Metabolic intensity and energy cost of Pilates exercises: an exploratory systematic review and meta-analysis of limited and heterogeneous evidence

Abstract

Background and objectives

Although Pilates exercises are widely practiced worldwide, the available evidence regarding their metabolic intensity and energy cost remains limited and inconsistent. This study aimed to explore the metabolic equivalent (MET) and energy cost of Pilates exercises. As a secondary objective, we examined other measures commonly used in professional practice to monitor exercise intensity, such as heart rate and the rating of perceived exertion (RPE).

Methods

A systematic search was conducted in PubMed, CENTRAL, Embase, Web of Science, SPORTDiscus, and CINAHL (May 2024). The methodological quality of the studies was assessed using the Rosendal scale. For the meta-analysis, weighted averages were calculated for the following outcomes: METs, energy cost (kcal·min⁻¹), oxygen consumption (V̇O₂, ml·kg⁻¹·min⁻¹), heart rate (bpm), and RPE (Borg 6–20). The certainty of the evidence was assessed using the GRADE approach.

Results

Six studies met the inclusion criteria. Very low- to low-certainty evidence suggests that a typical Pilates session elicits, on average, 3.7 (95% CI: 3.1–4.3) METs, 3.8 (95% CI: 3.1–4.5) kcal·min⁻¹, V̇O₂ of 10.3 (95% CI: 8.1–12.6) ml·kg⁻¹·min⁻¹, heart rate of 108.6 (95% CI: 104.1–113.0) bpm, and RPE of 10.6 (95% CI: 9.3–12.0). However, sensitivity analyses excluding two studies with high risk of bias and without indirect calorimetry yielded lower and more conservative estimates of 3.0 (95% CI: 2.3–3.6) METs and 2.9 (95% CI: 2.3–3.6) kcal·min⁻¹, which should be considered the most methodologically reliable estimates available to date. The rest interval between sets and exercises emerged as a potential moderator, with durations ≤ 60 s associated with higher metabolic intensity.

Conclusion

Within the limits of the current and highly heterogeneous evidence base, Pilates appears to elicit light-to-moderate metabolic intensity. These conclusions are exploratory, and the estimates require confirmation through future well-controlled primary studies.

Systematic review registration

Prospero registration number: CRD42022370937.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13643-025-03056-y.

Introduction

Pilates is characterized by localized muscular resistance and stretching exercises, which can be performed in two main formats: (1) mat Pilates, conducted on a gym mat using primarily body weight as resistance and, optionally, accessories such as Swiss balls and elastic bands; and (2) Pilates on equipment, using apparatus such as the Cadillac, Reformer, Step Chair, and Ladder Barrel, specifically designed for Pilates practice. In Pilates on equipment, resistance is provided not only by body weight but also through springs of varying intensities [1].

Although Pilates exercises were developed in the mid-twentieth century, they have only recently gained popularity worldwide. In the United States, the number of practitioners increased from approximately eight million in 2010 to nearly 12 million in 2023 [2]. In Brazil, a population survey grouped physical activities by affinity, and Pilates was classified in the same category as Yoga, stretching, and gymnastics. Among the general population, these activities are among the six most commonly practiced, and for adults over 60 years of age, they rank second only to walking [3].

Pilates has been prescribed for a variety of purposes, including improving muscle strength and endurance, flexibility, postural alignment, and postural control [4–7]. Exercises are performed following six fundamental principles: concentration, centering, control, fluidity, precision, and breathing [1]. Consequently, movements are executed more slowly than conventional exercises, resembling traditional mind–body modalities such as Yoga [8]. Due to this slower execution, Pilates exercises may not consistently reach moderate-to-vigorous intensity in terms of metabolic equivalents (METs) [9], although meta-analyses have demonstrated their potential to improve cardiorespiratory fitness and reduce body fat [10–13].

Understanding the metabolic intensity and energy cost of physical activities is important for exercise prescription, particularly when the goals include reducing body fat, improving cardiorespiratory fitness, and meeting current physical activity guidelines aimed at promoting and maintaining health [14]. In 2024, an update of the Compendium of Physical Activities for Adults was published, providing energy costs for 1,114 physical activities and specifying their intensity in METs, standardized by the resting metabolic rate (1 MET = 3.5 ml·kg⁻¹·min⁻¹ or 1 kcal·kg⁻¹·hour⁻¹, corresponding to the energy cost of sitting quietly) [15].

Adults are recommended to achieve a cumulative moderate-to-vigorous physical activity of 450–750 MET·min·wk⁻¹ (METs × cumulative daily minutes × number of sessions per week). METs are particularly useful, as they allow the comparison of energy cost and intensity across individuals with different body weights and fitness levels. Activities between 3 and 6 METs are considered moderate intensity, whereas those above 6 METs are classified as vigorous intensity [14]. In the 2024 Compendium of Physical Activities for Adults, Pilates was categorized into two groups: (1) traditional mat Pilates (1.8 METs) and (2) Pilates, general (2.8 METs) [16], corresponding to very light (< 2 METs) and light (2–2.9 METs) intensities, respectively [17].

However, studies investigating the metabolic intensity and energy cost of Pilates exercises have reported highly variable results, with METs ranging from 1.8 to 9.2. This variability may be influenced by potential moderators, including participants’ prior Pilates experience, rest intervals between sets/exercises, and the type of Pilates performed (mat Pilates vs. Pilates on equipment) [18–23]. Therefore, the primary objective of this exploratory systematic review and meta-analysis was to synthesize the available evidence and improve the understanding of factors influencing the metabolic intensity and energy cost of Pilates exercises. As a secondary objective, we examined additional variables commonly used in professional practice to monitor exercise intensity, such as heart rate and rating of perceived exertion (RPE).

Methods

This systematic review with meta-analysis was conducted following the PRISMA guidelines [24]. Methodological procedures adhered to the recommendations of the Cochrane Collaboration [25], and the protocol was prospectively registered in PROSPERO (CRD42022370937). The review was completed in accordance with the registered objectives, although the projected completion date was extended by approximately 12 months due to logistical constraints that delayed the initiation of the review process. No significant deviations from the registered protocol occurred. All planned analyses were conducted; however, a leave-one-out sensitivity analysis was added to strengthen the evaluation of the robustness of the findings, and although the planned subgroup analyses were performed, they were moved to the supplementary material due to the limited and heterogeneous evidence available. Both modifications were implemented in response to peer-review feedback and did not alter the methodological intent or scope of the review.

The inclusion criteria were as follows: (a) interventions involving Pilates exercises; (b) assessment of energy cost and exercise intensity; and (c) healthy participants, with no restrictions on age or physical activity level.

Databases and search strategy

The following databases were searched: PubMed (pubmed.gov, 1996 to May 9, 2024), Embase (embase.com, 1947 to May 9, 2024), CENTRAL (cochranelibrary.com, 1998 to May 9, 2024), CINAHL (via EBSCOhost, 1937 to May 9, 2024), Web of Science (webofscience.com, 1900 to May 9, 2024), and SPORTDiscus (via EBSCOhost, 1985 to May 9, 2024). In addition, clinical trial registries (clinicaltrials.gov and apps.who.int/trialsearch/) were consulted to identify potential unpublished studies. No restrictions were applied regarding publication date or language. The reference lists of all included studies were also manually screened to locate additional reports. The search strategies combined terms related to the intervention (Pilates) with the outcomes of interest (energy cost and exercise intensity) using Boolean operators (OR/AND). The complete search strategy for each database is provided in the supplementary material.

