Does ice affect healing after muscle injury? Ask me again next century

Debate over the benefits and harms of icing acute muscle injuries remains unresolved. Some contend that ice is ineffective or even harmful, while others promote cryotherapy as a universal remedy. Centrists, often academics, call for more high-quality randomized controlled trials (RCTs) to resolve the issue. This viewpoint reframes the debate around 3 key points: first, although ice produces analgesia, evidence for sustained pain relief, beyond the immediate post-treatment period, is limited; second, findings from human physiological studies suggest that effects on healing—beneficial or detrimental—are likely confined to superficial injuries in lean athletes; and third, given the limited progress in experimental research to date and the challenges of recruiting participants in the acute phase of injury, definitive evidence on ice’s effect on healing is unlikely to emerge in the foreseeable future. By exposing the physiological and methodological limits of existing research, this viewpoint urges clinicians to apply ice with realistic expectations, researchers to prioritize more feasible and targeted questions, and policymakers to recognize that definitive evidence on ice’s effects is unlikely to emerge in the near future.

1. Analgesic benefits: Immediate but ephemeral

Ice rapidly reduces skin temperature, and it is well established that this produces an immediate local analgesic effect. However, transient numbing does not translate into sustained pain relief. Excluding post-surgical studies—where pain outcomes are often confounded by concurrent use of oral analgesics—only 2 pilot trials^1,2 have directly compared the sustained analgesic effects of ice to no treatment following closed soft tissue injuries (n = 19 with calf muscle strain;¹ n = 30 with acute ankle sprain²). The largest effect favoring ice that was reported beyond the acute stage of injury (Day 6 post injury), was just 0.2 points on a 10-point scale (95% confidence interval (95%CI): −3.0 to 2.6),¹ corresponding to a standardized mean difference of 0.07, or a 2% change—an effect unlikely to be clinically meaningful in any population.

If there is a sustained reduction in pain from topical cooling it is likely to be indirect, mediated through proposed mechanisms such as reduced tissue metabolism or inflammation at the injury site. Critically, these downstream effects depend on achieving meaningful reductions in deep tissue temperature—something that has yet to be convincingly demonstrated in human studies.

2. Can we cool human muscle enough to matter?

There is ongoing debate around ice’s effect on inflammation and healing, particularly in relation to muscle injury. A 2024 critical review of 26 animal studies found evidence that icing post injury reduces cell metabolism and inflammation.³ The subsequent concern is that this could attenuate recovery after soft tissue injury, particularly if ice is used over a prolonged period. But concern is unfounded if the data derived from animal models are not ecologically valid to an injured human. It is now clear that the average human subject experiences modest reductions in muscle temperature compared to animal models.

Fig. 1 displays 79 data points representing the lowest mean muscle temperatures reported across 35 original studies, comprising 72 healthy human and seven injured animal intervention groups (Study data are summarized in Supplementary Table 1). Animal models consistently reported intramuscular temperatures lower than 15 °C (median = 13 °C, range: 8‒13.8 °C). Despite using similar modes (crushed ice or cold-water immersions) and durations of cooling (median = 20 min, range 5‒60 min), the median temperature reported from healthy human models was more than double (median = 29.5 °C, range: 15.0‒36.6 °C), that of the animal models, indicating that humans experienced around half of the cooling effect. Furthermore, 3 quarters of interventions (54/72) failed to cool human muscle below 27 °C.

Fig. 1.

Lowest mean muscle temperatures reported (Human vs. Animal models). It displays 79 data points representing the lowest mean muscle temperatures reported across 35 original studies, including 72 human and 7 animal groups treated with icing (Supplementary Table 1); x-axis is on log₁₀ scale for visualization.

To the best of my knowledge, no studies have directly measured intramuscular temperatures in injured human participants undergoing topical cooling. Fig. 2 illustrates how both probe depth and the thickness of overlying adipose tissue markedly influence the degree of muscle cooling observed in healthy individuals. In the 5 instances where a cooling intervention reduced average human muscle temperature below 20 °C (depicted by the solid blue dots), 3 consistent features were observed: temperature measurements were confined to superficial muscle regions (less than 1 cm beneath the adipose tissue), participants were predominantly male, and subcutaneous fat at the cooling site was low (mean adipose thickness ranging from 0.3 to 0.7 cm). Such minimal adipose levels are characteristic of an athletic phenotype; for example, male marathon runners exhibit average adipose thicknesses of 0.3 cm at the calf and 0.7 cm at the thigh,⁴ whereas data from a general population sample (n = 184, aged 18–78 years) report mean thigh adipose thicknesses of 1.4 cm in males and 2.2 cm in females.⁵

Fig. 2.

Temperature reductions in human muscle based on depth and overlying adipose. It displays data points representing the lowest mean muscle temperatures reported across 22/28 original studies in healthy humans; n = 6 did not provide adequate detail on participants’ adipose thickness (Supplementary Table 1). Data are jittered to avoid overlapping of data points. Vertical reference lines were added to indicate key x-axis values; the dashed line labelled “male marathon” depicts average thigh muscle adipose in male marathon runners (as reported by Friedrich et al.⁴); the dashed line labelled “Male/female gen. pop” depicts the mean thigh muscle adipose in a sample of male and female subjects from the general population (as reported by Torun et al.⁵). Horizontal reference lines were added to indicate key y-axis values; the solid line labelled “1 cm” depicts the approximate depth of muscle injuries involving the myofascia (as reported by Svensson et al.⁶); the solid line labelled 1.6 cm depicts the average muscle injury depth (as reported by Svensson et al.⁶). gen. pop = general population; temp = temperature.

