Study selection
Our team initially searched 594 articles from three Internet databases. After removing the duplicates, 497 articles remained. YCZ and DHC then screened the titles and abstracts, deleting 469 irrelevant studies. A further 28 articles were screened for full-text assessment: six studies were reviews or case studies and four studies were conference papers. In addition, four studies were excluded for not reporting the association between sarcopenia and mortality, as they only provided the association between skeletal muscle index (SMI) as a continuous variable and mortality 29-32. Therefore, 14 publications were finalised for analysis. (Shown in FigureS1).
Study summary
There were 14 studies with a total of 3,249 patients included in our meta-analysis. All of the studies that were included were retrospective cohort studies, with the exception of one34, which was a prospective cohort study. A total of five studies was conducted in the U.S. 10, 23, 24, 33, 34; while China 21, 22, Japan 17, 35, and the Netherlands11, 18 each had two studies, and Korea20, Australia25, and Brazil19 each had one study respectively. All studies used CT to detect sarcopenia. There were several outcomes reported in our meta-analysis. Seven studies used in-hospital mortality10,11,18,17,21,23,34, four studies used 30-day mortality19,22,25,35, and three studies used 1-year mortality20,24,33. The proportion of males among the studies that were included ranged from 51.40% to 61.40% (Table 1). TableS2 showed the result of each study with adjusted covariates. The pooled prevalence of sarcopenia among critically ill patients was 38% (95% CI: 36% -39%; p=0.000; I2=95.5%) (Figure 1).
Study quality
None of the studies was a randomised controlled study, and study quality was relatively moderate, ranging from 5 to 8 points NOS (TableS3)
Mortality
All studies that were included used mortality as the primary outcome. Our study showed that critically ill patients with sarcopenia have an increased risk of mortality when compared to those without sarcopenia (HR=2.22, 95%CI: 1.79-2.75; P<0.001; I2=0.0%) (Figure2). In addition, pooled data showed a significantly high risk of in-hospital mortality in critically ill patients with sarcopenia, compared to non-sarcopenic patients (HR=1.97, 95%CI:1.47-2.64;P<0.001; I2=0.0%), 30-day mortality (HR=2.08, 95%CI:1.36-3.19; P= 0.001; I2=0.0%), and 1-year mortality (HR=3.24, 95%CI:2.03-5.16; P<0.001; I2=0.0%) (shown in Figure3). Meanwhile, we examined the minimum sample size required by trial sequential analysis for meta-analysis and found that 1,105 participants were required. There were 3,249 participants included from these 14 studies. In addition, as we can see the Z line has crossed both information size and conventional boundaries, indicating that the association between sarcopenia and all-cause mortality in our analysis was reliable and robust. (Figure S2).
Subgroup analyses
Reasons for ICU admission
Six studies clearly reported that the reason for admission to ICU was trauma10, 23, 33-35, while three studies indicated sepsis as the reason for admission11,17, 22. The patients in the other studies showed complex reasons for ICU department admittance18-21,25. Therefore, we performed a subgroup analysis of the reasons for admission, and found that patients with both sepsis and sarcopenia had an increased mortality risk when compared to patients without sarcopenia (HR=2.20, 95% CI:1.53-3.18; p<0.001; I2=0.0%). The results were similar among trauma patients (HR=1.89, 95%CI:1.33-2.69; p<0.001; I2=0.0%) and patients admitted for other mixed reasons (HR=2.75, 95% CI:1.84-4.10; p<0.001; I2=21.7%) (Figure4).
Different definitions of sarcopenia
There are several methods to measure skeletal muscle mass, including total skeletal muscle area or psoas muscle area, as well as masseter cross-sectional area. Detailed information, including muscle measurement and cutoff values, are shown in Table2. Therefore, we performed a subgroup analysis based on different measures to detect whether there was a difference. Our results showed that critically ill patients with sarcopenia had an increased risk of mortality, compared with non-sarcopenic patients, when using SMI to define sarcopenia (HR=2.11, 95%CI:1.59-2.80;P <0.001; I2=0.0%). In addition, we found similar results when using TPA (HR=2.96, 95% CI: 1.72-5.11; P <0.001; I2=59.7%) or SMA (HR=2.11, 95%CI:1.33-3.33; P =0.001; I2=0.0%) to define sarcopenia. However, the association between sarcopenia, based on masseter cross-sectional area, and mortality was not significantly different (HR=2.00, 95%CI:0.82-4.90;P=0.129). (Figure5)
Subgroup analyses according to region
Five studies were conducted in Asian populations17,20-22,35, and nine studies in Western populations10,11,18,19,23-25,33,34. Therefore, we performed a subgroup analysis based on geographical region. The results showed that critically ill Asian patients with sarcopenia have an increased risk of mortality, compared to critically ill Asian patients without sarcopenia (HR=2.86, 95% CI:1.99-4.11; P =0.001; I2=0.0%). Similar results were found in Western populations (HR=1.93, 95% CI:1.48-2.52; P =0.001; I2=0.0%). Figure6 summarises the results.
Subgroup analysis by age
As age is an important confounding factor, we performed a subgroup analysis based on two age groups (more than or equal to 70 years versus less than 70 years). The results showed that the association between mortality and sarcopenia was observed in both of these age groups (HR =2.25, 95%CI:1.57-3.21; P <0.001; I2=0.0% versus HR=2.20, 95%CI:1.68-2.88; P =<0.001; I2=24.9%, respectively). See FigureS3.
Publication bias and sensitivity analysis
The results of Begg’s and Egger’s tests show no significant bias (P=0.760, P= 0.991, respectively) (Figure S4). In addition, sensitivity analysis results show that the pooled result did not result in significant change after one study was omitted each time (Figure S5).
Overall evidence quality
Our study indicates the quality of evidence was low due to the risk of bias, indirectness, and imprecision (TableS4).