Study selection

The database search was conducted by the team member with the greatest experience in search strategies (RGO), who subsequently identified and manually removed duplicates using reference management software (Rayyan, https://www.rayyan.ai/). According to Cochrane guidance [25], these procedural steps do not require duplicate independent assessment, provided that a predefined search strategy is applied. Title and abstract screening and full-text assessment were performed independently and blindly by two reviewers (GBBV and RAGP). Any disagreements between the two reviewers were first discussed in detail, and when consensus could not be reached, a third reviewer (RGO) adjudicated the decision.

Data extraction

Data extracted from each eligible study included the following: (a) study identification: author, year of publication, and country of origin; (b) participants: sample size, sex, age, height, body mass, and BMI; (c) experimental design: study type, number of sessions, interval between sessions, familiarization, and groups/type of intervention; (d) Pilates protocol: number and type of exercises, temperature and humidity control, session duration, sets and repetitions, and interval between sets and exercises; (e) preexperimental session control: fasting, diet, physical activity, and sleep; (f) baseline and postexperimental measurements; (g) devices used to measure energy cost and monitor exercise intensity; (h) main outcomes: METs, energy cost (kcal·min⁻¹), oxygen volume (V̇O₂, ml·kg⁻¹·min⁻¹), heart rate (bpm), lactate (mmol·L⁻¹), and rating of perceived exertion (RPE). Two blinded reviewers (GBBV and RAGP) used a standardized data extraction form. Any discrepancies between the reviewers were discussed, and if agreement could not be reached, a third reviewer (RGO) adjudicated the final decision.

When outcome data were not directly reported, estimates were derived from other available information, following procedures used in previous systematic reviews with similar objectives [8]. Specifically, METs values were calculated from V̇O₂ consumption [19, 22], assuming that 1 MET = 3.5 ml·kg⁻¹·min⁻¹: METs = V̇O₂ (ml·kg⁻¹·min⁻¹)/3.5. In two studies that did not measure V̇O₂ during the Pilates session [21, 23], METs were estimated from the reported energy cost, using the equivalence of 1 MET = 1 kcal·kg⁻¹·hour⁻¹ and the mean body mass of the sample: METs = (energy cost [kcal·min⁻¹] × 60)/body mass (kg). When V̇O₂ and energy cost were reported for the entire session duration, these values were converted to ml·kg⁻¹·min⁻¹ [22] and kcal·min⁻¹ [21, 22], respectively, to standardize units across studies. Because these variables are linearly related, both means and standard deviations were converted using the same multiplicative factor, in accordance with the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions [25]. For one study that presented results only in graphical form, the mean and standard deviation values were extracted using WebPlotDigitizer 4.7 [20].

Methodological quality of the included studies

The methodological quality of the included studies was assessed using the Rosendal scale [26], which was previously applied in a scoping review investigating methods for assessing the energy cost of resistance exercise [27]. This scale integrates items from the Jadad scale [28], PEDro scale [29], Delphi checklist [30], and CONSORT recommendations [31]. Each study’s score was calculated as the number of items met (“yes” answers) divided by the total number of relevant items. Studies with a final score below 60% were classified as having a high risk of bias. Two independent reviewers (GBBV and RAGP) conducted the assessment in a blinded manner, and any disagreements were discussed and, when consensus could not be reached, a third reviewer (RGO) adjudicated the final decision.

Data synthesis

For the meta-analysis, the effect measure was the weighted average across studies for each outcome of interest. For studies including multiple Pilates intervention arms, the arms were pooled, except in subgroup analyses comparing the type of Pilates (mat Pilates vs. Pilates on equipment) or the rest interval between sets/exercises (≤ 60 s vs. > 60 s), in which arms were either pooled or analyzed separately, as appropriate.

Heterogeneity was assessed using the Cochrane Q test, with p ≤ 0.10 considered statistically significant, and quantified with the I² statistic: 0–40% may not be important, 30–60% may represent moderate heterogeneity, 50–90% may represent high heterogeneity, and 75–100% may represent considerable heterogeneity [25]. Fixed effects models were used when heterogeneity was nonsignificant; otherwise, random effects models were applied.

Meta-analyses were conducted when at least two studies reported the same outcome. Subgroup differences were analyzed using ANOVA based on the Q test. Since no meta-analysis included 10 or more studies, funnel plot inspection was not performed [25, 32]. All analyses were performed using Comprehensive Meta-Analysis software (CMA, version 2.2.064; Biostat, NJ, USA).

Sensitivity and subgroup analyses

Sensitivity analyses were performed to assess potential changes in outcome measures by excluding studies at high risk of bias and those that did not estimate energy cost or metabolic intensity via indirect calorimetry. A leave-one-out sensitivity analysis was also performed to observe the individual impact of each study on the meta-analysis. Subgroup analyses were conducted to explore possible differences in outcomes for the following comparisons: (a) previous experience with Pilates (experienced vs. inexperienced participants); (b) rest interval between sets/exercises (≤ 60 s vs. > 60 s); and (c) type of Pilates (mat Pilates vs. Pilates on equipment).

Certainty of evidence

The certainty of evidence for each outcome was assessed using the GRADE approach (Grading of Recommendations, Assessment, Development, and Evaluations) [32]. Two independent reviewers (LCO and RGO) conducted blinded assessments, with disagreements resolved by consensus. The GRADE domains include risk of bias, inconsistency, indirectness, imprecision, and publication bias. This approach provides criteria for downgrading the certainty of evidence on the basis of limitations in these domains. Accordingly, the final rating for each outcome can be classified as follows: (a) high: new research is unlikely to change the estimate or confidence in the results; (b) moderate: future research is likely to affect confidence and may change the estimate; (c) low: future research is likely to have a significant impact on confidence and may change the estimate; and (d) very low: the results are highly uncertain.

Results

Initially, 1,191 potentially eligible reports were identified, in addition to three clinical trial registries. After removing duplicates, 652 reports remained for title and abstract screening, of which 636 were excluded for not meeting the eligibility criteria. The full texts of the remaining 16 reports were sought, but three could not be retrieved (Supplementary Table 1), leaving 13 reports for full-text review. Of these, eight were excluded (Supplementary Table 2) for the following reasons: (a) did not evaluate outcomes of interest (5 reports); (b) were conference abstracts (1 report); (c) were opinion articles (1 report); and (d) were registrations of a study already included (1 report). Therefore, five reports (each representing one study) met the eligibility criteria. By reviewing the references of the included studies, one additional report was identified and met the inclusion criteria, resulting in a total of six studies included in the systematic review and meta-analysis (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram

Table 1 summarizes the six studies included in the systematic review. The studies were published between 2014 and 2024, primarily in Brazil [19–21, 23], followed by Turkey [18] and South Korea [22]. The number of participants ranged from 10 to 33, with most studies including only women and two including both sexes. All participants were healthy young adults with a mean BMI within the normal weight range (18.5–24.9 kg·m^⁻2) and were inexperienced with Pilates, except in two studies that included experienced participants [21, 22]. The most common experimental design was crossover, whereas two studies used a single-group intervention [21, 23]. Most studies included familiarization sessions, ranging from one to eight sessions; however, two studies did not report this information [18, 21].

Table 1.

