The Effect of High-Intensity Interval Training Type on Body Fat Percentage, Fat and Fat-Free Mass: A Systematic Review and Meta-Analysis of Randomized Clinical Trials

This systematic review and meta-analysis of randomized controlled trials (RCTs) compared body compositional changes, including fat mass (FM), body fat percentage (BF%), and fat-free mass (FFM), between different types of high-intensity interval training (HIIT) (cycling vs. overground running vs. treadmill running) as well as to a control (i.e., no exercise) condition. Meta-analyses were carried out using a random-effects model. The I2 index was used to assess the heterogeneity of RCTs. Thirty-six RCTs lasting between 3 to 15 weeks were included in the current systematic review and meta-analysis. RCTs that examined the effect of HIIT type on FM, BF%, and FFM were sourced from online databases including PubMed, Scopus, Web of Science, and Google Scholar up to 21 June 2022. HIIT (all modalities combined) induced a significant reduction in FM (weighted mean difference [WMD]: −1.86 kg, 95% CI: −2.55 to −1.18, p = 0.001) despite a medium between-study heterogeneity (I2 = 63.3, p = 0.001). Subgroup analyses revealed cycling and overground running reduced FM (WMD: −1.72 kg, 95% CI: −2.41 to −1.30, p = 0.001 and WMD: −4.25 kg, 95% CI: −5.90 to −2.61, p = 0.001, respectively); however, there was no change with treadmill running (WMD: −1.10 kg, 95% CI: −2.82 to 0.62, p = 0.210). There was a significant reduction in BF% with HIIT (all modalities combined) compared to control (WMD: −1.53%, 95% CI: −2.13, −0.92, p = 0.001). All forms of HIIT also decreased BF%; however, overground running induced the largest overall effect (WMD: −2.80%, 95% CI: −3.89 to −1.71, p = 0.001). All types of HIIT combined also induced an overall significant improvement in FFM (WMD: 0.51 kg, 95% CI: 0.06 to 0.95, p = 0.025); however, only cycling interventions resulted in a significant increase in FFM compared to other exercise modalities (WMD: 0.63 kg, 95% CI: 0.17 to 1.09, p = 0.007). Additional subgroup analyses suggest that training for more than 8 weeks, at least 3 sessions per week, with work intervals less than 60 s duration and separated by ≤90 s active recovery are more effective for eliciting favorable body composition changes. Results from this meta-analysis demonstrate favorable body composition outcomes following HIIT (all modalities combined) with overall reductions in BF% and FM and improved FFM observed. Overall, cycling-based HIIT may confer the greatest effects on body composition due to its ability to reduce BF% and FM while increasing FFM.


