Does baseline strength influence adaptations to velocity loss threshold-based resistance training? Authors: Junjie Tao, Tianyuan He, Xinjun Chen, Lin Xie, Jiaxin He, Kelin Huo Published: 2026 ABSTRACT ObjectiveThis study examined whether baseline strength moderates the magnitude of adaptations to velocity loss (VL) threshold-based resistance training across different training populations.MethodsA systematic search was conducted in PubMed, Web of Science, SPORTDiscus (EBSCO), and Embase up to January 2026. Randomized controlled trials were included if they implemented resistance training prescribed using predefined VL thresholds, lasted at least 3 weeks, reported quantifiable baseline strength data, and provided pre- and post-intervention measures of strength and/or power outcomes. Effect sizes were calculated as standardized within-group pre-post changes (Hedges’ g). A three-level random-effects meta-regression model was applied to account for statistical dependence among multiple effect sizes within studies. Baseline strength (standardized 1RM) was examined as the primary moderator, with training status and VL threshold entered as additional moderators.ResultsSixteen studies providing 89 effect sizes were included. VL threshold-based resistance training produced a moderate overall within-group improvement in strength and power outcomes (g = 0.59, 95% CI [0.40, 0.77], p < 0.001). The omnibus test of moderators was not statistically significant (QM(4) = 1.09, p = 0.896). Baseline strength was not significantly associated with adaptation magnitude (β = -0.076, p = 0.355), and the interaction between baseline strength and training status was also not statistically significant (β = 0.045, p = 0.775).ConclusionVL threshold-based resistance training elicits moderate improvements in neuromuscular performance. Within the limitations of the available evidence, no statistically detectable moderation by baseline strength or categorical training status was observed. These findings suggest that current evidence does not clearly support modifying VL prescriptions solely on the basis of baseline strength; however, equivalence across populations cannot be definitively established.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13102-026-02067-z. Introduction Modern competitive sport increasingly requires athletes to develop high levels of strength and power, driving resistance training toward more precise and individualized prescription strategies [1, 2]. As a core method for improving neuromuscular performance, resistance training effectiveness depends largely on the appropriate management of training load and fatigue [3, 4]. Traditional prescription methods based on a fixed percentage of one-repetition maximum (1RM) are practical, but they may not adequately reflect day-to-day fluctuations in athletes’ readiness, potentially leading to either insufficient stimulus or excessive fatigue [5]. In this context, velocity-based training (VBT) has emerged as a practical approach for monitoring and adjusting training load through real-time feedback. Within VBT, the velocity loss (VL) threshold method—using the percentage decline in concentric velocity within a set as a criterion for termination—is an increasingly popular strategy to regulate fatigue and dose the training stimulus [6, 7]. Previous studies suggest that the relationship between VL and chronic adaptations may be non-linear [8], with moderate VL thresholds (e.g., 20–30%) often associated with favorable strength gains [9]. Recent meta-analytic work has synthesized the acute and chronic effects of implementing velocity loss thresholds, primarily characterizing average adaptation magnitudes across conditions [10]. However, these recommendations are largely derived from average effects across samples, leaving it unclear whether individual characteristics moderate the response to VL-based prescription. Evidence from the broader resistance training literature indicates that pre-training strength levels and training experience can influence adaptation magnitude, with less trained individuals often demonstrating greater relative improvements, whereas stronger or highly trained individuals may exhibit attenuated gains as they approach higher performance ceilings [11, 12]. Whether such moderation patterns persist when training is regulated using VL thresholds—designed to standardize relative effort and fatigue exposure—remains uncertain. While training status (e.g., trained vs. untrained) is commonly used as a categorical descriptor, it may reflect historical experience rather than current physiological capacity. Therefore, baseline strength assessed by absolute pre-intervention one-repetition maximum may provide a direct representation of an individual’s current neuromuscular capacity [13]. Unlike broad categorical labels, baseline strength provides a continuous indicator of functional performance capacity and neuromuscular characteristics, including differences in force-producing ability [14]. Examining baseline strength as a continuous variable allows for a more nuanced assessment of how proximity to an individual’s performance ceiling may relate to subsequent adaptation under VL-based training conditions. Modeling continuous variables rather than imposing arbitrary cut-off thresholds preserves statistical power and avoids information loss [15]. Consequently, prioritizing baseline strength as a primary phenotype may offer greater explanatory value for understanding inter-individual variability