Summary of studies that evaluated the energy cost and metabolic intensity of a Pilates exercise session

Author, year and country	Participants (n, sex, age, height, body mass and BMI)	Experimental Design	Pilates Protocol	Control of fasting, diet, PA and sleep	Assessment		Results* (METs, EC [kcal·min⁻1], V̇O2 [ml·kg⁻1·min⁻1], HR [bpm], Lactate [mmol·L⁻1], RPE)
					Baseline and postsession	Devices
Gultekin et al. [18] (2024) Turkey	31 (16♀; 15♂), healthy and without experience with Pilates 21.7 ± 3.7 years 168.7 ± 7.8 cm 63.6 ± 11.4 kg 22.2 ± 2.9 kg·m⁻2	Crossover (2 sessions): - Interval between sessions: NR - Rest between exercise protocols within the session: 30 min - Familiarization: NR Session A - 45 min of Mat Pilates - 10 min of walking at 3.2 km/h - 10 min of walking at 4.8 km/h Session B - 10 min of walking at 3.2 km/h - 10 min of walking at 4.8 km/h - 45 min of Mat Pilates	- 25 exercises (1 respiratory; 3 postural; 4 joint mobilization; 9 lumbopelvic stabilization; 8 core strengthening) - Temperature/humidity: NR - Session period/time: NR - Series x Repetitions: 3 X 10 - Series interval: 10 s - Exercise interval: 30 s - Session duration: 45 min	Abstain from food, drink, cigarettes, caffeine (4 h); and vigorous exercise (24 h)	Baseline: 20 min in supine position	Indirect calorimetry (Cosmed Quark CPET metabolic system [face mask]); calibration reported Wearable monitor (heart rate—Polar H10 [chest strap])	Mat Pilates: METs = 3.4 ± 0.5 EC = 3.7 ± 0.8 V̇O2 = 12.0 ± 1.7 HR = 112.5 ± 11.8 # Walking (3.2 km/h): METs = 3.3 ± 0.4 EC = 3.5 ± 0.5 V̇O2 = 11.9 ± 1.4 HR = 106.0 ± 12.3 Walking (4.8 km/h): METs = 4.1 ± 0.5 # EC = 3.7 ± 0.8 # V̇O2 = 14.5 ± 1.6 # HR = 113.1 ± 12.3 #
Almeida et al. [20] (2021) Brazil	33♀, healthy, physically inactive and inexperienced with Pilates 22.0 ± 3.3 years 163.0 ± 5.0 cm 58.9 ± 9.3 kg 22.0 ± 3.6 kg·m⁻2	Crossover (3 sessions): - Interval between sessions: 48 h - Familiarization: 5 sessions Session A Mat Pilates 1 (single set and long rest time) Session B Mat Pilates 2 (single set and short rest time) Session C Mat Pilates 3 (multiple sets and short rest time)	- 12 exercises (3 lumbopelvic stabilization; 6 core strengthening; 2 joint mobilization; 1 postural) - Temperature/humidity: 22°/40–50% - Session period/time: morning Protocol 1 (Mat Pilates 1) - Sets x Repetitions: 1 X 10 - Sets/exercise interval: 120 s - Session duration: 33 min Protocol 2 (Mat Pilates 2) - Sets x Repetitions: 1 X 20 - Sets/exercise interval: 30 s - Session duration: 23 min Protocol 3 (Mat Pilates 3) - Sets x Repetitions: 3 X 12 - Sets/exercise interval: 30 s - Session duration: 49 min	Abstain from exercise, caffeine and stimulants (48 h); and food and liquids, except water (8 h) Standard breakfast 321 kcal (30 min before sessions)	Baseline: 15 min in supine position Post session: 5 min in seated position	Indirect calorimetry (Metalyzer 3b Cortex metabolic system [face mask]); calibration reported Wearable monitor (heart rate—Polar RS800cx [chest strap])	Mat Pilates 1: METs = 1.8 ± 0.3 EC = 2.0 ± 0.4 V̇O2 = 6.5 ± 0.7 HR = 94.9 ± 4.6 Mat Pilates 2: METs = 2.8 ± 0.4 # EC = 2.6 ± 0.9 # V̇O2 = 9.8 ± 1.4 # HR = 114.2 ± 3.2 # Mat Pilates 3: METs = 3.1 ± 0.4 # EC = 3.0 ± 0.9 # V̇O2 = 10.4 ± 1.4 # HR = 114.8 ± 9.9 #
Andrade et al. [19] (2021) Brazil	18♀, healthy, physically inactive and inexperienced with Pilates 22.0 ± 3.0 years 1.6 ± 0.1 m 61.0 ± 10.0 kg 21.0 ± 3.0 kg·m⁻2	Crossover (2 sessions): - Interval between sessions: 48 h - Familiarization: 8 sessions Session A Mat Pilates Session B Pilates on the Reformer equipment	- 15 exercises (4 stretching; 5 lumbopelvic stabilization; 4 core strengthening; 2 joint mobilization) - Temperature/humidity: 22°/40–50% - Session period/time: morning - Series x Repetitions: 1 X 10 - Series/exercise interval: 120 s - Session duration: 90 min	Abstain from exercise, food and liquids, except water (8 h) Standard breakfast 321 kcal (30 min before sessions)	Baseline: 15 min in supine position Post session: 5 min in seated position	Indirect calorimetry (Metalyzer 3b Cortex metabolic system [face mask]); calibration reported Blood lactate (spectrophotometry using a YSI 2300 STAT Plus [capillary blood sample]) Wearable monitor (heart rate—Polar RS800cx [chest strap])	Mat Pilates: METs = 1.8 ± 0.2 EC = 1.9 ± 0.3 V̇O2 = 6.4 ± 0.7 HR = 92.3 ± 8.5 Lactato = 1.7 ± 0.4 Pilates on the Reformer: METs = 2.5 ± 0.4 # EC = 2.6 ± 0.5 # V̇O2 = 8.7 ± 1.2 # HR = 107.9 ± 5.8 # Lactato = 1.5 ± 0.4
Jung et al. [22] (2020) South Korea	10♀, healthy and experienced in Pilates 26.4 ± 3.0 years 162.2 ± 4.1 cm 50.8 ± 5.8 kg 19.3 ± 1.6 kg·m⁻2	Crossover (2 sessions): - Interval between sessions: 7 days - Familiarization: 1 session Session A Mat Pilates N (normoxic condition, with altitude at sea level [O2 = 20.9%]) Session B Mat Pilates H (hypoxic condition, with simulated altitude of 3,000 m [O2 = 14.5%])	- 25 exercises (4 stretching, 5 lower limb strengthening, 5 upper limb strengthening, 4 lumbopelvic stabilization, 7 core strengthening) - Temperature/humidity: 23°/50% - Session period/time: morning - Series x time/execution: 1 × 2 min - Interval between series/exercise: no rest intervals (continuous execution) - Session duration: 50 min	NR	Baseline: 30 min in ambient conditions	Indirect calorimetry (Cosmed K5 metabolic system [face mask]); calibration: NR Wearable monitor (cardiac function—PhysioFlow PF-05 [thoracic bioelectrical impedance])	Mat Pilates N: METs = 3.6 ± 0.7 EC = 3.2 ± 0.6 V̇O2 = 12.7 ± 2.2 HR = 102.8 ± 13.1 Mat Pilates H: METs = 3.7 ± 0.6 EC = 3.4 ± 0.5 V̇O2 = 13.0 ± 2.1 HR = 113.5 ± 11.1 #
Santo et al. [21] (2020) Brazil	10 (5♀; 5♂), healthy and experienced in Pilates (> 1 year) 26.3 ± 3.9 years 1.70 ± 0.13 m 71.7 ± 17.0 kg 24.9 ± 3.0 kg·m⁻2	Single group intervention (1 session) Single Mat Pilates session	- 21 exercises (3 stretching; 10 lumbopelvic stabilization; 8 core strengthening) - Temperature/humidity: NR - Session period/time: NR - Set x repetition: 1 × 10 - Set/exercise interval: no rest intervals (continuous execution) - Session duration: 45 min	NR	NR	Wearable monitor (heart rate—Polar T31 [chest strap]) RPE (Borg 6–20)	Mat Pilates: METs = 3.9 ± 1.4 EC = 4.7 ± 1.7 HR = 117.5 ± 16.0 RPE = 9.5 ± 3.5
Silva et al. [23] (2014) Brazil	10♀, healthy and physically inactive 19.0 ± 0.8 years 1.6 ± 0.1 m 53.7 ± 11.1 kg 19.8 ± 2.9 kg·m⁻2	Single group intervention (1 session) - Familiarization: 1 session Single Pilates session on equipment (Step-Chair, Reformer, Ladder Barrel, Cadillac and Wall-Unit)	- 5 exercises (1 stabilization and 4 strengthening) - Temperature/humidity: NR - Session period/time: NR - Series x repetition: 1 × 15 - Series/exercise interval: 60 s - Session duration: 30 min	NR	Baseline: 10 min in supine position Post session: 5th and 10th min	Wearable monitor (heart rate—Polar FT4 [chest strap]) RPE (Borg 6–20)	Pilates on equipment: METs = 9.2 ± 2.4 EC = 8.3 ± 2.2 HR = 109.7 ± 15.6 RPE = 11.3 ± 2.7

