Interventions to prevent and treat sarcopenia in a surgical population: a systematic review and meta-analysis

Abstract Background The aim of this systematic review was to summarize the results of trials evaluating interventions for the reduction of sarcopenia in patients undergoing surgery. Methods Searches were conducted using the Cochrane Central Register of Controlled Trials, MEDLINE and Embase. RCTs evaluating exercise, dietary or pharmacological interventions to address sarcopenia in the perioperative period were included. Treatment effect estimates were expressed as standardized mean differences (MDs) with confidence intervals, and heterogeneity was expressed as I2 values. Results Seventy trials including 3402 participants were selected for the data synthesis. Exercise interventions significantly increased muscle mass (MD 0.62, 95 per cent c.i. 0.34 to 0.90; P < 0.001), muscle strength (MD 0.55, 0.39 to 0.71; P < 0.001), measures of gait speed (MD 0.42, 0.05 to 0.79; P = 0.03), and reduced time for completion of set exercises (MD −0.76, −1.12 to −0.40; P < 0.001) compared with controls. Subgroup analysis showed that interventions in the early postoperative period were more likely to have a positive effect on muscle mass (MD 0.71, 0.35 to 1.07; P < 0.001) and timed tests (MD −0.70, −1.10 to −0.30; P = 0.005) than preoperative interventions. Treatment effects on muscle mass (MD 0.09, −0.31 to 0.49; P = 0.66) and strength (MD 0.46, −0.01 to 0.92; P = 0.05) were attenuated by the presence of cancer. Results of analyses restricted to nine trials at low risk of allocation concealment bias and fourteen trials at low risk of attrition bias were comparable to those of the primary analysis. Risk-of-bias assessment showed that most trials were at high risk of incomplete outcome and attrition bias, thus reducing the estimate of certainty of the evidence according to the GRADE assessment tool. Conclusion Exercise interventions appear beneficial in reducing the impact of sarcopenia. Because of the high risk of bias and low certainty of the current evidence, large RCTs using standardized measures of muscle mass should be undertaken.


Introduction
Sarcopenia, characterized by progressive and generalized muscle loss, is observed in over 20-40 per cent of patients recovering from major surgery, where it has been associated with higher rates of complications in sarcopenic compared with nonsarcopenic patients (45 versus 15 per cent), higher in-hospital mortality rates (23 versus 4 per cent) 1 , and longer hospital stay 2 . Sarcopenia is associated with advancing age and frailty, but can occur in younger patients 3 who have additional risk factors, including sedentary lifestyle, poor nutrition, chronic disease, and chronic inflammatory states 4 .
As the population ages, and the numbers of patients with frailty, multiple chronic conditions or cardiometabolic disease referred for surgery increases, sarcopenia will present an increasing challenge to clinicians and health systems. Research points towards a complex and multifactorial pathophysiology characterized by loss of mitochondrial function in skeletal muscle, chronic inflammatory changes, and exposure to oxidative stress 5 that may be modified by exercise, diet or pharmacological interventions. The aim of this review was to summarize the results of randomized trials of interventions aiming to attenuate sarcopenia in people undergoing surgery. The secondary aim was to evaluate the effectiveness of these interventions across a range of important subgroups defined by intervention type, age, disease type including cancer, and timing of the intervention. Finally, the strengths and limitations of different definitions of sarcopenia when measuring treatment effects in clinical trials were evaluated.

Methods
A systematic review of RCTs was performed according to the methods described in the Cochrane Handbook for Systematic Reviews of Interventions 6 . A protocol was registered prospectively on PROSPERO (CRD42020165325) 7 . The study adhered to PRISMA guidelines 8 .

Study eligibility
Studies were included if they fulfilled the inclusion criteria of RCTs in which an intervention was used to prevent or reverse sarcopenic changes in adult patients (over 18 years old) in a surgical population. Trials were excluded if they reported retrospective or observational studies, or included a significant proportion of people (over 50 per cent) with neuromuscular or neurodegenerative disease, cachexia, or chronic inflammatory conditions. Abstracts were reviewed and were included only if they were of high quality and adhered to CONSORT reporting criteria; recent studies have shown discrepancies between the data presented in conference abstracts and subsequent full-text publications 9 , or even between the abstract and main article text 10 . Furthermore, as the adherence of abstracts to the CONSORT reporting criteria has been reported as suboptimal 11 , these were only included if they adhered to CONSORT reporting guidelines.

