Acute Effects of Accentuated Eccentric Loading on Mean Concentric Velocity During Resistance Training: A Systematic Review and Meta-analysis Authors: Jiahao Yang, Tsuyoshi Nagatani, Paul Comfort, Kristina L Kendall, G Gregory Haff Published: 2026 ABSTRACT AbstractBackgroundAccentuated eccentric loading (AEL) may acutely enhance the concentric (CON) barbell velocity after the additional eccentric (ECC) load is removed, however there are inconsistent results regarding the effectiveness of this practice within the scientific literature.ObjectiveThe aims of the present study were to: (1) examine the acute effects of applying AEL, only on the first repetition of a set, on mean barbell velocity during the CON phase of the first 3 repetitions of the set and (2) investigate the influence of the magnitude of ECC load, loading differential and exercise selection on the acute effects of AEL.MethodsFour databases (PubMed, Web of Science, SPORTDiscus, and Google Scholar) were searched for peer-reviewed literature. Eligible studies included those that applied a greater external load during the ECC phase than during the CON phase in free-weight resistance training in an injury-free, strength-trained population. Risk of bias was assessed using the Cochrane risk of bias tool for randomized trials (RoB 2) adapted for crossover trials. The outcome of interest was the mean barbell CON velocity across the first three repetitions, comparing the AEL condition (AEL applied on the first repetition only) to the traditional (TRAD) loading condition. Correlated and hierarchical effects models, with robust variance estimation and small-sample adjustment, were fitted. Moderator analyses were used to examine whether ECC load (i.e., ECC load = vs. > 100% TRAD one-repetition maximum [1RM]), loading differential (≥ vs. < 40% TRAD 1RM), and exercise selection (back squat vs. bench press) influenced the acute effects of AEL across repetitions. Certainty of evidence was evaluated using the GRADE approach.ResultsThe final analysis included 8 studies (106 participants, male: n = 90, female: n = 16). AEL significantly impaired mean barbell velocity during the subsequent CON phase immediately after its removal (effect size [ES] = -0.25, 95% confidence intervals [CI]: [-0.33, -0.16], p < 0.001). The mean CON velocity impairment diminished by the second (ES = 0.01, 95% CI: [-0.10, 0.12], p = 0.822) and third (ES = 0.08, 95% CI: [-0.08, 0.23], p = 0.313) repetitions of the set. However, the certainty of evidence ranged from moderate to very low. Of all moderators examined, the application of AEL with the back squat resulted in greater velocity reductions during the first repetition only than those seen for the bench press (ES = -0.33 [back squat]; ES = -0.11 [bench press], p = 0.022).ConclusionAEL impaired the mean barbell velocity during the CON phase immediately after its removal compared with TRAD loading. This reduction in mean CON velocity appears greater in the back squat than in the bench press and is likely to dissipate across the subsequent repetitions following the removal of AEL.Protocol Registration The original protocol was prospectively registered (10.17605/OSF.IO/AYUJH) with the Open Science Framework. Introduction Muscular strength is one of the most critical factors contributing to sports performance [1, 2], and resistance training is the primary approach to improve it [3, 4]. During resistance training, strength and conditioning professionals commonly prescribe the identical load for both the eccentric (ECC) and concentric (CON) phases of an exercise based on a predetermined one-repetition maximum [1RM] [5]. However, since the 1RM is fundamentally representative of the individual’s maximal CON strength, the ECC phase of the lift is generally considered to be underloaded [6]. For example, during the performance of free-weight exercises, approximately 19.0 ± 4.5% and 18.7 ± 4.2% more load can be lifted during the ECC phase of the bench press [6, 7] and the back squat [6], respectively, when compared to the CON phase of each lift. To address this issue, several training tactics have been proposed to account for the greater force-generating capacity during the ECC phase of a resistance training exercise [8]. The most common strategy involves increasing the load that is applied during the ECC phase of the lift, which is commonly referred to as accentuated ECC loading (AEL) [9]. The application of this advanced programming tactic [8] is generally accomplished with the use of pneumatic devices [10] or with weight releasers [11, 12]. When this programming tactic is implemented, the additional ECC load is removed prior to the initiation of the CON phase of the lift, allowing the load applied to better match the force generating capacities associated with each phase of the lift [9]. Specifically, when weight releasers are used to apply the additional ECC load, to ensure athletes can release them at the ECC-CON turning point and to restabilize, a distance of 5–10 cm prior to the initiation of the CON phase of the lift [13] is recommended. Recent evidence [14–16] indicates that applying AEL on every repetition, with 3–5 sets of 4–6 reps, using 95–140% traditional (TRAD) 1RM during the ECC phase and 50–75% TRAD 1RM during the CON phase, performed 1–3 sessions per week, over 4–12 weeks, significantly increases the TRAD back squat 1RM. However, the number of intervention studies supporting this is limited, potentially because optimal AEL programming strategies remain undetermined from acute studies. Previous studies [17, 18] demonstrate that there may be an acute performance enhancement during the CON phase immediately after the removal of AEL (i.e., within-repetition effect). However, both acute within-repetition and within-set effects (i.e., the effects of AEL on subsequent repetitions in which TRAD loading [TL] was applied) remain inconsistently reported in the literature [12, 17–21]. For instance, Doan et al. [17] reported that the TRAD 1RM acutely increased by 2.27–6.80 kg (~ 3%) following the application of an ECC load equal to 105% of the TRAD 1RM during the performance of the bench press. Conversely, Ojasto and Häkkinen [20] reported a significant decrease in the heaviest load that could be lifted during the CON phase (TRAD 1RM) following the application of three different supramaximal AEL protocols (ECC loads equivalent to 105%, 110%, and 120% of the TRAD 1RM) during the bench press. Specifically, a significant decrease in performance was observed when an ECC load of 120% of the TRAD 1RM was applied, resulting in reductions in maximal performance from 108.4 ± 3.6 to 104.5 ± 3.3 kg [20]. This reduction in performance capacity is likely a result of an acute fatigue response associated with the increased ECC load, which was 20% above the TRAD 1RM. In addition to the TRAD 1RM, barbell velocities during the CON phase, after applying AEL, are common metrics used to assess the immediate effects of AEL, as they reflect athletes’ capacity to accelerate the barbell when maximal intent is applied. Munger et al. [18] reported a significant increase