The Effects of Sprint Interval Training on Physical Performance: A Systematic Review and Meta-Analysis

Abstract

Hall, AJ, Aspe, RR, Craig, TP, Kavaliauskas, M, Babraj, J, and Swinton, PA. The effects of sprint interval training on physical performance: a systematic review and meta-analysis. J Strength Cond Res 37(2): 457–481, 2023—The present study aimed to synthesize findings from published research and through meta-analysis quantify the effect of sprint interval training (SIT) and potential moderators on physical performance outcomes (categorized as aerobic, anaerobic, mixed aerobic-anaerobic, or muscular force) with healthy adults, in addition to assessing the methodological quality of included studies and the existence of small study effects. Fifty-five studies were included (50% moderate methodological quality, 42% low methodological quality), with 58% comprising an intervention duration of ≤4 weeks and an array of different training protocols. Bayesian’s meta-analysis of standardized mean differences (SMD) identified a medium effect of improved physical performance with SIT (ES_0.5 = 0.52; 95% credible intervals [CrI]: 0.42–0.62). Moderator analyses identified overlap between outcome types with the largest effects estimated for anaerobic outcomes (ES_0.5 = 0.61; 95% CrI: 0.48–0.75). Moderator effects were identified for intervention duration, sprint length, and number of sprints performed per session, with larger effects obtained for greater values of each moderator. A substantive number of very large effect sizes (41 SMDs > 2) were identified with additional evidence of extensive small study effects. This meta-analysis demonstrates that short-term SIT interventions are effective for developing moderate improvements in physical performance outcomes. However, extensive small study effects, likely influenced by researchers analyzing many outcomes, suggest potential overestimation of reported effects. Future research should analyze fewer a priori selected outcomes and investigate models to progress SIT interventions for longer-term performance improvements.

Introduction

Interval training is considered a time efficient approach to exercise that can produce physical performance benefits (⁸³) that are at least equivalent to those obtained with traditional endurance training (^42,68). Interval training involves repeated bouts of intense exercise interspersed with recovery periods of low-intensity activity or rest (⁸). Two of the most frequently investigated forms of interval training include high-intensity interval training (HIIT) and sprint interval training (SIT). High-intensity interval training has been defined as repeated short-to-long bouts of exercise performed at a power output or velocity within the severe-intensity domain between the second ventilatory threshold and maximal oxygen consumption (V̇o₂max) (⁸⁶). Therefore, HIIT requires “near maximal” efforts that elicit an intensity of ≥80% of maximal heart rate (HRmax) or V̇o₂max (¹⁰⁸). Also, it is frequently performed using a range of exercise modes, including running (³¹), cycling (⁹⁹), rowing (¹), and swimming (⁹⁷); resulting in wide applicability to trained and untrained populations. In contrast, SIT is defined by exercise performed at a power output or velocity above V̇o₂max (i.e., “all-out” efforts in the extreme-intensity domain) necessitating short bouts of exercise (⁸⁶). Within research studies, SIT is most often performed on a cycle ergometer, allowing a controlled application of training intensity through the application of substantive resistance over 6- to 30-second intervals (^2,62,111). The potential for SIT to generate large physiological improvements in a time efficient manner has resulted in uptake by athletes, thus becoming its own stand-alone training modality, where research findings have identified improvements in performance measures of competitive runners (⁵³), cyclists (⁶⁹), triathletes (³³), and ice hockey players (⁷⁴).

Although previous research has demonstrated that there may be overlap between adaptations produced from both HIIT and SIT (⁷⁷), differences are likely to exist for a selection of both aerobic and anaerobic outcomes, and delineation between the 2 training methods is important for future understanding. Additionally, Viana et al. (¹⁰⁷) identified the importance of careful evaluation of acute program variables when comparing findings from studies either within or across forms of interval training. Creation of an interval training session is complex and first involves the manipulation of several interconnected acute program variables, including interval intensity, work interval duration, recovery intensity, recovery duration, exercise modality, number of repetitions, number of series, series duration, time between series, and between series recovery intensity (¹⁴). Individual sessions can then be altered within a training microcycle and progressed over longer periods to create overload and continued adaptations. However, current knowledge regarding the effects of acute program variables and their interaction is in the early stages, particularly with regards to SIT and the range of different outcomes that may be of interest to athletes.

Previous systematic reviews and meta-analyses have attempted to synthesize an increasing evidence base focusing on SIT performed on a cycle ergometer and its effects on aerobic capacity (^{44,86,94,109,110}) and sprint power (¹¹⁰). These evidence synthesis studies have generally included data from healthy individuals between 18 and 45 years, who were either sedentary or engaged in moderate frequency recreational activities (^{44,86,94,109,110}). Additionally, previous evidence synthesis studies have chosen to focus on a restricted range of outcome variables and SIT protocols. Rosenblat et al. (⁸⁶) meta-analyzed results from studies directly comparing HIIT and SIT interventions with time-to-exhaustion tests. The analysis was restricted to 6 studies that met the inclusion criteria, with the primary analysis identifying no differences between the forms of interval training. Sloth et al. (⁹⁴) and Gist et al. (⁴⁴) both investigated the effectiveness of SIT interventions to improve V̇o₂max. Gist et al. (⁴⁴) restricted their analysis to SIT interventions employing the popular repeated Wingate protocol comprising 4–6 “all-out” 30-second sprints with approximately 4 minutes of recovery. The meta-analysis included 16 studies and compared SIT interventions with either traditional endurance training or no-exercise controls. When analyzed separately, the results demonstrated a moderate effect (d = 0.69 [95% CI: 0.46–0.93]) of SIT compared with no-exercise controls, and no effect (d = 0.04 [95% CI: −0.17 to 0.24]) of SIT when compared with traditional endurance training. Similar findings were obtained in the meta-analysis conducted by Sloth et al. (⁹⁴), which included 21 studies with a wider range of SIT protocols (10- to 30-s sprints) incorporating either noncontrolled or no-exercise-controlled designs. Sloth et al. (⁹⁴) also reported a moderate effect (d = 0.63 [95% CI: 0.39–0.87]) of SIT to improve V̇o₂max. In contrast, Weston et al. (¹¹⁰) acknowledged that SIT performed on a cycle ergometer had the potential to improve sprint performance and aerobic capacity. The authors' meta-analyzed results from 16 studies, including either controlled or noncontrolled designs that measured power produced during a maximum 30-second sprint. The analysis was restricted to SIT studies employing the repeated Wingate protocol, with results demonstrating that SIT interventions had an unclear effect on improvements in peak (+1.8% [90% CL: ±5.0]) and mean (+2.2% [90% CL: ±10.3]) power.

Given the work of previous meta-analyses, Vollaard et al. (¹⁰⁹) stated that it was surprising that there had been minimal attempt to identify “optimal” protocols. As a result, the authors investigated the modifying effects of maximum number of sprints, intervention duration, number of sessions, sprint duration, recovery time, and sprint resistance on V̇o₂max in 34 studies. The results indicated a possible small modifying effect of the maximum number of sprints, with decreased improvements with additional sprints (¹⁰⁹). All other program variables were found to exert unclear or trivial effects (¹⁰⁹). However, the meta-analysis conducted by Vollaard et al. (¹⁰⁹) included a limited number of data points and only focused on a single outcome variable. The inclusion of a limited number of outcome measures in previous meta-analyses is no longer reflective of the research area with studies investigating a range of outcomes including those that assess anaerobic (¹¹³), neural (²⁶), and force production systems (¹⁰⁶).

Given the recent increase in the number of diverse SIT protocols to improve a range of fitness parameters associated with physical performance and sporting activity, there is a need to synthesize the available evidence and identify the protocols that are most effective. This would provide athletes, practitioners, and researchers with a practical framework for SIT prescriptions targeting specific training outcomes. Therefore, the aim of this systematic review and meta-analysis was to perform a comprehensive synthesis of the published research and quantify the effect of SIT and potential moderators on a range of physical performance measures collected from healthy adults. Additionally, assessment of the overall research quality was made to combine with the meta-analytic findings to better inform current practice and future research.

Methods

Search Strategy

This review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (⁷⁰). A 3-step search strategy was employed; first, an initial limited search was performed in MEDLINE and SportDiscuss followed by an analysis of the text words contained in the resulting titles, abstracts, keywords, and index terms used to describe the publications. Second, a search strategy tailored to each information source (MEDLINE, Web of Science, SPORTDiscuss) was developed based on the identified keywords and index terms (e.g., “sprint interval training,” “high intensity interval training,” “high intensity intermittent training,” “HIIT,” “interval exercise,” “high intensity training,” “high intensity exercise,” “high intensity aerobic interval training,” and “aerobic interval training”) and executed. Finally, the reference lists of all included studies as well as forward citation tracking using Google Scholar were searched for additional sources. Searches were limited to the years 2000–February 2020 and to the English language. The full electronic search strategy, including limits, can be found in Electronic Supplementary Material (see Appendix S1, https://links.lww.com/JSCR/A325).

Inclusion Criteria

Studies were included in this review if they satisfied the population, intervention, comparator, outcome (PICO) criteria—Population: young to middle-aged adults (mean age, 18–45 years), not suffering from any acute or chronic disease. Studies that specifically recruited overweight or obese subjects were excluded. Intervention: a minimum of 2 weeks of duration comprising maximum-intensity (“all-out”) sprints on a resistance bike. Interventions comprising sprints greater than 30-second duration were not considered maximum intensity and therefore were excluded. Studies incorporating combined training interventions (e.g., SIT plus strength training or aerobic training) or combined supplement interventions (e.g., SIT plus creatine supplementation) were excluded. Comparator: cohort, nonrandomized, and randomized-controlled (no exercise or habitual physical activity) designs. Where studies compared relevant SIT groups with other modes of exercise, data from SIT groups were included. Outcome: Three-outcome categories were defined and included: (a) aerobic (e.g., V̇o₂max, incremental time, oxygen consumption [O₂], respiratory exchange ration [RER]), (b) anaerobic (e.g., Wingate peak power, Wingate mean power), and (c) mixed aerobic-anaerobic (e.g., total work across critical power test measures, peak power across repeated tests).

Study Selection

Search results were imported to Proquest Refworks and duplicates were removed. Titles and abstracts of all sources were screened by 2 independent reviewers (A.J.H. and T.P.C.) for relevance to the review questions. Full-text manuscripts were retrieved for articles that potentially met the inclusion criteria and were screened independently by the same 2 reviewers. At each stage of the screening process, disagreements were resolved by discussion and inclusion of a third reviewer (P.A.S.). Articles identified from hand searching of reference lists were assessed for relevance based on their titles and abstracts with those meeting inclusion criteria added to the full-text screening stage. Full-text studies that did not meet the inclusion criteria were excluded, and reasons for their exclusion were documented (See Appendix S2, Electronic Supplementary Material, https://links.lww.com/JSCR/A325).

Data Extraction

A bespoke data extraction tool was piloted on 10 studies by 2 independent reviewers and discussed within the research team before full data extraction. Data extraction was completed independently by 2 authors (A.J.H. and T.P.C.) and discrepancies resolved through discussion. Reviewers were not blinded to manuscript authors or journals. Data regarding study type (controlled, uncontrolled), subject characteristics (sex, age, body mass, number), training parameters (intervention duration, total number of training sessions, exercise bike used, number of sprint repetitions within training intervention, sprint duration, recovery duration, applied sprint resistance), and training outcome measures were extracted. All extracted outcome data were assigned to a single-outcome category. Where data were presented in figures or percentage change units, the corresponding author was contacted for the original information. Where this was not made available, data within figures were extracted through graph digitizer software (DigitizeIt, Germany), with data expressed in percentage change units omitted from extraction.