in response to velocity-based training. From a practical perspective, although higher baseline strength is often associated with greater absolute loading capacity, the critical question within a VBT framework is whether the association between initial strength and adaptation magnitude differs across training backgrounds. If baseline strength moderates responses to specific VL thresholds differently for novices and experienced athletes, training prescriptions might require stratification. Conversely, if no meaningful moderation is observed, this would suggest that VL thresholds may operate as a broadly applicable autoregulatory mechanism across populations. Despite the practical relevance of this issue, no quantitative synthesis has systematically examined whether baseline strength and training status interact to influence adaptations to VL-based resistance training within a multilevel meta-analytic framework. Furthermore, the existing literature presents methodological challenges. Many primary studies report multiple outcomes (e.g., 1RM and jump performance) from the same sample, resulting in statistically dependent effect sizes. Conventional two-level meta-analyses often ignore this nesting structure, potentially underestimating standard errors and inflating Type I error rates [16]. To appropriately model the hierarchical structure of the available data and account for statistical dependence among effect sizes, the present study employed a three-level meta-regression model combined with robust variance estimation (RVE) to examine: (1) the overall effect of VL-based training on strength and power outcomes; and (2) whether baseline strength and training status explain variability in these adaptations. By doing so, this study aims to provide a rigorous evidence-based foundation for evaluating whether VBT prescriptions require modification according to initial strength characteristics. Materials and methods Protocol registration This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [17]. The study protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) in January 2026 (registration number: CRD420261294520). Eligibility criteria The eligibility criteria were defined according to the PICOS framework (participants, intervention, comparison, outcomes, and study design). P: Studies were eligible if they included healthy adults and reported quantifiable baseline strength data (e.g., pre-intervention one-repetition maximum, 1RM). Descriptions of participants’ training status were extracted for moderator coding purposes but were not sufficient for inclusion in the absence of measurable baseline strength.I: Eligible studies had to implement resistance training prescribed using a predefined velocity loss (VL) threshold as the within-set criterion for terminating repetitions. To ensure that only chronic training adaptations were considered, the intervention period had to last at least 3 weeks.C: Studies were included if they incorporated at least one VL-based resistance training arm. When present, comparator groups (e.g., traditional training, non-training controls, or alternative VL thresholds) were described but were not directly analyzed in the primary within-group meta-analysis.O: Eligible studies were required to report at least one outcome reflecting chronic adaptations in strength- or power-related performance. These outcomes included maximal strength measures (e.g., squat or bench press 1RM), jumping performance (e.g., countermovement jump height), and sprint performance.S: Only randomized controlled trials were included. P: Studies were eligible if they included healthy adults and reported quantifiable baseline strength data (e.g., pre-intervention one-repetition maximum, 1RM). Descriptions of participants’ training status were extracted for moderator coding purposes but were not sufficient for inclusion in the absence of measurable baseline strength. P: Studies were eligible if they included healthy adults and reported quantifiable baseline strength data (e.g., pre-intervention one-repetition maximum, 1RM). Descriptions of participants’ training status were extracted for moderator coding purposes but were not sufficient for inclusion in the absence of measurable baseline strength. I: Eligible studies had to implement resistance training prescribed using a predefined velocity loss (VL) threshold as the within-set criterion for terminating repetitions. To ensure that only chronic training adaptations were considered, the intervention period had to last at least 3 weeks. I: Eligible studies had to implement resistance training prescribed using a predefined velocity loss (VL) threshold as the within-set criterion for terminating repetitions. To ensure that only chronic training adaptations were considered, the intervention period had to last at least 3 weeks. C: Studies were included if they incorporated at least one VL-based resistance training arm. When present, comparator groups (e.g., traditional training, non-training controls, or alternative VL thresholds) were described but were not directly analyzed in the primary within-group meta-analysis. C: Studies were included if they incorporated at least one VL-based resistance training arm. When present, comparator groups (e.g., traditional training, non-training controls, or alternative VL thresholds) were described but were not directly analyzed in the primary within-group meta-analysis. O: Eligible studies were required to report at least one outcome reflecting chronic adaptations