in peak barbell velocity during the CON phase of the front squat (CON load equivalent to 90% of the TRAD 1RM) following the application of a supramaximal AEL (ECC load equivalent to 105%, 110%, and 120% of the TRAD 1RM). Conversely, when a supramaximal AEL is applied during the ECC phase of the back squat, peak barbell velocity during the CON phase of the lift is not consistently increased [19, 22]. It is also important to note that in some instances, peak barbell velocity during the CON phase of the bench press (CON loads equivalent to 50% and 65% of the TRAD 1RM) has been reported to be impaired after the acute application of a supramaximal AEL (ECC load equivalent to 120% of the TRAD 1RM) [12]. While it appears that the impact of AEL tactics is influenced by the loads applied in an exercise-specific manner, based on the available data, this has yet to be fully explored within the scientific literature. The lack of consensus regarding the acute effects of using AEL during resistance training is likely associated with the methodologies used to prescribe the AEL. Specifically, programming parameters, such as the magnitude of the ECC load, differences between the ECC and CON loads, and exercise selection, may influence the effectiveness of this programming tactic in enhancing subsequent performance. Additionally, how the acute effects associated with the application of the AEL are determined may also impact the performance responses reported within the scientific literature. For instance, previous researchers investigating the acute effects of AEL have primarily used two approaches for determining the performance impact of AEL. The first approach is used to examine how AEL affects the CON phase of the lift immediately after the removal of the AEL (i.e., single repetition within each set [17, 18, 20, 23, 24]). The second approach is to investigate how AEL impacts CON performance across a series of repetitions (the within-set effects), using different loading strategies such as applying AEL on every repetition of the set [11, 21, 25–27] or selectively applying AEL on specific repetition[s] [11, 12, 19, 21, 22, 27–30]). Compounding the impact of these two approaches is the fact that some researchers have only reported the CON performance as the average of all repetitions contained within the set, which may influence the interpretation of the acute effects associated with AEL [11, 27, 28, 30]. It is plausible that this approach may fail to adequately capture the acute effects of AEL on each individual repetition, especially when AEL is applied only on the first repetition of a set. As such, alternative methods for evaluating acute responses to the application of AEL are warranted to provide coaches with more detailed programming information. Given that meta-analytic techniques are powerful statistical tools that are widely used to synthesize results from systematically screened studies with similar study designs, quantify variance, and consider the implications of variances to estimate the true effect sizes of a training method [31, 32], a meta-analysis is warranted to provide summary data to assist practitioners in designing resistance training programs using AEL and to inform further research. Within the AEL literature, there are a wide range of performance metrics (i.e., mean CON velocity [12, 19, 21, 22, 24, 27, 28, 30, 33], peak CON velocity and force [18], TRAD 1RM [17, 20], electromyography activity [25, 26, 34]) that have been examined to determine the effects of this programming tactic. Among these metrics, the mean CON barbell velocity is the most consistently reported outcome and closely reflects the acute CON performance of interest. To minimize heterogeneity and strengthen the precision of the meta-analytic estimates [35], the present meta-analysis therefore focused exclusively on mean CON barbell velocity. The present study aimed to: (1) examine the acute effects of applying AEL on the first repetition of a set on the mean CON barbell velocity across the first 3 repetitions of the set, and (2) investigate the influence of the magnitude of ECC load, loading differential and exercise selection on the acute performance effects resulting from the application of AEL. Methods This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [36]. The protocol was registered with the Open Science Framework (OSF; 10.17605/OSF.IO/AYUJH) prior to the analysis of any data. Search Strategy A comprehensive literature search was conducted using multiple approaches through October 31, 2025. Initially, systematic searches were performed in three major electronic databases: PubMed, Web of Science, and SPORTDiscus. Key search terms were grouped and searched using the search conjunction ‘OR’ in PubMed, Web of Science, and SPORTDiscus databases. Combinations of the following phrases were searched as exact phrases in the titles and abstracts of all major electronic databases: ‘Accentuated Eccentric Load’, ‘Accentuated Eccentric Loading’, ‘Accentuated Eccentric Loads’, ‘AEL’, ‘Accentuated Eccentric Overloading’, ‘Supramaximally Loaded’, ‘Supra-maximal Loading’, ‘Additional Eccentric Loads’, ‘Increased Eccentric Loading’, ‘Eccentric Overload’, ‘Augmented Eccentric Loading’. To ensure comprehensive coverage, we manually examined the reference lists of included articles. Additionally, supplementary searches were conducted using Google Scholar to identify relevant publications. The full reproducible search queries and inclusion filters used for each database are available in the OSF repository (see Supplemental file 1, https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60). Selection Process All titles and abstracts obtained from the database searches were screened independently by two authors (JY and TN) and those potentially meeting the inclusion criteria (Table 1) were included for full text analysis. Duplicates were automatically removed using an online systematic review tool (Covidence, https://www.covidence.org/). Disagreements in eligibility decisions were resolved initially through the discussion between two authors (JY and TN), and the third investigator (GGH) was available to make a final decision if a consensus could not be reached. Table 1Inclusion and exclusion criteria for the present meta-analysisCategoryInclusion criteriaExclusion criteriaPopulationParticipants with strength training experiencesParticipants with injuriesIntervention(a) Application of a greater external load during the eccentric phase of the lift when compared with the concentric phase(a) Using inertial flywheel devices(b) Applied AEL on resistance training exercises(b) Applied AEL on plyometrics and ballistic exercises (e.g., drop jumps, countermovement jumps)(c) Using free weights to apply AEL(c) Using other pathways to apply AEL (e.g., computer-driven device, pneumatic device)ComparatorTraditional loading condition: utilized identical loads in both the eccentric and concentric phases of the lift.Absence of traditional loading condition as the comparatorOutcomeThe mean velocity during the concentric phase of the lift.Other variablesStudy