Evaluation of Methodological Quality

Methodological quality and risk of bias were evaluated by 3 independent reviewers (T.P.C., R.R.A., M.K.) with agreement reached on each item by at least 2 reviewers. The quality of each review outcome category was assigned using a strategy based on the recommendations of the Grading of Recommendations Assessment Development and Evaluation (GRADE) working group (⁴⁵). Each individual study was initially appraised using a modified version of the Downs and Black Checklist (²⁸), which was specifically tailored for use in this study (See Appendix S3, Electronic Supplementary Material, https://links.lww.com/JSCR/A325). The modified checklist comprised 23 outcomes for comparator studies and 19 outcomes for noncomparator studies after removal of items relating to group differences. A total of 4 domains were evaluated, including (a) reporting, (b) internal validity—bias, (c) internal validity—confounding, and (d) statistical power. A total of 5 items were added to both checklists and included the following: “Were familiarization sessions of training completed?” “Were familiarization sessions of testing completed?” (internal validity—bias); “Were number of sessions attended reported?” “Was a minimum number of sessions for inclusion reported” (reporting); and “If a power calculation was completed, was this adjusted to account for multiple outcome variables?” (statistical power). Scoring for the additional questions employed the same protocol for the original questions: Yes = 1, No = 0, Unable to determine = 0. These additional items were included because they were considered fundamental in determining precision of the effects of an intervention and associated statistical rigor. Individual studies were assigned a rating based on the percentage of items scored positively with the following criteria used: “high” (80% +), “moderate” (60–79%), “low” (40–59%), or “very low” (0–39%). For each of the primary meta-analysis outcomes, an overall quality rating was assigned based on the mode rating of individual studies contributing data.

Statistical Analyses

A Bayesian’s framework was chosen over frequentist methods to provide a more flexible modeling approach and enable results to be interpreted intuitively through reporting of subjective probabilities (⁵⁵). The effects of SIT on included outcomes were quantified by calculating effect sizes in the form of standardized mean differences (SMDs). Magnitude-based SMDs obtained by dividing the mean difference by the preintervention standard deviation are the most popular form of effect size used in meta-analyses in sport and exercise science and are informative when considering the change an individual can be expected to make relative to a population pre- to postintervention (²¹). Most studies did not include a no-exercise control, and so intervention group–only effect sizes were used for primary analyses. Sensitivity analyses were conducted where possible with effect sizes incorporating data from no-exercise controls (⁷²). Within-study variance of effect sizes were calculated according to standard distributions with bias correction for small samples applied to both the effect size estimate and its variance (²¹). Standardized mean difference effect sizes are equal to Dz effect sizes calculated using a pre-post correlation of 0.5. Dz effect sizes account for the pre-post correlation, and generally result in larger effect sizes than SMD effect sizes, since the pre-post correlations are typically larger than 0.5. Dz effect sizes can be calculated using the SD of the difference scores or mathematically accounting for the pre-post correlation, if the correlation is reported. However, such distributions are influenced by pre-post correlations that are generally not reported (⁸⁷). Therefore, within-study variances were calculated and inputted using a standard value of 0.7 (³²), with an additional error term included to enable individual study values to vary. An informative Gaussian prior was placed on the error term such that the overall distribution of values matched the within-study variance distribution obtained from correlation values ranging from approximately 0.5 to 0.9 (³²). Three-level random-effects Bayesian’s hierarchical models were used to pool effect sizes and model average effects, variance between studies, and covariance of multiple outcomes reported in the same study (e.g., reporting of a single outcome across multiple time points or reporting values from multiple outcomes). The overall analysis approach was determined a priori and included an initial pooling of all effect sizes, followed by investigation of average effects by outcome category and training status. Meta-regressions were then performed to investigate associations between effect sizes and intervention duration, training intensity, training volume, and training work-to-rest ratios. Meta-regressions were only performed where there were sufficient data including a minimum of 4 data points per category level or 10 data points for continuous variables (³⁷).

Noninformative priors were used for all model parameters other than the within study variance correlations. Inferences from all analyses were performed on posterior samples generated using the Hamiltonian Markov Chain Monte Carlo method with 4 chains for 20,000 iterations with a burn-in period of 10,000. Interpretations were based on the median value (ES_0.5: 0.5-quantile), the range within credible intervals (CrIs) and calculation of probabilities that the magnitude of the average effect size exceeded commonly used qualitative thresholds (e.g., small: 0.2, medium: 0.5, and large: 0.8) (²⁵). Bayesian's CrIs can be interpreted probabilistically, such that with a 95% CrI, there is a 95% probability that the true (unknown) estimate would lie within the interval given the priors implemented and the evidence provided by the observed data. Additionally, the ES_0.5 represents the center of the posterior such that values close to this point are generally more probable. Analyses were performed using the R wrapper package brms (example code presented in Appendix S4, Electronic Supplementary Material, https://links.lww.com/JSCR/A325) interfaced with Stan to perform sampling (¹⁹). Convergence of parameter estimates was obtained for all models with Gelman-Rubin R-hat values below 1.1 (³⁹). All values presented in tables and figures include analyses conducted on data post removal of outliers, except for the association between controlled and no-exercise-controlled effect sizes presented in Figure 3, which include all data. Small-study effects (publication bias, etc.) were visually inspected with funnel plots and quantified with a multilevel extension of Egger's regression with effect sizes regressed on within-study variances and weights obtained from the reciprocal of the within- and between-study variances (^34,71).

Results

Search Results

Figure 1 illustrates the studies identified and selected included based on the search strategy and screening process. A total of 139 studies were screened at full text and 84 excluded primarily because of sprint duration being greater than 30 seconds or sprints not completed at an “all out” intensity. A total of 55 studies were included in the review (^{11–14,25–27}), (^42–69), (^69–89).

Figure 1.: PRISMA flow diagram detailing the results of each search and screening stage. A final number of 55 studies were included in the review.

Study Characteristics

Details of the 55 studies included in this review are shown in Table 1, with 25 (45%) of the studies including a nonexercise control group. Thirty studies (55%) comprised all male subjects, 2 studies (4%) comprised all female subjects, and 23 studies (41%) comprised both male and female subjects, with 3 of the studies reporting male and female data separately. In total, 589 subjects were allocated to a SIT intervention (median group size = 9 [IQR: 8–11]), with 257 subjects allocated to a nonexercise control group (median group size = 8 [IQR: 7–9]). Most studies (42 studies, 76%) recruited recreational subjects, with 10 studies (18%) recruiting sedentary subjects and 3 studies (6%) recruiting competitive athletes. In total, 617 outcomes were extracted demonstrating large variation in the number obtained from individual studies with the median equal to 7 (IQR: 3.5–15). Fifty percent of the outcomes were categorized as aerobic, whereas 12% were categorized as anaerobic and 38% categorized as mixed.