in strength- or power-related performance. These outcomes included maximal strength measures (e.g., squat or bench press 1RM), jumping performance (e.g., countermovement jump height), and sprint performance. O: Eligible studies were required to report at least one outcome reflecting chronic adaptations in strength- or power-related performance. These outcomes included maximal strength measures (e.g., squat or bench press 1RM), jumping performance (e.g., countermovement jump height), and sprint performance. S: Only randomized controlled trials were included. S: Only randomized controlled trials were included. Studies were excluded if they: examined only acute responses to a single training session;included clinical populations, para-athletes, animals, or in vitro designs;used non-traditional resistance modalities in which VL was derived from sled, flywheel, elastic-band, or other non-gravity-based loading systems;were reviews, meta-analyses, conference abstracts, case reports, or articles with insufficient data for effect size calculation. Studies that did not report a specific VL threshold were also excluded. examined only acute responses to a single training session; examined only acute responses to a single training session; included clinical populations, para-athletes, animals, or in vitro designs; included clinical populations, para-athletes, animals, or in vitro designs; used non-traditional resistance modalities in which VL was derived from sled, flywheel, elastic-band, or other non-gravity-based loading systems; used non-traditional resistance modalities in which VL was derived from sled, flywheel, elastic-band, or other non-gravity-based loading systems; were reviews, meta-analyses, conference abstracts, case reports, or articles with insufficient data for effect size calculation. Studies that did not report a specific VL threshold were also excluded. were reviews, meta-analyses, conference abstracts, case reports, or articles with insufficient data for effect size calculation. Studies that did not report a specific VL threshold were also excluded. Search strategy and study selection A systematic literature search was performed in PubMed, Web of Science, SPORTDiscus (EBSCO), and Embase from database inception to 29 January 2026. The detailed search strategies and specific terms used for each database are provided in the Supplementary Material (Table S1). Search terms were developed around three main concepts: (1) velocity loss and velocity-based training; (2) resistance training; and (3) strength- or power-related outcomes. In addition, forward citation tracking was performed using Google Scholar, and the reference lists of eligible studies and relevant reviews were manually screened to identify additional studies. Only peer-reviewed articles published in English were considered eligible. All retrieved records were imported into Zotero (version 7.0), and duplicate records were removed automatically and manually. Study selection was performed independently by two reviewers in two stages: title/abstract screening and full-text screening. Disagreements were resolved through discussion, and a third reviewer was consulted when consensus could not be reached. Data extraction Data extraction was conducted independently by two reviewers using a predesigned standardized Excel form. The following information was extracted from each study: author and publication year; sample characteristics (sample size, sex, training experience, body mass, and pre-intervention one-repetition maximum [1RM] as baseline strength); intervention characteristics (VL threshold, training frequency, exercise type, loading intensity, and intervention duration); and outcome data for eligible performance measures (e.g., 1RM, countermovement jump, and sprint time), including means and standard deviations. When numerical data were not directly reported in the text or tables, values were extracted from figures using WebPlotDigitizer. All extracted data were checked by a second reviewer to minimize transcription errors. Risk of bias assessment Risk of bias was assessed using the Cochrane Risk of Bias 2.0 (RoB 2) tool for randomized trials [18]. The following domains were evaluated: (1) bias arising from the randomization process; (2) bias due to deviations from intended interventions; (3) bias due to missing outcome data; (4) bias in measurement of the outcome; and (5) bias in selection of the reported result. Each domain was judged as “low risk”, “some concerns”, or “high risk”, and an overall risk-of-bias judgment was assigned accordingly. Because participant and coach blinding is generally not feasible in exercise training interventions, the domain related to deviations from intended interventions was expected to raise some concerns unless specific control procedures were reported. Any disagreement between reviewers was resolved through discussion with a third reviewer. Sensitivity analyses excluding studies rated as high overall risk of bias were conducted to evaluate the robustness of the findings. Statistical analysis Effect size calculation Effect sizes were calculated as standardized within-group pre–post changes using Hedges’g to correct for small-sample bias. Each velocity loss (VL) training arm was treated as an independent intervention condition, and no between-group comparisons were included in the primary meta-analysis. Therefore, the pooled effect represents the average magnitude of adaptation observed following VL-based resistance training rather