^*results reported as the mean and standard deviation; # significantly superior to the unlabeled comparison group (intergroup comparisons within the same study); BMI: body mass index; EC: energy cost; HR: heart rate; MET: metabolic equivalent; NR: not reported; PA: physical activity; RPE: rating of perceived exertion (Borg 6–20); V̇O₂: oxygen volume

The type of Pilates was mainly mat Pilates, with the exception of one study that used Pilates on equipment [23] and another that compared mat Pilates with Pilates on the Reformer [19]. In the crossover studies, the comparisons were as follows: (a) mat Pilates vs. walking [18]; (b) mat Pilates with different rest intervals and sets [20]; (c) mat Pilates vs. Pilates on the reformer [19]; and (d) mat Pilates under normoxic vs. hypoxic conditions [22]. The Pilates protocols varied from five [23] to 25 exercises [18, 22]. Only three studies reported the control of temperature, humidity, and time of day during which the sessions took place [19, 20, 22]. Two studies used multiple sets per exercise [18, 20], whereas the others used a single set. The interval between sets/exercises ranged from 0 to 120 s, and the session duration varied from 30 to 90 min.

Control of preexperimental conditions — including fasting, diet, and physical activity — was reported in only three studies [18–20], of which two provided a standardized breakfast before the session [19, 20]. No study controlled for sleep on the night preceding the experimental session. Baseline measurements of the variables of interest were obtained in five studies [18–20, 22, 23], whereas postintervention measurements were obtained in only three studies [19, 20, 23].

Regarding the measuring instruments, most studies assessed energy cost via indirect calorimetry and used wearable monitors to record heart rate [18–20, 22]. One study did not report calibration procedures for the metabolic system [22]. Two studies estimated energy cost using predictive formulas based on heart rate obtained from wearable monitors and also assessed RPE using the Borg 6–20 scale [21, 23]. Only one study measured blood lactate concentrations before and after the experimental session [19].

Among the results reported for Pilates sessions, METs ranged from 1.8 to 9.2, energy cost from 1.9 to 8.3 kcal·min⁻¹, V̇O₂ from 6.4 to 13.0 ml·kg⁻¹·min⁻¹, heart rate from 92.3 to 117.5 bpm, and RPE (Borg 6–20 scale) from 9.5 to 11.3, across the studies that assessed these parameters. The only study that measured blood lactate reported concentrations increasing from 0.62 mmol·L⁻¹ preintervention to 1.5 mmol·L⁻¹ after Pilates on the Reformer and 1.7 mmol·L⁻¹ after mat Pilates.

To facilitate interpretation of the substantial heterogeneity observed across outcomes, an additional table was included to categorize the studies according to key protocol characteristics that may influence metabolic responses to exercise, including exercise format, session duration, rest intervals, participant experience with Pilates, and methods used to assess energy cost (Table 2). This table complements the detailed methodological information provided in Table 1 by offering a concise, cross-study comparison of protocol features that likely contributed to the observed between-study variability.

Table 2.

Key protocol characteristics contributing to heterogeneity across included studies

Study	Type of Pilates	Session duration (min)	Rest intervals (s)	No. of exercises	Practitioner's experience with Pilates	Energy cost assessment
Gultekin et al. [18]	Mat	45	10–30	25	Novice	Indirect calorimetry
Almeida et al. [20]	Mat	23, 33 and 49	30 and 120	12	Novice	Indirect calorimetry
Andrade et al. [19]	Mat and Equipment	90	120	15	Novice	Indirect calorimetry
Jung et al. [22]	Mat	50	no rest intervals	25	Experienced	Indirect calorimetry
Santo et al. [21]	Mat	45	no rest intervals	21	Experienced	Heart rate
Silva et al. [23]	Equipment	30	60	5	Novice	Heart rate

Multiple values/information are reported for studies that included more than one experimental group. Detailed methodological information is presented in Table 1

Methodological quality of studies

The assessment of the methodological quality of the studies included in this systematic review is summarized in Table 3. Among the six included studies, two were classified as having a high risk of bias (score < 60%). The mean methodological quality score across all studies was 66.3 ± 0.2%. It is noteworthy that several important methodological elements were not carried out or not reported across studies. The most frequent omissions included failure to describe the method used to generate the random allocation sequence (item 3), lack of sample size justification (item 4), absence of control of pre-trial conditions (item 5), lack of reporting on participants who did not complete the study (item 10), and failure to conduct familiarization testing (item 16).

Table 3.