Introduction
High-intensity interval training (HIIT) involves repeated bouts of high-intensity work (exercise) performed at or near maximal oxygen uptake (VO 2max ) or above lactate threshold interspersed with low-intensity exercise or periods of rest [1][2][3]. Conceivably coined for the first time in 1912 by Hannes Kolehmainen, the 10,000 m Olympic champion runner [4], HIIT received little attention from athletes until the middle of the 20th century. Gradually since that time, HIIT has become increasingly popular within training regimens for athletes and coaches across sporting disciplines, both for amateurs and professionals alike [4]. Yet research into the effectiveness of HIIT as a viable means to improve fitness began much later when sports scientists compared physiological outcomes to single-component exercises (e.g., endurance exercise alone) [5][6][7]. Unsurprisingly, earlier studies with athletes revealed HIIT promoted significant physiological and performance improvements in submaximal heart rate, VO 2max , repeated sprint ability, and running/jumping performance [6,8]. More recently, attention has turned to other population cohorts, such as non-athletes, sedentary individuals, and those with chronic diseases such as obesity and cardiovascular disease, to assess whether HIIT can similarly promote beneficial body composition changes compared to performance benefits [9,10]. In this regard, HIIT can incorporate a wide range of exercise modalities (including, but not limited to, cycling, overground running, or treadmill running), intensity, recovery periods (active vs. passive), volume, repetitions, and sets that can be modified to accommodate a variety of populations [10]. Moreover, HIIT provides a time-efficient strategy to meet minimum activity levels [11] and may promote overcoming the often-cited lack of time barrier to initiating an exercise program. Therefore, incorporating HIIT into the exercise training programs of individuals across healthy and clinical cohorts represents a practical and feasible strategy to improve health and fitness. Despite the fact that HIIT has been endorsed by numerous health and fitness professionals [5] as well as the lay media [12][13][14][15][16] as a practical and time-efficient intervention for body composition improvements, meta-analytic investigations on this topic have produced conflicting results. One such meta-analysis found that low-volume HIIT (≤500 metabolic equivalent minutes per week [MET-min/week]) did not improve both lean and fat mass (FM) compared to continuous endurance training and non-exercise controls [17]. However, the authors did not compare different modes of HIIT (cycling, overground running, or treadmill running), which is relevant as body composition changes may be affected by varied movement and muscle activation patterns. On the other hand, a meta-analysis by Wu et al. found that HIIT involving a combination of cycling and whole-body circuit training led to significant improvements in body fat percentage (BF%) and lean mass compared to continuous endurance training in older adults [18]. Moreover, Wewege et al. reported that HIIT incorporating a cycle ergometer and treadmill significantly reduced FM in overweight and obese adults [19]. Based on these inconsistent findings, there is no clear consensus regarding the utility of which type of HIIT induces the greatest improvements in body composition. Therefore, the present systematic review and meta-analysis of randomized controlled trials (RCTs) were conducted to compare the effects of HIIT type, namely cycling, overground running, or treadmill running, on changes in body fat (both FM and BF%) and fat-free mass (FFM). Body compositional changes induced with these different types of HIIT were also compared to a control/no exercise condition. Because overground running causes higher muscular activation compared to other exercise modes [20], we hypothesize that HIIT using overground running may result in improved body composition compared to other modalities. A greater understanding of the type of HIIT that may promote the most beneficial body compositional changes is important to provide robust evidence for practitioners and coaches for helping develop effective and practical HIIT programs.

Methods
The present study is based on the PRISMA protocol for reporting systematic reviews and meta-analyses [21] and has been registered with PROSPERO (CRD42022323444).

Data Sources and Search Strategies
A comprehensive database search was performed by FK utilizing PubMed, Scopus, Web of Science, and Google Scholar up to 21 June 2022. The following keywords were used in the search strategy: ("High-intensity interval training" or "High-intensity intermittent exercise" or "Aerobic interval training" or "HIIT") and ("Body composition" or "Fat free mass" or "Skeletal muscle" or "Fat mass" or "Fat percentage"). No language or date of publication restrictions were applied, nor were unpublished investigations considered. Reference lists of relevant studies were manually screened to avoid oversight of eligible investigations not captured in database searches.

Study Selection and Inclusion/Exclusion Criteria
Studies meeting the following eligibility criteria were analyzed for inclusion: (1) RCTs, (2) adult population cohorts (≥18 years), (3) trials reporting mean (SD) alterations to body composition (e.g., subcutaneous skinfold caliper, air displacement plethysmography [ADP], Bioelectrical Impedance Analysis [BIA], and dual-energy X-ray absorptiometry [DEXA]) for both intervention and control (i.e., no exercise intervention) groups or that presented adequate information for calculation of overall effect size, (4) RCTs with a minimum two week (14-day) intervention duration, and (5) trials reporting outcomes in multiple manuscripts from the same data set (e.g., study participants); the original or most complete datasets were included for analysis. It should be noted that RCTs utilizing multiple intervention groups (i.e., more than one HIIT group with different intensities and/or durations) were considered different outcomes, and data were pooled separately in these cases. Investigations and clinical trials that utilized animal models or were performed in children and teenagers (years < 18), were review and/or observational in nature, lacked a control group, or involved population cohorts with a known disease were excluded from the meta-analysis. The acronym PICOS (Population, Intervention, Comparison, Outcomes, and Study Design) was used to develop a focused question and establish inclusion and exclusion criteria for this overview. The PICOS criteria are listed in Table 1.

