free-text words for key concepts related to caffeine and endurance running performance (see Appendix A). No filters or limits were used for the search. The sources of information were obtained by searching for studies using MEDLINE (via OVID), Scopus, Web of Science, and SPORTDiscus (via EBSCO). The search was established from the start of construction to 5 October 2022. The search strategy was different for each database, and it can be consulted in Appendix A. All titles and abstracts from the search were downloaded to Endnote 20 (Clarivate Analytics, London, UK) and manual cross-referencing was performed to identify duplicates and any potential missing studies. Titles and abstracts were then screened for a subsequent full-text review. The search for published studies was independently performed by two authors (Z.W. and B.Q.) and disagreements were resolved through discussion.

We defined the inclusion criteria according to PICOS principles [28]. We only incorporated in the review studies with crossover experimental designs in which the ingestion of caffeine was compared to a placebo in a single- or double-blind randomized manner and the outcomes were associated with endurance running performance. We considered any form of caffeine intake to be included in the review, but only if the effect of caffeine on endurance performance could be isolated; meanwhile, we excluded dietary supplements and foodstuffs where caffeine was provided in addition to other ergogenic aids (e.g., caffeinated energy drinks). Data from the studies in which a comparison of caffeine and placebo was made with an emphasis on placebo effects [29] or after sleep deprivation [30] were removed, as this may exaggerate or reduce the true effect of caffeine and is also inconsistent with single- or double-blind trials. We only considered samples that were categorized as healthy runners irrespective of their fitness level, and participants were excluded if they reported regular participation in other sports in addition to running (e.g., football, triathlon, etc.). Systematic reviews and meta-analyses were excluded, in addition to those original studies with no full text available, non-peer-reviewed articles, opinion pieces, commentaries, case reports and editorials. See Table 1 below for details of the inclusion criteria.

We performed data extraction in the following topics: (a) study design; (b) sample characteristics and participants’ running experience; (c) training status; (d) caffeine dose and administration form; (e) time of ingestion with respect to exercise onset; (f) running conditions (i.e., on a standard 400 m track, road, or treadmill in laboratory conditions); and (g) endurance performance outcomes, as the time to exhaustion or time employed to complete a trial with fixed distance or amount of work. Two authors performed the data extraction separately (Z.W. and B.Q.). In instances where data were presented in a graphical format, images were enlarged to improve the precision of the data estimates and data were extracted from studies’ figures using WebPlotDigitizer [31]. In the event of disagreements between the two assessors regarding data extraction, they were resolved through discussion and consultation with a third party when necessary.

The Cochrane’s risk-of-bias tool (RoB 2) was used to rate the methodological quality of the included studies [32]. The seven areas covered by this tool are: random sequence generation, allocation concealment, participant and personnel blinding, outcome assessor blinding, insufficient outcome data, selective reporting of results, and other possible sources of bias. The application of RoB 2 for the included studies was performed by two separate authors (Z.W. and B.Q.), and disagreements were resolved through discussion and consultation with a third party when necessary.

The Comprehensive Meta-Analysis software (version 3; Biostat, Englewood, NJ, USA) was used to compare the effects of caffeine vs. placebo ingestion on study outcomes using standardized mean differences (SMDs), estimated by Hedges’ g and 95% confidence intervals (CI). All the meta-analyses were performed using the random-effects model. We considered the SMD of ≤0.20, 0.20–0.49, 0.50–0.79, and ≥0.8 as small, medium, large and very large, respectively, based on a previous categorization [33]. The statistical significance threshold was set at p < 0.050. For each outcome, the SMD was calculated using mean and standard deviation values from the placebo and caffeine trials, the sample size from each study, and the correlations between the trials. Only two studies [34,35] reported correlation values for the performance data between the placebo and caffeine conditions, with a 0.87 correlation for Whalley et al. [34] and 0.86 correlation for Bell et al. [35], respectively. For the remaining studies, a 0.5 correlation was assumed, as recommended by Follmann et al. [36]. For each outcome, a minimum of three studies were required to perform the meta-analysis [37]. In studies with several caffeine–placebo comparisons, such as different doses or forms of caffeine administration, we obtained data independently for each caffeine–placebo comparison. We assumed that the sample size was the same in each separate caffeine–placebo comparison, as each participant completed all testing conditions in multi-arm studies [38]. A subgroup analysis was conducted on the effects of caffeine on endurance running performance based on the participants’ training levels (trained vs. recreational). The assignation of each study’s sample to the subgroup of trained vs. recreational runners was based on the participants’ characteristics reported in the study. For instance, when the sample was described as “elite or sub-elite”, “varsity”, “competitive” or “well-trained”, the participants were included in the subgroup of trained runners. If the sample was described as “amateur” or “recreational”, the participants were included in the subgroup of recreational runners. An additional subgroup analysis was conducted on the effects of caffeine on endurance running performance based on the distance covered during the running trial. The outcomes of running trials between 800 and 5000 m were included in the subgroup of middle-distance, and running trials covering more than 5000 m were included in the subgroup of long-distance. Finally, the magnitude of caffeine’s effect on endurance performance was presented as the percentage of change calculated by the formula proposed by Lopez-Gonzalez et al. [39].