Search methods
Electronic searches were conducted in the Cochrane Central Register of Controlled Trials, MEDLINE, and Embase using the following search terms: sarcopenia, muscle mass, dietary proteins, exercise therapy, testosterone or androgen or growth hormone and related terms. A full description of the search terms is available in Appendix S1. The final search was undertaken on 19 December 2019.

Study selection
Title and abstract screening were carried out independently by two authors using the Rayyan QCRI web app (Qatar Computer Research Institute, Hamad Bin Khalifa University, Doha, Qatar). Selected references were managed using Endnote TM X9 (Clarivate Analytics, Philadelphia, Pennsylvania, USA). Full-text screening was carried out and the reference lists of included papers were also screened for suitable articles. Excluded studies and the reason for exclusion were recorded. Disagreements were resolved by discussion or, where this was not possible, by a third author.

Assessment of risk of bias in included studies
Included trials were appraised using the Cochrane risk-of-bias tool version 8 12 . Two authors assessed each outcome of interest as being at either at low, high or unclear risk of bias for each domain. Disagreements were resolved as above.

Data extraction
Data were extracted by two reviewers and managed using Excel TM 2016 (Microsoft, Redmond, Washington, USA). This included year, study type, setting, sample size, participant demographics, baseline characteristics, type of surgery, details of interventions, outcomes, and risk-of-bias assessments. The primary outcome of this review was measures of sarcopenia, evaluated either by functional tests or imaging. A large variety of measures were used in the included trials; direct measures of muscle mass, such as cross-sectional area of lumbar spine or quadriceps muscle assessed by direct imaging (dual-energy x-ray absorptiometry (DEXA), CT, MRI), and muscle strength, such as hand-grip strength or equivalent, were included. Furthermore, in view of the definition of sarcopenia as a decline in muscle quantity and quality, and in keeping with an international consensus guideline 3 , the measures of muscle function were included. These can be broadly split into two categories: functional assessments which record the time taken to complete a specific task such as the timed-get-upand go (TUG) test and the sit-to-stand test, and time-based tests in which the numbers of metres walked or repetitions of an exercise were measured, such as the 6-minute-walk test (6MWT).
For the purposes of meta-analyses, these measures were grouped into four categories. Category A comprised measures of muscle mass, including appendicular skeletal muscle mass evaluated by DEXA, CT or MRI; whole-body skeletal muscle mass by DEXA, CT or MRI; mid-thigh cross-sectional area by CT or MRI; lumbar muscle cross-sectional area; bioelectrical impedance analysis; measurement of muscle thickness by ultrasonography; measurements of muscle volume by ultrasound imaging, CT or MRI; or measurement of skinfold thickness . Category B included  measures of muscle strength, including hand-grip strength,  quadriceps muscle strength, lower limb (any muscle group) resistance test, and upper limb (any muscle group) resistance test not including hand-grip strength. Category C comprised tests for the completion of set exercises including chair stand test (sit to stand), TUG test, and 10-m walk test. As these trials measure the time for completion of the exercises, beneficial treatment effects are negative in these trials. Category D consisted of repetition-based tests in which the number of metres walked or repetitions of an exercise in a set timeframe were assessed, including gait speed, 6MWT, and 30-s chair-rise test.
Secondary outcomes included self-reported quality of life, anaemia, mortality at 30 days, rates of readmission, duration of hospital stay, and rates of admission to a more intensive place of care either upon hospital discharge or from the previous place of residence.
Subgroup analysis was undertaken for type of surgery (orthopaedic, cardiothoracic, general, gynaecological, urological, bariatric, breast, transplant, and trauma), timing of the intervention (preoperative, perioperative, early postoperative, late postoperative), age of the participants (Aged 65 or under or aged over 65), cancer status of the participants.