A temperature probe at 1 cm below the adipose tissue is positioned approximately at the superficial myofascial layer of muscle (see horizontal reference lines on Fig. 2).⁶ The chart shows that this region cools most readily, especially when overlying body fat is minimal. However, injuries affecting this superficial layer are relatively uncommon. Svenson’s⁶ analysis of 59 lower limb muscle injuries reported a mean injury depth of 1.6 cm (see horizontal reference line labelled on Fig. 2). At this depth and beyond, achieving therapeutic temperature reductions through topical cooling becomes increasingly difficult. Indeed, most studies assessing muscle temperatures at 1.6 cm depth or greater, failed to reduce average intramuscular temperatures below 30 °C, casting doubt on the plausibility of cryotherapy exerting a direct physiological effect on a substantial proportion of muscle injuries. Another caveat is that all existing evidence on topical muscle cooling comes from healthy participants. Since acute injury elevates local temperature and perfusion, the reductions shown in Fig. 1, Fig. 2 likely represent a best-case scenario. The net effect of elevated baseline temperature on cooling efficiency remains unknown.

Debates surrounding the efficacy of cryotherapy often focus on secondary mechanisms—such as inflammation and metabolism—while overlooking a fundamental physiological constraint: if cooling fails to meaningfully lower the temperature at the site of injury, then any proposed downstream effects (whether positive or negative) become theoretical. A recent suggestion by Racinais et al.³ to “apply cryotherapy with caution” is judicious and appears grounded in the animal data generated from their review. However, if we assume that a threshold level of tissue cooling is required before physiological effects ensue, and that this threshold is akin to the large reductions observed in animal models (Fig. 1), then such effects are unlikely to occur in humans, except for the rarer cases involving lean athletes with superficial injuries. To better understand the relevance of cryotherapy, and the level of caution required when applying, fundamental questions remain: are the therapeutic effects of cooling governed by an all-or-none temperature response? Or do they follow a more linear dose–response pattern?

3. Why we may never get a definitive answer on ice?

As icing is a physiotherapeutic intervention, robust conclusions on its effectiveness are best informed through well-designed, adequately powered experimental studies. However, progress in this area has been remarkably slow. Since the first controlled trial investigating post-injury cryotherapy was published in 1976,⁷ aggregate data from just 682 participants with closed acute soft tissue injuries (excluding exercise induced muscle damage) have been analyzed across 15 randomized controlled trials,^2,8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 an average recruitment rate of approximately 13 participants per year, over 5 decades of research. For context, a 2020 Cochrane review included 20 RCTs examining non-steroidal anti-inflammatory drugs (NSAIDs) for acute soft tissue injuries, analyzing aggregate data from 3305 participants,²¹ almost 5 times the number included in the entire cryotherapy literature to date.

Ice research seems even more paltry when we consider that, across 15 RCTs and 682 participants with acute soft tissue injury, fewer than 3% (19/682) presented with a muscle injury, and only 3 trials^1,2,10 compared ice with no ice. The remaining studies were focused almost exclusively on ankle sprains, pursuing narrower questions—such as dosing schedules or the additive effects of compression or exercise—while bypassing the fundamental issue of whether icing is better than no treatment at all. This represents a critical error in research sequencing: the field has focused on adjusting application parameters and delivery methods before establishing ice’s basic efficacy. In doing so, 5 decades of work have produced a literature that is fragmented, underpowered, and largely silent on the central clinical question.

A reasonable starting hypothesis for future work in this area, is that any true effect of icing on muscle healing—positive or negative—is likely to be small or moderate at best. If we assume a small effect of icing (Cohen’s d = 0.2) and standard α (5%) and β (20%) thresholds, a single adequately powered RCT in this field (with high quality methodology and no dropouts) would require 785 participants, with an additional 785 needed for independent replication. At the current recruitment rate of 13 participants per year, it would take 120 years to reach the required sample size and obtain a definitive answer—placing completion around the year 2145 (for science fiction fans, that puts us in the epoch between Alien and Avatar). If we instead assume a moderate treatment effect (Cohen’s d = 0.5), the required sample size drops to a more manageable 252 participants (including a replication study), but the projected completion date is still 2045.

The low total number of research participants in this field is perhaps unsurprising, given the narrow recruitment window in acute soft tissue injury studies. Identifying, consenting, and enrolling eligible participants within this brief timeframe is inherently challenging. Securing substantial government funding for this type of work is equally challenging; there is also little commercial incentive, perhaps because physics dictates that a simple bag of ice is the most economical and effective way to extract heat from the body.

On a positive note, for the most superficial soft tissue injuries—such as ligament sprains of the ankle, knee, or elbow (e.g., Medial Collateral Ligament)—meaningful cooling may be more readily achievable, making this a logical focus for short-term research. Similarly, in joint injuries, when ice can be used to complement therapeutic exercise, only modest reductions in tissue temperature may be required, although this approach also remains underexplored in RCTs.

This viewpoint sought to cut through the debate on the relative merits and harms of icing after muscle injury. A crucial point is that informed and fruitful discussions on clinical effectiveness are usually grounded in data from replicated, well-powered randomized trials. As such evidence remains distant, perhaps it is best to put this debate “on ice”—at least until the next century.

Declaration of competing interest

The author declares that he has no competing interests.