than a comparison with traditional or non-training control groups. When available, mean change scores and corresponding standard deviations were used. If change-score standard deviations were not reported, they were estimated from pre- and post-intervention standard deviations using the following formula:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g=d\times\left(1-\frac{3}{4\left({n}_{1}+{n}_{2}\right)-9}\right)$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{S}\mathrm{D}}_{\mathrm{change}}=\sqrt{{\mathrm{SD}}_{\mathrm{pre}}^{2}+{\mathrm{S}\mathrm{D}}_{\mathrm{post}}^{2}-2\mathrm{r}\cdot\:{\mathrm{SD}}_{\mathrm{pre}}\cdot\:{\mathrm{SD}}_{\mathrm{post}}}$$\end{document} where r represents the assumed pre–post correlation coefficient. In accordance with the Cochrane Handbook, r was conservatively set at 0.50 [19]. Sampling variances were calculated using standard meta-analytic formulas. Effect size directions were aligned so that positive values consistently indicated performance improvements. For time-based outcomes (e.g., sprint time), the sign of the effect size was reversed prior to analysis. All statistical analyses were conducted in R (version 4.5.2). Three-level meta-analytic model Because several included studies reported multiple outcomes from the same sample, effect sizes were not statistically independent. Therefore, a three-level random-effects model was used to account for the nested structure of the data [20]. Variance was partitioned into sampling variance (level 1), within-study variance between effect sizes (level 2), and between-study variance (level 3). Model parameters were estimated using restricted maximum likelihood (REML). When the number of studies was limited or the clustering structure was unbalanced, robust variance estimation (RVE) was applied to obtain more reliable standard errors [21].\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned}{g}_{ijk}&={\upbeta}_{0}+{\upbeta}_{1}{X}_{ijk}+{u}_{0k}\\&+{v}_{0jk}+{e}_{ijk}\end{aligned}$$\end{document} Moderator variables The primary moderator of interest was baseline strength, operationalized as absolute pre-intervention one-repetition maximum (1RM, kg). Baseline strength was selected as the primary moderator because the objective of this study was to examine whether initial strength level influences adaptation magnitude following velocity loss threshold-based resistance training. For each intervention arm, baseline strength corresponded to the primary trained exercise. In studies reporting both upper- and lower-body 1RM values (e.g., squat and bench press), the measure most directly aligned with the primary training modality was selected. To improve comparability across studies and account for differences in absolute load scales, baseline 1RM values were standardized (z-scored) across the meta-analytic dataset prior to analysis. Training status and velocity loss threshold were examined as additional moderators. Training status was operationally defined based on participant characteristics reported in each primary study. Studies were categorized into two groups: (1) High training level, comprising trained athletes, resistance-trained individuals, or experienced practitioners with documented resistance training backgrounds; and (2) Low training level, comprising untrained students, novice exercisers, or individuals without structured resistance training experience [22]. Furthermore, the interaction between baseline strength and training status was examined to determine whether the association between baseline strength and adaptation magnitude differed according to training background. Velocity loss threshold was treated as a continuous moderator and expressed as a percentage ranging from 0% to 50%. Continuous moderators were mean-centered to facilitate interpretation of interaction terms. Multicollinearity among moderators was assessed using variance inflation factors (VIF), with higher values indicating greater concern for collinearity. Sensitivity analyses Several sensitivity analyses were conducted to evaluate the robustness of findings. First, robust variance estimation was employed to obtain cluster-robust standard errors that account for the dependency structure of effect sizes [23]. Second, Wild Bootstrap resampling was implemented to generate empirical p-values under non-normal conditions [24]. Third, an outcome-specific sensitivity analysis was conducted by restricting the dataset to countermovement jump (CMJ) outcomes to examine whether the moderation findings were consistent for explosive power adaptations. Fourth, studies involving soccer players were excluded in a sensitivity analysis given the potential for sport-specific adaptations that may influence training responsiveness. Finally, leave-one-study-out analyses were performed to identify influential outliers [25]. Additionally, sensitivity analyses excluding studies rated as high overall risk of bias were conducted. Publication bias assessment Publication bias was assessed at the study level using one effect size per study to avoid statistical dependence among multiple outcomes. Funnel plot asymmetry was evaluated visually and statistically using Egger’s regression test [26]. Given the limited number of included studies (k = 16), results were interpreted cautiously, as asymmetry tests have reduced reliability in small meta-analyses. Results Study selection and characteristics of included studies The literature search identified 1,254 records across PubMed, Web of