The study protocol was registered on PROSPERO (Registration number: CRD420251171826). The systematic review and meta-analysis followed the PRISMA guidelines strictly (23). This ensured transparency and consistency. According to the classification by Ato et al. (2013), this study is a theoretical study (24). The methodology leader (ZY) independently designed the process and monitored quality to ensure scientific rigor.

Computer searches were conducted in databases including PubMed, Web of Science, Cochrane Library, and Embase. The literature search period spanned from the inception of each database to October 20, 2025. In addition to the electronic database searches, we manually screened the reference lists of all included studies and relevant systematic reviews to identify further eligible studies. A strategy combining subject headings with free-text terms was employed for literature retrieval. Using PubMed as an example, the search strategy is illustrated in Figure 1, while detailed search strategies for the remaining databases are provided in the Supplementary Materials.

The inclusion and exclusion criteria for literature were established according to the Participants, Interventions, Comparisons, Outcomes, and Study Design (PICOS) framework (25).

Search results were managed using EndNote X9 software, with duplicates removed by an independent reviewer (YZ). Subsequently, two reviewers (LL, TL) independently screened titles and abstracts against the inclusion criteria. Studies failing to meet eligibility standards were excluded, and full-text articles were retrieved for detailed evaluation. Any discrepancies were resolved through discussion to reach consensus, with a third reviewer (ZY) adjudicating if necessary. Two reviewers independently extracted and cross-checked the data; inconsistencies were verified by a third reviewer. In instances of missing or incomplete data, corresponding authors were contacted to ensure data integrity.

Literature data extraction was performed independently by two researchers using a double-blind method. A standardized template created in Microsoft Excel was employed to extract the following information: 1. Basic details: First author, publication date, study type, study population (sample size, gender, age, height, weight); 2. Study design: Exercise type and protocol, CWI protocol, outcome measures, intervention duration.

The risk of bias was assessed using the Cochrane Risk of Bias Tool 2 (RoB 2) (27). The assessment encompassed five key domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Disagreements were resolved through discussion; failing consensus, a third reviewer served as an arbitrator. Risk of bias visualization plots were generated using the robvis Shiny web app (28).

To supplement the quality assessment, the Physiotherapy Evidence Database (PEDro) scale was utilized (29). The PEDro scale has a total of 10 points. The criteria were: ≥6 points (high quality), 4–5 points (moderate quality), and ≤3 points (low quality). PEDro scores were employed to quantify bias risk and study design quality, thereby facilitating the interpretation of the robustness of the pooled results.

All statistical analyses were conducted using Stata-MP 18. To minimize confounding from baseline differences, change scores (i.e., the mean differences before and after the intervention along with their standard deviations) were used as the effect size input. Standardized mean differences [SMD (30)] were calculated, and Hedges' g was applied to correct for small-sample bias. All pooled effects are reported as Hedges' g with corresponding 95% confidence intervals. Effect sizes were interpreted according to Cohen's classification (31): large effect (g > 0.8), moderate effect (0.5 ≤ g ≤ 0.8), small effect (0.2 ≤ g ≤ 0.5), and trivial effect (g ≤ 0.2). Descriptive statistics for continuous variables are presented as mean ± standard deviation (mean ± SD). Statistical significance was set at a two-sided P < 0.05.

Heterogeneity was assessed using Cochran's Q test and the I² statistic. Significant heterogeneity was defined as p < 0.10 or I² > 50%, in which case a random-effects model was employed; otherwise, a fixed-effect model was applied. To explore potential sources of heterogeneity, stratified subgroup analyses were conducted based on immersion regions at specific follow-up time points (0 h, 24 h, 48 h, and 72 h). Sensitivity analyses were conducted using a leave-one-out approach to evaluate the robustness of pooled effects (32). Publication bias was assessed both qualitatively, via funnel plot inspection, and quantitatively, using Egger's regression test. Potential publication bias was considered present if Egger's test P < 0.05 or if funnel plot asymmetry was evident. To further validate the reliability of the results, the trim-and-fill method proposed by Duval and Tweedie was applied to adjust for possible publication bias and recalculate the effect sizes (33).