Table 1 - Overview of studies included in systematic review and meta-analysis.^*

Author	Aim	SIT intervention population	Study design	Intervention variables	Extracted outcomes	Summary of findings	Quality rating
Astorino et al. (2)	To compare differences in adaptations to short-term high-intensity training in active men and women matched for age and V̇o 2max	Recreational males (n = 11) and females (n = 9)	Nonexercise control	2 wk (6 sessions), 4–6 × 30 s sprints, 300-s recovery, 7.5% BM resistance	Peak power (W·kg−1), mean power (W·kg−1), and minimum power (W·kg−1), from a Wingate test; V̇o 2max (L·min−1; ml·kg−1·min−1), V̇co 2 (L·min−1), VE (L·min−1), O2 pulse at V̇o 2max (ml·beat−1), and RER from an incremental exercise test to exhaustion on a cycle ergometer	Similar improvements in power output and oxygen kinetics occurred between sexes matched for V̇o 2max and physical activity.	18/24 75% Moderate
Babraj et al. (3)	To determine if low-volume high-intensity interval exercise involving ∼250 kcal work improves glycemic control in sedentary young adults	Sedentary males (n = 16)	Nonexercise control	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	250 kJ cycle time trial (s)	Low-volume high-intensity interval exercise increases glycemic control and 250 kJ cycle time trial performance increased	12/24 50% Low
Bailey et al. (4)	To determine the effect of work-matched repeated sprint training and endurance training on the kinetics of V̇o 2, HR, and muscle deoxygenation during moderate- and severe-intensity exercise and tolerance in recreationally active subjects	Recreational males (n = 5) and females (n = 3)	Exercise comparator (endurance training), and Nonexercise control	2 wk (6 sessions), 4–7 × 30 s sprints, 240-s recovery, 7.5% BM resistance	Total work done (kJ) within each training session; V̇o 2peak (L·min−1; ml·kg−1·min−1), V̇o 2 (L·min−1), and work rate (W) at gas exchange threshold, and peak work rate (W) from an incremental exercise test to exhaustion on a cycle ergometer V̇o 2peak (L·min−1) and time to exhaustion (s) during a moderate and severe cycle step test	Repeated sprint training accelerated V̇o 2 kinetics during transitions to moderate and severe intensity exercise and enhanced exercise tolerance compared with endurance training	17/24 70.8% Moderate
Barnett et al. (5)	To compare enzymatic and histochemical adaptations to sprint training with sprint performance and exercise-induced changes in high-energy phosphagens, muscle glycogen, and lactate	Recreational (n = 8)	Nonexercise control	8 wk (24 sessions), 3–6 × 30 s sprints, 180-s recovery, 8.87 flywheel revolutions per pedal crank revolution gear ratio resistance	V̇o 2peak (L·min−1) from an incremental exercise test to exhaustion on a cycle ergometer Peak power (W) and mean power (W) during 10-s sprint	Sprint training improved sprint and V̇o 2peak performance, and lowered net ATP degradation during sprint exercise	10/24 41.7% Low
Bayati et al. (6)	To compare the established SIT protocol versus a modified type of high-intensity training on both aerobic and anaerobic performance measures	Recreational males (n = 8)	Exercise comparator (sprint training at 125% power at V̇o 2max) and Nonexercise control	4 wk (12 sessions), 3–5 × 30 s sprints, 240-s recovery, 7.5% BM resistance	V̇o 2max (ml·kg−1·min−1) from an incremental exercise test to exhaustion on a cycle ergometer Power at V̇o 2max (W), total work (kJ), and time to exhaustion at power at V̇o 2max (s) from a time to exhaustion at power at V̇o 2max test Peak power (W), mean power (W), and total work (kJ) from a Wingate test	Aerobic and anaerobic performance similarly improved across both protocols, except for mean power output, which only improved within the SIT protocol	13/24 54.2% Low
Benítez-Flores et al. (7)	To determine the combined effects of resistance and sprint training, with very short efforts (5 s), on aerobic and anaerobic performances and cardiometabolic health-related parameters in young healthy adults	Recreational males (n = 4) and females (n = 4)	Exercise comparator (undulating periodized resistance training), and (concurrent resistance training and SIT), and nonexercise control	2 wk (6 sessions), 6–12 × 5 s sprints, 24-s recovery, 0.7 N·m resistance	V̇o 2max (ml·kg−1·min−1), power at V̇o 2max (W), and RERmax from an incremental exercise test to exhaustion on a cycle ergometer Peak power (W), total work (kJ), and maximum pedalling rate (rpm) from 2 × 5 s sprints countermovement jump (CMJ) height (cm) Mean velocity (m·s−1), mean power (W), mean force (N) from an isoinertial squat test	Concurrent training promotes improvements in lower-body strength and aerobic capacity similar to resistance training and SIT interventions	21/24 87.5% High
Broatch et al. (13)	To determine the effects of regular postexercise cold water immersion on key markers of mitochondrial biogenesis following 6 wk of SIT	Recreational males (n = 8)	Exercise comparator (cold water immersion)	6 wk (18 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5–9.5% BM resistance	2 km cycle time trial (s) and mean power (W) 20 km cycle time trial (s), lactate threshold ,and peak aerobic power (W) from an intermittent graded exercise test; V̇o 2peak (ml·kg−1·min−1) from a steady-state cycle to fatigue at supramaximal power output	Cold water immersion administered following 6 wk of SIT had limited effects on endurance performance, mitochondrial biogenesis, or changes in mitochondrial content and function	19/24 79.2% Moderate
Burgomaster et al. (15)	To determine the effects of 6 sessions of SIT on muscle oxidative potential, V̇o 2peak, and endurance time to fatigue during cycling at an intensity equivalent to 80% V̇o 2peak	Recreational males (n = 6) and females (n = 2)	Nonexercise control	2 wk (6 sessions), 4–7 × 30 s sprints, 240-s recovery, 7.5% BM resistance	O2 uptake (L·min−1), expired ventilation (L·min−1), RER, V̇o 2peak (ml·kg−1·min−1), and time to fatigue (min) from an incremental exercise test to exhaustion on a cycle ergometer Peak power (W) and mean power (W) across 4 consecutive Wingate tests	SIT increased citrate synthase maximal activity and doubled endurance capacity during cycling exercise at 80% V̇o 2peak in recreationally active subjects	12/24 50% Low
Burgomaster et al. (16)	To determine the effects of 2 wk of SIT on carbohydrate metabolism during submaximal exercise	Recreational males (n = 8)	Nonexercise control	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	Peak power (W) and mean power (W) from a Wingate test; 250 kJ time trial (s), and mean power (W) V̇o 2 (L·min−1) at 60% V̇o 2peak and 90% V̇o 2peak during a 2 stage cycling test	SIT decreased net muscle glycogenolysis and lactate accumulation, increased pyruvate oxidation capacity, and decreased 250 kJ time trial time	15/24 62.5% Moderate
Burgomaster et al. (17)	To determine the time course for adaptations in metabolite transport proteins following SIT	Recreational males (n = 8)	Nonexercise control	6 wk (18 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	250 kJ cycle time trial (min) and mean power (W)	Muscle oxidative potential and proteins associated with glucose and lactate/H+ transport, GLUT4 and MCT4, increased following 1 wk of SIT, and MCT1 increased following 6 wk of SIT	12/24 50% Low
Burgomaster et al. (18)	To compare the effects of endurance training and SIT on adaptations of metabolic markers	Recreational males (n = 5) and females (n = 5)	Exercise comparator (endurance training)	2 wk (6 sessions), 4–6 × 30 s sprints, 270-s recovery, ∼500 W resistance	V̇o 2peak (ml·kg−1·min−1; L·min−1) from an incremental exercise test to exhaustion on a cycle ergometer, and V̇o 2 (L·min−1), RER, and ventilation (L·min−1) at 65% V̇o 2max	SIT elicits comparable adaptations in markers of skeletal muscle carbohydrate and lipid metabolism, and metabolic control, as endurance training despite a lower training duration	12/24 50% Low
Camacho-Cardenosa et al. (22)	To determine the effects of maximal intensity interval training in hypoxia in active adults	Recreational subjects (n = 8)	Exercise comparator (hypoxia), and nonexercise control	4 wk (8 sessions), 2 sets of 5 × 10-s sprints, 20–600 s recovery, no resistance stated	V̇o 2max (ml·kg−1·min−1), peak power (W), mean power (W), mean cadence (rpm), maximum torque (N·m) from a 3-min all-out test	Eight sessions of maximal intensity interval training in hypoxia is enough to decrease the percentage of fat mass, improve hematocrit (HCT) and Hb parameters, and mean muscle power in healthy and active adults	15/24 62.5% Moderate
Cochran et al. (23)	To determine if β-alanine (ALA) supplementation or a placebo would improve physiological and performance adaptations following SIT	Recreational males (n = 12)	Exercise comparator (β-ALA supplement & SIT)	6 wk (18 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	V̇o 2peak (ml·kg−1·min−1) and peak power (W) from an incremental exercise test to exhaustion on a cycle ergometer 250 kJ time trial mean power (W) and time (s) Mean power (W) from a repeated Wingate test	SIT with β-ALA supplementation did not augment performance measures, training workload, or improvements in skeletal muscle oxidative capacity in comparison with a SIT with placebo intervention	19/24 79.2% Moderate
Cocks et al. (24)	To determine the effects of 6 wk of traditional endurance training and SIT on skeletal muscle microvascular density and microvascular enzyme content (eNOS and NOX2) in previously sedentary men	Sedentary males (n = 8)	Exercise comparator (endurance training)	6 wk (18 sessions), 4–6 × 30 s sprints, 270-s recovery, 7.5% BM resistance	V̇o 2peak (ml·kg−1·min−1), and peak aerobic power output (W) from an incremental exercise test to exhaustion on a cycle ergometer	Muscle microvascular density and eNOS protein content increased following endurance training and sprint interval training in sedentary males	15/24 62.5% Moderate
Creer et al. (26)	To determine the effects of short term, high-intensity sprint training on the root-mean-squared and median frequency derived from electromyography (EMG), peak power, mean power, total work, and plasma lactate levels during a series of 30-s maximal sprints compared with endurance training alone in trained cyclists	Competitive males (n = 10)	Nonexercise control	4 wk (8 sessions), 4–10 × 30 s sprints, 240-s recovery, no resistance stated	V̇o 2max (L·min−1) from an incremental exercise test to exhaustion on a cycle ergometer Peak power (W), mean power (W), and total work (kJ) from a Wingate test	SIT increased motor unit recruitment and total work compared with endurance training alone	12/24 50% Low
Forbes et al. (36)	To determine whether a short-term high-intensity interval cycling training program increases the rate of PCr recovery following moderate-intensity exercise in which pH changes are minimal	Recreational males (n = 4) and females (n = 3)	Nonexercise control	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s recovery, 6.5–7.5% BM resistance	Leg extension peak force (N) Mean power (W) and mean peak power (W) in training sessions 1 and 6	Short-term SIT increases PCr recovery following moderate-intensity exercise, indicating an improvement in oxidative capacity	16/24 66.7% Moderate
Gibala et al. (40)	To compare changes in exercise capacity, and molecular and cellular adaptations in skeletal muscle after low-volume SIT and high-volume endurance training	Recreational males (n = 8)	Exercise comparator (endurance training)	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	750 kJ cycle time trial (s) and mean power (W); 50 kJ cycle time trial (s) and mean power (W)	Low-volume SIT or traditional high-volume endurance training induced similar improvements in muscle oxidative capacity, muscle buffering capacity, and exercise performance	14/24 58.3% Low
Gillen et al. (43)	To determine whether SIT was a time-efficient exercise strategy to improve insulin sensitivity and other indices of cardiometabolic health to the same extent as traditional moderate-intensity continuous training	Sedentary males (n = 9)	Exercise comparator (moderate-intensity continuous training) and a nonexercise control	12 wk (31 sessions), 3 × 20 s sprints, 120-s recovery, 5% BM resistance	V̇o 2peak (ml·kg−1·min−1; L·min−1) and maximum workload (W) from an incremental exercise test to exhaustion on a cycle ergometer	SIT improved insulin sensitivity, cardiorespiratory fitness, and skeletal muscle mitochondrial content to the same extent as moderate-intensity continuous training, despite a 5-fold lower exercise volume and training time commitment	18/24 75% Moderate
Harmer et al. (46)	To determine the effects of sprint training on respiratory, metabolic, and ionic perturbations during intense exercise conducted at an identical power output in 2 separate tests: one test matched for duration in pre- and posttraining trials and the other continued until exhaustion	Recreational males (n = 7)	No control	7 wk (21 sessions), 4–10 × 30 s sprints, 180–240-s recovery, 7.5% BM resistance	Peak, mean, and relative expired ventilation (L·min−1), peak, mean, and relative O2 uptake (L·min−1), peak, mean, and relative CO2 output (L·min−1), peak RER, accumulated V̇o 2 (mmol.kg), total work (kJ) and time to exhaustion (s) from a test to exhaustion at 130% V̇o 2peak; V̇o 2peak (L·min−1), power output (W) and time to exhaustion (s) from an incremental exercise test to exhaustion on a cycle ergometer; peak power (W) and total work (kJ) from a 30-s all-out sprint	Sprint training reduces metabolic and ionic perturbations within tissue during intense exercise matched for power output and work production, although indexes of anaerobic metabolism were not augmented during exhaustive exercise after training, despite the increased exercise duration, suggesting the importance of aerobic adaptations to performance after sprint training	7/19 36.8% Very low