Science, EBSCO, and Embase. After removal of 320 duplicate records, 934 records were screened by title and abstract, of which 896 were excluded. Full texts were sought for 38 reports, and 7 reports could not be retrieved. A total of 31 full-text reports from database searches were assessed for eligibility. An additional 5 reports were sought through other sources. Of these, 1 could not be retrieved and 4 were assessed for eligibility. One report was subsequently excluded because of missing key data. In total, 16 randomized controlled trials met all eligibility criteria and were included in the quantitative synthesis [27], contributing 89 effect sizes (Fig. 1). Fig. 1PRISMA flow diagram of the study selection process PRISMA flow diagram of the study selection process The included samples covered a relatively broad range of training backgrounds. Ten studies (48 effect sizes) were classified as high training status, including professional soccer players, competitive boxers, rowers, and resistance-trained men. Six studies (41 effect sizes) were classified as lower training status and mainly involved physical education students and recreationally active individuals. The interventions primarily involved core resistance exercises such as the squat and bench press. VL thresholds ranged from very low fatigue levels (0–10%) to higher fatigue conditions (40–50%), and the mean intervention duration was 7.4 weeks (Table 1). Table 1Characteristics of the included studiesStudyVLSample_sizeBody mass (kg)exerciseLoad(%1RM)Duration/FrequencyOutcomeBaseline Strength (kg)Rodríguez-Rosell et al.2021 A [27]10%1170.7 ± 5.1Squat55 ~ 70%8wk2sess/wkCMJ96.1 ± 13.930%1273.7 ± 9.4T1097.4 ± 19.245%1272.1 ± 9.6T2096.8 ± 15.11RMRojas-Jaramillo et al.2024 [28]10%1073.3 ± 4.7Squat45 ~ 60%CMJ77.63 ± 14.728 wkT1030%1073.6 ± 4.32sess/wkT2081.1 ± 17.31RMPérez-Castilla et al.2017 [29]10%1073.9 ± 17.1loaded CMJ40%4 wk1RM148 ± 27.820%1080.1 ± 15.22sess/wkT15Pareja-Blanco et al.2017 A [30]20%1275.8 ± 7.0Squat70 ~ 85%8 wk1RM106.5 ± 12.240%1075.8 ± 7.02sess/wkCMJ104.5 ± 15.1T20Pareja-Blanco et al.2017B [31]15%1075.5 ± 8.6Squat50 ~ 70%6 wk1RM101.3 ± 18.830%875.5 ± 8.63sess/wkCMJ100.2 ± 20.3T30Pareja-Blanco et al.2020 A [32]0%1575.5 ± 9.7Bench press70 ~ 85%1RM65.5 ± 9.615%1675.5 ± 9.78 wk64.5 ± 17.425%1575.5 ± 9.72sess/wk68.4 ± 18.550%1675.5 ± 9.770 ± 16.9Galiano et al.2022 [33]5%1572.5 ± 11.3Squat50%7 wk1RM97.7 ± 13.720%1375.7 ± 9.42sess/wkCMJ97 ± 13.9T20Rodiles-Guerrero et al.2022 [34]0%1273.9 ± 10.4Bench press50 ~ 70%1RM67.3 ± 10.515%1373.9 ± 10.48 wk69.1 ± 1825%1373.9 ± 10.42sess/wk68 ± 9.750%1273.9 ± 10.468.3 ± 13.6Han et al.2025 [35]10%1477.9 ± 9.1Squat70%CMJ135 ± 32.68 wkSLJ2sess/wkT301RMPareja-Blanco et al.2020B [36]0%1475.5 ± 9.7Squat70 ~ 85%1RM99.9 ± 19.110%1475.5 ± 9.78 wk96.7 ± 15.320%1375.5 ± 9.72sess/wk97.8 ± 23.240%1475.5 ± 9.794.5 ± 17.5Rissanen et al.2022 [37]20%1281.5 ± 8.065 ~ 75%1RMSquat:112.3 ± 28.2/111.2 ± 14/67.8 ± 12.3/66 ± 18.8Bench:80.6 ± 17.4/78.3 ± 13.6/40.2 ± 9.9/39.3 ± 9.940%1182.3 ± 14.3Squat &8 wk20%1161.2 ± 4.6Bench press2sess/wk40%1160.1 ± 7.5Rodiles-Guerrero et al.2026 [38]0%1761.3 ± 9.5Bench press70 ~ 85%8 wk1RM32.8 ± 7.425%1861.3 ± 9.52sess/wk32.5 ± 7.950%1461.3 ± 9.533.5 ± 8.7Sánchez-Moreno et al.2021 [39]15%1173.9 ± 7.2Squat60 ~ 80%T1086.3 ± 7.38 wkT2045%1173.9 ± 7.22sess/wkCMJ86.1 ± 10.61RMHeld et al.2021 [40]10%575.8 ± 8.6Squat80%8 wk1RM123.3 ± 8.62sess/wkRodríguez-Rosell et al.2021B [41]10%1275.1 ± 10.3Squat70 ~ 85%CMJ100.8 ± 24.68 wkT1030%1374.0 ± 9.12sess/wkT2096.6 ± 14.71RMRodiles-Guerrero et al.2020 [42]10%1573.3 ± 5.9Bench press65 ~ 85%5 wk1RM73.7 ± 1230%1573.3 ± 5.93sess/wk80.4 ± 16.150%1573.3 ± 5.976.9 ± 16.3 Table 1Characteristics of the included studiesStudyVLSample_sizeBody mass (kg)exerciseLoad(%1RM)Duration/FrequencyOutcomeBaseline Strength (kg)Rodríguez-Rosell et al.2021 A [27]10%1170.7 ± 5.1Squat55 ~ 70%8wk2sess/wkCMJ96.1 ± 13.930%1273.7 ± 9.4T1097.4 ± 19.245%1272.1 ± 9.6T2096.8 ± 15.11RMRojas-Jaramillo et al.2024 [28]10%1073.3 ± 4.7Squat45 ~ 60%CMJ77.63 ± 14.728 wkT1030%1073.6 ± 4.32sess/wkT2081.1 ± 17.31RMPérez-Castilla et al.2017 [29]10%1073.9 ± 17.1loaded CMJ40%4 wk1RM148 ± 27.820%1080.1 ± 15.22sess/wkT15Pareja-Blanco et al.2017 A [30]20%1275.8 ± 7.0Squat70 ~ 85%8 wk1RM106.5 ± 12.240%1075.8 ± 7.02sess/wkCMJ104.5 ± 15.1T20Pareja-Blanco et al.2017B [31]15%1075.5 ± 8.6Squat50 ~ 70%6 wk1RM101.3 ± 18.830%875.5 ± 8.63sess/wkCMJ100.2 ± 20.3T30Pareja-Blanco et al.2020 A [32]0%1575.5 ± 9.7Bench press70 ~ 85%1RM65.5 ± 9.615%1675.5 ± 9.78 wk64.5 ± 17.425%1575.5 ± 9.72sess/wk68.4 ± 18.550%1675.5 ± 9.770 ± 16.9Galiano et al.2022 [33]5%1572.5 ± 11.3Squat50%7 wk1RM97.7 ± 13.720%1375.7 ± 9.42sess/wkCMJ97 ± 13.9T20Rodiles-Guerrero et al.2022 [34]0%1273.9 ± 10.4Bench press50 ~ 70%1RM67.3 ± 10.515%1373.9 ± 10.48 wk69.1 ± 1825%1373.9 ± 10.42sess/wk68 ± 9.750%1273.9 ± 10.468.3 ± 13.6Han et al.2025 [35]10%1477.9 ± 9.1Squat70%CMJ135 ± 32.68 wkSLJ2sess/wkT301RMPareja-Blanco et al.2020B [36]0%1475.5 ± 9.7Squat70 ~ 85%1RM99.9 ± 19.110%1475.5 ± 9.78 wk96.7 ± 15.320%1375.5 ± 9.72sess/wk97.8 ± 23.240%1475.5 ± 9.794.5 ± 17.5Rissanen et al.2022 [37]20%1281.5 ± 8.065 ~ 75%1RMSquat:112.3 ± 28.2/111.2 ± 14/67.8 ± 12.3/66 ± 18.8Bench:80.6 ± 17.4/78.3 ± 13.6/40.2 ± 9.9/39.3 ± 9.940%1182.3 ± 14.3Squat &8 wk20%1161.2 ± 4.6Bench press2sess/wk40%1160.1 ± 7.5Rodiles-Guerrero et al.2026 [38]0%1761.3 ± 9.5Bench press70 ~ 85%8 wk1RM32.8 ± 7.425%1861.3 ± 9.52sess/wk32.5 ± 7.950%1461.3 ± 9.533.5 ± 8.7Sánchez-Moreno et al.2021 [39]15%1173.9 ± 7.2Squat60 ~ 80%T1086.3 ± 