Cold water immersion protocol optimization across exercise modalities: a systematic review and network meta-analysis of resistance training, endurance exercise, and team sport applications

Abstract Background Cold water immersion (CWI) is a common method of post-exercise recovery, but the optimal protocols are not well understood. The lack of exercise-specific guidelines has caused confusion on how CWI should be implemented in various training settings. Purpose To identify optimal CWI protocols in resistance training, endurance exercise, and team sport modalities and measure modality-specific effects on performance and physiological recovery indicators. Methods We conducted a systematic search of four databases (PubMed, Scopus, SPORTDiscus, Web of Science) of randomized controlled trials investigating the use of post-exercise CWI (≤ 20 C) in healthy athletes between January 2014 and June 2025. CWI protocols were classified based on temperature (cold: 5–8 C, moderate: 9–12 C, mild: 13–20 C ) and length (short: less than 10 min, medium: 10–15 min, long: more than 15 min). Random-effects network meta-analyses were performed using standardized mean differences (Hedges’ g) when 10 or more studies provided an outcome. Primary outcomes were modality-specific performance measures: strength and power in resistance training (confirmatory); aerobic performance and recovery in endurance exercise (confirmatory); power and strength in team sports (confirmatory). Secondary outcomes were muscle damage biomarkers and perceived recovery measures across modalities (exploratory). Surface Under the Cumulative Ranking Curve (SUCRA) values were used to evaluate treatment rankings based on effect estimates and 95% credible intervals. Results Eighty-seven studies comprising 2,313 participants were considered. Passive recovery was highest in resistance training strength (24 studies, 598 participants, moderate certainty) with worse results in CWI protocols of medium duration and moderate temperature (Hedges’ g = 0.00, 95% CrI: -0.38 to 0.38 vs. passive). Cold-temperature long-duration protocols showed potential benefits in resistance training power (17 studies, 389 participants, LOW certainty, Hedges’ g = 0.80, 95% CrI: 0.03 to 1.57, SUCRA = 85.0%). In endurance aerobic performance (11 studies, 235 subjects, LOW certainty), cold-temperature medium-duration protocols were the most favored (SUCRA = 75.1%) but with uncertain effects (Hedges’ g = 0.17, 95% CrI: -0.71 to 1.05). In endurance recovery (17 studies, 463 participants, LOW certainty), moderate-temperature protocols exhibited non-significant trends (Hedges’ g = -0.47, 95% CrI: -0.97 to 0.04 vs. passive). In team sport power (18 studies, 418 participants, LOW certainty), cold-temperature long-duration protocols were effective (Hedges’ g = 1.00, 95% CrI: 0.10 to 1.89, SUCRA = 93.3%). Secondary exploratory findings demonstrated that moderate-temperature medium-duration protocols positively affected muscle damage biomarkers across modalities (37 studies, 862 participants; Hedges’ g = -0.51, 95% CrI: -0.86 to -0.16) and perceived recovery (47 studies, 1,114 participants; Hedges’ g = -0.66, 95% CrI: -0.98 to -0.33). Heterogeneity between studies was moderate to high across networks (τ = 0.00-0.51), and 69–85% of comparisons were based on indirect evidence using star-shaped network structures. Conclusions Ideal CWI protocols vary significantly among exercise modalities (MODERATE to LOW certainty evidence). Passive recovery seems to be more desirable than CWI when it comes to resistance training aimed at increasing strength without disrupting adaptive responses. In endurance exercise, moderate-temperature exercise protocols (9–12 C, 10–15 min) can offer recovery effects, but the evidence is of low certainty. Uses of team sports demonstrate fluctuating impacts necessitating personalized strategies. These results question universal CWI application strategies and argue in favor of modality-specific protocol choice, but significant evidence gaps exist, especially in relation to female athletes (85% male subjects), long-term adaptations, and head-to-head protocol comparisons. Protocol registration INPLASY202570101 (registered July 24, 2025, prior to literature search).