designWithin-participants crossover study designReview articlesAEL accentuated eccentric loading Inclusion and exclusion criteria for the present meta-analysis AEL accentuated eccentric loading Inclusion and Exclusion Criteria Only peer-reviewed studies written in English were eligible for inclusion according to the participants, intervention, comparator, outcome, and study design (PICOS) criteria (Table 1). Since the ECC load during flywheel inertial training depends on the kinetic energy generated in the CON phase and does not always ensure greater mechanical loading in the ECC phase compared to the CON phase [37], related studies were excluded from the present meta-analysis. Data Extraction Two authors (JY and TN) independently read and extracted all studies which met the inclusion criteria. Extracted data from each full-text article included (i) study information (author(s), year of publication); (ii) sample size; (iii) participant characteristics (i.e., age, height, body mass, relative 1RM and absolute 1RM [if applicable]); (iv) exercise applied AEL with; (v) applied ECC load; (vi) applied CON load; (vii) applied AEL strategy (applied AEL on every repetition or only applied AEL on the first repetition of a set); (viii) applied ECC tempo; (ix) applied inter-repetition and inter-set rest; (x) mean and standard deviation (SD) values of the mean barbell velocity of CON phase of individual repetition(s) under the TL condition; and (xi) mean and SD values of the mean barbell velocity of CON phase of individual repetition(s) under AEL condition(s). For data not reported in the articles, the research team initially attempted to directly email the original authors to request the necessary data. In cases where data could not be obtained directly from the original authors [12, 19, 24], two authors (JY and TN) independently extracted data from figures presented in the corresponding articles using WebPlot Digitizer [38]. The means of the values extracted by the two authors were used for further analysis. Risk of Bias, Publication Bias, and Certainty Assessment The risk of bias of each study was assessed using the Cochrane risk of bias tool for randomized trials (RoB 2) adapted version for crossover trials [39]. The risk of bias assessment was independently conducted by two authors (JY and TN). A consensus was reached after discussions between the two authors (JY and TN) or consultation with the third investigator (GGH). Publication bias was planned to be assessed using funnel plot asymmetry and Egger’s test when the total number of included studies exceeded 10 [40, 41]. The certainty of evidence was assessed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) tool [40]. Statistical Analyses In the present study, the standardized mean difference (SMD) of the mean CON velocity of the lift between TL and AEL conditions was selected as the effect size (ES) of interest. Therefore, the within-group SMD ( for each study were calculated. To correct the small sample size bias [42] commonly existing in sports science literature [43], Hedges’ g was subsequently calculated. Given that the data used to calculate is not independent when the mean CON velocity measured under the TL and the mean CON velocity measured under the AEL come from the same group of participants, the correlation between the two mean CON velocity values is used to calculate the standard error (SE) of [35]. The full set of formulas used is available in the OSF repository (see Supplemental file 2, https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60). As mentioned previously, a dependency issue existed in the included primary studies, as each study contained at least two comparisons between different AEL protocols and paired TL within the same group of participants (i.e., ESs and sampling errors are correlated). For instance, 12 ESs are attributed to the investigation conducted by Taber et al. [24], which resulted in these ESs being correlated to one another because they come from the same group of participants. Given that the statistical independence is one of the core assumptions when pooling ESs in a meta-analysis [44], three correlated and hierarchical effects (CHE) models [45] and robust variance estimation (RVE) were employed for the first three repetitions of the set, respectively (where applicable). The variance-covariance matrices used to fit CHE models were constructed using the ‘impute_covariance_matrix’ function. The within-study correlation among ESs was informed using participant-level data where available (Repetition 1 model: n = 5 [19, 23, 24, 28, 30]; Repetition 2 [19, 28, 30] and 3 [19, 28, 30] models: n = 3). The detailed procedures used to estimate the correlations are provided in the R script available in the OSF repository (see Supplemental file 3, https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60). Several sensitivity analyses were subsequently conducted using a range of assumed sampling correlations (r = 0.5, 0.6, 0.7, 0.8) [35]. The “CR2” small-sample adjustments for p-values were calculated in RVE to account for the correlation between ESs and sampling errors. ESs were interpreted where values of < 0.20, 0.20–0.59, 0.60–1.19, and 1.20-2.0 represent the trivial, small, moderate, large, and very large effect, respectively [46]. The (restricted) maximum likelihood procedure was used to calculate the heterogeneity variance of the distribution of true ESs. The I2 statistic was used to quantify the heterogeneity, which represents the percentage of variability in ESs not caused by sampling error [47]. The I2 was calculated as the ratio of true heterogeneity variance (sum of within-study and between-study variance components) to the total variance (the sum of true heterogeneity and the sampling variance), expressed as a percentage. Values of < 25%, 25–50%, 51–75%, and > 75% indicate low, low to moderate, moderate to substantial, and substantial heterogeneity, respectively [48]. The formulas used to calculate I2 in different levels are provided in the OSF repository (see Supplemental file 2, https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60). Given that different AEL protocol characteristics may influence the observed acute effects [12, 19, 21, 25], subgroup analyses were performed. Specifically, for each repetition within the set of lifts, the present meta-analysis incorporated each potential moderator as a covariate in the corresponding CHE model. Studies were grouped into the following categories to conduct subgroup analyses: (1) ECC load (maximal AEL: = 100% TRAD 1RM; and supramaximal AEL: > 100% TRAD 1RM), (2) loading differential between the ECC and CON load (≥ 40% TRAD 1RM; < 40% TRAD 1RM), (3) exercise selection (back squat; bench press). The 40% threshold was chosen based upon the work of Merrigan et al. [19] where it was noted that loading differentials > 40% resulted in greater mean CON velocities. All analyses were performed in R (version 4.2.0), with the alpha level set as 0.05. CHE models were fitted using the ‘rma.mv’ function in the metafor package (Version 4.4-0). RVE was conducted alongside