Statistical analysis
Standardized mean differences (MDs) with 95 per cent confidence intervals and P values were estimated for treatment effect measures as continuous outcomes using an inverse-variance random-effects method. Risk ratios with 95 per cent confidence intervals were estimated for dichotomous outcomes using the Mantel-Haenszel random-effects method. All analyses were carried out using Review Manager (RevMan) version 5.3 (The Nordic Cochrane Centre, Copenhagen, Denmark).
The heterogeneity of treatment effects was explored using a prespecified subgroup analysis for the following criteria: type of surgery, cancer status, age, and timing of interventions. The test for subgroup differences in the Cochrane software was used to identify significant treatment-subgroup interactions. Sensitivity analyses excluded studies with high risk of bias in two domainsallocation concealment and incomplete outcome data-as it was predicted that these would be the most likely sources of bias in this review. Heterogeneity within each meta-analysis was explored by using a v 2 test with significance set at a P value of 0.10, and was expressed as percentage heterogeneity due to variation rather than to chance (I 2 ). An I 2 value of 0-40 per cent indicated no or mild heterogeneity; 41-80 per cent indicated moderate heterogeneity; and over 80 per cent represented severe heterogeneity.
Publication bias for the primary outcome was assessed using funnel plots, where 10 or more studies contributed to an outcome. The quality of evidence was assessed using Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology in GRADEPRO GDT software (https:// gdt.gradepro.org) 13,14 .

Assessment of bias
The results of the risk-of-bias assessments are shown in Fig. S1. Overall, the methodological quality of the included trials was low. A total of nine trials were at low risk of allocation concealment and fourteen trials were considered to be at low risk of attrition bias.

Data synthesis
For the primary analyses, the studies were grouped by one of the four methods of sarcopenia assessment. In addition, the treatment effect was also determined for each of the three therapeutic approaches to sarcopenia: exercise interventions, dietary  Table 1). When stratified by type of treatment, exercise (17 trials, 606 patients; MD  Fig. 3 and Table 1). In analyses stratified by type of treatment, exercise significantly increased muscle strength (18 trials, 670 patients; MD 0.55, 0.39 to 0.71; P < 0.001; I 2 ¼ 0 per cent), whereas dietary interventions did not (3 trials, 123 patients; MD 0.19, À0.17 to 0.55; P ¼ 0.30; I 2 ¼ 0 per cent). No pharmacological trials reported these outcomes.
Other treatment effects of interventions overall, and stratified by each of three main intervention groups are reported in Table 1. Adams et al. 15 (a) Adams et al. 15 (b) Battaglini et al. 17 Friedman-Bette et al. 29 Grapar Zargi et al. 33 Hasegawa et al. 35 Iversen et al. 40 Karelis et al. 42 Liao et al. 45 Lima et al. 46 Shaarani et al. 61 Suetta et al. 66 (a) Suetta et al. 66 (b) Takarada et al. 67 Tsukagoshi et al. 69 (a) Tsukagoshi et al. 69 (b) Dreyer et al. 26 Lattanzi et al. 44 Malafarina et al. 49 Nishizaki et al. 51   interventions, or for one trial of a pharmacological intervention ( Table 1). Summary treatment estimates for other secondary outcomes were limited by small numbers of trials reporting each outcome ( Table 1). No data were reported on levels of anaemia after surgery.

Subgroup analyses
The results of subgroup analyses for the primary outcome are reported in Table S2 For timing of the intervention (preoperative, perioperative, early postoperative, late postoperative) there was a treatmentsubgroup interaction with significant treatment effects for interventions in the early postoperative period aimed at increasing muscle mass (4 trials, 308 patients; MD 0.71, 0.35 to 1.07; P < 0.001; I 2 ¼ 49 per cent) and timed tests (7 trials, 367 patients; MD À0.70, À1.10 to À0.30; P ¼ 0.005; I 2 ¼ 59 per cent). There was also a treatment-subgroup interaction with significant treatment effects for late postoperative interventions aimed at reducing timed-test scores (2 trials, 65 patients; MD À0.64, À1.17 to À0.11; P ¼ 0.02; I 2 ¼ 0 per cent).

Sensitivity analysis
Sensitivity analyses for the primary outcome (Table S4) Table S4).

GRADE assessments
GRADE assessment judged the certainty of the effect estimates for exercise interventions on muscle mass or strength as low, and those for timed walking tests, or gait speed or equivalent assessments as moderate ( Tables 2-5). The certainty of the effect estimates for pharmacological or dietary interventions were low or very low for all measures of sarcopenia.