Methodological quality of the studies included in the systematic review

Study	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	Score‡
Gultekin et al. [18]	+	+	+	+	+	+	NA	NA	NA	+	+	+	+	-	+	-	84.6%
Almeida et al. [20]	+	+	-	+	+	+	NA	NA	NA	+	+	+	+	-	+	+	84.6%
Andrade et al. [19]	+	+	-	+	+	+	NA	NA	NA	-	+	+	+	-	+	+	76.9%
Jung et al. [22]	+	+	-	-	-	+	NA	NA	NA	-	+	+	+	-	+	+	61.5%
Santo et al. [21]	+	NA	NA	+	-	+	NA	NA	NA	-	-	+	NA	-	+	-	50.0%
Silva et al. [23]	+	NA	NA	-	-	+	NA	NA	NA	-	+	+	NA	-	-	-	40.0%

 + : Yes; -: No; NA: Not applicable; ‡ Percentage of relevant items met

Scale items: 1) A clear description of the inclusion and exclusion criteria was provided? 2) The trials were randomized? 3) The method used to generate the random allocation sequence, including details of any restrictions was described? 4) Sample size was justified? 5) Attempts were made to control and/or monitor pre-trial conditions? 6) Design incorporated measures of important baseline variables? 7) There was blinding of all subjects? 8) There was blinding of all investigators involved in the trials? 9) Both the method of blinding and the evaluation of the successfulness of blinding were described? 10) Details were provided regarding the inability of a subject to complete study requirements? 11) Statistical methods used to compare groups for primary outcome measure(s), and methods for additional analyses, such as subgroup analyses and adjusted analyses, were described? 12) Both point measures and measures of variability for the primary outcome measure(s) were provided? 13) The results of between-group statistical comparisons were reported for the primary outcome measure(s), and its precision? 14) The method used to assess adverse effects was described? 15) Reproducibility of the primary outcome measure(s) was reported? 16) If a performance test was used, a familiarization trial was conducted?

Importantly, none of the studies assessed adverse effects (item 14), representing a critical gap in safety monitoring. On the other hand, measures of reproducibility were unclear in only one study [23]. As expected for single-group designs, items related to randomization (items 2 and 3) and between-group comparisons (item 13) were marked as not applicable. Likewise, in all studies, blinding of participants, researchers, and outcome assessors (items 7, 8, and 9) was not possible given the characteristics of the study designs.

Meta-analysis and certainty of evidence

Table 4 summarizes the assessment of the certainty of evidence, whereas Figs. 2 and 3 display the meta-analytic estimates for each outcome. Across all analyses, the certainty of evidence ranged from very low to low. Statistical heterogeneity was assessed using Cochran’s Q test (p-values reported) and quantified using the I² statistic.

Table 4.

Analysis of the certainty of evidence via the GRADE system

Certainty assessment					№ of participants	Mean (95% CI)	Certainty of evidence
№ of studies	Risk of bias	Inconsistency	Indirectness	Imprecision
Metabolic equivalents (METs)
6	seriousa	seriousb	not serious	seriousc	237	3.7 (3.1–4.3)	⨁◯◯◯
							Very lowa,b,c
Energy cost (kcal·min⁻1)
6	seriousa	seriousb	not serious	seriousc	237	3.8 (3.1–4.5)	⨁◯◯◯
							Very lowa,b,c
Oxygen volume – V̇O2 (ml·kg⁻1·min⁻1)
4	not serious	seriousb	not serious	seriousc	217	10.3 (8.1–12.6)	⨁⨁◯◯
							Lowb,c
Heart rate (bpm)
6	seriousa	seriousb	not serious	seriousc	237	108.6 (104.1–113.0)	⨁◯◯◯
							Very lowa,b,c
Rating of perceived exertion – RPE (Borg 6–20)
2	very seriousd	not serious	not serious	seriousc	20	10.6 (9.3–12.0)	⨁◯◯◯
							Very lowc,d

CI: confidence interval. Explanations: a. One-third of the studies presented a high risk of bias; b. There was high heterogeneity among the studies; c. The sample size was below the optimal information size; d. All studies presented a high risk of bias

Fig. 2.

Meta-analysis of the weighted average among studies that evaluated the intensity and energy cost of Pilates exercises: a metabolic equivalents (METs); b energy cost (kcal·min⁻¹); c oxygen volume – V̇O₂ (ml·kg⁻¹·min⁻.¹); d heart rate (bmp); e) rating of perceived excretion – RPE (Borg 6–20)

Fig. 3.

Sensitivity analysis excluding studies with a high risk of bias and that did not measure energy cost and metabolic intensity via indirect calorimetry: a metabolic equivalents (METs); b energy cost (kcal·min⁻.¹); c heart rate (bpm)

The primary meta-analysis indicated that Pilates exercises elicited a weighted mean of 3.7 METs (95% CI: 3.1–4.3; n = 237, studies = 6; I² = 98%, p < 0.001) and an energy cost of 3.8 kcal·min⁻¹ (95% CI: 3.1–4.5; n = 237, studies = 6; I² = 98%, p < 0.001). However, sensitivity analyses excluding studies with a high risk of bias and those that did not use indirect calorimetry yielded lower and more conservative estimates, with pooled values of 3.0 METs (95% CI: 2.3–3.6; n = 217, studies = 4; I² = 98%, p < 0.001) and 2.9 kcal·min⁻¹ (95% CI: 2.3–3.6; n = 217, studies = 4; I² = 98%, p < 0.001), respectively. No statistically significant differences were detected between studies with low versus high risk of bias for METs (p = 0.184) or energy cost (p = 0.054).

For oxygen consumption, the pooled V̇O₂ was 10.3 ml·kg⁻¹·min⁻¹ (95% CI: 8.1–12.6; n = 217, studies = 4; I² = 99%, p < 0.001). Sensitivity analyses were not conducted for this outcome, as all included studies presented similar risk-of-bias profiles.

The pooled heart rate response was 108.6 bpm (95% CI: 104.1–113.0; n = 237, studies = 6; I² = 85%, p < 0.001). When studies with a high risk of bias were excluded, a slight reduction was observed (107.2 bpm; 95% CI: 102.2–112.3; n = 217, studies = 4; I² = 89%, p < 0.001). No statistically significant differences were detected between studies with low versus high risk of bias (p = 0.178).

The pooled rating of perceived exertion was 10.6 on the Borg 6–20 scale (95% CI: 9.3–12.0; n = 20, studies = 2; I² = 40%, p = 0.198). Sensitivity analyses were not performed for RPE, as all studies contributing to this outcome shared the same risk-of-bias classification.

The leave-one-out sensitivity analysis revealed considerable instability in the pooled estimates across all outcomes (Table 5). When each study was sequentially removed, MET values ranged from 3.1 to 4.1 METs, energy cost estimates ranged from 3.2 to 4.1 kcal·min⁻¹, and V̇O₂ estimates ranged from 9.5 to 11.2 ml·kg⁻¹·min⁻¹. The removal of Silva et al. [23] produced the most substantial reduction in the pooled MET estimate (3.1 [2.5–3.7] vs. 3.7 [3.1–4.3]) and energy cost (3.2 [2.6–3.8] vs. 3.8 [3.1–4.5]). Taken together, these fluctuations indicate that the pooled effects are highly sensitive to the inclusion or exclusion of individual datasets, underscoring the fragility of the current evidence base. Furthermore, heterogeneity remained consistently high across all outcomes, except for the analysis based on the two studies assessing RPE.

Table 5.