We used the I² statistic to assess the degree of heterogeneity, with values ≤50% indicating low heterogeneity, 50–75% indicating moderate heterogeneity, and >75% indicating high heterogeneity [40]. Potential asymmetries in funnel plots and Egger’s linear regression test were employed to detect publication bias [41].

We obtained 6193 studies through database searching. After deleting duplicates and screening the titles and abstracts of all the remaining original papers, we evaluated 127 full-text articles for inclusion in our research. We eliminated 106 papers for various reasons, leaving only 21 studies to be included in the systematic review (Figure 1). The experiments that met the inclusion criteria were published between 1991 and 2022.

In the 21 studies included in this systematic review [7,23,24,25,26,30,34,35,42,43,44,45,46,47,48,49,50,51,52,53,54], there was a total sample of 254 participants, including 220 men, 19 women and 15 participants with no information about gender. The participants were all runners, of which 167 were categorized as amateur and 87 were categorized as trained runners. All studies were crossover randomized controlled trials. A total of 18 studies [7,23,24,25,26,30,35,42,43,45,46,48,49,50,51,52,53,54] provided caffeine in liquid or capsule form, with doses normalized by participants’ body mass. In these studies, the doses of caffeine administered ranged from 3 to 9 mg/kg. A total of three studies [34,44,47] provided absolute doses of caffeine in the form of caffeine powder, gum, and mouth strips with doses ranging from 200 to 300 mg. The general data of the experiments included in this systematic review are depicted in Table 2.

Figure 2 displays the categorization for each RoB 2 item for each included study. Regarding selection bias, we judged only one study [53] as low risk because it reported the appropriate method of participant randomization sequences. We defined all other studies as unclear because none provided sufficient information for this item. In addition, the allocation concealment process was not detailed in any of the studies; therefore, we rated their risk as unclear for allocation concealment. We considered that four studies were unclear regarding the blinding process of participants and researchers [35,47,50,51]. The remaining studies were all low risk for this item. Only two studies [7,26] described the blinding of outcome assessors, and they were rated as low-risk. All other studies did not provide sufficient information; therefore, they were rated as having an unclear risk. Most of the included studies [7,23,24,26,30,35,43,44,46,47,49,50,52,54] had no reported participant dropouts, and therefore they were rated unclear. None of the included studies had trial registration protocols. We identified no other sources of bias in the included studies. Thus, we categorized the overall methodological quality as with an unclear-to-low risk of bias.

Both time-to-exhaustion and time-trial exercise test protocols are commonly used to examine the influence of experimental interventions on endurance performance. In the context of endurance running, time-to-exhaustion protocols represent an unusual scenario with respect to real-environment running [22], as most endurance running events consist of completing a given distance as fast as possible. However, time-to-exhaustion runs provide a measure of an athlete’s endurance capacity and they are considered a valid measure of endurance performance [56]. In the current systematic review, seven studies, including 12 unique placebo–caffeine comparisons, tested the effect of caffeine on running protocols that entailed participants performing submaximal intensity running until they could no longer maintain the required speed. Overall, the values of time to exhaustion were higher with caffeine than in the placebo situation, and the meta-analysis revealed a statistically significant effect of caffeine in this performance variable (Figure 3). Caffeine promotes the production of ß-endorphins and dopamine [57], which can lessen perceived effort and discomfort [58,59]. These factors can explain the ergogenic effect of caffeine on time to exhaustion runs, as the “exhaustion time” when runners decide to stop running is associated with feelings of fatigue and muscle pain. Although running events consisting of “run as long as you can” are scarce, the higher time to exhaustion induced by caffeine may apply to ultra-endurance running events where athletes are required to run in scenarios close to exhaustion, such as 24 h races and 100-mile races or mountain running events. It is important to note that performance in time-to-exhaustion trials is habitually more variable than in time trials, especially for those with a lower training background [60,61]. However, our subgroup analyses indicated that caffeine ingestion significantly increased run-to-fatigue time to a moderate degree in both recreational and trained runners. In summary, it seems safe to conclude that caffeine is an effective substance to enhance running performance in situations that entail running until volitional fatigue.

