Two authors (Y-R and Y-BG) independently performed a computerized literature search from inception to May 15, 2024, using PubMed, Embase, and Cochrane Library. The search strategy included medical subject headings and free text terms (HGM OR beta-hydroxy-beta-methylbutyrate OR hydroxy methylbutyrate OR) AND (critically ill OR critical care OR intensive care) without language restrictions. The details of the full search strategy are summarized in Supplementary Data Sheet. The recruited articles were also examined in any of the eligible studies.

The inclusion criteria were as follows: (1) population: critically ill patients aged 18 years or older; (2) intervention: HMB was used alone or in combination with other supplements, regardless of dose, route of administration, treatment course, and exercise (HMB group); (3) comparators: placebo or other HMB-free supplements (control group); (4) design: randomized controlled trials (RCTs); and (5) outcomes: studies reporting any efficacy and safety outcomes as defined by each author of the included studies.

Two authors independently extracted the relevant data from the included RCTs. The data can be extracted from tables, figures, or text. These variables included study characteristics (first author’s name, year of publication, study design, and country), patient characteristics (age, gender, patient population, disease severity, and body mass index), HGM and control protocols, and predefined outcomes.

The primary outcome was all-cause mortality. Secondary outcomes included the duration of mechanical ventilation (MV), length of stay in the ICU or hospital, muscle measures (i.e., muscle thickness or area changes, defined by each author), and adverse events.

Y-R and Y-BG independently evaluated the quality of each included RCT using the Cochrane Risk of Bias tool (version 2) (25). Publication bias was evaluated using visual inspection funnel plots when ten or more studies were included. We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system to evaluate the quality of evidence (26). Disagreements between the two authors were resolved by consulting with a third author (H-BH).

The results from all relevant studies were combined to estimate pooled odds ratios (ORs) and associated 95% confidence intervals (CIs) for dichotomous outcomes. For continuous outcomes, we estimated the mean differences (MD) and 95% CIs as effective results. For studies that reported the median with an accompanying interquartile range (IQR) but not standard deviations (SD), we estimated the mean from the median and SD from the IQR based on the methods described in the Cochrane Handbook (27).

To analyze each predefined outcome, we conducted a meta-analyses of relevant trials. Meta-analyses were conducted when at least two studies could be pooled. To test the robustness of the outcomes and explore potential influencing factors, we performed sensitivity analyses to identify the influence of each study on the overall pooled estimate of the outcome of interest. Additionally, subgroup analyses were conducted separately by pooling studies based on (1) exercise: with or without; (2) location: Asia and non-Asia; (3) patient population: specific ICU patients or mixed ICU patients; (4) route of intake: oral or tube feeding; and (5) double-blind or single-blind study design.

The I² statistic was used to test for heterogeneity, with values of I² < 50% and I² > 50% indicating low and high heterogeneity, respectively. A fixed-effects model was used when I² > 50%, and a random-effects model was used when I² < 50%, using the Mantel-Haenszel method (28). The significance level for P-values was set at 0.05. Review Manager (version 5.4) was used for all analyses.

The search strategy identified 181 records from databases and additional searches. After removing duplicates, 117 records were available for title and abstract screening. Of these, 15 were retrieved for full-text screenings, with nine RCTs eligible for inclusion in the final analyses (19–23, 29–32) (Figure 1).

Tables 1, 2 present the main characteristics and study strategies of the included RCTs. These trials were published from 2006 to 2024, with study durations ranging from 999 weeks to 1 year. 99 participants were analyzed, 99 in the HMB group and 99 in the control group (19–23, 29–32). The included studies, except Wu et al. (32), were all single-center studies. These trials focused on unselected ICU populations (23, 29, 31, 32), severe trauma (20, 22), major surgery (30), and severe brain injuries (19). Five of the included RCTs used HMB as a single supplement (19, 21–23, 32), two used HMB in combination with arginine and glutamine (29, 30), and the remaining two used both HMB alone and the HMB combination formula (19, 21, 31). All studies provided 3g/day of HMB (1.5 g, twice a day). Three trials conducted exercises with HMB interventions including early electrical muscle stimulation (30), early rehabilitation (31), and resistance training (32). All studies reported on follow-ups, with the timing of outcome assessment varying between days 4 and 3 months post-intervention. The details in HMB strategies are summarized in Supplementary Data Sheet.

We evaluated the risk of bias in the included studies using the Cochrane risk of bias tool for RCTs (Supplementary Data Sheet). The risk of bias in the RCTs was low in all critical domains. The assessment of publication bias using visually inspected funnel plots showed no potential publication bias in the included studies (Supplementary Data Sheet). Using the GRADE methodology, we rated the evidence for pooled data for intubation rate, mortality, and ICU stay as moderate, moderate, and very low, respectively (Supplementary Data Sheet).