Based on the predefined search strategy, a total of 928 articles were retrieved from multiple databases. Through manual reference verification, an additional 6 articles were included, bringing the total to 934. After deduplication using EndNote X9 software, 363 articles were excluded, leaving 571. Following screening of titles and abstracts, 473 articles were excluded, resulting in 98 articles advancing to full-text review. During full-text review, 60 studies were excluded. An additional 8 studies were excluded due to unavailability of full-text. Ultimately, 30 studies met the inclusion criteria and were used for quantitative meta-analysis (Figure 2).

This study includes a total of 30 research articles, published between 2008 and 2025. The total sample size across all studies is 527 participants (498 male, 32 female), with sample sizes ranging from a minimum of 7 participants (34) to a maximum of 60 participants (35). The age range of participants spans from 15.6 to 50.3 years. All studies included are randomized controlled trials, with 15 studies employing a parallel-group design (3, 35–48), and 15 studies utilizing a crossover design (4, 34, 49–61). In the crossover studies, a washout period was implemented to minimize carryover effects, ensuring that participants' physiological states, muscle damage markers, and subjective symptoms returned to baseline or stabilized before entering the next intervention phase. This approach ensured the comparability of results across intervention periods (62).

Among the included studies, 12 focused on healthy, non-athlete participants (3, 4, 35–37, 42, 43, 47–49, 56, 60), while the remaining 18 involved professional athletes (34, 38–41, 44–46, 50–55, 57–59, 61). The distribution of studies by type of exercise is as follows: 12 studies investigated eccentric or eccentric-concentric resistance training (3, 34–37, 39, 41, 43, 48, 52, 56, 59); 15 studies focused on high-intensity interval training (HIIT), including sprint running, LIST/Yo-Yo tests, and simulated competition/training (4, 38, 44–46, 49–51, 53–55, 57, 58, 60, 61); and 3 studies examined aerobic endurance exercises (40, 42, 47). All studies reported adverse events, with specific details provided in Tables 1, 2.

While the RoB 2 assessment indicated a high overall risk of bias due to the infeasibility of blinding participants (Domain 2), the structural quality of the included studies remained high. All studies showed low risk regarding data completeness and selective reporting, and the majority (approx. 75%) employed blinded assessors to mitigate detection bias. Consistent with this, PEDro scores of 6–8 confirmed a generally high methodological standard. Therefore, despite the unavoidable limitation of performance bias, the synthesized evidence is considered reliable (Figures 3, 4; Table 3).

The acute and delayed effects of CWI at different body regions (whole-body vs. partial) on post-exercise muscle damage recovery were quantitatively evaluated through a systematic review and meta-analysis. The main findings were as follows: First, post-exercise CK levels were significantly reduced and DOMS was alleviated by CWI compared to seated rest. However, the restorative effect on neuromuscular function was limited. No significant improvement in MVIC was observed. A significant inhibitory effect on immediate post-exercise CMJ performance was shown. Notably, while the results for DOMS were robust, statistical significance for CK disappeared after adjustment using the trim-and-fill method. This suggested that its efficacy might have been overestimated due to publication bias. Second, subgroup analysis focusing on the “anatomical dimension” confirmed no statistical difference between whole-body and partial immersion across all outcome measures (P_m > 0.05). This result revised the traditional hypothesis based on hydrodynamic (hydrostatic pressure) and neurophysiological differences mentioned in the Introduction (19). It indicated that recovery benefits meeting the physiological threshold are generated as long as the immersion range covers the affected muscle groups. Whole-body immersion is not required. Finally, the benefits of CWI showed significant time-dependency. They were primarily concentrated within the 24-hour post-exercise window. Conversely, immediate post-exercise partial cold therapy showed significant inhibition of explosive power. This suggested that caution regarding the temporary impairment of neuromuscular function is needed during acute application.