Harris et al. (47)	To determine and compare the effects of work matched SIT with a less time committing sprint continuous protocol on brachial artery endothelial function, arterial stiffness, cardiorespiratory fitness, and circulating angiogenic cell number and function	Recreational females (n = 6)	Exercise comparator (sprint continuous training)	4 wk (12 sessions), 4 × 30 s sprints, 270-s recovery, 7.5% BM resistance	V̇o 2max (ml·kg−1·min−1; L·min−1), lactate threshold (ml·min−1·kg−1), peak work rate (W), and time (min) from an incremental step exercise test on a cycle ergometer	Sprint continuous training improved cardiorespiratory fitness to a similar extent as SIT, with a trend for brachial artery flow-mediated dilation (FMD) increase following SIT but not sprint continuous training	18/24 75% Moderate
Hazell et al. (48)	To determine whether 10-s or 30-s SIT bouts with 2- or 4-min recovery periods can improve aerobic and anaerobic performance	Recreational males (n = 6) and females (n = 6) in each of the 3 SIT groups	Nonexercise control	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s recovery, 100 g·kg−1·BM−1 resistance 2 wk (6 sessions), 4–6 × 10 s sprints, 240-s recovery, 100 g·kg−1·BM−1 resistance 2 wk (6 sessions), 4–6 × 10 s sprints, 120-s recovery, 100 g·kg−1·BM−1 resistance	V̇o 2max (ml·kg−1·min−1) from an incremental exercise test to exhaustion on a cycle ergometer; 5-km time trial (s); peak power (W·kg−1) and mean power (W·kg−1) from a 30-s Wingate test	The 10-s SIT protocols produced similar improvements in V̇o 2max and 5-km time trial performance compared with the established 30 s SIT protocol	14/24 58.3% Low
Hommel et al. (49)	To determine and compare the influence of SIT and endurance training on calculated power in maximal lactate steady state and maximal oxygen uptake.	Recreational males (n = 10)	Exercise comparator (endurance training), and nonexercise control	6 wk (18 sessions), 4–6 × 30 s sprints, 270 s recovery, 7.5% BM resistance	V̇o 2max (ml·min−1·kg−1) from an incremental exercise test to exhaustion on a cycle ergometer; power in lactate steady state (W); peak anaerobic power (W) from a modified sprint test.	SIT and endurance training improve calculated power in maximal lactate steady state through differently influencing maximal lactate production rate and V̇o 2max.	15/24 62.5% Moderate
Ijichi et al. (50)	To compare the effects of sprint training on exercise performance between sprint training twice every second day and sprint training once daily, with the same total number of training sessions	Recreational males (n = 20) SIT once every day (n = 10) SIT twice every second day (n = 10)	No control	SIT daily: 5 d per week ×4 weeks (20 sessions), 3 × 30-s sprints, 10-min recovery, 5% BM resistance SIT twice: 2–3 sessions per week × 4 wk (20 sessions total), 3 × 30-s sprints, 10-min recovery, 5% BM resistance	V̇o 2max (ml·min−1·kg−1; L·min−1), peak aerobic power (W) and onset of blood lactate accumulation (W) from an incremental exercise test to exhaustion on a cycle ergometer Time to fatigue (s) from a submaximal cycling test at 90% V̇o 2max Peak power (W) and mean power (W) from 2 × 30 s maximal sprint tests	Similar improvements in peak and mean power output during 30-s sprint tests and anaerobic endurance capacity occurred between the groups, although SIT every second day improved the onset of blood lactate accumulation to a greater extent in physically active males	12/24 50% Low
Ikutomo et al. (51)	To determine the influence of inserted long rest periods during repeated sprint training on performance adaptations in competitive athletes	Competitive male (n = 17) and female (n = 4) sprinters Short recovery (n = 10) Long recovery group (n = 11)	No control	Short recovery: 3 wk (9 sessions), 2 sets of 12 × 6 s sprints, 24-s recovery, 20 min between the sets, 7.5% BM resistance Long recovery: 3 wk (9 sessions), 2 sets of 12 × 6 s sprints, 24-s recovery—with an additional 7 min recovery every third sprint, 20 min between the sets, 7.5% BM resistance	V̇o 2max (ml·min−1·kg−1) from an incremental exercise test to exhaustion on a cycle ergometer Time to exhaustion (s) at 80% V̇o 2max Peak power (W·kg−1) per sprint, 10 min, and 30 min following a repeated sprint test	Repeated sprint training with longer rest periods is an efficient strategy for improving power output compared with shorter rest periods alone	10/24 41.7% Low
Jakeman et al. (52)	To determine whether shorter-duration high-intensity training involving 6-s sprints and totalling 60 s of exercise per session could elicit improvements in performance	Recreational males (n = 6)	Nonexercise control	2 wk (6 sessions), 10 × 6 s sprints, 60-s recovery, 7.5% BM resistance	Time to exhaustion (s) and the onset of blood lactate accumulation (s) from an incremental exercise test to exhaustion on a cycle ergometer; 10-km time trial (s); peak power output (W) for each training session	Shorter duration SIT repeated over 2 wk improves aerobic performance and produces an attenuation of blood lactate accumulation normally seen with longer duration sprints or longer training interventions	11/24 45.8% Low
Kavaliauskas et al. (53)	To determine the effectiveness of cycling based high intensity training with different work-to-rest ratios for long-distance running.	Competitive males (n = 14) and females (n = 18) 1:3 group: Males (n = 3) and females (n = 5) 1:8 group: Males (n = 3) and females (n = 5) 1:12 group: Males (n = 4) and females (n = 4)	Nonexercise control	1:3 group: 2 wk (6 sessions), 6 × 10 s sprints, 30 s recovery, 7.5% BM resistance 1:8 group: 2 wk (6 sessions), 6 × 10 s sprints, 80 s recovery, 7.5% BM resistance 1:12 group: 2 wk (6 sessions), 6 × 10 s sprints, 120 s recovery, 7.5% BM resistance	3 km running time trial (s); V̇o 2peak (ml·min−1·kg−1) and time to exhaustion (s) from an incremental exercise test to exhaustion on a cycle ergometer; Peak power (W·kg−1) and mean power (W·kg−1) from a Wingate test	SIT with a lower work-to-rest ratio provides a sufficient training stimulus for improving running performance, with nonspecific training contributing to running performance in runners who regularly undergo endurance training.	12/24 50% Low
Kavaliauskas et al. (54)	To determine the effects of SIT on cardiorespiratory fitness and aerobic performance measures in young females	Recreational females (n = 8)	Nonexercise control (subjects acted as own controls)	4 wk (8 sessions), 4 × 30 s sprints, 240-s recovery, 7% BM resistance	V̇o 2peak (ml·min−1·kg−1) and time to exhaustion (s) from an incremental exercise test to exhaustion on a cycle ergometer 10-km time trial (s) 3-min critical power (W·kg−1) Peak power (W), mean power (W), sum of peak power (W), and sum of mean power (W) during training sessions	SIT performed twice per week improves aerobic performance measures in young, untrained females	12/19 63.2% Moderate
Larsen et al. (56)	To determine the acute and short-term effects of high-intensity training on human skeletal muscle energetics in vivo using phosphorus magnetic resonance spectroscopy	Recreational males (n = 8)	No control	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	V̇o 2peak (ml·min−1·kg−1), time to exhaustion (s) and peak workload (W) from an incremental exercise test to exhaustion on a cycle ergometer Knee extension maximal force (N) Peak power (W) and mean power (W) during training sessions	6 sessions of high-intensity training alter in vivo muscle energetics likely contributing to increased exercise capacity	14/19 73.7% Moderate
Lewis et al. (57)	To determine the neuromuscular adaptations to SIT	Recreational males (n = 7)	No control	2 wk (6 sessions), 4–7 × 30 s sprints, 240-s recovery, 7.5% BM resistance	10-km time trial (s) Peak power (W) and mean power (W) during training sessions Quadriceps maximal voluntary contraction (N) pre and post sprints	SIT improved performance measures without measurable neuromuscular adaptations	18/24 75% Moderate
Little et al. (60)	To determine if sprint snacks increased V̇o 2peak and aerobic exercise performance in healthy individuals	Recreational males (n = 14) and females (n = 14) Sprint snacks: Males (n = 5) and females (n = 7) Traditional SIT: Males (n = 9) and females (n = 7)	No control	Sprint snacks: 6 wk (18 sessions), 3 × 20 s sprints, 1–4 h of recovery, 0.21 N m·kg−1 resistance Traditional SIT: 6 wk (18 sessions), 3 × 20 s sprints, 180-s recovery, 0.21 N m·kg−1 resistance	V̇o 2peak (ml·min−1·kg−1; L·min−1), peak power W), and time to exhaustion (min) from an incremental exercise test to exhaustion on a cycle ergometer 150 kJ time trial (min) Peak power (W), mean power (W) and total work (kJ) across each training session	Sprint snacks improved V̇o 2peak, peak aerobic power, and 150 kJ time trial performance to the same extent as traditional SIT	19/24 79.2% Moderate
Lloyd Jones et al. (62)	To determine whether repeated 6-s sprint bouts with differing work-to-rest ratios resulted in different training adaptations	Recreational males (n = 18) and females (n = 9) 1:8 group: Males (n = 6) and females (n = 3) 1:10 group: Males (n = 6) and females (n = 3) 1:12 group: Males (n = 6) and females (n = 3)	Nonexercise control	1:8 group: 2 wk (6 sessions), 10 × 6 s sprints, 48-s recovery, 7.5% BM resistance 1:10 group: 2 wk (6 sessions), 10 × 6 s sprints, 60-s recovery, 7.5% BM resistance 1:12 group: 2 wk (6 sessions), 10 × 6 s sprints, 72 s recovery, 7.5% BM resistance	10-km time trial (s) Peak power (W), mean power (W), and session work (kJ) across each training session	All SIT conditions resulted in significant improvements in performance with no significant differences in improvement across any of the groups	12/24 50% Low
McGarr et al. (64)	To determine and compare any improvements in heat adaptation from short-term endurance training and SIT in moderately fit individuals	Recreational males (n = 6) and females (n = 2)	Exercise comparator (endurance training)	2 wk (8 sessions), 4–5 × 30 s sprints, 240-s recovery between each sprint, 7.5% BM resistance	V̇o 2peak (ml·kg−1·min−1) from an incremental exercise test to exhaustion on a cycle ergometer	Short-term endurance and SIT Improved aerobic fitness and attenuated cardiovascular strain during exercise in a hot environment, although neither training modality increased heat loss responses nor in minimized thermal strain	17/24 70.8% Moderate
Metcalfe et al. (65)	To determine the effects of a reduced exertion high-intensity training exercise intervention on insulin sensitivity and aerobic capacity	Sedentary males (n = 7) and females (n = 8)	Nonexercise control	6 wk (18 sessions), 2 × 10–20 s sprints, 200–220 s of recovery, 7.5% BM resistance	V̇o 2peak (L·min−1; ml·min−1·kg−1) from an incremental exercise test to exhaustion on a cycle ergometer	SIT is associated with improved insulin sensitivity in sedentary young men and improved aerobic capacity in men and women	16/24 66.7% Moderate
Metcalfe et al. (67)	To determine whether there is a true sex difference in response to reduced exertion high-intensity interval training or if these findings can be explained by the large interindividual variability response inherent to all exercise training	Sedentary males (n = 17) and females (n = 18)	No control	6 wk (18 sessions), 1–2 × 10–20 s sprints, 200–220 s of recovery, 5% BM resistance	V̇o 2peak (ml·min−1·kg−1) and V̇o 2max (L·min−1; ml·min−1·kg−1) from an incremental exercise test to exhaustion on a cycle ergometer	Reduced exertion high-intensity interval training presented substantial interindividual variability for all parameters with no sex differences evidenced	18/24 75% Moderate
Muggeridge et al. (73)	To determine the effects of dietary nitrate on the response to 3 wk of SIT	Recreational males (n = 10)	Exercise comparator (SIT with nitrate) and a nonexercise control	3 wk (9 sessions), 4–6 × 15 s sprints, 240-s recovery, 7% BM/5–10 air brake resistance	V̇o 2max (ml·min−1·kg−1), ventilatory threshold (W) and maximal workrate (W) from an incremental exercise test to exhaustion on a cycle ergometer Peak power (W) and mean power (W) from each sprint within training sessions 1 and 9	SIT improved performance parameters, although no additional benefit was gained from the administration of dietary nitrate supplementation	18/24 75% Moderate
Nalçakan (76)	To determine and compare the effects of SIT and continuous endurance training on anthropometric, aerobic, and anaerobic performance indices, mechanical gross efficiency, blood lipids, inflammation, skeletal muscle damage, and myocardial cell injury in healthy young males	Recreational males (n = 8)	Exercise comparator (endurance training)	7 wk (21 sessions), 4–6 × 30 s sprints, 270-s recovery, 7.5% BM resistance	Peak power (W), mean power (W), time to peak power (s), and power drop (%) from a Wingate test Mechanical gross efficiency from a submaximal cycle test at 60% V̇o 2max	SIT improved body composition and performance measures to the same extent as continuous endurance training, although no changes occurred in lipid profile, serum levels of inflammatory markers, myocardial cell injury markers, or skeletal muscle damage markers following training	15/24 62.5% Moderate
Nalçakan et al. (75)	To determine whether reducing the sprint duration in the reduced exertion high-intensity training protocol from 20 to 10 s per sprint influences acute affective responses and the change in V̇o 2max following training	Recreational males (n = 19) and females (n = 17) 20-s sprint group: Males (n = 8) and females (n = 10) 10-s sprint group: Males (n = 11) and females (n = 7)	No control group	20 s group: 6 wk (18 sessions), 2 × 10–20 s sprint, 220–240 s of recovery, 7.5% BM resistance 10 s group: 6 wk (18 sessions), 2 × 5–10 s sprint, 220–230 s of recovery, 7.5% BM resistance	V̇o 2max (L·min−1) from an incremental exercise test to exhaustion on a cycle ergometer	SIT involving 20- s sprints reported greater improvements in V̇o 2max compared with 10-s sprints	16/19 84.2% High