All-cause mortality was reported in five RCTs (20, 22, 23, 29, 32). We found no significant difference in mortality between the two groups (RR = 0.96; 95% CI, 0.44 to 2.08; I² = 0%, P = 0.92) (Figure 2). To investigate the sources of heterogeneity, we performed stratified analyses based on predefined main study characteristics and clinical conditions. Excluding any single study from the sensitivity analysis showed results similar to the overall combined OR (P-values ranged from 0.37 to 0.93, with all I² = 0%). Subgroup analyses were also conducted. These results suggested similar mortality risks when studies with predefined characteristics were pooled (Table 3).

The pooled estimates showed that compared with the control group, the HMB group showed no differences in the duration of MV (MD = −0.40; 95% CI, −0.91 to 0.12; I² = 64%; P = 0.13; Figure 3A) (20, 29, 31, 32), length of ICU stay (MD = −0.61 days; 95% CI, −3.59 to 2.38; I² = 93%; P = 0.69; Figure 3B) (20, 22, 29, 30, 32), and length of hospital stay (MD = 1.52 days; 95% CI, −1.18 to 4.22; I² = 81%; P = 0.27; Figure 3C) (20, 22, 29, 30, 32).

Three studies reported changes in body weight/body mass index after treatment. The pooled results showed that HMB did not affect the changes in body weight (MD = −0.47 kg; 95% CI, −1.96 to 1.01; I² = 0%; P = 0.53) or body mass index (MD = 0.21 kg/m²; 95% CI, −0.48 to 0.89; I² = 0%; P = 0.56). Two studies described changes in quadriceps thickness and suggested no differences between the HMB and control groups (MD = 0.52; 95% CI, −2.61 to 3.56; I² = 95%; P = 0.74). The outcomes of changes in skeletal muscle area (P = 0.95) and percenters in muscle loss (P = 0.16) were similar between the two groups. In addition, only one study by Meng et al. showed that the HMB group had a significantly lower muscle mass change score (P < 0.0001) than in the control group.

Our findings were unexpected, as previous studies have shown that HMB could safely mitigate muscle loss in older adults, cancer patients, and human immunodeficiency virus/acquired immune deficiency syndrome patients (14–16, 33). There is also evidence that HMB complexes can improve muscle strength in patients with postoperative malnutrition or rheumatoid cachexia (18). In addition to muscle measurements, HMB used after discharge in patients with chronic lung disease and heart disease significantly reduced mortality (34). Similarly, in another study, HMB administration to malnourished elderly patients reduced their 90-day mortality and improved their nutritional status (34). The success of HMB in other settings has stimulated several clinical attempts in critically ill patients (18).

However, our meta-analysis did not support the benefits of HMB alone or in combination with its complexes in critically ill patients. Most of the included trials focused on HMB’s effect on different clinical outcomes (20–23, 29–32), including mortality, and indicated no benefits from HMB. In the study by Kuhls et al., HMB resulted in a longer MV duration and ICU and hospital stays (20). Conversely, Norouzid et al. found that perioperative HMB supplementation significantly reduced hospital stay in patients undergoing cardiac surgery (30).

Approximately half of the included trials focused on the effects of HMB on muscle metabolism (19, 21–23, 29, 31). These trials evaluated different muscle metrics and most showed that HMB supplementation during the acute phase of critical care did not prevent muscle loss. Additionally, the included trials suggested that HMB can reduce inflammation and catabolism (21) and improve amino acid metabolism (19), nitrogen balance (20), and nutritional status (19). However, these effects were not due to a reducing in muscle protein turnover rates, which was initially hypothesized. A recent study found that a combination of resistance training and HMB, but not HMB alone, improved physical function and muscle strength in medical ICU patients but had no effect on muscle quality, quality of life, or 60-day mortality (32).

The negative results of this study may be attributed to several factors. Insufficient study power is a classic explanation for this finding. For example, most RCTs included too few patients to statistically detect the effect of the intervention. However, the negative results can also be due to different HMB strategies in the included trials, not just the statistical conditions. In this situation, increasing the sample size may not have an additional effect. These interventions are heterogeneous for rehabilitation exercises, HMB dosage, timing of outcome assessment, nutritional adequacy, and primary outcome. Exploring this heterogeneity is important because it not only helps us explain our results but also provides insights into future research designs that explore the clinical efficacy of HMB for critically ill patients.