A significant efficacy of CWI in reducing post-exercise CK levels was confirmed by this study. This is consistent with previous research findings (36, 63). CK is widely regarded clinically as a key marker for monitoring skeletal muscle damage. Elevated levels are typically associated with muscle cell membrane damage and myofiber leakage induced by strenuous exercise (64, 65). Mechanistically, the increase in membrane permeability is inhibited by single acute CWI through the rapid reduction of muscle temperature. Consequently, CK leakage into the blood is reduced (17). Simultaneously, physical pressure is exerted on the muscle by the hydrostatic pressure effect of immersion. Cell swelling is limited and metabolic waste clearance is promoted. The structural integrity of muscle fibers is further protected (60). Furthermore, the infiltration and diffusion of inflammatory cells and pro-inflammatory mediators (such as IL-6) are effectively inhibited by inducing local vasoconstriction. Acute inflammatory responses and tissue edema are alleviated. An optimal environment for muscle regeneration is created (37, 66). Subsequently, blood redistribution is facilitated during the reperfusion phase following vasoconstriction. Metabolic product clearance is accelerated and reoxygenation of affected tissues is improved. Nutrient and oxygen supply to the muscle is increased by enhanced vasodilation. The repair process is accelerated and normal physiological function is restored (67). Additionally, oxidative stress is triggered by strenuous exercise. Free radical generation is accelerated, exacerbating muscle damage (51). However, the activity of antioxidant enzymes such as superoxide dismutase and glutathione peroxidase is significantly enhanced by cold therapy. The clearance of excess reactive oxygen species is accelerated. Damage to muscle cells from oxidative stress is mitigated (54).

However, no significant difference was found between whole-body and partial CWI in reducing CK levels. This may be primarily because lower-limb dominant exercise protocols, such as plyometric jumps or shuttle runs, were employed in some of the included literature (37, 39, 40, 47). EIMD was mainly concentrated in muscle groups such as the quadriceps and triceps sura. Complete immersion of the affected target tissues was achieved by partial immersion with water levels up to the iliac crest. Local metabolic inhibition was sufficiently induced via transcutaneous thermal conduction (68). Although a larger central blood volume shift is indeed induced by whole-body immersion, the pressure gradient required to counteract gravitational edema and promote tissue fluid return to the lymphatic system is sufficiently generated by the hydrostatic pressure exerted on the lower limbs during partial immersion (69). Therefore, a “saturation effect” was produced by partial immersion for lower limb muscle damage. Further expansion of the immersion area did not translate into additional biochemical clearance benefits. Moreover, excessive systemic stress associated with whole-body cold exposure was avoided (70).

Significant and relatively robust efficacy of CWI in alleviating post-exercise DOMS was confirmed. It was consistently indicated by previous randomized controlled trials and systematic reviews that pain thresholds are increased and subjective pain sensation is reduced after CWI intervention (13, 26). DOMS typically peaks 24–72 h post-exercise. Its mechanism is primarily related to mechanical muscle fiber damage and the subsequently induced local inflammatory response and metabolite accumulation (71). It is commonly assessed clinically using Visual Analog Scale scores (72). The analgesic mechanism of CWI involves complex physiological and psychological multidimensional interactions. Local metabolic rate and oxygen consumption are reduced by vasoconstriction and tissue temperature decrease caused by cold stimulation. The accumulation of metabolites (such as lactate, prostaglandins, bradykinin) is reduced. Consequently, chemical pain stimulation is alleviated (26). Exudation of inflammatory cells and release of cytokines (IL-6, TNF-α, IL-1β) are inhibited by low temperature. Secondary tissue damage and pain signal transmission are limited (60). Most importantly, nerve conduction velocity is reduced and pain threshold is increased by cold stimulation, thereby directly relieving local pain. Tissue fluid return is promoted and edema and mechanical tenderness are reduced by hydrostatic pressure effects and hemodynamic changes (73). Furthermore, the analgesic effect of CWI stems partly from the immediate improvement of subjective recovery and emotional state. This subjective and emotional improvement is associated with sympathetic activation induced by cold exposure. It is manifested as increased plasma norepinephrine/epinephrine levels. Alertness and vitality are enhanced, producing a positive placebo effect (74, 75).

However, a clear time-dependent decay is presented by this analgesic effect. The analgesic effect of a single CWI session (partial or whole-body) tended to disappear 48 h post-exercise. This indicated that its residual effect is extremely limited. Physiologically, this period corresponds to the peak active period of muscle damage repair and the inflammatory cascade. Cytokine signaling and pain drivers are at a high stage at this time (76). Klich et al. further elucidated the distinction between acute and chronic alterations; they demonstrated that while CWI induced immediate improvements in pressure pain thresholds (PPT), these analgesic benefits did not uniformly persist into the delayed recovery phase, particularly across muscle groups with distinct fatigue characteristics (77). Consequently, the positive effects induced by a single acute CWI session are insufficient to counteract this continuous and complex physiological process over the long term (78). This finding suggests that the optimal window for a single cold therapy session is limited to the acute phase. If efficacy needs to be prolonged, a single intervention is clearly insufficient. Repeated immersion or combination with other recovery strategies needs to be explored in the future to alter the long-term repair trajectory (5, 79).