Data Extraction
Data extraction from each eligible full-text study followed a similar format that included the first author's name, publication year, sample size (control and intervention groups), participant characteristics (mean age and body mass index [BMI]), duration of intervention, type of HIIT, and mean (SD) percentage and mass (in kg) changes in fat and FFM for both intervention and control groups. In a couple of cases when published data was unclear or ambiguous (i.e., unclear from the publication's statistical methods whether standard errors, 95% confidence intervals, or SD were used), attempts were made to contact corresponding authors for clarification. Data reported from RCTs using units other than the International System of Units (SI) were converted as applicable.

Risk of Bias Assessment
The Cochrane quality assessment tool was used to evaluate the quality of RCTs [22] in seven domains, including random sequence generation, allocation concealment, reporting bias, performance bias, detection bias, attrition bias, and other sources of bias. Any domain was scored a "high risk" if a study included methodological defects that may have affected outcomes, "low risk" if it did not, and "unclear risk" if the information provided was not sufficient to determine the impact. The overall risk of bias for a randomized controlled trial (RCT) was considered 1: high-quality if four or more domains had "low risk", 2: moderatequality if two or three domains had "low risk", and 3: low-quality if one or less domains had "low risk" [18]. The risk of bias assessment was undertaken independently by two reviewers.

Statistical Analysis
Data compiled for meta-analysis were analyzed using the Stata software, version 14 (StataCorp, College Station, TX, USA), with significant differences considered at p ≤ 0.05. The mean change (±SD) of relevant outcomes was used to estimate the overall effect size. When mean changes were not reported (e.g., only body FM in kg pre-and post-intervention was provided), such alterations in body composition (BF% and FFM) were calculated. Standard errors (SEs), 95% confidence intervals (CIs), and interquartile ranges (IQRs) were converted to SDs for subsequent analysis of mean changes in body composition outcomes using the method of Hozo et al. [23]. If the outcome measures were only reported in figures, we used software (GetData Graph Digitizer) to estimate the value. The SD change was calculated based on the following formula: A random-effects model was applied to account for between-study variations to obtain overall effect sizes. Heterogeneity among RCTs was determined by the I 2 statistic and Cochrane's Q test. An I 2 value > 50% or p < 0.05 for the Q-test was considered significant between-study heterogeneity. To find probable sources of heterogeneity, subgroup analyses were performed according to predefined variables, including the type of intervention (cycling vs. overground running vs. treadmill running), duration of the intervention (>8 vs. ≤8 weeks), participant sex (female vs. male) and BMI (<25 vs. 25-30 vs. >30 kg·m −2 ). Sensitivity analysis was conducted by removing RCTs one by one and re-analyzing pooled effects. Publication bias was assessed by examining the asymmetry of funnel plots using the Egger [24] and Begg tests [25].

Certainty Assessment
The general certainty of evidence across RCTs was rated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) working group guidelines. According to the related assessment criteria, the quality of evidence was categorized into four classes: high, moderate, low, and very low [26].

Characteristics of the Included RCTs
Characteristics of the 36 RCTs included in the current systematic review and metaanalysis are illustrated in Table 2 [31,44,45,47,55,59,60,[63][64][65]67,70]. The total sample size included 1130 individuals, 551 of whom were in a HIIT exercise intervention group, and 579 served as controls. The mean age of participants was between 18 and 57 years. According to Cochrane scores, as described, nine studies were classified as high-quality, 27 were classified as moderatequality, and no study was deemed to be low quality. The results of the quality assessment are reported in Supplementary File S1.

