Are cold plunges good for recovery?
Healthy exercising adults; cold-water immersion after exercise
Very Low
High (randomised) −1 risk of bias (serious) −1 inconsistency (serious) −1 indirectness (serious) −1 imprecision (serious) → Very Low
How this level was reached
Each row is a recorded judgment with the reason given for it. The level is arithmetic over these — change a judgment and the level changes on the next read, with nothing stored to migrate.
Started at High (randomised) → Very Low
| Domain | Judgment | Reason given |
|---|---|---|
| Risk of bias | -1 serious | 4 primary studies assessed (Some concerns, Some concerns, Some concerns, Some concerns) |
| Inconsistency | -1 serious | Findings pull in opposite directions and vary by outcome and protocol — meta-analyses report only trivial-to-small benefits for perceived fatigue and performance, while the strength-training trial found cold-water immersion actively blunted anabolic signalling and long-term adaptation relative to active recovery. |
| Indirectness | -1 serious | Much of the body reports surrogate or mechanistic endpoints (inflammatory markers, anabolic signalling, thermoregulatory rationale) rather than recovery outcomes that matter to the exerciser, and one pooled synthesis is restricted to adolescent athletes while reviews note reliance on untrained participants and non-whole-body immersion protocols. |
| Imprecision | -1 serious | The primary trials are very small (20 junior soccer players, 24 active men, single-session crossovers) and even the largest pooled analysis contains only 357 participants, so effect estimates around a trivial-to-small mean could shift with a single further trial. |
| Publication bias | 0 not serious | The literature is not uniformly positive — null and adverse findings (blunted training adaptation, trivial pooled effects) are published — and there is no signal of selective industry-sponsored reporting. |
| Large effect | 0 no | Pooled effects are characterised as trivial to small, nowhere near the magnitude that would survive plausible bias on its own. |
| Dose-response gradient | 0 no | Protocol comparisons of differing immersion durations and temperatures did not establish a consistent gradient of greater cooling producing greater recovery. |
| Residual confounding | 0 no | The body is randomised, and in any case the plausible biases (unblinded participants expecting benefit from an unpleasant, salient intervention) would inflate rather than suppress the apparent effect. |
Not applicable: a randomised body starts at High and cannot be upgraded.
The risk-of-bias row is derived from the 4 primary studies, not asked again at body level — 4 read from full text. The rule that turns those into a downgrade is ours and unsigned.
Summary of findings
| Outcome | Importance | Certainty | Studies |
|---|---|---|---|
| post-exercise muscle soreness and recovery of muscle function | Critical | Very Low | 10 |
The body of evidence
10 studies entered.