O'Driscoll et al. (78)	To determine the combined adaptations of the cardiac autonomic nervous system and myocardial functional and mechanical parameters to high-intensity interval training	Sedentary males (n = 40)	Nonexercise control (subjects acted as own controls)	2 wk (6 sessions), 3 × 30 s sprints, 120-s recovery, 7.5% BM resistance	V̇o 2peak (ml·min−1·kg−1; ml·min−1) and ventilatory equivalent (ml·min−1) from an incremental exercise test to exhaustion on a cycle ergometer	SIT improves cardiac autonomic modulation, myocardial function, and myocardial mechanics	20/24 83.3% High
Ørtenblad et al. (80)	To determine the effects of 5 wk of sprint training on intermittent exercise performance, sarcoplasmic reticulum (SR) Ca2+ sequestration, and release function and SR ryanodine binding	Recreational males (n = 9)	Nonexercise control	5 wk (15 sessions), 20 × 10 s sprints, 50-s recovery, 8.25% BM resistance	V̇o 2peak (ml·min−1·kg−1) from an incremental exercise test to exhaustion on a cycle ergometer Total work (kJ) and mean power (W·kg−1) from a 10 × 8 s sprint test Mean power (W·kg−1) across each training session sprint and each second of each sprint	High-intensity intermittent training increases the peak rate of AgNO3-stimulated SR Ca2+ release	12/24 50% Low
Parra et al. (81)	To determine the effect of 2 different SIT protocols on muscle metabolic response and performance	Recreational males (n = 10) No recovery program (n = 5) Two-day recovery program (n = 5)	SIT groups only	2 wk (14 sessions), 4–14 × 15–30 s sprints, 45 s to 12 min of recovery between sprints, 7.5% BM resistance (no recovery days between sessions) 6 wk (14 sessions), 4–14 × 15–30 s sprints, 45 s to 12 min of recovery between sprints, 7.5% BM resistance (2 days recovery between sessions)	Peak power (W) and mean power (W) from a Wingate test	During high-intensity training, shorter rest periods between sessions induced greater biochemical adaptations in human muscle compared with longer rest periods	11/24 45.8% Low
Rakobowchuk et al. (82)	To determine whether 6 wk of high-intensity, low-volume, SIT improves central (carotid) artery distensibility, peripheral (popliteal) artery distensibility and endothelial function in the trained legs to the same extent as high-volume, moderate-intensity endurance training	Sedentary males (n = 5) and females (n = 5)	Exercise comparator (endurance training)	6 wk (18 sessions), 4–6 × 30 s sprints, 270-s recovery, 7.5% BM resistance	V̇o 2peak (ml·min−1·kg−1) from an incremental exercise test to exhaustion on a cycle ergometer, peak power output (W) from a Wingate test	SIT elicits similar improvements in peripheral vascular structure and function to endurance training, although central artery distensibility may require a longer training stimuli or greater initial vascular stiffness	14/19 73.7% Moderate
Richardson and Gibson (84)	To determine the effects of hypoxic SIT on aerobic capacity	Recreational males (n = 6) and females (n = 3)	Nonexercise control	2 wk (6 sessions), 4–7 × 30 s sprints, 240-s recovery, 7.5% BM resistance	V̇o 2peak (L·min−1) from an incremental exercise test to exhaustion on a cycle ergometer, time to exhaustion (min) from an incremental exercise test to exhaustion on a cycle ergometer at 80% V̇o 2peak power output, and mean power output (W·kg−1) across the first 4 sprints in sessions 1 and 6	V̇o 2peak and time to exhaustion improved following hypoxic and normoxic SIT compared with a control, although hypoxia did not provide any additional improvements in endurance performance	13/24 54.2% Low
Rodas et al. (85)	To determine the changes in aerobic and anaerobic metabolism produced by a new incremental training program of “all-out” loads, repeated daily for 2 wk, and with long recovery periods	Recreational males (n = 5)	No control	2 wk (14 sessions), 4–14 × 15–30 s sprints, 45–720 s of recovery, 7.5% BM resistance	V̇o 2 (ml·kg−1·min−1) and power output (W) from an incremental exercise test to exhaustion on a cycle ergometer, V̇o 2 (ml·kg−1·min−1) and peak and mean power output (W) from a Wingate test, and pedalling rate (rpm) across each training session	Enzymatic activities of energetic pathways improve in a short time following short-duration, high-load, and long recovery period “all-out” sprints	9/19 47.4% Low
Scalzo et al. (88)	To determine changes in endurance exercise performance after SIT and to measure the integrated muscle protein synthesis response, mitochondrial biogenesis, and proteome kinetics in males and females over the course of 3 wk of SIT.	Recreational males (n = 11) and females (n = 10)	No control	3 wk (9 sessions), 4–8 × 30 s sprints, 240-s recovery, 7.5% BM resistance	V̇o 2max (ml·kg−1·min−1) from an incremental exercise test to exhaustion on a cycle ergometer, 40-km time trial (s) and mean power output (W; W·kg−1 fat free mass) across each sprint for sessions 1 and 9	Greater synthesis rates of muscle protein synthesis and mitochondrial biogenesis were observed in males than females during SIT, although there were no differences in V̇o 2max, time trial or power output when normalized to fat free mass	12/19 63.2% Moderate
Schlittler et al. (89)	To determine the effects of 3 weeks of SIT on high-intensity cycling performance, ryanodine receptor modifications, and the recovery of isometric force in recreationally active human subjects	Recreational males (n = 8)	No control	3 wk (9 sessions), 4–6 × 30 s sprints, 240-s recovery, 0.7 N·m·kg−1·BM−1 resistance	Maximal power (W) from an incremental exercise test to exhaustion on a cycle ergometer Total work (kJ) and peak power (W·kg−1) across 6 Wingate cycles Isometric knee extension maximal voluntary contraction (N) pre and post training session	SIT did not accelerate the recovery of isometric force, although did provide incomplete protection against RyR1 alteration	10/19 52.6% Low
Shenouda et al. (91)	To determine the effects of 6 and 12 wk of moderate-intensity continuous training and low-volume SIT on brachial and popliteal artery endothelial function and diameter, and central and lower limb arterial stiffness in sedentary, healthy men compared with nontraining controls	Sedentary males (n = 9)	Exercise comparator (moderate-intensity continuous training), and a nonexercise control	12 wk (31 sessions), 3 × 20 s sprints, 120-s recovery, 5% BM resistance	V̇o 2peak (ml·kg−1·min−1) from an incremental exercise test to exhaustion on a cycle ergometer	Brachial artery responses to SIT may follow a different time course not captured by a 6- and 12-wk intervention, although these are observed with moderate-intensity continuous training	17/24 70.8% Moderate
Shepherd et al. (92)	To determine whether SIT induces improvements in insulin sensitivity and net intramuscular triglyceride (IMTG) breakdown and to investigate the underlying mechanisms	Sedentary males (n = 8)	Exercise comparator (endurance training)	6 wk (18 sessions), 4–6 × 30 s sprints, 270-s recovery, 7.5% BM resistance	V̇o 2peak (L·min−1; L·kg−1·min−1) and peak Workload (W) from an incremental exercise test to exhaustion on a cycle ergometer, and V̇o 2 (L·min−1), V̇co 2 (L·min−1) and RER from a 60 min cycle at 65% V̇o 2peak	6 wk of SIT and endurance training improve insulin sensitivity through mechanisms involved with increased PLIN2, PLIN5, and IMTG utilization during exercise	17/19 89.5% High
Songsorn et al. (95)	To determine whether a single 20-s cycle sprint per training session can provide a sufficient stimulus for improving V̇o 2max	Recreational males (n = 5) and females (n = 10)	Nonexercise control	4 wk (12 sessions), 1 × 20 s sprints, 7.5% BM resistance	V̇o 2max (L·min−1) and peak power output (W) from an incremental exercise test to exhaustion on a cycle ergometer	A single 20-s cycle sprint per training session is not a sufficient stimulus for improving V̇o 2max.	20/24 83.3% High
Terada et al. (103)	To determine the effects of SIT with exogenous carbohydrate supplementation and SIT following overnight fast on aerobic capacity and high-intensity aerobic endurance	Recreational males (n = 11)	Exercise comparator (SIT with exogenous carbohydrate)	4 wk (12 sessions), 4–7 × 30 s sprints, 240-s recovery, 7.5% BM resistance	V̇o 2peak (ml·O2 −1·kg−1·min−1) from an incremental exercise test to exhaustion on a cycle ergometer, cycling time to exhaustion (s) at 85% V̇o 2peak, and mechanical work (Joules·kg−1) and peak power output (W·kg−1) across each training week	Fasted SIT compromises exercise intensity and volume but can increase the ability to sustain high-intensity aerobic endurance exercise compared with SIT with exogenous carbohydrate supplementation	19/19 100% High
Thompson et al. (104)	To determine the independent and combined performance and physiological effects of SIT and NO3 - supplementation during a 4 wk intervention.	Recreational males (n = 6) and females (n = 6)	Nonexercise control (with concurrent NO3 - beetroot juice) and exercise comparator (SIT with concurrent NO3 - beetroot juice)	4 wk (14 sessions), 4–5 × 30 s sprints, 240 s recovery, 7.5% BM resistance	V̇o 2peak (L·min−1) and peak work rate (W) from an incremental exercise test to exhaustion on a cycle ergometer, and V̇o 2peak (L·min−1) and work rate (W) at gas exchange threshold	NO3 - supplementation reduced the O2 cost of submaximal exercise, resulting in a greater improvement in incremental exercise performance and muscle metabolic adaptations to training compared with a placebo.	18/19 94.7% High
Thompson et al. (105)	To compare the physiological and exercise performance adaptations to 4 wk of SIT accompanied by concurrent supplementation with NO3 - beetroot juice, or potassium NO3 - or SIT undertaken without dietary NO3 -.	Recreational males (n = 6) and females (n = 6)	Exercise comparators (SIT with concurrent NO3 - beetroot juice) and (SIT with concurrent potassium NO3 -)	4 wk (14 sessions), 4–5 × 30 s sprints, 240 s recovery, 7.5% BM resistance	V̇o 2peak (L·min−1) and peak work rate (W) from an incremental exercise test to exhaustion on a cycle ergometer, V̇o 2peak (L·min−1) and time to task failure (s) during a moderate and severe cycle step test	4 wk of sprint interval training with concurrent NO3 - beetroot juice supplementation results in greater exercise capacity adaptations compared with sprint interval training alone or sprint interval training with concurrent potassium NO3 - supplementation.	19/19 100% High
Vera-Ibanez et al. (106)	To determine the neural adaptations associated with a low-volume Wingate-based high-intensity interval training	Recreational males (n = 7)	Nonexercise control	4 wk (12 sessions), 3–6 × 30 s sprints, 240-s recovery, 7.5% BM resistance	Peak power (W; W·kg−1) from a Wingate test, plantar flexor maximum voluntary contraction (MVC) (N) on a soleus isolation machine	Wingate-based training increased peak power and higher spinal excitability, with no changes in volitional wave or MVC	14/24 58.3% Low
Yamagishi et al. (111)	To determine the time course of training adaptations to 2 different SIT programs with the same sprint: Rest ratio (1:8) but different sprint duration	Recreational males (n = 13) and females (n = 5) 15-s sprint group (n = 9) males (n = 7) and females (n = 2) 30-s sprints group (n = 8) males (n = 5) and females (n = 3)	Nonexercise control	9 wk (18 sessions), 4–6 × 15 s sprints, 120 s recovery, 7% BM resistance 9 wk (18 sessions), 4–6 × 30 s sprints, 240 s recovery, 7% BM resistance	V̇o 2peak (ml·min−1·kg−1; L·min−1), O2 pulse (ml·beat−1·kg−1) and time to exhaustion (s) from an incremental exercise test to exhaustion on a cycle ergometer, 10 km time trial (s), critical power (W) from a 3-min critical power test, peak power output (W·kg−1) and total work (kJ) across training sessions 6, 12 and 18.	A 50% reduction in sprint duration does not diminish overall training adaptations over 9 wk, although cardiorespiratory function plateaus within several weeks of sprint interval training with endurance capacity more sensitive to training over a longer timeframe.	13/24 54.2% Low
Yamagishi et al. (112)	To determine the effects of recovery intensity on endurance adaptations during SIT.	Recreational males (n = 9) and females (n = 5) 30 s sprints group (n = 7) males (n = 4) and females (n = 3) Recreational males (n = 5) and females (n = 2)	No control	2 wk (6 sessions), 4–6 × 30 s sprints, 240-s active recovery at 40% V̇o 2peak, 7.5% BM resistance 2 wk (6 sessions), 4–6 × 30 s sprints, 240-s passive recovery, 7.5% BM resistance	V̇o 2peak (ml·min−1·kg−1; L·min−1) and peak power (W) from an incremental exercise test to exhaustion on a cycle ergometer 10-km cycle time trial (s) Critical power (W) from a 3-min critical power test Total work (kJ), peak V̇o 2peak (L·min−1) and mean V̇o 2peak (L·min−1) for total test and across every 30 s from a 3-min critical power test Total work (kJ), peak power (W·kg−1), peak and mean power reproducibility (%) across every training session Mean V̇o 2 (L·min−1) over 4 sprints, and 4 rest periods within sessions 1 and 6	Greater endurance adaptations occurred with active recovery when performing SIT over a short time frame, without increasing total training commitment time	10/24 41.7% Low
Zelt et al. (113)	To determine the effect of reducing SIT work interval duration on increases in maximal and submaximal performance	Recreational males (n = 23) 30-s sprint group (n = 11) 15-s sprint group (n = 12)	Exercise comparator (endurance training)	4 wk (12 sessions), 4–6 × 30 s sprints, 270-s recovery, 7.5% BM resistance 4 wk (12 sessions), 4–6 × 15 s sprints, 285-s recovery, 7.5% BM resistance	V̇o 2peak (ml·min−1), lactate threshold (mmol·L−1), relative lactate threshold (%V̇o 2peak) and peak O2 pulse (mlO2·beat−1), from an incremental exercise test to exhaustion on a cycle ergometer, peak power (W), and mean power (W) from a Wingate test, critical power (W) from a 3-min critical power test	Reducing SIT work interval from 30 to 15 s does not impact training-induced increases in either aerobic or anaerobic power, absolute lactate threshold, or critical power	15/19 78.9% Moderate