Leave-one-out sensitivity analysis

Removed study	METs	Energy cost (kcal·min⁻1)	V̇O2 (ml·kg⁻1·min⁻1)	Heart rate (bpm)	RPE (Borg 6–20)
None (all studies)	3.7 (3.1–4.3)	3.8 (3.1–4.5)	10.3 (8.1–12.6)	108.6 (104.1–113.0)	10.6 (9.3–12.0)
	I2 = 98%, p < 0.001	I2 = 98%, p < 0.001	I2 = 99%, p < 0.001	I2 = 85%, p < 0.001	I2 = 40%, p = 0.198
Gultekin et al. [18]	3.8 (3.0–4.6)	3.7 (3.0–4.5)	9.8 (7.5–12.0)	107.6 (102.6–112.5)	–
	I2 = 98%, p < 0.001	I2 = 97%, p < 0.001	I2 = 98%, p < 0.001	I2 = 80%, p < 0.001
Almeida et al. [20]	4.1 (3.2–4.9)	4.1 (3.3–5.0)	10.8 (7.5–14.1)	109.1 (102.5–115.6)	–
	I2 = 98%, p < 0.001	I2 = 98%, p < 0.001	I2 = 99%, p < 0.001	I2 = 88%, p < 0.001
Andrade et al. [19]	4.0 (3.3–4.7)	4.1 (3.3–4.9)	11.2 (8.8–13.7)	110.3 (107.3–113.3)	–
	I2 = 97%, p < 0.001	I2 = 97%, p < 0.001	I2 = 98%, p < 0.001	I2 = 52%, p = 0.079
Jung et al. [22]	3.7 (3.0–4.4)	3.9 (3.1–4.8)	9.5 (7.0–12.0)	108.7 (103.5–113.9)	–
	I2 = 98%, p < 0.001	I2 = 98%, p < 0.001	I2 = 99%, p < 0.001	I2 = 88%, p < 0.001
Santo et al. [21]	3.7 (3.0–4.3)	3.6 (2.9–4.3)	–	107.5 (102.9–112.1)	11.3 (9.6–13.0)
	I2 = 98%, p < 0.001	I2 = 98%, p < 0.001		I2 = 86%, p < 0.001	not applicable
Silva et al. [23]	3.1 (2.5–3.7)	3.2 (2.6–3.8)	–	108.5 (103.6–113.3)	9.5 (7.3–11.7)
	I2 = 98%, p < 0.001	I2 = 97%, p < 0.001		I2 = 88%, p < 0.001	not applicable

Data expressed as means, 95% confidence intervals, followed by heterogeneity expressed as I² (%) and p-values from Cochran’s Q test; MET: metabolic equivalent; RPE: rating of perceived exertion; V̇O2: oxygen volume

Attempted subgroup analyses were precluded by the severely limited and unbalanced number of studies in each category. No reliable conclusions can be drawn from them, and they are presented in Supplementary Table 3 for transparency only. Initially, the rest interval between sets and exercises, when ≤ 60 s, appears to be a moderator, although this finding should be interpreted cautiously given the fragility of the subgroup analyses.

Discussion

Summary of main findings and key limitations

This exploratory systematic review and meta-analysis sought to identify the metabolic intensity and energy cost of Pilates exercises, as well as the influence of potential moderators. It should be clear to the reader that, due to the heterogeneity of protocols in the studies included in this meta-analysis, the pooled mean is a statistical artifact that may not represent an actual Pilates session. Therefore, the main value of this meta-analysis is to highlight the inadequacy of the current evidence base, and not to provide a definitive estimate.

The high heterogeneity observed across outcomes is largely attributable to substantial variability in Pilates protocols among the included studies. Marked differences were observed in the number and type of exercises performed, session duration, rest intervals between sets and exercises, use of mat-based versus apparatus-based Pilates, and methods used to assess energy cost. Pooling results across such heterogeneous protocols may therefore produce summary estimates that are difficult to interpret and potentially misleading if viewed as representative of a “typical” Pilates session.

For some outcomes, the limited number of available studies combined with pronounced methodological diversity suggests that a narrative synthesis may be equally, or in some cases more, appropriate. Nevertheless, we retained a meta-analytic approach to provide an exploratory quantitative summary of the available evidence, while explicitly acknowledging the instability of the pooled estimates and the need for cautious interpretation.

Sensitivity analyses and robustness of pooled estimates

In the primary analysis including all eligible studies, the mean metabolic intensity was 3.7 METs. However, the sensitivity analysis — excluding studies with low methodological quality and those that did not assess metabolic responses via indirect calorimetry — yielded a lower mean of 3.0 METs. This estimate is likely more representative, as the pooled result of the primary analysis was strongly influenced by a single outlier (9.2 METs) [23], reflecting an atypically high value for this modality. The extreme value presented by the study by Silva et al. [23] possibly occurred due to two factors: (a) the observed energy cost was derived from a predictive formula based on heart rate, which may have led to an overestimated value; and (b) the sample size of only 10 participants may also have been a limiting factor.

Although sensitivity analysis initially resolves this issue, it should be clear that the estimate still derives from only four studies, which are highly heterogeneous in terms of protocol (session duration, exercises, and participant characteristics). Therefore, while 3.0 METs is methodologically more reliable due to the use of indirect calorimetry, it is not necessarily a clinically generalizable value, being only the best estimate available at this time, but not a robust and prescriptive reference parameter.

The leave-one-out analysis also highlighted the fragility of the pooled estimates. MET values varied between 3.1 and 4.1 METs, and energy cost between 3.2 and 4.1 kcal·min⁻¹, depending on the study removed. Silva et al. [23] had the strongest influence, markedly reducing MET and energy cost estimates when excluded. Excluding any of the remaining studies produced smaller but still noticeable fluctuations in the pooled effects. These fluctuations confirm that the conclusions are highly dependent on individual datasets, which reinforces the exploratory nature of the present findings.

Comparison with previous literature

Regarding energy cost, our initial analysis revealed an average value of 3.8 kcal·min⁻¹, which decreased to 2.9 kcal·min⁻¹ after the sensitivity analysis. For descriptive purposes only, a typical 60-min Pilates session would correspond to an estimated cost of approximately 174 kcal. The mean V̇O₂ was 10.3 ml·kg⁻¹·min⁻¹, the mean heart rate was 108.6 bpm, and the mean RPE was 10.6 on the Borg 6–20 scale. All these outcomes exhibited substantial variability according to the rest interval duration between sets and exercises (> vs. ≤ 60 s). Subgroup analyses also suggested that rest interval duration may act as a relevant moderator: protocols with intervals longer than 60 s were associated with approximately 2.0 METs, whereas those with ≤ 60 s reached about 3.4 METs. Importantly, the certainty of evidence for all estimates was low to very low, indicating that future studies employing more rigorous and standardized methodologies are likely to modify these results. Consequently, these results, derived from only six studies, should be interpreted with caution and regarded as exploratory.

According to the 2024 Compendium of Physical Activities, Pilates typically ranges from 1.8 METs (traditional/mat Pilates) to 2.8 METs (general Pilates). In the present review, however, only two of the six included studies ([19, 20], both by the same authors) reported individual results below 3.0 METs, which can be largely explained by differences in rest intervals between sets and exercises. Specifically, across five intervention groups in these studies: (a) three groups with longer intervals (120 s) achieved intensities between 1.8 and 2.5 METs; and (b) two groups with shorter intervals (30 s) reached intensities between 2.8 and 3.1 METs. These findings suggest that recovery duration may be a key moderator of metabolic intensity in Pilates sessions.

Indeed, the recovery interval between sets and exercises appears to exert a substantial influence on the observed outcomes. In the present analysis — after excluding studies with a high risk of bias — a significant difference was found when comparing protocols with rest intervals longer than 60 s versus those with rest intervals of 60 s or less, with mean values of 2.0 and 3.4 METs, respectively (p < 0.001). A similar pattern has been reported in a previous systematic review investigating factors that affect the energy cost of conventional resistance exercise, where the energy cost ranged from 4.9 to 5.6 kcal·min⁻¹ for intervals > 60 s and from 6.2 to 7.3 kcal·min⁻¹ for intervals ≤ 60 s [33]. These consistent findings reinforce the potential role of rest interval duration as a key determinant of metabolic intensity in exercise settings.