CHE models to obtain more robust p-values using the ‘coef_test’ function, and more robust estimated SE and CI values using the ‘conf_int’ function in the clubSandwich package (Version 0.5.10). Prediction intervals (PI) were calculated from the fitted meta-analytic models using the ‘predict’ function in the metafor package, based on the estimated between-study variance. Results Study Selection A total of 1287 records were identified with the search strategy implemented in the current study. Once duplicate records and articles were excluded after review of titles and/or abstracts, 41 studies were assessed for eligibility (Fig. 1). After reading the full text of each document, 33 studies were excluded due to: not an eligible study design [14, 15, 49–57]; comparator not meeting inclusion criteria [58]; intervention or experimental exposure not meeting inclusion criteria [59–64]; outcomes not relevant to the review questions [17, 18, 20, 26, 34, 65–71]; insufficient data for extraction [11, 27, 33]. Therefore, 8 studies [12, 19, 21–24, 28, 30] were included in the final meta-analyses. Four primary studies provided additional data upon request [22, 23, 28, 30]. Among these, authors from three studies provided participant-level data [23, 28, 30], and authors from one study provided group-level mean and SD values for each condition [22]. PRISMA flow diagram of the study selection process Study Characteristics A summary of included participant and protocol characteristics is presented in Tables 2 and 3, respectively. A total of 106 (90 males and 16 females) participants were included. Inclusion criteria of the included primary studies required participants to be able to bench press at least 1.0 times their body mass or back squat 1.5 times their body mass. Among the included primary studies, four studies [12, 19, 21, 22] reported mean values averaged across multiple sets for each repetition. In contrast, one study [23] reported the mean of two repetitions in which AEL was applied in both repetitions, separated by a 3-minute rest interval. Table 2Participant characteristics of included studiesStudy N Age(yrs)Body Mass(kg)Height(m)1RM(kg)Relative 1RM (1RM/BM)ExerciseExperience with the Chosen Exercise (yrs)Training Frequency (times/week)Merrigan et al. [12]8 (M)2 (F)M: 26 ± 4F: 23 ± 2M: 87.9 ± 11.1 F: 59.1 ± 2.5M: 1.75 ± 0.07F: 1.58 ± 0.03M: 131.0 ± 27.0F: 61.4 ± 9.6M: 1.5 ± 0.2F: 1.0 ± 0.1BP≥ 1≥ 1Merrigan et al. [19]21(M)24.0 ± 4.285.5 ± 11.11.75 ± 0.07169.8 ± 29.02.0 ± 0.3BS≥ 21–3Wagle et al. [21]11 (M)26.1 ± 4.192.5 ± 10.51.84 ± 0.04Not reported1.8 ± 0.3BS≥ 1≥ 1Chae et al. [22]12 (M)25.6 ± 4.481.7 ± 11.41.77 ± 0.06152.4 ± 27.81.9 ± 0.4BS7.8 ± 4.2Not reportedCastro et al. [23]12 (M)25.8 ± 6.186.0 ± 12.51.74 ± 0.09134 ± 33Not reportedBP≥ 1≥ 3Taber et al. [24]10 (M)23 ± 382.3 ± 9.21.76 ± 0.06124.3 ± 19.41.5 ± 0.2BP≥ 1Not reportedSuchomel et al. [28]16 (M)24.4 ± 3.885.8 ± 12.31.78 ± 0.08168 ± 23.72.0 ± 0.3BS≥ 1≥ 3Campbell et al. [30]14 (F)23.6 ± 2.670.3 ± 8.41.67 ± 0.07Not reported1.50 ± 0.19BS≥ 1≥ 3M male, F female, BM body mass, 1RM one repetition maximum, BP bench press, BS back squat Participant characteristics of included studies 8 (M) 2 (F) M: 26 ± 4 F: 23 ± 2 M: 87.9 ± 11.1 F: 59.1 ± 2.5 M: 1.75 ± 0.07 F: 1.58 ± 0.03 M: 131.0 ± 27.0 F: 61.4 ± 9.6 M: 1.5 ± 0.2 F: 1.0 ± 0.1 M male, F female, BM body mass, 1RM one repetition maximum, BP bench press, BS back squat Table 3Protocol characteristics of included comparisons in meta-analysis modelsStudyExerciseTL Scheme(ECC/CON, % of TRAD 1RM)AEL Scheme(ECC/CON, % of TRAD 1RM)Set × RepetitionLoading StrategyRest(Inter-rep / Inter-set)ECC TempoMerrigan et al. [12]Bench Press50/50; 65/65120/50; 120/654 × 5AEL only on the first repetition0/3 minSelf-paceMerrigan et al. [19]Back Squat65/65; 80/80120/65; 120/803 × 5–65/65 and 120/65;3 × 3–80/80 and 120/80AEL only on the first repetition0/3 minNot reportedWagle et al. [21]Back Squat80/80105/803 × 5AEL only on the first repetition0/3 minNot reportedChae et al. [22]Back Squat60/60110/603 × 10AEL on every 5 repetitions (first and sixth of the set)0 for the first 5/135 sSelf-paceCastro et al. [23]Bench Press30/30; 80/80100/30; 100/801 × 2AEL on every repetition3 min/3 minIn a controlled mannerTaber et al. [24]Bench Press30/30; 40/40; 50/50; 60/60; 70/70; 80/80100/30; 100/40; 100/50; 100/60; 100/70; 100/80;110/30; 110/40; 110/50; 110/60; 110/70; 110/801 × 1 for each loading schemeAEL on every repetition0/5 minSelf-paceSuchomel et al. [28]Back Squat50/50; 60/60; 70/70; 80/80100/50; 100/60; 100/70; 100/80;110/50; 110/60; 110/70; 110/801 × 3 for each loading schemeAEL only on the first repetition0/3 minSelf-paceCampbell et al. [30]Back Squat50/50; 60/60; 70/70; 80/80100/50; 100/60; 100/70; 100/80;110/50; 110/60; 110/70; 110/801 × 3 for each loading schemeAEL only on the first repetition0/3 minSelf-paceTL traditional loading, AEL accentuated eccentric loading, ECC eccentric, CON concentric, TRAD 1RM traditional one-repetition maximum Protocol characteristics of included comparisons in meta-analysis models 3 × 5–65/65 and 120/65; 3 × 3–80/80 and 120/80 100/30; 100/40; 100/50; 100/60; 100/70; 100/80; 110/30; 110/40; 110/50; 110/60; 110/70; 110/80 100/50; 100/60; 100/70; 100/80; 110/50; 110/60; 110/70; 110/80 100/50; 100/60; 100/70; 100/80; 110/50; 110/60; 110/70; 110/80 TL traditional loading, AEL accentuated eccentric loading, ECC eccentric, CON concentric, TRAD 1RM traditional one-repetition maximum A total of eight studies [12, 19, 21–24, 28, 30] provided 36 ESs (the SMD of mean CON velocity between TL and AEL conditions) for the repetitions where AEL was applied, which was the first repetition of the set. Five studies (73 participants, male: n = 57, female: n = 16) contributed 21 ESs for the second repetition within the set, where no inter-repetition rest was given following the initial application of AEL. Similarly, six studies (84 participants, male: n = 68, female: n = 16) contributed 22 ESs for the third repetition of the set. The discrepancy in the number of studies where data were reported for the second and third repetitions is due to one study [21] reporting only the mean CON velocity of the first, third, and fifth repetitions. Six studies included at least two pairs of conditions [12, 19, 23, 24, 28, 30]. Two studies [21, 22] included a single pair of conditions (i.e., matched CON loads, with one condition applying AEL and the other TL) but examined various set structures (i.e., straight sets, cluster sets, and rest redistribution). To eliminate the potential confounding effects of differing set configurations, the present analysis included only comparisons involving TRAD straight sets and sets where AEL was only applied to the first repetition of the set [21] or first and fifth repetition of sets of 10 repetitions [22], and analyzed across the first three repetitions when multiple repetitions were performed. Risk of Bias, Publication Bias, and Certainty Assessment The result of risk of bias assessment is presented in Fig. 2. Two studies were rated as “low risk of bias” [22, 24], and 6 studies [12, 19, 21, 23, 28, 