7.38 wkT2045%1173.9 ± 7.22sess/wkCMJ86.1 ± 10.61RMHeld et al.2021 [40]10%575.8 ± 8.6Squat80%8 wk1RM123.3 ± 8.62sess/wkRodríguez-Rosell et al.2021B [41]10%1275.1 ± 10.3Squat70 ~ 85%CMJ100.8 ± 24.68 wkT1030%1374.0 ± 9.12sess/wkT2096.6 ± 14.71RMRodiles-Guerrero et al.2020 [42]10%1573.3 ± 5.9Bench press65 ~ 85%5 wk1RM73.7 ± 1230%1573.3 ± 5.93sess/wk80.4 ± 16.150%1573.3 ± 5.976.9 ± 16.3 Characteristics of the included studies 8wk 2sess/wk Squat: 112.3 ± 28.2/111.2 ± 14/67.8 ± 12.3/66 ± 18.8 Bench: 80.6 ± 17.4/78.3 ± 13.6/40.2 ± 9.9/39.3 ± 9.9 Risk of bias Risk-of-bias assessment using the RoB 2 tool showed that most studies were rated as having “some concerns”, mainly in the domains of the randomization process and selection of the reported result. These judgments were primarily attributable to insufficient reporting of allocation concealment and related methodological details (Fig. 2). Fig. 2Summary of risk-of-bias judgments across domains Summary of risk-of-bias judgments across domains Main effects A three-level random-effects model was used to synthesize 89 within-group effect sizes from 16 studies. The analysis demonstrated a significant overall pre–post improvement following velocity loss-based resistance training (Hedges’ g = 0.59, 95% CI [0.42, 0.75], p < 0.001). Variance components indicated heterogeneity at both the study level (σ² = 0.024) and the within-study level (σ² = 0.173). The test for residual heterogeneity was statistically significant (QE(84) = 221.62, p < 0.001), suggesting that variability remained beyond sampling error. The forest plot (Fig. 3) illustrates the distribution of effect sizes across studies. Fig. 3Forest plot of study-level within-group effects of velocity loss-based resistance training on maximal strength outcomes Forest plot of study-level within-group effects of velocity loss-based resistance training on maximal strength outcomes Moderation by baseline strength A three-level meta-regression was conducted to examine whether baseline strength, training status, velocity loss threshold, or their interaction explained variability in adaptation magnitude. The omnibus test of moderators was not statistically significant (QM(4) = 1.09, p = 0.896), indicating that the included moderators did not account for a significant proportion of residual heterogeneity.Variance inflation factor analysis indicated no evidence of problematic multicollinearity between baseline strength and training status (VIF = 1.19). Specifically, baseline strength was not significantly associated with adaptation magnitude (β = -0.076, SE = 0.082, 95% CI [-0.237, 0.085], p = 0.355). Training status was also not a significant predictor (β = 0.003, SE = 0.202, 95% CI [-0.392, 0.398], p = 0.988), nor was velocity loss threshold (β = 0.001, SE = 0.004, 95% CI [-0.007, 0.009], p = 0.807). The interaction between baseline strength and training status was small and non-significant (β = 0.045, SE = 0.159, 95% CI [-0.266, 0.357], p = 0.775), suggesting no statistically detectable difference in the strength–adaptation relationship across categorical training levels. However, the absence of statistical significance should not be interpreted as evidence of equivalence.The distribution of effect sizes across baseline strength levels is illustrated in Fig. 4. Fig. 4Association between baseline strength and within-group adaptation Association between baseline strength and within-group adaptation Robustness and sensitivity analyses To ensure the stability of the primary findings regarding the moderating role of baseline strength, several robustness and sensitivity checks were conducted: Statistical robustness of the interaction The non-significant interaction between training status and baseline strength was consistently observed across estimation methods. Specifically, the interaction coefficient remained small and non-significant under Restricted Maximum Likelihood, Robust Variance Estimation, and Wild Bootstrap procedures (see Supplementary Table S4). These findings indicate that the absence of a statistically detectable interaction was not driven by model specification or dependency structure. Outcome-specific sensitivity To further evaluate whether the primary findings were influenced by outcome type, an outcome-specific sensitivity analysis was conducted by restricting the dataset to explosive power outcomes (CMJ). The omnibus test of moderators was not statistically significant (QM(4) = 1.50, p = 0.827), indicating that baseline strength, training status, velocity loss threshold, and their interaction did not explain a significant proportion of residual heterogeneity within this subset. The interaction between baseline strength and training status remained small and non-significant (β = 0.038, 95% CI [-1.894, 1.971], p = 0.969; see Supplementary Table S7). These results suggest that the absence of a statistically detectable moderation effect was not confined to maximal strength outcomes and was similarly observed in explosive power performance measures. Influence and outlier analysis A leave-one-study-out sensitivity analysis was performed to evaluate the impact of individual studies. The removal of any single study did not materially change the magnitude of the main effect or the non-significance of the interaction term, indicating that the findings were not driven by influential outliers. Risk-of-bias sensitivity analysis To evaluate the potential influence of methodological quality, a sensitivity analysis was conducted excluding three