Although Pilates is characterized primarily as a form of localized muscular resistance exercise, its execution according to six fundamental principles results in movements that are typically slower and more controlled than those performed in conventional resistance training. This slower tempo likely contributes to the lower energy cost observed in Pilates. Indeed, previous research has shown that the speed of movement in resistance exercises has a significant impact on energy cost [34]. Moreover, Pilates is often classified as a mind–body exercise, as it emphasizes breathing control, concentration, precision, and fluidity of movement—features also shared by other modalities such as Tai Chi Chuan and Yoga.

For comparison, a systematic review on Yoga reported an average metabolic intensity of 3.3 METs, which decreased to 2.9 METs after excluding an outlier. The intensity can be as low as 1.3 METs during pranayamas (breathing control exercises) [8]. Consistently, the 2024 Compendium of Physical Activities lists eight categories of Yoga, with intensities ranging from 2.0 METs for breathing control to 8.0 METs for high-intensity practice [16]. By analogy, Pilates may also encompass a wide range of intensities depending on the training protocol, level of effort, and population studied. This possibility underscores the need for further detailed investigations to establish specific MET classifications for different Pilates modalities.

Methodological considerations

Importantly, in the present review, the estimates of METs and energy cost (kcal) were derived primarily from V̇O₂ measurements, which averaged 10.5 ml·kg⁻¹·min⁻¹ across studies [18–20, 22]. Indirect calorimetry enables the estimation of energy cost and metabolic intensity through direct analysis of inspired and expired gases, offering greater precision than predictive equations based on heart rate. However, this method may underestimate the energy cost in activities with an intermittent nature, such as Pilates, which typically includes rest intervals between sets and exercises. Indirect calorimetry provides more accurate estimates during continuous exercise (e.g., aerobic activities), where a steady state can be achieved and energy production relies predominantly on aerobic metabolism. In contrast, during localized resistance exercises, the steady state is not reached, and anaerobic contributions become more prominent, which may lead to an underestimation of the total energy cost [27].

One potential approach to improve the estimation of energy cost during Pilates exercise is the incorporation of methods capable of capturing the glycolytic contribution, such as blood lactate measurements. An energy equivalent of 3 ml O₂·kg⁻¹ per mmol·L⁻¹ of lactate has been established for increases in lactate concentration [35]. In the present review, only one study assessed blood lactate before and after the intervention, reporting increases from 0.62 mmol·L⁻¹ at baseline in both groups to 1.68 mmol·L⁻¹ following mat Pilates and 1.53 mmol·L⁻¹ following reformer-based Pilates [19]. However, given the intermittent and resistance-based nature of Pilates exercise, isolated pre- and post-session lactate measurements may fail to capture transient peaks in anaerobic metabolism. Ideally, lactate should be assessed after individual exercise bouts or sets to provide a more accurate representation of anaerobic energy contribution during such protocols [27].

The studies included in this review also commonly assessed heart rate and rating of perceived exertion (RPE) during Pilates sessions, as these measures are widely used in professional and clinical practice due to their low cost and ease of application. Nevertheless, the present findings indicate that heart rate may not validly reflect metabolic intensity in this context. Specifically, the mean heart rate of 108.6 bpm observed in our meta-analysis corresponds to just over 50% of the estimated maximum heart rate for young adults (mean age range: 19–26 years), a value that appears inconsistent with the metabolic intensity suggested by MET-based estimates.

Similarly, the pooled mean RPE of 10.6 on the Borg 6–20 scale corresponds to a perception of light intensity (RPE 9–11) [17], further suggesting an underestimation of the metabolic demand observed in the meta-analysis. Collectively, these findings indicate that both heart rate and RPE may provide only approximate indicators of exercise intensity during Pilates and should be interpreted with caution.

From a physiological perspective, this limitation is not unexpected. Pilates shares key characteristics with localized muscular resistance exercise, in which rapid fluctuations in workload, intermittent contractions, and limited active muscle mass often result in attenuated cardiovascular responses. Under such conditions, heart rate may not reach a steady state and therefore may not track oxygen consumption or energy expenditure accurately, in contrast to continuous aerobic exercise [36].

Likewise, the Borg RPE 6–20 scale, although extensively validated for aerobic exercise, may be suboptimal for assessing perceived effort during resistance-based activities, as it may not adequately capture localized muscular fatigue and effort [37]. In contrast, it is possible that alternative scales specifically developed for resistance exercise, such as the Borg CR-10 [38] and the OMNI-RES scale [39], may offer a more appropriate and sensitive assessment of perceived exertion during Pilates protocols.

It is also important to consider the findings from other subgroup analyses (available in Supplementary Table 4), which evaluated the effects of practitioners’ prior experience with Pilates and the type of Pilates performed (mat versus equipment-based). Regarding previous experience, our initial hypothesis was that experienced practitioners might incur greater energy cost, as greater familiarity with the technique could allow for increased workload and, consequently, greater intensity during the session. This expectation is consistent with conventional resistance training, in which the energy cost generally increases proportionally with exercise intensity [27, 33].

Our hypothesis was partially supported, as differences were observed for METs and V̇O₂, but not for energy cost, heart rate, or RPE. The limited number of significant findings may be explained by the small sample size of experienced practitioners compared with inexperienced participants. Consequently, these results should be interpreted with caution, and future studies including larger numbers of experienced participants are needed to better elucidate the impact of prior Pilates experience on these variables.

A similar pattern was observed for comparisons between mat-based and equipment-based Pilates, as relatively few participants performed exercises on equipment, rendering the results preliminary rather than definitive. One study included in this review reported that practicing Pilates on the Reformer generated significantly greater metabolic intensity and energy cost compared with mat Pilates [19]. This difference may be explained by the greater adjustability of effort offered by equipment, such as springs of varying resistance, which could increase the overall energy cost of the session. Future research should further investigate these differences to determine how exercise modality and equipment use influence metabolic responses in Pilates.

Future studies should also consider factors that have been underexplored or not addressed in the studies included in this review, including: (a) a detailed description of the exercises performed, as movements involving a greater amount of muscle mass are likely to incur higher energy cost [27]; (b) enhanced control of exercise intensity, since exercises performed near muscle failure may lead to greater energy cost [27]; (c) assessment of postexercise oxygen consumption (EPOC), as the recovery phase in predominantly anaerobic activities, can make a substantial contribution to total energy cost [33]; (d) rigorous control of fasting, diet, sleep, and physical activity during the experimental period; (e) inclusion or investigation of other populations, such as older adults and athletes, for whom the energy cost may differ substantially; and (f) a direct comparison of different Pilates styles, particularly mat versus equipment-based Pilates, is warranted.

Implications for clinical practice

Based on limited data from heterogeneous studies, the metabolic intensity of Pilates exercises appears to fall within the light-to-moderate range. However, practitioners should be aware that this estimate is highly dependent on the specific training protocol—particularly the duration of rest intervals—and therefore should not rely on a single MET value when prescribing Pilates exercise. The current evidence remains insufficient to support generalizable reference parameters for clinical or practical use, and existing classifications such as those provided by the Compendium of Physical Activities (1,8–2,8 METs) may therefore remain appropriate for broad intensity categorization at this stage. Although rest interval duration emerged as a potential moderator, this observation derives from an exploratory subgroup analysis with an imbalanced comparison (five vs. two studies) and must be interpreted cautiously until confirmed by future research. Moreover, the available data were derived from healthy young adults and cannot be generalized to other populations, such as older adults or individuals undergoing rehabilitation.