30] were rated as “some concerns”. In Domain S (risk of bias arising from period and carryover effects), five studies [12, 19, 21, 28, 30] only reported using a randomized order of different programming tactics but did not indicate whether it was also counterbalanced. This may have caused learning effects, particularly in the AEL condition. For example, a sequence, such as ‘Session 1 (AEL protocol 1) – Session 2 (TRAD protocol 1) – Session 3 (AEL protocol 2) – Session 4 (TRAD protocol 2)’, may result in a greater mean CON velocity than another sequence, like ‘Session 1 (TRAD protocol 1) – Session 2 (TRAD protocol 2) – Session 3 (AEL protocol 1) – Session 4 (AEL protocol 2)’. Such learning effects may be amplified when participants are unfamiliar with weight releasers. Risk of bias assessment for included studies In Domain 4 (measurement of the outcome), seven studies [12, 19, 21, 22, 24, 28, 30] used linear position transducers to measure barbell velocity and directly assessed the 1RM for the selected exercises. One study [23] used an 8-camera 3-dimensional motion analysis system to capture barbell velocity, but estimated the bench press 1RM using the two-point load velocity profile method and required participants to place their feet on the bench during lifts. Although this method of measuring the outcome is inappropriate, and the algorithm of judgement suggested ‘high risk’, all between-tactic comparisons in this study were made within-participant, which mitigated some of the concerns regarding the outcome measurement. Therefore, the judgement was downgrade to ‘some concerns’ based on consensus among three authors (JY, TN, and GGH). In Domain 5 (risk of bias in selection of the reported result), none of the included studies indicated whether the analyses were in accordance with a pre-specified analysis plan finalized before unblinded outcome data. The algorithm of judgment suggested ‘some concern’, however, three authors considered the mean CON velocity was unlikely to be influenced by this. As such, ‘low risk’ was assigned to all studies for this domain. Based on the available scientific literature on this topic [40, 41], publication bias was not assessed because the total number of included studies was less than 10. The overall certainty of evidence was rated as moderate for the first-repetition model, low for the second-repetition model, and very low for the third-repetition model (Table 4). Detailed GRADE assessments are provided in Supplemental File 4, available in the OSF repository (https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60). Table 4GRADE assessment for the certainty of evidenceCertainty assessmentNumber of ParticipantsEffect estimate (Hedges’ g, 95% CI)Overall Certainty of EvidenceNumber of StudiesRisk of BiasInconsistencyIndirectnessImprecisionPublication BiasFirst Repetition8SeriousNot SeriousNot SeriousNot SeriousNot assessed due to insufficient number of studies106-0.25 (-0.33, -0.16)ModerateSecond Repetition5SeriousNot SeriousNot SeriousSerious730.01 (-0.10, 0.12)LowThird Repetition6SeriousSeriousNot SeriousSerious840.08 (-0.08, 0.23)Very LowGRADE Grading of Recommendations Assessment, Development, and Evaluation, CI Confidential intervals GRADE assessment for the certainty of evidence GRADE Grading of Recommendations Assessment, Development, and Evaluation, CI Confidential intervals Sensitivity Analysis Sensitivity analyses were conducted using different assumed constant sampling correlations within each primary study (r = 0.5, 0.6, 0.7, and 0.8) for constructing the variance-covariance matrix in CHE models (Table 5). The results were robust across different correlation coefficients (including the coefficient imputed from available participant-level data), showing consistent direction and magnitude of estimated pooled ESs. Table 5Summary of sensitivity analyses using different assumed constant sampling correlationsES [95% CI]Imputed correlation based on available participant-level dataAssumed constant sampling correlation (r)0.50.60.70.8First Repetition-0.25* [-0.33, -0.16] (r = 0.37)-0.25* [-0.33, -0.16]-0.25* [-0.33, -0.16]-0.25* [-0.33, -0.16]-0.25* [-0.33, -0.16]Second Repetition0.01 [-0.10, 0.12] (r = 0.41)0.01 [-0.10, 0.12]0.01 [-0.10, 0.12]0.01 [-0.10, 0.12]0.01 [-0.10, 0.12]Third Repetition0.08 [-0.08, 0.23] (r = 0.37)0.08 [-0.08, 0.23]0.08 [-0.08, 0.23]0.08 [-0.08, 0.23]0.08 [-0.08, 0.23]* Indicates significant difference (p < 0.05) between accentuated eccentric loading and traditional loading conditions in the mean concentric velocity of the specified repetition. ES effect size, CI confidence intervals Summary of sensitivity analyses using different assumed constant sampling correlations * Indicates significant difference (p < 0.05) between accentuated eccentric loading and traditional loading conditions in the mean concentric velocity of the specified repetition. ES effect size, CI confidence intervals Main Effects First Repetition after Removal of the AEL Eight studies provided a total of 36 ESs (Fig. 3). Based on the pooled ES, the application of AEL resulted in a significant small negative effect on the mean CON velocity after the removal of the AEL compared to TL (p < 0.001, ES = -0.25, 95% CI: [-0.33, -0.16], 95% PI: [-0.34, -0.15]). Residual heterogeneity was not statistically significant (Q [df = 35] = 32.08, p = 0.610), with none of the observed variability attributable to true heterogeneity (I2-total = 0%). Forest plots of the effects of AEL on mean concentric velocity during the first repetition of the set. ECC eccentric, CON concentric, TRAD traditional, 1RM one repetition maximum, SMD standardized mean difference, TL traditional loading, AEL accentuated eccentric loading, BP bench press, BS back squat. → indicates the magnitude of ES is greater than 1, ← indicates the magnitude of ES is lower than − 1 Subgroup analyses indicated that neither the loading differential between the ECC load and CON load (≥ 40% vs. < 40% of TRAD 1RM) nor the magnitude of the ECC load (= 100% of TRAD 1RM vs. > 100% of TRAD 1RM) significantly influence the mean CON velocity after the removal of the AEL (Fig. 4). Specifically, applying AEL with a greater or smaller ECC to CON loading differential did not significantly influence the barbell velocity during the CON phase immediately after the removal of the AEL compared to TL (QM [df = 1] = 0.86, p = 0.353, ES = -0.18 [< 40% TRAD 1RM]; ES = -0.28 [≥ 40% TRAD 1RM]). Likewise, maximal and supramaximal AEL protocols produced small and trivial non-significant negative effects on mean CON velocity after AEL removal compared to TL (QM [df = 1] = 0.01, p = 0.908, ES = -0.24 [= 100% of TRAD 1RM]; ES = -0.25 [> 100% of TRAD 1RM]). In contrast, exercise selection significantly affected the velocity decrement during the CON phase immediately following AEL removal, relative to TL (QM [df = 1] = 5.23, p = 0.022, ES = -0.33 [back squat]; ES = -0.11 [bench press]). Subgroup analysis of magnitude of ECC load, loading differential between the ECC and CON load, and exercise selection during specific repetitions. * Indicates significant difference (p < 0.05) between two subgroups during the specified repetition. ● represents each individual effect size. The ridgeline illustrates the distribution of individual effect sizes. Note: Bench press data from the second and third repetitions were omitted from individual plots due to the small number of effect sizes (n = 2), to ensure the robustness of the visual density. ES pooled effect size, TRAD traditional, 1RM one repetition maximum Second Repetition Five studies contributed a total of 21 ESs (Fig. 5). Based on the pooled results, the application of AEL resulted in a comparable mean CON velocity to TL during the second repetition of the set (p = 0.822, ES = 0.01, 95% CI: [-0.10, 0.12], 95% PI: [-0.26, 0.29]). Residual heterogeneity was not statistically significant (Q [df = 20] = 28.76, p = 0.093). A low amount of total variance was attributed to true heterogeneity (I2-total = 27.02%), with 13.51% attributed to within-study heterogeneity (I2-level 2) and 13.51% attributed to between-study heterogeneity (I2-level 3). Forest plots of the effects of AEL on mean concentric velocity during the second repetition of the set. ECC eccentric, CON concentric, TRAD traditional, 1RM one repetition maximum, SMD standardized mean difference, TL traditional loading, AEL accentuated eccentric loading, BP bench press, BS back squat None of the selected moderators significantly influenced the mean CON velocity of the second repetition (Fig. 4). While the back squat subgroup presented a higher pooled estimate than the bench press subgroup, suggesting a greater mean CON velocity when AEL was applied with the back squat relative to TL, this difference did not reach statistical significance (QM [df = 1] = 3.78, p = 0.052, ES = 0.04 [back squat]; ES = -0.35 [bench press]). Similarly, different loading differentials between the ECC and CON load did not significantly affect the mean CON velocity of the second repetition of the set (QM [df = 1] = 0.33, p = 0.565, ES = -0.03 [< 40% TRAD 1RM]; ES = 0.03 [≥ 40% TRAD 1RM]). Likewise, the magnitude of ECC load also did not significantly influence the mean CON velocity of the second repetition of the set (QM [df = 1] = 0.21, p = 0.651, ES = -0.02 [= 100% of TRAD 1RM]; ES = 0.03 [> 100% of TRAD 1RM]). Third Repetition Six studies provided a total of 22 ESs (Fig. 6). Based on the pooled results, the application of an AEL resulted in a trivial positive effect on the mean CON velocity compared to TL during the third repetition of the set (p = 0.313, ES = 0.08, 95% CI: [-0.08, 0.23], 95% PI: [-0.48, 0.63]). Residual heterogeneity was statistically significant (Q [df = 21] = 53.69, p < 0.001), and a moderate to substantial amount of total variance was attributed to true heterogeneity (I2-total = 59.93%), with 29.97% attributed to within-study heterogeneity (I2-level 2) and 29.97% attributed to between-study heterogeneity (I2-level 3). Forest plots of the effects of AEL on mean CON velocity during the third repetition of the set. ECC eccentric, CON concentric, TRAD traditional, 1RM one repetition maximum, SMD standardized mean difference, TL traditional loading, AEL accentuated eccentric loading, BP bench press, BS back squat. → indicates the magnitude of ES is above 1 Subgroup analyses indicated that none of the selected moderators significantly influenced the mean velocity of the third repetition of the set (Fig. 4). Although there were no significant differences between the two subgroups, the application of AEL with a loading differential ≥ 40% of the TRAD 1RM induced a more favorable effect on the mean CON velocity during the third repetition than a loading differential < 40% of the TRAD 1RM (QM [df = 1] = 1.42, p = 0.234, ES = -0.05 [< 40% of TRAD 1RM]; ES = 0.13 [≥ 40% of TRAD 1RM]). While the magnitude of ECC load did not significantly influence the mean CON velocity of the third repetition of the set, the supramaximal AEL induced a slightly more favorable pooled effect than the maximal AEL (QM [df = 1] = 0.62, p = 0.431, ES = -0.004 [= 100% of TRAD 1RM]; ES = 0.12 [> 100% of TRAD 1RM]). Additionally, the exercise selection did not significantly alter the acute effects on mean CON velocity during the third repetition of the set (QM [df = 1] = 1.72, p = 0.190, ES = 0.11 [back squat]; ES = -0.24 [bench press]). Residual heterogeneity during the third repetition remained substantial after accounting for all moderators, suggesting that none of them was a major source of between-study variability (exercise type: I2-total = 59.54%, I2-level 2 = 29.77%, I2-level 3 = 29.77%; loading differential: I2-total = 59.15%, I2-level 2 = 29.57%, I2-level 3 = 29.57%; magnitude of ECC load: I2-total = 60.86%, I2-level 2 = 30.43%, I2-level 3 = 30.43%). Discussion This meta-analysis was conducted to examine the impact of AEL on the mean barbell velocity during the CON phase of the bench press and the back squat to inform future AEL programming. The primary finding of the current study was that AEL significantly impaired mean barbell velocity during the CON phase immediately after the removal of the AEL. However, the reduction in mean barbell velocity appears to dissipate during the CON phase of the second and third repetitions following the removal of AEL. Based upon the moderator analyses, the magnitude of the ECC load, loading differential, and exercise selection did not significantly affect the acute effects of AEL across the first three repetitions. The only exception to these findings occurred in response to the application of AEL during the back squat, where there were significantly greater reductions in CON barbell movement velocity immediately after the removal of the AEL when compared to AEL application during the bench press (p < 0.05). First Repetition after Removal of AEL When examining the pooled ES, the application of AEL resulted in a small negative effect on the mean CON barbell velocity of the first repetition after the removal of the additional ECC load when compared to TL. These data provide evidence that the application of AEL likely impairs the CON mean barbell velocity immediately after its removal. The findings from the present study contrast with the within-repetition performance enhancements reported in previous studies, where the TRAD 1RM and the peak velocity were acutely enhanced during the subsequent CON phase of the bench press [17] and the front squat, respectively [18]. These discrepancies may be partially explained by the fact that peak velocity only accounts for a small percentage of the CON phase, and differences in the outcome variables assessed. For instance, while the application of AEL protocols resulted in greater mean CON force after the removal of AEL compared with the TL strategy [19, 22], the mean CON barbell velocity was not enhanced. It is plausible that the magnitude of residual force enhancement (i.e., an increase in force caused by active stretching [72]) may occur following the application of AEL, however, its magnitude may be insufficient to consistently accelerate the CON load faster through the full range of motion. Additional explanations for the