Discussion
This systematic review of RCTs of interventions that aim to reduce postoperative sarcopenia identified seventy trials, and the vast majority had important methodological limitations; only nine trials were at low risk of allocation concealment bias. Exercise interventions were shown consistently to improve measures of sarcopenia, whether defined by muscle mass, muscle strength, timed tests or gait speed, whereas dietary interventions did not. One analysis of two trials of pharmacological interventions suggested an improvement in muscle mass attributable to the interventions. Further subgroup analyses indicated that interventions in the early and possibly late postoperative periods were most likely to have a positive effect than those undertaken before surgery. Treatment effects were independent of age. Treatment effects on muscle mass and muscle strength, but not timed tests or gait speed, were attenuated by the presence of cancer. The findings of sensitivity analyses restricted to trials at low risk of allocation concealment bias or attrition bias were comparable to the results of the primary analysis.
The review used contemporary, standardized review methods to test a prespecified hypothesis described in a prospectively registered protocol. However, it is limited by the quality and small sample size of the included studies, and the potential heterogeneity of outcome measures included in the analyses. Where possible these outcome measures were grouped according to the type of sarcopenia measure: muscle mass, muscle strength, timed test results, or gait speed (or equivalent). The results provided useful insights into a clinical problem where progress is limited by a lack of standardization and consensus definitions of outcomes.
Exercise interventions, targeted at all patient regardless of age, in the early, or possibly late, postoperative phase could reduce sarcopenia. Postoperative exercise interventions are not part of standard perioperative care, presumably owing to gaps in knowledge. On this basis, a trial of postoperative exercise interventions may be warranted. Preoperative exercise, often as part of a prehabilitation programme, is increasingly being advocated as part of enhanced recovery for people undergoing surgery. However, only two trials with small sample sizes evaluating preoperative interventions were included in the present analysis. The evaluation of treatment effects by surgical specialty was also limited by small numbers of trials in many of the prespecified subgroups. These are further knowledge gaps identified by the present review. Evaluation of the evidence using the GRADE assessment tool identified the need to downgrade the certainty of evidence to low or very low, except for evidence presented for exercise interventions. Common reasons can be identified across all primary and secondary outcomes. First, there was a serious risk of bias across most trials. This was mainly attributable to the uncertainty around the blinding of participants, personnel and assessors across the included trials; as most of the interventions required that participants completed an exercise or nutritional programme, the lack of blinding of participants could have affected concordance and affected outcomes. Furthermore, as some assessments of muscle mass, such as the 6MWT, require volitional effort, lack of blinding could have affected the results of the assessments of such measures.
Another relevant source of bias was the large proportion of trials reporting incomplete outcomes, with high rates of participants excluded from the final analysis or lost to follow-up. This is a significant issue not only because of the volitional nature of the interventions, but also because exclusion of participants who had developed complications or required readmission to ICU may have led to significant overestimation of the benefits of intervention.
The impact of small numbers of participants on the certainty of the evidence was less severe in trials of exercise interventions, thus increasing confidence in the treatment effects of exercise programmes.
A tertiary aim of the analysis was to review the use of different methods for measurement of sarcopenia in RCTs. Muscle mass and muscle strength offer objective and highly reproducible measures of sarcopenia; however, these may miss more qualitative aspects, including changes in motivation or cognition, that may influence the outcomes of functional and semiquantitative measures such as timed tests or gait speed. Heterogeneity of effect across different measures of sarcopenia was observed in multiple analyses, but this provided limited insights into the best measurement for clinical trials. For example, in the cancer subgroup analysis, there was no treatment effect on muscle mass or strength for people with cancer; however, for measures of gait speed, a benefit was observed for patients with cancer.The reasons for this are unclear. One approach to establishing the value of different measures of sarcopenia is to look for associations with clinically important outcomes. This was not possible in the present analysis owing to the limited number of trials that reported any clinical outcomes. There was general consistency of treatment effects between primary and some secondary outcomes, with agreement between treatment effects on muscle strength and mass in most analyses (primary analysis, type of treatment, timing of intervention, cancer versus no cancer). Discordance in other analyses was generally an issue of precision of the estimate rather than concerning the direction of the treatment effect. These were short-term assessments of well-being, however, and the association between measures of sarcopenia and long-term outcomes and quality of life is a research priority identified by this review.
Although this systematic review of RCTs indicates that exercise interventions are likely to reduce the severity of sarcopenia after surgery, this issue should be evaluated further. Other areas of uncertainty identified by this work include the need for validation of commonly used measures with respect to long-term outcomes, the role of exercise intervention in patients with cancer, and the role of preoperative exercise interventions on sarcopenia and long-term outcomes.