The use of time trials offers a better scenario to study the effect of caffeine on endurance running performance, as runners’ performance is more reproducible in time-trial runs than in running to exhaustion events [56]. This likely explains why 14 studies used time trials while only seven studies used time-to-exhaustion runs to determine the effect of caffeine on running performance. The meta-analysis revealed that caffeine has a small effect but significant ergogenic effect to reduce time in endurance running for a given distance (Figure 4). The ergogenic effect of caffeine on running time trials was lower compared to the effect of this substance on time-to-exhaustion runs (0.71% ± 0.83% vs. 16.97% ± 14.65%, respectively). The explanation for the different magnitude of caffeine’s benefits on these two types of endurance performance running trials may be associated with the characteristics of each running protocol. In time-to-exhaustion runs, the intensity is fixed, and runners cannot adjust the pace during the running trial. However, in time trials, like in real running races, runners may adjust their pace depending on their feelings of fatigue and the distance left to complete the trial [62]. As an adenosine antagonist [5], caffeine can decrease fatigue by crossing the blood–brain barrier and inhibiting A₁ and A₂ monoadenosine receptors [5,9]. So, caffeine may have an inhibitory effect on perceptual response during exercise, which may give participants an increased ability to tolerate the discomfort associated with fatigue during exercise, effectively masking the sensation of fatigue [58]. Therefore, caffeine intake may produce a faster running speed than placebo intake with the same perceptual response [63], thereby influencing running pacing strategies [64]. Independently of the type of test employed, caffeine produced statistically significant benefits for both time-to-exhaustion runs and time trials, suggesting that caffeine is a substance with the potential for enhancing performance in different endurance running scenarios.

Interestingly, when analyzing the ergogenic effect of caffeine on time-trial runs according to participants’ training status, the benefit of caffeine was not statistically significant for trained runners or recreational runners (Table 3). Previous studies indicated that caffeine is more effective in reducing the time to complete specific running distances in trained athletes, and trained runners are more reliable [60,61]. Similar results have been found in other types of exercise, such as swimming, as it has been suggested that caffeine has a facilitative effect on trained but not “active” swimmers [65]. Although more studies are granted on this topic, we can assume that training status is not a modifying factor for the ergogenic effect of caffeine on endurance performance. Likewise, when considering the distance of the running time trials, caffeine did not reduce the time to complete middle-distance or long-distance trials (Table 3), despite it previously being suggested that caffeine is more effective in long endurance events than in short but intense endurance events [38,66]. Overall, the effect of caffeine on improving running time trial performance was statistically significant but small, which demonstrated that the potential ergogenic effect of caffeine was the same across subgroups of training status and distance.

Although we only selected blinded randomized controlled studies, the placebo choice appears to be diverse in some studies (e.g., sugar-free, sugary, and decaffeinated), and caffeine was also taken with other compounds (e.g., carbohydrates and artificial sweeteners). Thus, it is possible that the actual effects of caffeine may be masked or exaggerated by some of these substances. In addition, different sources of caffeine (capsules, tablets, gums, beverages, etc.) may affect the pharmacokinetics of caffeine [11], thus influencing the results of different studies. Another limitation is that, due to the differences in participants’ characteristics and form of caffeine administration, it was not possible to meta-analyze the effect of participants’ sex, form of caffeine intake or caffeine dose on the benefits obtained with caffeine supplementation. In this regard, only 7.5% of the pooled sample of this systematic review were women, suggesting that the outcomes of this study possess a sex bias. For this reason, it is perhaps more accurate to indicate that caffeine’s ergogenicity on endurance running performance in men is supported by the results of this study, while the benefits of this substance for women endurance runners should be further investigated. Last, in the current analysis, we only focused on running performance variables, namely the time to exhaustion and the time to complete a given distance. The study of the effect of caffeine on physiological and perceptual responses during endurance running such as heart rate, blood lactate concentration and ratings of perceived fatigue requires further investigation and analysis, as these responses may contribute to explaining the ergogenic effect of caffeine. From a methodological perspective, it should be mentioned that most of the studies were classified as having low risk of bias, as they all utilized randomized crossover designs. Therefore, the outcomes of the current study are not influenced by the inclusion of studies with poor methodological quality, which strengthens the main conclusions.

The data presented in this study are available on request from the corresponding author (J.G.).