MVIC and CMJ are commonly used muscle strength recovery assessment indicators in sports science. They represent the recovery of maximal strength and explosive power, respectively (80, 81). It was shown in this paper that post-exercise MVIC of affected muscle groups was neither significantly improved nor attenuated by CWI. No significant difference was found at other follow-up time points. This result is consistent with the randomized controlled trial by Lee et al. (82). The recovery trajectory of MVIC was not altered by either whole-body or lower-body CWI. This indicates that the analgesic effect (perceptual recovery) of CWI is not synchronously translated into immediate functional improvement (83). The reason may be that low temperature cannot reverse excitation-contraction coupling impairment caused by ultrastructural damage to muscle fibers (such as Z-line disruption) in a short time (57). In contrast, a significant inhibitory effect on explosive power was produced by immediate CWI. The meta-analysis by Moore et al. also confirmed that dynamic/rate-dependent strength (such as CMJ, rate of force development, RTD) is more sensitive to CWI than static peak force (MVIC) (5).

Physiologically, tissue temperature is sharply reduced after acute cold therapy. Nerve conduction velocity (NCV) and motor unit firing rate are slowed. Muscle fiber contraction kinetics are altered. Consequently, RTD and instantaneous explosive power are impaired (84, 85). Empirical evidence from Dixon et al. (86) found that the short-term adverse effects of CWI on CMJ could be partially reversed by increasing muscle temperature through dynamic warm-up. This suggests that immediate performance impairment is closely related to acute changes in muscle temperature and nerve conduction velocity (87). Notably, subgroup analysis further confirmed no difference in functional impact between whole-body and partial immersion. As long as the immersion range covers the main working muscle groups, a cooling effect reaching the inhibition threshold is sufficiently induced by partial CWI (73). Therefore, caution should be exercised with both whole-body and partial immersion during rest periods involving explosive power output (12). Alternatively, sufficient active rewarming strategies must be combined to counteract the aforementioned physiological inhibition.

Although a rigorous systematic review process was followed, several limitations must be considered. First is the risk of bias. Due to the nature of CWI, blinding of subjects was impossible. This may introduce psychological expectations or placebo effects. The efficacy of subjective indicators (DOMS) may be particularly overestimated. Moreover, sensitivity analysis suggested potential publication bias for CK results. A conservative attitude towards its true effect size is required. Second is the high heterogeneity of DOMS (I² = 76.02%). Although it was reduced to a moderate level after excluding outlier studies, “multivariable meta-regression” could not be employed to further control for the interactive effects of covariates such as water temperature, duration, and exercise type due to the limited number of studies included in each subgroup. Third is taxonomic ambiguity. Although strict classification criteria based on anatomical landmarks (iliac crest) were established, studies with water levels ranging from the xiphoid process to the neck were still included in the “whole-body immersion” group. These subtle differences in anatomical coverage may have diluted the inter-group effects. Finally, the vast majority of included subjects were male athletes. Given the physiological differences in body fat distribution and thermoregulatory responses in women, the applicability of the conclusions to the female population remains to be verified. Furthermore, this study focused only on acute recovery (≤72 h). The potential interference effects of CWI on long-term training adaptations (such as muscle hypertrophy signaling pathways) were not discussed.

Based on this, future research should focus on: (1) Filling the gender data gap. Research targeting the female population is urgently needed to clarify the potential impact of the menstrual cycle on cold therapy efficacy. (2) Refining dosage comparison. Multi-arm trials directly comparing differences in water levels are recommended to establish optimal anatomical dosages. (3) Focusing on long-term adaptation. The long-term interference of different cold therapy regions on muscle strength and hypertrophy adaptation needs to be further elucidated. This will assist practitioners in balancing acute recovery with long-term training benefits.