lack of performance enhancement immediately after the removal of the additional ECC load may be partially explained by the prolonged ECC duration and a potential technique breakdown (e.g., the restabilization after the removal of the additional ECC load). Specifically, when the AEL is applied using weight releasers, the duration of the ECC phase has been reported to be prolonged [12, 19, 25, 26], which may result in an attenuation of the ability to utilize the stretch–shortening cycle (SSC) as the athlete transitions from the ECC to CON phase of the lift [73]. This is often noted as the ECC phase is typically performed at a slower speed to allow the athlete to better control their movement while the additional ECC load is applied. Additionally, applying a large loading differential (i.e., 70% of TRAD 1RM) between the ECC and CON loads may disrupt bench press technique immediately after the removal of the AEL [23]. Therefore, it is plausible that there is an attenuated ability to maximize SSC utilization, which is compounded by alterations to lifting technique resulting in an inability to enhance mean barbell velocity during the subsequent CON of the lift. Second Repetition After Removal of AEL Based on the pooled results from the meta-analytic model, the mean CON velocity of the second repetition of the set is likely to be equivalent to the velocities observed under TL conditions when the AEL was applied only on the first repetition of the set. For example, when an AEL was applied on every other repetition across a set of 10 repetitions (i.e., 5 cluster sets of 2 repetitions), the mean barbell velocity during the CON phase of the back squat exhibited an alternating pattern, decreasing during the CON phase immediately after the removal of the AEL and returning to levels similar to those associated with TL [22]. None of the moderators examined (ECC load: >100% of TRAD 1RM vs. =100% TRAD 1RM, loading differential between ECC and CON loads: ≥40% of TRAD 1RM vs. <40% of TRAD 1RM, and exercise selection: back squat vs. bench press) significantly influenced the acute effects of the AEL on the mean CON barbell velocity during second repetition (p > 0.05). However, it is still worth noting that using the same AEL protocol (ECC load equivalent to 120% of TRAD 1RM, CON load equivalent to 65% of TRAD 1RM) during the performance of the bench press or back squat induced different acute mean CON barbell velocity responses during the second repetition after the removal of the AEL (back squat: g = 0.50, bench press: g = -0.57) [12, 19]. Based upon these data, applying AEL with an ECC load equivalent to 120% of the TRAD 1RM during the ECC phase of the bench press is likely ineffective, while this loading application may be effective during the back squat. However, more research is warranted to verify this contention. Based on the results from the moderator analysis, applying loading differentials greater than 40% of the TRAD 1RM (pooled ES = 0.03) likely results in a trivial increase in mean CON velocity compared with smaller loading differentials (i.e., < 40% of 1RM) (pooled ES = -0.03) (Fig. 4). Given the negligible difference between pooled ESs, these findings may not be practically meaningful. However, there is some evidence that supports the implementation of loading differentials greater than 40% of the TRAD 1RM. For example, Suchomel et al. [28] report that supramaximal AEL protocols (ECC load equivalent to 110% of TRAD 1RM) tended to produce more favourable mean CON velocities (CON load equivalent to 60, 70, and 80% of TRAD 1RM) when compared to maximal AEL (ECC load equivalent to 100% of TRAD 1RM) and TL protocols (Fig. 5). Therefore, based upon the current findings and those of Suchomel et al. [28] it is generally recommended that to minimize the negative impact of the application of AEL on mean CON barbell velocity during the second repetition of the set, a supramaximal AEL load should be applied in conjunction with a loading differential greater than 40% of the TRAD 1RM (i.e., difference between ECC and CON load). While we were unable to conduct a moderator analysis on the aggregated relative strength data [74], there is some evidence demonstrating that individuals who are able to back squat ≥2 x body mass (BM) are more likely to display an enhance CON movement velocity during the second repetition after the removal of the AEL when performing the back squat [19, 28]. Specifically, two previous studies [28, 30] were performed with identical AEL protocols during the back squat but observed different acute mean velocity responses during the CON phase of the second repetition (Fig. 5). Although the biological sex difference exists between these two primary studies [28, 30], which may represent an additional factor contributing to the observed differences in mean CON velocity, differences in lower-body performance between biological sexes have been shown to be largely explained by relative strength [75–77]. Therefore, this review focuses on the potential influence of relative strength on mean CON velocity. Specifically, faster mean CON velocities were observed under AEL conditions in individuals with greater relative strength (2.0 ± 0.3 times BM) [28]. In contrast, individuals with lower relative strength (1.50 ± 0.19 times of BM) exhibited faster mean CON velocities under TL conditions [30]. Furthermore, based on a hierarchical linear model analysis, Merrigan et al. [12] have reported that stronger individuals demonstrated greater mean CON velocity during the bench press performed with AEL when compared to TL conditions when relative strength level was included as a fixed factor. Therefore, it is plausible that stronger athletes are better able to better utilize the SSC to enhance the mean CON barbell velocity during the second repetition following the removal of the AEL when compared to the TL condition. Third Repetition after Removal of AEL Based on the model results, the pooled ES indicated a trivial positive effect (ES = 0.08, 95% CI: [-0.08, 0.23], p > 0.05) on the mean CON barbell velocity during the third repetition of the set when AEL was applied only on the first repetition of the set. However, it is worth noting that 95% of the CI crossed zero, indicating that it is possible that there may still be a chance of the application AEL resulting in a slower mean CON velocity during the third repetition after the removal of the AEL when compared to TL. The pooled trivial positive effect may be partially explained by greater utilization of the SSC, supported by the observation of faster mean barbell velocities during the ECC phase of the third and fourth repetitions compared to TL reported within a previous study [19]. Alternatively, by the third repetition, the acute neuromuscular fatigue associated with the application of AEL may have attenuated sufficiently, allowing a performance enhancement to be observed. This is consistent with the conceptual framework proposed by Seitz and Haff [78], whereby performance following a conditioning activity depends on the net balance between fatigue and potentiation. It is