Recommendations for future research

Future research in this area would benefit from improved methodological standardization and adequate statistical power. Although this meta-analysis was not designed to support formal prospective sample size calculations, the substantial imprecision, wide confidence intervals, and instability observed across pooled estimates indicate that future studies should recruit larger samples than those commonly used to date. Based on the observed variability and the marked sensitivity of the results to individual studies, samples of at least 20–30 participants per group are recommended as a pragmatic minimum for both exploratory and confirmatory analyses in this field.

In addition, the adoption of more standardized Pilates research protocols is recommended, including clear reporting of exercise selection, session duration that reflects clinical practice (typically ~60 min), and the inclusion of participants with at least minimal prior experience in Pilates (e.g., 3–6 months of practice). It is also important to compare rest interval durations (e.g., 60 s vs. 120 s) that are realistic for clinical settings, as some studies implemented no rest intervals—an approach that is uncommon and may not be feasible for real-world exercise prescription. Another relevant comparison involves mat-based versus equipment-based Pilates, to better elucidate differences between the two primary Pilates modalities. Such standardization would facilitate comparisons across studies and help reduce unexplained heterogeneity.

From a measurement perspective, future studies should incorporate test–retest reliability assessments of metabolic and perceptual measures, include post-exercise oxygen consumption (EPOC) to better capture the total energy cost of intermittent exercise, and consider repeated blood lactate measurements to quantify glycolytic contributions.

Finally, as the available evidence is derived exclusively from healthy young adults, there is a clear need for studies involving older adults and clinical populations, such as individuals undergoing rehabilitation, to improve the external validity and clinical relevance of future findings.

Certainty of evidence

The certainty of evidence was evaluated using the GRADE system and ranged from very low to low across all analyses. The main limitations were related to the risk of bias, inconsistency, and imprecision. Regarding the risk of bias, which was assessed via the Rosendal scale, two studies scored below 60% and were considered high risk. A critical issue across studies was the lack of reporting of potential adverse events, an important aspect in intervention research. Additional concerns included insufficient description of the randomization process in crossover studies; inadequate control of preintervention conditions such as diet, sleep, and physical activity; limited reporting of participants’ inability to complete study requirements; and lack of familiarization prior to the experimental protocol. Future studies should prioritize addressing these methodological limitations to improve the certainty of evidence in this field.

With respect to inconsistency, our meta-analyses showed high heterogeneity for most outcomes, with the exception of RPE, which led most analyses to be conducted using random-effects models. This approach inherently increases variability in the results owing to differences in observed effects across studies. Imprecision was also a key factor in lowering the certainty of evidence, as none of the analyses reached the optimal information size recommended by the GRADE system (n ≥ 400) [32]. This limitation largely reflects the small sample sizes of the included studies, which ranged from 10 to 33 participants per group. Future research should therefore aim for larger sample sizes to improve precision and reduce uncertainty in the estimates.

Possible biases in the review process

It is important to acknowledge potential biases in the review process. The literature search was not conducted across all available databases, which may have led to the omission of some eligible studies. Nevertheless, major databases relevant to the outcomes of interest were thoroughly searched. In addition, two clinical trial registries were consulted to identify unpublished studies, and the reference lists of all included studies were manually checked. These steps likely reduced the probability of missing relevant studies, although the possibility cannot be completely excluded.

Regarding the results, it should be noted that some outcome measures were estimated based on available data from the original publications, rather than being directly reported. These estimations may have led to slight alterations in the means and standard deviations for certain variables in the present study.

Notably, our main analysis included two studies with a high risk of bias that estimated energy cost using heart rate, a method that may increase imprecision [21, 23]. To address this limitation, we conducted a sensitivity analysis excluding these studies, following recommendations from the Cochrane Collaboration [25]. As a result, the estimated metabolic intensity decreased from 3.7 to 3.0 METs, which is likely a more accurate reflection of the true value, although uncertainty remains. Nevertheless, this 3.0 MET estimate is characterized by wide confidence intervals (2.3–3.6), spanning both light and moderate intensity classifications, and should therefore be interpreted with caution. Additionally, it should be noted that indirect calorimetry may underestimate total energy expenditure in intermittent exercise modalities, such as Pilates, which represents an important methodological limitation.

Another limitation worth highlighting is the use of heart rate and RPE as indicators of Pilates exercise intensity. Although these measures are widely applied in exercise science, heart rate is generally more appropriate for steady-state aerobic exercise and may not accurately reflect metabolic demands during resistance-based or intermittent modalities such as Pilates. Likewise, the 20-point RPE scales used in the studies included in this review (Borg CR-20) are typically better suited for aerobic exercise, whereas 10-point scales such as the OMNI-RES and Borg CR-10 are considered more appropriate for resistance exercise. This mismatch may have reduced the accuracy of perceived effort and heart rate–based estimates of metabolic intensity in the context of Pilates. Because primary studies have commonly used these markers, our objective in including them was precisely to discuss the potential limitations associated with their use.

Finally, it should be emphasized that this meta-analysis is inherently fragile due to the small number of included studies (n = 6), which limits the stability and robustness of the pooled estimates. The leave-one-out analysis identified the study by Silva et al. [23] as a clear outlier, exerting a disproportionate influence on the overall results, mainly due to methodological differences and the absence of indirect calorimetry. The analysis further showed that removing this single study substantially reduced the pooled estimates of metabolic intensity (from 3.7 to 3.1 METs), indicating that the conclusions are highly dependent on the inclusion or exclusion of specific datasets. Given this instability, the findings should be interpreted with caution, and more homogeneous, well-controlled primary studies are needed to generate more robust and reliable estimates.

Conclusion

Within the limits of the current and highly heterogeneous evidence base, the findings of this exploratory analysis suggest that Pilates may elicit light-to-moderate metabolic intensity. However, the certainty of the evidence is very low, and the estimates appear to be strongly influenced by protocol characteristics, particularly the duration of rest intervals. The more conservative sensitivity analysis (≈3.0 METs) likely provides the most reliable estimate at this stage. Therefore, these results should be interpreted with caution and confirmed through future studies employing rigorous, standardized, and adequately controlled methodologies.

Abbreviations

BMI

Body mass index

MET

Metabolic equivalent

RPE

Rating of perceived exertion

V̇O₂

Oxygen volume

Authors’ contributions

Gleice Beatriz Batista Vitor: Data curation; Investigation; Visualization; Roles/Writing—original draft. Laís Campos de Oliveira: Conceptualization; Methodology; Supervision; Validation; and Writing—review & editing. Rafaela Almeida Gonçalves Pessôa: Data curation; Investigation; Visualization; and Roles/Writing—original draft. Andreo Fernando Aguiar: Conceptualization; Formal analysis; Project administration; and Writing—review & editing. Raphael Gonçalves de Oliveira: Conceptualization; Formal analysis; Methodology; Project administration; Supervision; and Writing—review & editing.

Funding

Not applicable.

Data availability

The dataset(s) supporting the conclusions of this article is(are) included within the article (and its additional file(s)).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.