Findings from the Meta-Analysis
Overall, 47 effect sizes from 36 studies following the systematic review were included for meta-analysis. These trials had a total sample size of 1130 individuals, and the mean age was 26.27 ± 5.42 years.
As illustrated in Figure 4, only HIIT involving cycling interventions resulted in a significant increase in FFM compared to other exercise modalities (WMD: 0.63 kg, 95% CI: 0.17 to 1.09, p = 0.007). As with BF%, studies utilizing interventions longer than 8 weeks produced larger effects than trials of shorter (i.e., less than 8 weeks) durations (WMD: 0.63 kg, 95% CI: 0.17 to 1.10, p = 0.008). The training frequency subgroup showed that  Table 3. Sensitivity analysis revealed that after excluding individual RCTs, the removal of one study [51] in particular eliminated any observed significant overall effect of HIIT on FFM (WMD: −0.46, 95% CI: −1.38 to 0.44). Moreover, the Begg and Egger test rejected our hypothesis about the presence of substantial publication bias (p = 0.244 and p = 0.079, respectively).

Quality of Evidence
The GRADE guideline was employed to assess the quality of evidence for outcomes, which indicated the effect of FFM to be of high quality. However, the evidence about FM and BF% was downgraded to medium and low levels, respectively (Table 4).

Discussion
The current meta-analysis investigated whether, as a whole or defined aspect of a HIIT program (e.g., exercise modality, duration of intervention, participant sex, and BMI) had differing effects on body composition, including BF%, FM, and FFM. We observed an overall significant pooled effect that HIIT reduces BF% by 1.53% (p = 0.001). Subgroup analysis based on the exercise mode of HIIT showed that overground running induced the highest reduction in BF% (WMD = −2.80%; p = 0.001) and FM (WMD: −4.25 kg; p = 0.001) compared to other modalities. Additionally, FFM increased only in studies utilizing a HIIT cycling intervention (WMD = 0.63 kg; p = 0.007), while treadmill running and overground running did not increase FFM. Subgroup analysis investigating the intervention length revealed significant changes occurred for BF%, FM, and FFM in studies longer than 8 weeks duration, high frequency (3 sessions/week), with a work training bout duration of ≤60 s and ≤90 s, and active recovery time.
To the best of our knowledge, this is the first meta-analysis to demonstrate the positive effects of different HIIT types on measures of body composition, including BF%, FM, and FFM in an adult population. A previous meta-analysis of RCTs in overweight and obese participants (as opposed to the general population in our current work) showed significant reductions in BF% following HIIT, particularly when incorporating running modalities compared to cycling [19]. However, a significant limitation of this work included a small pool of RCTs (n = 13) and thus an inability to perform subgroup analyses, and a lack of investigation into the effects of HIIT on FFM. We observed an overall significant reduction of BF% in all modes of HIIT, and a subgroup analysis of the exercise mode (overground running, cycling, or treadmill running) showed a magnitude effect with the overground running protocol. Accordingly, we speculate that the magnitude of changes in BF% following HIIT is dependent on exercise modality, equipment utilized, and/or relative exercise intensity in healthy individuals. Indeed, our findings suggest cycling or treadmill running does not provide the same exercise stimulus (intensity) as overground running. It is plausible overground running, particularly with terrain variations (i.e., diverse surfaces and inclines/declines), may activate larger and more numerous muscle groups compared to treadmill running or stationary cycle ergometers to more effectively increase metabolic rate. Thus, while our findings suggest overground running is a preferential HIIT modality for reducing BF% in healthy adults, it should be recognized that this may not be applicable to clinical cohorts. For instance, it has been established that overweight and individuals with obesity exhibit altered joint loads, absolute ground reaction forces, and forefoot pressures compared to non-obese adults [75], indicating the biomechanical stress of overground running may be more detrimental to joint health in these individuals compared to weight-bearing activities such as cycling. Nevertheless, previous research has shown that a 1% reduction in BF% can lead to a 3% decrease in the risk of type 2 diabetes [76], and our findings suggest that HIIT may be an effective intervention for achieving these reductions. Thus, incorporating HIIT as a routine may be a promising approach with important clinical implications.