Early HMB administration may prevent the development of sepsis-associated muscle dysfunction (10, 35). It blocks the reduction in protein synthesis induced by early sepsis (36). Theoretically, HMB administered after muscle weakness does not improve muscle function. There are several reasons why patients experience muscle weakness in the ICU. The muscle fiber cross-sectional area in ICU patients decreases by 1–2% one day. After seven days, the rectus femoris cross-sectional area decreased by 10.3% (3). Most of the included studies recruited patients on MV (20, 21, 29, 31, 32) or patients with severe trauma (19, 20, 22). This indicated that muscle dysfunction was present before enrollment. For example, in the study by Supinski et al. (31), patients were on MV for an average of 6 days before receiving HMB supplementation. This delay may have hindered the beneficial effects of treatment.

HMB was administrated at a dose of 3g/day in all included studies. This standard dose has been used in many previous studies of non-critically ill populations (14–16). It has shown beneficial effects on muscle function and tolerability without significant side effects (14–18). However, it is unclear whether the total dose of HMB prescribed for critically ill patients is appropriate. In this population, gastrointestinal dysfunction, fasting, and gastric decompression are frequent, which may impair drug absorption and limit the effectiveness of HMB in improving muscle function (37). To further evaluate the role of HMB, dose-response trials are required.

Three studies reported on early rehabilitation (29, 31, 32). One of these studies reported inadequate training (<10 min/day) (31). In the study by Wu et al. (32), resistance training (RT) was conducted throughout the hospitalization of medical ICU patients from ICU admission to discharge. This could explain the significant improvements in physical function, muscle strength, and physical activity observed in the RT and HMB + RT groups. However, HMB alone did not produce these effects, highlighting the importance of RT (32). Early rehabilitation is a non-pharmacological intervention that can directly or indirectly support muscle protein turnover (38, 39). It can benefit critically ill patients as it may reduce muscle wasting while enhancing muscle strength (40, 41). HMB alone may have limited effects due to restricted physical activity and insufficient exercise-induced stimulation of muscle protein synthesis (32). In another study focusing on the effects of HMB in ICU patients, the authors suggested that their negative results could be due to the active early rehabilitation applied in both patient groups (29).

HMB treatment should be on adequate nutritional therapy. However, only five included studies reported nutritional treatment regimens (20, 22, 23, 29, 32). Only one of these studies used indirect calorimetry for caloric intake (23), whereas the other four relied on weight-based predictive formulas (i.e., 20–30 kcal/kg/d) (20, 22, 29, 32). Thus, these patients may be at risk for over- or under-caloric intake. The protein doses used in these five trials were within the guidelines (i.e., 1.2–2.0 g/kg/d) (20, 22, 23, 29, 32). However, none of these protein regimens have addressed the need for individualized nutrition in critically ill patients. All groups in Kuhls’s study had a negative nitrogen balance during the study period (20). Viana et al. found that the average daily protein intake of both HMB and the control groups fell below 1.2 g/kg at day 4, with 0.90 g/kg/d and 0.82 g/kg/d respectively (23). Therefore, future research should focus on improving the effectiveness of HMB interventions by ensuring nutritional adequacy.

Four studies in the current meta-analysis combined HMB and other immune supplements, such as arginine, glutamine, and eicosapentaenoic acid (20, 29–31). It was impossible to confirm that such HMB complexes had any effect owing to the small sample size. Previous studies have shown that arginine and glutamine supplements may aid muscle synthesis and benefit critically ill patients (42, 43). In contrary, in their meta-analysis, Heyland et al. found that arginine supplementation increased mortality in critically ill patients (44). When analyzing studies with higher methodological quality scores, they also found that immune nutrition was associated with a significantly higher mortality rate (44). In one of the included studies, Kuhls reported that the average nitrogen balance for the control group, HMB combined with arginine or glutamine group, and HMB group was −9, −10.9, and −6.5 g/d, respectively (P < 0.05) (20). Interestingly, the HMB combination group maintained in a more negative nitrogen balance throughout the study (20). This suggests that the addition of arginine or glutamine to HMB offsets its benefits.

Our study is the first to evaluate HMB in critically ill patients. However, this meta-analysis had some limitations. First, all included trials had a small sample size and may have introduced bias. Second, the HMB strategies varied among the included studies, such as dosage, timing, route, and duration, which might also have prognostic value. However, the original trying to perform subgroup analyses to explore their influences according to such diversities was hindered by insufficient data. Third, only a few trials have reported the effects of HMB on muscles, resulting in a lack of statistical power. This study also failed to measure important metrics such as muscle strength, grip power, and functional status, which are crucial in treating patients in this population. Fourth, these studies used ultrasound rather than the traditional “gold-standard” method for measuring muscles, which requires careful interpretation. Finally, HMB could affect critically ill patients differently, depending on their cause. However, owing to the limited sample size in the existing trials, we could not fully analyze their effects related to specific causes.