also plausible that participants require the second repetition to re-establish the lifting rhythm after the removal of AEL. By the third repetition, an improved lifting rhythm may have allowed them to transition more smoothly from the ECC phase to the CON phase, thereby enhancing CON performance. The conflicting acute responses presented in the included primary studies are likely exercise-specific and ECC load-dependent (Fig. 6). For example, Suchomel et al. [28] and Campbell et al. [30] have reported conflicting acute effects on the mean CON barbell velocity during the third repetition. These conflicting results are likely explained by the relative strength of the participants. Participants with higher relative strength (2.0 ± 0.3 times BM) levels were able to generate higher CON barbell velocities during the third repetition after the removal of the AEL [28]. Conversely, those with lower relative strength levels (1.50 ± 0.19 times BM) generated lower CON barbell velocities during the third repetition after the removal of the AEL [30]. These results are further supported by Merrigan et al. [19], who also reported an enhanced performance during the third repetition after the removal of the AEL when examining individuals who were able to squat double their body weight. Ultimately, based upon the available evidence, it is plausible that stronger individuals are more likely to exhibit a CON performance enhancement during the third repetition after the removal of the AEL. Limitations and Considerations There are several limitations that should be acknowledged when interpreting the findings from the present meta-analysis. First, due to the limited number of acute AEL studies and inconsistent outcome variables reported across primary studies, we focused solely on mean barbell velocity during the CON phase of the lift to minimize heterogeneity. Second, despite efforts to reduce between-study heterogeneity, significant heterogeneity remained in models examining the third repetition. However, the variance was primarily attributable to variations in AEL protocols (within-study level) during the first repetition. Third, while our analysis included 36 ESs for the first repetition, 21 ESs for the second repetition, and 22 ESs for the third repetition after the removal of the AEL, these ESs were primarily derived from only 8 primary studies. To account for the nested structure of these studies and the correlations between ESs from the same primary study, RVE with the CHE models were considered. Future researchers examining the impact of AEL on performance should consider reporting a wider array of kinetic and kinematic variables, such as barbell acceleration, mean force, and mean velocity during the CON phase of the lift. In addition, studies using repeated-measures designs are encouraged to examine and report the correlations between outcome variables measured under AEL and TL conditions. When applying AEL to a set of lifts, researchers should examine and report acute effects on each individual repetition rather than averaged values across the entire set. This detailed information may increase coaches’ confidence in programming AEL into resistance training and interacting with AEL with alternative set configurations, such as cluster sets or rest redistribution. Notably, the increased mechanical load induced by AEL, despite reductions in mean CON velocity, may contribute to greater maximal strength adaptations compared to TL during the strength block [15, 16, 79]. However, optimal AEL protocols should still be further examined to guide long-term training interventions and improve their implementation across different phases of periodized training. Conclusions Based on the results of this meta-analysis, the application of AEL is likely to impair the mean barbell velocity during the CON phase of the first repetition following its removal, compared to the TL. This reduction appears to be greater in the back squat than in the bench press. However, this impairment tends to dissipate during the CON phase of the second and third repetitions after the removal of AEL in the first repetition. The positive impact on barbell velocity may also be exercise-specific and load-dependent. Although aggregated means of relative strength were not suitable for conducting moderator analyses, evidence from the included primary studies suggests that AEL may be better suited for stronger individuals. However, further research is warranted to validate this recommendation. Given the lower certainty of evidence for subsequent repetitions, caution is warranted when applying these findings to AEL programming when AEL is applied only on the first repetition of the set. Further well-controlled studies are needed to reduce uncertainty and better define the effects across repetitions. Acknowledgements Not applicable. Abbreviations Accentuated eccentric loading Bench press Back squat Body mass Correlated and hierarchical effects Confidence intervals Concentric Eccentric Effect size Grading of Recommendations Assessment, Development, and Evaluation One-repetition maximum Prediction intervals Participants, intervention, comparator, outcome, and study design Preferred Reporting Items for Systematic Reviews and Meta-Analyses Robust variance estimation Standard deviation Standard error Standardized mean difference Stretch–shortening cycle Traditional Traditional loading Author Contributions All authors contributed to the conception and design of the study. JY conducted the literature search. JY and TN screened the literature, assessed the risk of bias of primary studies, and extracted the data from primary studies. JY performed all statistical analyses, visualized the data, and drafted the initial manuscript. GGH, PC, KLK, and TN critically reviewed and revised multiple versions of the manuscript. All authors read and approved the final version. Funding Open Access funding enabled and organized by CAUL and its Member Institutions. No funding was received for this project. Data Availability The R script used for effect size calculation and model fitting, as well as the portion of the extracted dataset derived from published studies, are available on the Open Science Framework (OSF): [https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60]. The full extracted dataset also contains aggregate data provided directly by the original study authors, which cannot be shared publicly due to data-use restrictions. Declarations Ethics Approval and Consent to Participate Not applicable. Consent for Publication Not applicable. Competing Interests All authors declare that they have no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The R script used for effect size calculation and model fitting, as well as the portion of the extracted dataset derived from published studies, are available on the Open Science Framework (OSF): [https://osf.io/fme9c/overview? view_only=06c37407c81049b39b0a76eadbb1ae60]. The full extracted dataset also contains aggregate data provided directly by the original study authors, which cannot be shared publicly due to data-use restrictions.