^*SIT = sprint interval training; BM = body mass; RER = respiratory exchange ration; PCr = phosphocreatine.

Methodological Quality

The overall quality ratings and ratings across the 4 domains evaluated are presented in Table 2. The mean (±SD) percentage quality rating score was 63% for comparator studies and 56% for noncomparator studies (Table 2). Quality ratings were highest for the reporting domain with all studies clearly describing characteristics of subjects and providing estimates of random variability in the data. Reporting of the number of sessions attended and a minimum number of sessions for inclusion were identified as important requirements given the small number of sessions typically performed and the subsequent influence this could have on estimates of effectiveness. It was identified that less than half of studies (44 and 41%, respectively) reported this information.

Table 2 - Overall and domain specific methodological quality ratings.

Research design	Reporting	Internal validity bias	Internal validity confounding	Statistical power	Overall rating
Comparator	80%	43%	67%	10%	High: 6% Moderate: 51% Low: 43%
Noncomparator	78%	43%	49%	11%	Moderate: 50% Low: 33% Very low: 17%
Both comparator and noncomparator	79%	43%	63%	11%	High: 6% Moderate: 50% Low: 42% V. Low: 2%

Following reporting, the domain with the highest-quality ratings was internal validity—confounding. Within this section, the highest scoring item (94%) was “were the subjects in different intervention groups recruited from the same population,” and the lowest scoring item (49%) was “were losses of subjects to follow-up taken into account?” As part of the methodological evaluation, the present review also considered studies use of familiarization for both SIT sessions and outcome assessment methods because these were identified as important sources of bias. Evaluation showed that 39% of studies included familiarization sessions for training sessions, and 53% of studies included familiarization sessions for outcome assessments. Finally, only 9 studies demonstrated a priori sufficient power for their statistical analysis (^{7,24,47,57,60,66,75,89,95}). Additionally, only 3 of these studies adjusted power calculations to account for inclusion of multiple outcome variables (^24,89,95).

Training Intervention Description

Most interventions were very short in duration, with 17 studies (31%) comprising interventions of 2 weeks, a total of 32 studies (58%) comprising interventions of 4 weeks or less, and the longest duration equal to 12 weeks. To describe the training interventions implemented, selected training data were extracted and summary statistics calculated to quantify frequency, intensity, volume, energy system specificity, and periodization. Training frequency was quantified by extracting the number of sessions performed per week. The most common training frequency employed across the studies was 3 training sessions per week (43 studies, 78%), with a range from 2 (6 studies, 11%) to 7 (2 studies, 4%).

The intensity of the training stimulus was quantified by the external resistance applied during cycling and the average duration of sprints performed (shorter duration sprints characterized by higher intensities). Forty-seven studies (85%) reported applying external resistance as a percentage of body mass. Of these studies, 8 (17%) applied loads less than 7.5% of body mass (as low as 5% body mass), 37 (79%) applied a load equal to 7.5% of body mass, and 2 (4%) applied a load up to 9.5% of body mass. Similarly, a standard value was frequently applied to sprint duration, with 37 studies (67%) including average sprint durations of 30 seconds. The percentage of studies that included average sprint durations up to 10 seconds, between 10 and 20 seconds, and between 20 and 30 seconds was equal to 13, 16, and 4%, respectively.

The volume of training was quantified by the average number of sprints per session and total sprint time per session. Eighteen studies (33%) included interventions comprising on average between 1 and 4 sprints per session, approximately half (28 studies 51%) comprising 5 or 6 sprints per session, and 9 studies (16%) comprising 6 or more with a maximum of 24. Total sprint time per session ranged from 17.5 to 210 seconds, with a median value of 150 seconds (IQR: 75–150 seconds). Energy system specificity was quantified by calculating work-to-rest ratios, with values across studies ranging from 1:100 to 1:3, with a median ratio of 1:8 (IQR: 1:9 to 1:8). Finally, the extent to which interventions employed periodization was quantified by examining variation in training frequency, volume, and intensity. Only 4 studies (7%) altered training frequency, and only 5 studies (9%) altered intensity (as quantified by sprint duration). In contrast, most studies (41 studies, 75%) included variation in training volume, calculated by differences in the number of sprints performed each session. In those studies that included variation, the median absolute change in the number of sprints was 2 (IQR: 2–3), with values ranging from 1 to 10. Periodization models altering training volume tended to be simple and included linear increases in the number of sprints performed per session. Several studies integrated a taper week through a decreased sprint number before postintervention testing.

Meta-Analysis

Of the 619 outcomes selected for extraction, 436 outcomes from 52 studies included sufficient before and after data from sprint intervention groups to be included in the meta-analysis. In contrast, only 114 outcomes from 24 studies included sufficient data from both sprint and a no-exercise control group to be included in sensitivity analyses. The primary meta-analysis conducted across all outcome types estimated a medium pooled effect demonstrating improved physical performance following SIT intervention (ES_0.5 = 0.52 [95% CrI: 0.42–0.62]; Figure 2). Relatively large between-study variance was identified τ_0.5 = 0.32 (75% CrI: 0.27 to 0.38) with central estimates indicating very low intraclass correlation ICC_0.5 = 0.02 (75% CrI: 0.00–0.09) of multiple outcomes reported within the same studies. When categorized by outcome type, the analysis provided some evidence of differences. The greatest effects were obtained for anaerobic outcomes (ES_0.5 = 0.61 [95% CrI: 0.48–0.75]), followed by mixed aerobic-anaerobic (ES_0.5 = 0.50 [95% CrI: 0.30–0.70]) and aerobic (ES_0.5 = 0.49 [95% CrI: 0.39–0.60]) outcomes. Sensitivity analyses of the main meta-analytic findings were conducted with effect sizes adjusted for no-exercise control group data. Initial assessment comparing both noncontrolled and no-exercise-controlled effect sizes demonstrated a close association (Figure 3) with a small positive bias identified for noncontrolled effect sizes (β_0:0.5 = 0.09 [95% CrI: −0.04 to 0.23]; β_1:0.5 = 1.00 [95% CrI: 0.94–1.07]; Figure 3). Sensitivity analyses conducted for the pooled data and split by outcome category resulted in no substantive changes (Table 3).

Figure 2.: Bayesian’s forest plot of multilevel meta-analysis conducted on noncontrolled effect sizes. Results from individual studies represent shrunken estimates based on the random effects model fitting and borrowing of information across studies to reduce uncertainty. Circles represent the pooled estimate from individual studies and across studies (average), generated with Bayesian’s inference along with the 95% credible intervals (95% CrI). Positive values describe improvements in outcomes based on SIT intervention. SIT = sprint interval training.

Figure 3.: Comparison of noncontrolled and no-exercise-controlled effect sizes. Solid line is the unity line, and dashed line is the best fit line illustrating positive bias of noncontrolled effect sizes.

Table 3 - Results from primary analyses conducted on noncontrolled effect sizes and sensitivity analyses conducted on controlled effect sizes from studies including nonexercise control groups.^*†

Analysis	Analysis details	Effect size/probability of medium effect	Sensitivity analysis	Sensitivity analysis details	Effect size/probability of medium effect
Noncontrolled effect sizes: all outcomes	432 effect sizes from 52 studies (mode quality = moderate: 50%)	0.52 [95% CrI: 0.42–0.62; d ≥ 0.5: Pr = 64%]	Controlled effect sizes: all outcomes	111 effect sizes from 24 studies (mode quality = low: 58%)	0.51 [95% CrI: 0.27–0.76; d ≥ 0.5: Pr = 55%]
Noncontrolled effect sizes: aerobic outcomes	259 effect sizes from 49 studies (mode quality = moderate: 51%)	0.49 [95% CrI: 0.39–0.60; d ≥ 0.5: Pr = 41%]	Controlled effect sizes: aerobic outcomes	76 effect sizes from 22 studies (mode quality = low: 50%)	0.45 [95% CrI: 0.32–0.70; d ≥ 0.5: Pr = 39%]
Noncontrolled effect sizes: anaerobic outcomes	59 effect sizes from 20 studies (mode quality = low: 50%)	0.61 [95% CrI: 0.48–0.75; d ≥ 0.5: Pr = 93%]	Controlled effect sizes: anaerobic outcomes	23 effect sizes from 8 studies (mode quality = low: 63%)	0.59 [95% CrI: 0.21–0.91; d ≥ 0.5: Pr = 73%]
Noncontrolled effect sizes: mixed aerobic-anaerobic outcomes	114 effect sizes from 18 studies (mode quality = moderate: 53%)	0.50 [95% CrI: 0.30–0.70; d ≥ 0.5: Pr = 50%]	Controlled effect sizes: mixed aerobic-anaerobic outcomes	12 effect sizes from 2 studies (mode quality = moderate: 50%)	0.40 [95% CrI: = 0.12–0.726; d ≥ 0.5: Pr = 32%]

^*Effect sizes are magnitude-based standardized mean differences.

^†Results are from multilevel random effects models with median parameter estimates and 95% credible intervals (95% CrI). Pr expresses the proportion of the pooled effect size posterior sample that is greater or equal to a moderate effect (d ≥ 0.5).