We also observed a greater magnitude decrease in FM in both the overground running and cycle groups, although the treadmill subgroup did not show a significant change, likely due to the small number of studies. In contrast, a systematic review and meta-analysis by Wewege et al. [19] found no significant effect of cycling on measures of body composition (whole-body FM) in overweight and obese adults. While the physiological basis of such findings is unclear, potential mechanisms that may explain the discrepancy between cycling and other exercise modalities include muscle recruitment patterns (e.g., cycling, as a none weight-bearing activity, may preferentially recruit slow-twitch muscle fibers), body position during exercise such as standing vs. sitting, and total caloric expenditure during HIIT. In this regard, previous work has indicated that the more muscle mass recruited during exercise at any given exercise intensity, the greater the energy expenditure [19,20], which is all the more relevant when HIIT involves repeated exercise bouts of near or at maximal intensities. Moreover, the skeletal muscle pump, which facilitates venous return to the heart and intravenous injection of skeletal muscle, has been shown to be more efficient during running than cycling [20]. Several factors influence the activity of the skeletal muscle pump (and thus blood flow), including the frequency of contraction, which is directly influenced by the rate of running steps or cycling revolutions per minute [20]; body position, where standing promotes involuntary muscle contraction vs. sitting; and type of muscle contraction, such as the stretch-shortening cycle during exercise [77]. All these factors are seemingly advantageous in weight-bearing activities like running. Further, the greater muscle mass involvement observed during running leads to increased blood flow which may play a role in increasing the rate of fat oxidation and fat oxidation kinetics, ultimately leading to more significant reductions in FM and BF% over time [78]. In addition, the kinematics of overground and treadmill running are different such that treadmill platforms provide a lower impact stimulus and potentially less muscle activation compared to overground running. For example, the rectus femoris and biceps femoris muscles of the anterior and posterior regions of the quadriceps, respectively, have been shown by electromyography (EMG) to be more active during overground running than in treadmill running [79]. While such enhanced lipid metabolism may be due to the exercise stimulus itself, it is likely that even though total energy expenditure is relatively low during HIIT training, greater reductions in body fat come from elevated post-exercise oxygen consumption (EPOC) observed with activities involving more muscle mass [80]. Future investigations should endeavor to determine the most effective HIIT exercise modality under well-controlled and similar conditions (e.g., intervention duration, relative/absolute intensity, participant characteristics, etc.), notwithstanding any participant limitations.
Overall, HIIT appeared to significantly and effectively influence FFM, yet such differences were only noted in RCTs utilizing a cycling intervention (WMD = 0.63 kg; p = 0.007) compared to other exercise modalities investigated. The reasons for such disparity in the response of FFM to exercise modality are not fully known. Intuitively, this may be due to the relative specificity of recruiting large muscle groups of the lower extremity necessary for pedaling [20,81] and resistance imposed by increases in cycle ergometer wattage/power output that are distinct from the gravitational forces imposed by running. Heydari et al. [51] similarly observed significant increases in FFM following 12 weeks of HIIT in young males, while we have also demonstrated 6 weeks of cycle ergometer HIIT to significantly increase lean mass (1.0 kg), albeit in conjunction with a high protein diet [82]. More research is required on the possible mechanism of enhanced cycling-induced FFM improvements during HIIT compared to other modalities before definitive conclusions may be reached.