studies rated as high overall risk of bias. After exclusion, the omnibus test of moderators remained non-significant (QM(4) = 0.38, p = 0.984), and the interaction between baseline strength and training status remained small and non-significant (β = 0.004, p = 0.984). The direction and magnitude of the main parameter estimates were largely unchanged, suggesting that the findings were not driven by studies at high risk of bias. Publication bias Publication bias was assessed at the study level using one effect size per study (k = 16) to avoid statistical dependence among multiple outcomes(Fig. 5). The pooled study-level effect remained significant (g = 0.59, 95% CI [0.40, 0.77], p < 0.001), with low-to-moderate heterogeneity (I² = 22.6%). Egger’s regression test indicated significant funnel plot asymmetry (z = 3.39, p = 0.0007), which may reflect small-study effects. However, given the limited number of included studies and the use of within-group effect sizes, this finding should be interpreted cautiously. Tests for funnel plot asymmetry have limited reliability in small meta-analyses, and asymmetry may also arise from between-study methodological variation rather than selective publication alone. Fig. 5Funnel plot of study-level effect sizes (k = 16) Funnel plot of study-level effect sizes (k = 16) Discussion Baseline strength and training status as moderators of adaptation The primary objective of this study was to examine whether baseline strength moderates neuromuscular adaptations to VL-based resistance training across different training backgrounds. Contrary to our initial hypothesis, the present meta-regression did not identify a statistically significant association between baseline strength and adaptation magnitude. Furthermore, no significant interaction was detected between baseline strength and categorical training status. These findings suggest that, within the limits of the available data, baseline strength does not appear to systematically influence the magnitude of within-group adaptations to VL-based resistance training. Importantly, the absence of a statistically detectable moderation effect should not be interpreted as evidence of equivalence across populations, but rather as an indication that current evidence does not demonstrate differential responsiveness based on initial strength or training status. Physiological mechanisms of self-regulation Although the present analysis did not identify statistically significant moderation by baseline strength or training status, it is theoretically relevant to consider how velocity-based fatigue management may influence adaptation patterns across individuals. In traditional fixed-load training (e.g., percentage of 1RM), stronger individuals often handle substantially greater absolute loads, which may result in higher absolute mechanical stress and fatigue demands [10]. In contrast, velocity loss thresholds regulate set termination based on a relative decline in movement velocity. This autoregulatory approach may help align the training stimulus with an individual’s momentary capacity, potentially reducing excessive fatigue accumulation and standardizing relative effort across participants. However, direct evidence demonstrating that identical VL thresholds produce equivalent physiological responses across different strength levels remains limited. While it is plausible that similar percentage reductions in velocity reflect comparable levels of neuromuscular fatigue, this assumption has not been definitively established. Therefore, the present findings should not be interpreted as confirming mechanistic equivalence across populations, but rather as indicating that current empirical data do not demonstrate statistically detectable differences in adaptation magnitude. Future mechanistic studies incorporating physiological and neuromuscular measures are needed to clarify whether VL thresholds produce comparable internal load responses across individuals with divergent strength profiles. Sensitivity across outcomes The sensitivity analyses across outcome domains further support the primary findings. When the analysis was restricted to explosive power outcomes (CMJ), no statistically significant interaction between baseline strength and training status was observed (p = 0.969). This suggests that the absence of a detectable moderation effect was not limited to maximal strength outcomes. However, these findings should be interpreted cautiously. The lack of statistical significance does not confirm equivalence of responses across different neuromuscular qualities, but rather indicates that current evidence does not demonstrate differential adaptation patterns based on baseline strength. Given the limited number of available studies and the short intervention durations, further research is required to determine whether subtle or long-term differences may exist across physical qualities. From a practical standpoint, the present data do not provide clear statistical support for modifying velocity loss thresholds solely on the basis of baseline strength or categorical training status. Nevertheless, practitioners should continue to individualize prescription according to training objectives, athlete characteristics, and longitudinal progression. Practical implications From a practical perspective, the present findings do not provide clear statistical support for modifying velocity loss thresholds solely on the basis of baseline strength or categorical training status. Within the limitations