Meta-regressions were performed to assess the effects of demographic factors and training-related variables on pooled effect sizes. An initial meta-regression was performed across all outcome types to assess the effect of intervention duration with time in weeks included as a covariate. It was estimated that the pooled effect size increased by 0.03 per week (β₁: ES_0.5 = 0.03 [95% CrI: 0.00–0.06]), with the covariate added to all subsequent meta-regressions. Substantive variation was identified in the number of outcomes extracted across sedentary (23), recreationally active (350), and competitive (31) populations. No clear population differences were obtained for pooled effects sizes obtained in a meta-regression across all outcomes types, with large uncertainty in estimates identified (β_{recreational:competitive}: ES_0.5 = 0.21 [95% CrI: −0.20 to 0.61], β_{recreational:sedentary}: ES_0.5 = 0.13 [95% CrI: −0.16 to 0.38]).

The effects of training-related variables were assessed pooling across all outcomes while controlling for intervention duration and outcome type (through inclusion of fixed effects) and were assessed for each outcome separately (Table 4). To assess the influence of training intensity, meta-regressions were performed separately with sprint duration and external load expressed as categorical variables. Sprint duration was categorized as short (5–10 seconds), medium (10–20 seconds), and long (+20 seconds). Pooled effects sizes obtained across all outcomes provided evidence of a reduced effect with short-duration sprints (β_long:short: ES_0.5 = −0.15 [95% CrI: −0.42 to 0.08]), and no evidence of a difference between medium and long (β_long:mid: ES_0.5 = 0.04 [95% CrI: −0.12 to 0.19]) duration. External load was expressed as a binary variable and categorized as low (≤7% body mass) and high (+7% body mass). Results indicated a similar pooled effect size irrespective of the external load (β_high:low: ES_0.5 = −0.10 [95% CrI: −0.30 to 0.18]).

Table 4 - Results from meta-regressions conducted on training variables across all outcomes and individual outcome categories.^*†‡

	All outcomes ES0.5 [95% CrI]	Aerobic ES0.5 [95% CrI]	Mixed ES0.5 [95% CrI]	Anaerobic ES0.5 [95% CrI]
Training intensity
Sprint duration
Long (+20 s): short (5–10 s)	−0.15 [−0.42 to 0.08] Number of effects: (302/75)	−0.24 [−0.51 to −0.01] Number of effects: sizes: (195/25)	−0.02 [−0.44 to 0.66] Number of effects: (58/42)	−0.26 [−0.96 to 0.44] Number of effects: (49/8)
Long (+20 s): medium (10–20 s)	0.04 [−0.12 to 0.19] Number of effects: (302/55)	−0.03 [−0.21 to 0.15] Number of effects: (195/39)	0.22 [−0.19 to 0.64] Number of effects: (58/14)	Analysis not completed because of sample size
External load
High (>7% BM): low (≤7% BM)	−0.10 [−0.30 to 0.18] Number of effects: (311/79)	−0.10 [−0.44 to 0.22] Number of effects: (186/47)	−0.10 [−0.82 to 0.68] Number of effects: (78/32)	Analysis not completed because of sample size
Training volume
No. of sprints per session
High (+6 sprints): medium (5–6 sprints)	−0.14 [−0.42 to 0.12] Number of effects: (97/280)	−0.06 [−0.40 to 0.27] Number of effects: (41/174)	−0.22 [−0.61 to 0.10] Number of effects: (42/64)	−0.19 [−0.52 to 0.15] Number of effects: (14/42)
High (+6 sprints): low (1–4 sprints)	−0.20 [−0.51 to 0.13] Number of effects: (97/55)	−0.16 [−0.47 to 0.14] Number of effects: (41/44)	−0.39 [−0.71 to 0.15] Number of effects: (42/8)	Analysis not completed because of sample size
Total sprint time per session (standardised)	0.05 [0.00 to 0.11] Number of effects: 432	0.05 [−0.05 to 0.11] Number of effects: 259	0.07 [−0.21 to 0.34] Number of effects: 114	−0.04 [−0.29 to 0.20] Number of effects: 59
Work-to-rest ratio
Work-to-rest ratio (standardized)	−0.00 [−0.06 to 0.06] Number of effects: 432	0.06 [−0.01 to 0.12] Number of effects: 259	−0.03 [−0.35 to 0.29] Number of effects: 114	−0.08 [−0.19 to 0.04] Number of effects: 59

^*BM = body mass.

^†Effect sizes are magnitude-based standardized mean differences.

^‡Results are from multilevel random effects models with median parameter estimates and 95% credible intervals [95% CrI].

To assess the influence of training volume, meta-regressions were performed separately with the number of sprints performed in each session expressed as categorical variable and total sprint time per session expressed as a covariate. Number of sprints performed in each session were categorized as low (1–4 sprints), medium (5–6 sprints), and high (+6 sprints). Pooled effects sizes obtained across all outcomes provided evidence of an increased effect with higher volume, reflected in both the high versus medium (β_high:medium: ES_0.5 = −0.14 [95% CrI: −0.42 to 0.12]) and high versus low (β_high:low: ES_0.5 = −0.20 [95% CrI: −0.51 to 0.13]) comparisons. Similarly, results demonstrated greater effects with increased total sprint time per session. The median estimate obtained using all outcome types indicated a 0.05 increase in pooled effect size for each standard deviation increase in total sprint time per session (β₁: ES_0.5 = 0.05 [95% CrI: 0.00–0.11]).

The final training variable assessed was work-to-rest ratio. The analysis conducted across all outcome variables identified no evidence of a change in pooled effect size when expressed in standard deviation units (β₁: ES_0.5 = −0.00 [95% CrI: −0.06 to 0.06]). However, when the analysis was conducted on individual outcome types, the results indicated that lower work-to-rest ratios were more effective for aerobic outcomes, and increased work-to-rest ratios were more effective for anaerobic outcomes (Table 4). Meta-regressions performed for work-to-rest ratio were the only analyses that demonstrated clear evidence of contrasting results across outcome types (Table 4).

Small Study Effects

Evidence of extensive small study effects were identified visually from funnel plot asymmetry of noncontrolled effect sizes (Figure 4) and through the multilevel extension of Egger's regression (β₀: ES_0.5 = −0.77 [95% CrI: −0.90 to −0.64]). Results demonstrated that studies with small subject numbers (n ≤ 10) were much more likely to report very large effect sizes (ES > 1) but rarely reported small or negative effect sizes.

Figure 4.: Funnel plot of noncontrolled effect sizes and their standard errors. Highlighted blue region illustrates pooled effect size estimate and 95% credible interval. Red line illustrates a null effect.

Discussion

The aim of this review was to quantify the effects of SIT and potential moderators on a range of physical performance measures using aggregate data from published studies. The results demonstrated that healthy individuals engaging in SIT were most likely to experience moderate improvements across a range of physical performance outcomes. The largest pooled effect size was estimated for anaerobic outcomes; however, effect estimates were also similar for aerobic and mixed aerobic-anaerobic outcomes. Substantive variation in training protocols was identified with regards to primarily sprint duration (intensity) and the number of sprints performed (volume). The results from meta-regressions identified that intervention protocols with longer sprint durations and more sprints resulted in greater improvements. Most interventions included were very short in duration, with 17 studies (31%) comprising interventions of 2 weeks and 32 studies (58%) comprising interventions of 4 weeks or less. Collectively, the findings indicate that SIT training can improve a range of performance outcomes dependent on both the aerobic and anaerobic energy systems over short intervention periods, indicating that the training strategy may be effective for improving sport performance and provide multiple opportunities to include the training within broader training plans. However, it is noteworthy that only 6% of the data included in the analysis were obtained from competitive athletes, which limits application of findings to this population. Additionally, analysis of methodological quality of studies and identification of extensive small study effects indicates limitations of the research base that likely overestimates the effectiveness of interventions and suggests areas for future development.

The finding that SIT interventions generate medium effects on physical performance outcomes is consistent with previous meta-analyses. Sloth et al. (⁹⁴) and Gist et al. (⁴⁴) reported pooled effects sizes of 0.69 and 0.63, respectively, for intervention only and nonexercise-controlled comparisons. In the present review, Bayesian meta-analyses were conducted generating posterior distributions for pooled effect sizes that can be readily interpreted probabilistically. Across all outcomes, the present review estimated a median pooled effect size of 0.52 with the probability that the value was greater than small (d ≥ 0.2) almost equal to 1 and the probability that the value was greater than medium (d ≥ 0.5) equal to 0.64. The quality rating of the included studies generating this overall outcome estimate was moderate. In agreement with previous reviews, the findings of the current meta-analysis identified the existence of potential moderators. Weston et al. (¹¹⁰) reported that subjects initial training status was the most influential moderator with the largest pooled effects estimated for sedentary individuals. In the current review, there was a large skew toward studies conducted with recreationally active subjects (76%) compared with sedentary subjects (18%) or competitive athletes (6%). No clear population differences were identified for the pooled effect size. In the current meta-analysis, there were overlaps in effect estimates between the 3 outcome domains, with central values indicating that the largest values were obtained for anaerobic outcomes (Table 3). The most common measures included in this category were related to anaerobic power (e.g., peak power) and capacity (e.g., mean power, total work). Improvements in anaerobic fitness following SIT can be attributed to improvements in both anaerobic and aerobic metabolism. Previous research has demonstrated a range of enzymatic adaptations to SIT including increased activity in creatine kinase and key glycolytic enzymes, such as phosphofructokinase, lactate dehydrogenase, glycogen phosphorylase, and aldose (^58,81). Additionally, research has established that adaptations to SIT can include greater muscular glycogen concentration and enhanced muscle buffering capacity (^17,81). Across sprint durations representative of SIT (i.e., 6–30 seconds), the contributions of phosphocreatine (PCr) and anaerobic glycolysis to ATP turnover are similar (^11,38,96), leading to consistent increases in peak and mean power outputs. The ability to sustain a higher power output following SIT indicates greater fatigue resistance and enhanced exercise capacity (³⁶). However, the anaerobic adenosine triphosphate (ATP) utilization rate is reduced during the second half of a 20-second sprint when compared with the first half (¹²). Therefore, improvements in successive sprints are likely to be dependent on improvements in aerobic metabolism as demonstrated by a greater increase in aerobic ATP provision as multiple-sprints exercise progresses (³⁵). Therefore, although improvements in successive sprints may be dependent on improvements in aerobic metabolism as demonstrated by a greater increase in aerobic ATP provision as multiple-sprints exercise progresses (³⁵), the extent of this is dependent on the program variables used, which may explain the greater increases in anaerobic measures.

Few studies have investigated the relative importance of intervention duration, training volume, and training frequency in mediating the magnitude and time course of physiological adaptations following SIT. The results of the current review identified a positive association between the pooled effect size and training volume (Table 4) quantified by the average number of sprints performed per session. This finding contradicts a previous meta-analysis conducted by Vollaard et al. (¹⁰⁹) who did not find a clear relationship between the number of sprints performed in each session and change in V̇o₂max, but estimated that the relationship was most likely to be negative with improvements maximized by performing only 2 sprints per session. The dose-response relationship to SIT is likely to be determined by complex interactions between several factors with multiple ways to accumulate higher training volumes that could influence outcomes. Previous research by Stavrinou et al. (⁹⁸) identified that increasing training volume through increased frequency from 2 to 3 interval sessions per week resulted in greater increases in a range of outcome variables and altered the time course over which positive improvements were obtained. At present, there is limited research to summarize the relationship between improvements in SIT and training volume. Both the meta-regressions conducted in the present review and by Vollard et al. (¹⁰⁹) were linear in nature. However, it is unlikely that the underlying relationship would be linear and consistent with other training modalities an initial positive relationship that plateaus and then reverses may be most likely.