Moreover, subgroup analysis in our current study indicated that a more than 8 weeks duration of HIIT proves more beneficial for BF%, FM, and FFM than shorter duration programs, indicating the need for a prolonged stimulus to reduce body fat and promote muscle mass remodeling. It was also found that three sessions per week were more effective on body composition than two sessions per week in the training frequency subgroup. Consistent with the findings of a study by Stavrinou et al. [57], it was concluded that a higher frequency of exercise during a week has effective results in reducing BF%. There is evidence that high-frequency exercise (≥3 sessions) can promote fat loss due to significant catecholamine response, increased concentration of β-adrenergic receptors, and increased fat oxidation versus low frequency [57]. Our analysis of the existing data revealed that a lower time of training (≤60 s) per repetition and shorter rest (≤90 s) period was more effective for improving body composition than a longer training schedule. A short work bout duration of ≤60 s per repetition increased FFM while a longer work bout (>60 s) had no significant effect. In agreement with our finding, multiple studies with short-term intervals have reported reduced body fat [14,63,83,84]. However, a recent study showed both short and long intervals reduced BF% and FM (kg), although short-term intervals (≤60 s) induced a greater reduction. In this regard, Rønnestad et al. [85] reported that the short-intervals protocol (30 s work intervals separated by 15 s recovery) achieved a larger relative improvement in VO 2max than the long intervals group (4 × 5-min work intervals separated by 2.5-min recovery periods). Another novel finding from our current work was that active and short rest (≤90 s) is more effective for promoting body composition changes, especially FFM. Both active and passive recovery showed a reduction in BF% and FM. There is a paucity of knowledge regarding the mechanistic basis for these positive effects on body composition with active and passive recovery. Findings by Spencer et al. [86] showed active recovery promotes an increase in muscle lactate removal compared to the passive recovery condition. Presumably, in active recovery, blood lactate concentration was decreased due to an increase in muscle lactate metabolism, rather than a greater release of lactate. High blood lactate levels down-regulate the use of glucose and free fatty acids (FFA). Furthermore, lactate accumulation upregulates the expression of monocarboxylate transporter 1 (MCT1) which may serve a role in the efflux of lactate [87]. Nonetheless, in our results, regardless of rest mode, BF% and FM (kg) decreased significantly, while only active rest could increase FFM (WMD: 0.68; p = 0.004). Future studies are required to better understand the effect of HIIT with active recovery on FFM changes.
Several limitations in the present meta-analysis are worth noting when interpreting our results. Primarily, a majority of investigations were conducted on overweight or obese inactive adults as opposed to active normal body mass or BMI adults. Hence, it remains challenging to draw generalized conclusions on how different types of HIIT influence body composition outcomes in these populations (active and inactive adults). In addition, significant heterogeneity (I 2 ) was found across studies, particularly with BF%, as noted, which likely relates to the large variability in study design (e.g., exercise modalities, duration, participant sex, age range, and training intensity) between studies. Furthermore, BF%, FM, and FFM were measured using different methodologies (e.g., subcutaneous skinfold caliper, ADP, BIA, and DEXA), pre-test guidelines, dietary control, participant hydration status, etc., which may give different results when tracking the changes in body composition variables [88].
In conclusion, the results of this meta-analysis demonstrate favorable body composition outcomes following HIIT, including overall reductions in BF%, FM, and improved FFM. Such improvements in BF% were enhanced following HIIT involving overground running compared to other exercise modalities. While similar overall improvements as BF% were noted for FFM following HIIT, in this case, cycling proved to be the only effective mode of HIIT. These findings indicate that individuals who want to reduce their BF% may benefit more from overground running, while those who want to increase their FFM may benefit more from cycling. In all cases, studies lasting 8 weeks or more provided the necessary stimulus to promote improvements in BF%, FM, and FFM when pooled together. These data suggest that training for more than 8 weeks, at least three sessions per week, lower than 60 s work intervals separated by ≤90 s recovery periods, and active rest (instead of passive) are more effective on body composition. In light of the public health implications of maintaining proper body composition for health and well-being, and considering the increasing popularity of HIIT exercises, future research should aim to determine ideal training models for maximizing improvements in body fat and FFM in the general population.