of the available evidence, adaptations to VL-based resistance training did not appear to differ systematically according to initial strength level. However, these results should not be interpreted as definitive evidence of equivalence across populations. Practitioners should continue to individualize training prescription according to specific performance goals, exercise selection, recovery capacity, sport demands, and long-term progression. Velocity loss thresholds may serve as a useful autoregulatory framework, but decisions regarding threshold selection should be guided by comprehensive contextual considerations rather than baseline strength alone. Methodological implications and limitations Methodologically, this study employed a three-level meta-regression framework with RVE and Wild Bootstrap procedures to account for statistical dependence among effect sizes and improve inference accuracy. Nevertheless, several limitations warrant cautious interpretation. First, although no statistically significant interaction between baseline strength and training status was detected, absence of statistical significance should not be interpreted as evidence of equivalence. With k = 16 studies, statistical power to detect small or subtle moderation effects is inherently limited, and the possibility of Type II error cannot be excluded. Moreover, meta-regression analyses operate at the study level and may be subject to ecological bias when extrapolating findings to individual adaptation patterns. Second, the included interventions were relatively short in duration (typically 7–8 weeks), which may be insufficient to capture long-term divergences in adaptation, particularly as athletes approach higher performance levels or potential physiological ceilings. Third, most included samples were male-dominated and involved bilateral barbell exercises conducted in controlled research settings. Therefore, generalizability to female athletes, unilateral training modalities, applied field environments, or sport-specific contexts may be limited. Fourth, publication bias analyses indicated potential small-study effects at the study level. Although this finding should be interpreted cautiously given the limited number of included studies, it may influence confidence in the magnitude of pooled effects. Fifth, baseline strength values were derived from exercise-specific 1RM measures, including both upper- and lower-body exercises. Although these values were standardized prior to analysis, combining different movement patterns may introduce construct heterogeneity, which should be considered when interpreting the findings. Although relative strength (1RM/body mass) may theoretically provide additional information regarding strength capacity independent of body size, body mass data were not consistently available across all included studies. Therefore, absolute baseline 1RM was selected as the primary strength indicator to maximize data availability and maintain comparability across studies. Sixth, only English-language publications were included in the review. This language restriction may introduce language bias and potentially limit the comprehensiveness of the evidence base. Finally, unmeasured confounders—such as nutritional status, recovery practices, genetic predispositions, and motivational factors—were not accounted for and may influence adaptation magnitude. Overall, while the present findings do not demonstrate statistically detectable moderation by baseline strength or training status, further well-powered and longitudinal investigations are required to clarify potential boundary conditions and individual-level response patterns. Future directions Future research should address several key areas. First, longer-term longitudinal interventions exceeding 10 weeks are needed to determine whether baseline strength influences adaptation patterns over extended training periods, particularly as athletes approach higher performance levels. Second, studies involving more diverse populations—including female athletes, youth cohorts, and sport-specific contexts—are required to improve external validity and clarify potential population-specific responses. Third, mechanistic investigations incorporating advanced neuromuscular and physiological assessments (e.g., high-density EMG, muscle architecture imaging, or metabolic profiling) are warranted to better understand how velocity loss thresholds regulate internal load and adaptation processes across individuals with different strength characteristics. Conclusion Velocity loss threshold-based resistance training produces moderate within-group improvements in strength and power outcomes. In the present meta-regression, no statistically significant association was observed between baseline strength and adaptation magnitude, and no interaction was detected between baseline strength and categorical training status. These findings indicate that current evidence does not demonstrate differential adaptation patterns based on initial strength level or training status. However, the absence of statistical significance should not be interpreted as definitive evidence of equivalence across populations. Given the limited number of studies and short intervention durations, further research—particularly longer-term and individual-level investigations—is required to clarify potential boundary conditions and mechanistic pathways underlying adaptations to VL-based training. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.