The meta-regressions conduced in the current review also demonstrated increased effectiveness with long duration sprints (>20 seconds) compared with short duration sprints (<10 seconds) across all outcome categories (Table 4). Most outcomes (72%) included in the meta-analysis were extracted from 30-second sprints, reflecting seminal research conducted on Wingate-based protocols (^10,38,63). However, based on criticisms that repeated 30-second sprints may not be time efficient overall (^48,61), there has been an increasing number studies investigating shorter duration sprints. The finding that increased effect sizes may be obtained with longer-duration sprints is consistent with previous research demonstrating greater oxidative contributions to ATP turnover because of PCr depletion and glycolytic inhibition (³⁸), resulting in increased mitochondrial biogenesis, mitochondrial enzyme activity, and skeletal muscle capillarization (¹⁵). Additionally, during longer-duration sprints, the number of muscular contraction cycles (i.e., cross-bridge attachments and detachments) are increased leading to greater disturbances in metabolic environments, potentially augmenting the response. Although this mechanism is yet to be fully explored, increased cross-bridge cycling will promote greater movement of calcium ions (Ca²⁺), increased levels of adenosine monophosphate (AMP) and AMP-activated protein kinase activation, and subsequent rate of peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) expression (^{30,41,59,90,102}). Following repeated 30-second sprints, PGC-1α expression has been shown to increase 7-fold (¹⁰²) compared with only 2-fold following 4-second sprints (⁹⁰), with associations demonstrated between PGC-1α expression and mitochondrial adaptations and improvements in physical performance (^30,41,59). However, it is important to note that the aforementioned measures were taken 4 and 3 hours, respectively, following sprints, with measures taken from messenger RNA and not protein changes (^90,102). Further research is required to identify the influence of sprint duration on adaptations to SIT interventions and the underling mechanisms that may be responsible.

The only meta-regressions to exhibit clearer difference in moderating effect across outcome types were obtained with work-to-rest ratios (Table 4). When analyzed across all outcomes the meta-regression identified no evidence of an effect. However, when split by outcome type, the results identified that work-to-rest ratios with shorter rest periods were more effective for aerobic outcomes, whereas ratios with longer rest periods were more effective for anaerobic outcomes. These findings are consistent with those reported by Kavaliauskas et al. (⁵³) who investigated the effects of a SIT intervention comprising 10-second sprints interspersed with either a 30-, 80-, or 120-second recovery. The aerobic performance measures demonstrating greater improvements with shorter rest periods included V̇o₂peak, incremental time to exhaustion, and 3-km time trial, whereas the anaerobic performance measures demonstrating greater improvements with longer rest periods included peak and mean power output (⁵³). These adaptations were suggested to occur because of reduced rest periods eliciting a greater aerobic challenge from a lack of PCr replenishment and glycolytic inhibition and longer rest periods enabling greater PCr resynthesis and increased power output to stimulate anaerobic adaptations (⁵³). This hypothesis is supported by a recent acute comparison study demonstrating a stronger cardiorespiratory response with higher VEpeak, V̇o₂peak, and HRpeak values in subjects performing 10-second sprints interspersed with 30 seconds vs. 4 minutes of recovery periods (²⁷). In contrast, no significant differences were reported by Lloyd Jones et al. (⁶²) for groups performing 6-second sprints interspersed with 48, 60, or 72 seconds of recovery periods for either aerobic (10-km time trial) or anaerobic (peak power output, mean power output) performance measures. Similarly, Olek et al. (⁷⁹) reported no significant differences in aerobic, anaerobic, or skeletal muscle enzyme activity following 2 weeks of 10 seconds SIT matched for total sprint time but with 2 different recovery times (1 vs. 4 minutes). Contrasting results between studies may be because of a range of factors, including the length of interventions, different work-to-rest ranges investigated across groups, and the interrelation between factors, such as sprint duration and the nature of the recovery (active vs. passive). The results from this meta-analysis and the subsequent increased statistical power obtained support the hypothesis that work-to-rest ratios can be altered to more effectively target aerobic- or anaerobic-based outcomes. However, the modifying effects reported must be treated with caution as aggregate analyses made over studies may not hold at the individual level. In addition to statistical heterogeneity because of compounding intervention differences (e.g., types of subjects, training stimulus, and setting), methodological differences (e.g., control over bias) can also act to confound moderator analyses.

Consistent with the findings reported by Vollard et al. (¹⁰⁹), the current meta-analysis identified no moderating effects of sprint resistance. Clustering of loads may have influenced results with most studies (79%) applying resistance as a percentage of body mass and selecting a load of 7.5%. Given the consistent use of 7.5% body mass as a resistance, only 2 authors in the included studies that selected an alternative value provided justification. Broatch et al. (¹³) stated that the selection of an increased resistance was made to reduce power output to 20 W·s⁻¹, whereas Kavaliauskas et al. (⁵⁴) selected a reduced resistance for female subjects because of lower expected muscle mass compared with male subjects. Although scaling resistance to body mass is less challenging practically, scaling to muscle mass may represent a more standardized stimulus for training prescription and represents an area for future research.

In the current review, methodological quality of studies was assessed using a modified version of the Downs and Black checklist. Most studies were classified as moderate (51%) or low (42%) in methodological quality. The average score obtained was 62% with the highest scoring study achieving 88% (^7,13) and the lowest 37% (⁴⁶), which was the only study to be classified as very low. The most notable methodological limitations identified in the present review included failure to blind outcome assessors, a lack of statistical power, and limited reporting of familiarization sessions. Similar findings have been reported by Sultana et al. (¹⁰¹) who also used a modified Downs and Black checklist and Rosenblat et al. (⁸⁶) who used the PEDro scale. Previous authors identified similar limitations and noted the substantive risk of bias in comparison studies where outcome assessors were not blinded to allocation. The methodological limitations identified in the present review may also have contributed to the finding of extensive funnel plot asymmetry. Often wrongly attributed solely to publication bias, funnel plot asymmetry can be caused by a range of phenomenon collectively referred to as small study effects (¹⁰⁰). Statistical heterogeneity and methodological differences can be causes of small study effects if they induce correlations between sampling error and intervention effects. However, sample size across the included studies was consistent with the interquartile range restricted to between 8 and 11 subjects such that statistical heterogeneity may not be the most influential factor explaining the asymmetry. Previous meta-analyses investigating SIT interventions have not identified any small study effects. Funnel plots and associated null hypothesis tests were presented by Gist et al. (⁴⁴), Vollaard et al. (¹⁰⁹), and Rosenblat et al. (⁸⁶) with authors reporting nonsignificant results and no clear asymmetry in the visual plots. Across these reviews, the number of data points investigated was low ranging from 9 to 38, reflecting the narrow focus of the reviews to either V̇o₂max or V̇o₂peak. In contrast, in the present meta-analysis, a total of 411 outcomes were included in the primary meta-analysis across an extensive range of variables reflecting many different physical outcomes relevant to sporting performance. The included studies featured interventions conducted on cycle ergometers popular within exercise science laboratories. Software packages connected to cycle ergometers automatically calculate numerous variables that can be analyzed and presented as absolute or relative values. Additionally, researchers often compare these variables across multiple repetitions and time points, thereby increasing the reported number of outcomes. Across the included studies in the present review, the median number of outcomes extracted was 7, with 15 or more outcomes extracted from more than 25% of the studies. Based on the performance focus of this review, additional variables such as metabolic markers and muscle fiber measures, which are also frequently reported in SIT intervention studies, were not included in this summary such that the actual number of variables analyzed by authors was even higher. The potential for researchers to retrospectively select among an extensive pool of variables and publish multiple outcomes is a common problem in sport and exercise science, which leads to overestimation of effects (²⁰), and may be a primary factor creating the small study effects identified in the present review. With many variables and small subject numbers, by chance, effects for many variables will be overestimated, and in a relatively small number of cases, overestimations will be extremely large.

An additional source of bias that has the potential to overestimate effects reported in this current review is the lack of familiarization with testing procedures. Approximately half (53%) of the included studies integrated familiarization with testing protocols. Where this was included, it was mainly limited to just 1 familiarization session or minimal information was provided regarding the procedures adopted. Connected to the issue of familiarization, the present review identified that only 20% of the included studies reported the reliability of main outcome measures, further limiting the confidence that can be placed on accuracy. To assess the effects of potential sources of bias, a comparison between effect sizes calculated with and without nonexercise controls was included. The results identified a small positive bias with noncontrolled effect sizes; however, some data points demonstrated very large positive errors (Figure 4). Additionally, it may be expected that studies that include a control group are generally of a higher overall quality and are less likely to exhibit large systematic biases. Collectively, instances of very large differences between noncontrolled and controlled effect sizes, and the presence of many very large effect sizes (41 effects > 2), despite interventions generally lasting between 2 and 4 weeks, demonstrate that the small study effects identified present a challenge for accurately estimating the benefits of SIT interventions and the most important moderators to generate optimum protocols.

The findings from the present review suggest several important areas for future research. First, the review highlights that a large array of both performance outcomes and SIT protocols have been investigated, with the results generally demonstrating moderate improvements. Reflecting the moderator analyses conducted here, future research should continue to tease out combinations of protocols that maximize specific predetermined adaptations. However, greater emphasis is required on establishing effective methods to progress SIT training and obtain greater improvements over longer periods. Most studies investigating SIT interventions are extremely short in duration and often feature no variation or progression in the training stimulus. Where progression has been included, it has generally been restricted to small increases in the number of sprints performed per session. However, given the complexity and interrelated nature of SIT training variables, progression could be achieved through many different options. It is recommended that future studies focus on longer duration interventions guided by periodization structures and research designs investigated with other training modalities, such as resistance training. Second, the review highlights limitations of the evidence base that are consistent with other areas in sports science. Most notably, the review identified extensive small study effects that are suggested to be caused primarily by a posteriori selection of outcome variables and data reduction procedures. It is recommended that where possible, future research should be hypothesis driven with clear and defined outcome measures that best match the aims and hypothesis of the research. It is also recommended that studies select a priori a smaller number outcomes that demonstrate appropriate validity, reliability (only 20% of studies reported reliability of outcome measures), and practical relevance. To address issues of statistical power and precision of effect size estimates, it is suggested that more collaborative work featuring multicenter data collection be considered. Given the ability to standardize training protocols on cycle ergometers and the consistency of equipment used, SIT research may provide an effective model for prospective multicenter collaboration. With regards to improvement of overall methodological quality, prospective reference tools, such as the Downs and Black checklist, the Consolidated Standards of Reporting Trials (CONSORT), and Consensus on Exercise Reporting Template (CERT), can assist with study design and address common limitations, including use of small sample sizes, omission of control groups, and insufficient use of familiarization sessions (^9,29,93).

Practical Applications

Short-term SIT interventions can be used to create medium improvements across a range of physical performance outcomes in healthy individuals.Training protocols comprising longer sprint durations and more sprints result in greater improvements in performance outcomes. These outcomes can be affected by the work-to-rest ratio, with shorter rest periods more effective for aerobic outcomes, whereas longer rest periods were more effective for anaerobic outcomes.Future SIT interventions studies should be designed and conducted in accordance with the proposed methodological guidelines identified within this review. It is recommended that before data collection, researchers select a limited number of outcomes that match the research hypothesis and select data reduction procedures that are appropriate and adequately statistical powered accounting for multiplicity issues. Use of research evaluation tools (Downs and Black, CONSORT & CERT) should be used to inform study design.