1 INTRODUCTION

Worldwide, the rate of obesity has almost tripled since 1975 with over 650 million adults, or 13% of the population, having obesity in 2016.1 The rising incidence of obesity has resulted in more people classified with obesity (defined as a body mass index [BMI] of 30 kg/m2 or greater) before undergoing elective surgery. For example, in the United States, the rates of extreme obesity (BMI ≥40 kg/m2) among those undergoing knee arthroplasty have doubled from 1993 to 2003.2 Increasing obesity is problematic because people with obesity are at a greater risk of certain postoperative complications, with greater health and financial burdens to the patient and the health system.3 Specifically, obesity is associated with elevated risks for revision surgery,4 wound complications,4-6 venous thromboembolism,6 pulmonary emboli,4 urinary tract infection,5 with worse long-term patient-reported functional outcomes,7 reduced mobility7 and inadequate activity levels postoperatively; all of which increase risk of chronic disease. Consequently, health professionals recommend preoperative weight loss to improve postoperative outcomes. That said, obesity is not always associated with worse postoperative outcomes because for instance, lower mortality rates may occur among people with obesity when compared to people who are underweight.8 Thus, it remains unclear whether weight loss before surgery should be recommended.

Recently, some health services worldwide have restricted criteria for undergoing elective surgeries based on weight. For example, the National Health Service in England reported that 31% of Clinical Commissioning Groups, who are responsible for service provision, have at least one mandatory policy on BMI level and weight management before elective surgery.9 Similarly, the surgeons at Logan Hospital in Queensland, Australia declined to operate on people with obesity unless they lost 10% of their body weight.10 In this case, a dietitian-led presurgical weight management programme was implemented for their patients to achieve this target. Weight-loss diets are a safe option, with potential cost benefits to the individual and health-care system. However, little is known about the evidence in support of weight-loss diets before any elective surgery to decrease adverse events postoperatively in populations with obesity. Thus, the objective of this review is to determine the effect of preoperative weight-loss diets on postoperative clinical and service outcomes in people with obesity undergoing elective surgery.

2 METHODS

This systematic review was performed using methods from the Cochrane Handbook for Systematic Reviews of Interventions11 and according to the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines.12 It was registered prior to commencement on Open Science Framework in February 2020 (https://osf.io/dgf3t) and PROSPERO (CRD42020154074).

2.1 Inclusion criteria

Eligibility criteria for included studies are summarised in Table 1.

TABLE 1. Eligibility criteria for inclusion of studies Parameter Inclusion criteria Exclusion criteria Participants Adults aged ≥18 years BMI ≥30 kg/m2 Awaiting elective surgery including (but not limited to): orthopaedic procedures, cardiac surgery and gastrointestinal surgeries Bariatric or cancer-related surgeries were excluded because postoperative outcomes can be confounded by unique postoperative complications associated with such procedures or the underlying condition Intervention Weight-loss diets prior to surgery Including (but not limited to): dietary modification, caloric restriction, meal replacement Pharmacological weight loss Exercise alone Bariatric surgery as the weight-loss intervention prior to elective surgery Weight loss prior to bariatric surgery Comparator Eligible intervention comparators (control groups) included elective surgical waiting lists where participants underwent usual or standard care This may include receiving general advice about healthy eating provided by a preadmission clinic or GP Control groups that prescribed specific preoperative weight-loss interventions were excluded Outcomes

Primary outcome

Postoperative complications to 90 days

Secondary outcomes

Amount of weight loss Acute length of hospital stay Discharge destination Duration of inpatient rehabilitation Patient-reported outcomes for pain, function and quality of life Time take to return to work in any capacity Time taken to return to full work duties Study design Prospective studies including randomised controlled trials and non-randomised controlled trials, cohort studies Retrospective studies Studies retrospectively assessing registry data Studies published in languages other than English 2.2 Search strategy

A preliminary limited search of MEDLINE, CINAHL and Scopus databases was performed to identify relevant keywords contained in study titles, abstracts, and subject descriptors, and their synonyms to inform an extensive search strategy, informed by a librarian (Supplementary File 1). Electronic databases searched from inception to the 14th of May 2021 included: MEDLINE, EMBASE, Cochrane Database of Systematic Reviews (references from eligible reviews were reviewed for eligible studies), Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, Scopus, ClinicalTrials.gov and WHO International Clinical Trials Registry Platform (WHO ICTRP). There were no limits for the year of publication or publication status. Reference lists of included studies and grey literature were reviewed to identify additional studies.

2.3 Study selection

Search results were merged using EndNote and duplicate records were removed. Two reviewers (NP, RB) independently screened articles for relevance and excluded irrelevant articles based on titles and abstracts. Multiple reports of the same study were identified, with the most recent version included for review. Full texts of the remaining articles were independently assessed by NP and RB for inclusion against selection criteria. Disagreements were resolved through discussion with a third reviewer (JN).

2.4 Risk of bias assessment

Selected studies were independently assessed for methodological validity by two reviewers (NP, RB). The Cochrane Handbook's Risk of Bias (RoB) Version 2 checklist13 was used to assess individual outcomes from randomised controlled trials (RCTs) according to five domains of bias (randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result). A priori, it was planned to use the Cochrane Handbook ROBINS-I tool14 to assess the risk of bias in prospective non-randomised trials. Discrepancies between reviewers were resolved through discussion. All studies, regardless of methodological quality, underwent data extraction and synthesis where possible.

2.5 Data extraction

Standardised items in an Excel spreadsheet were used by two independent reviewers (NP, RB) to record the following extracted data from eligible studies: title; author; year of publication; journal; study design; setting; participant characteristics; recruitment procedures utilised; trial size; preoperative weight-loss intervention characteristics; details of the control; type of surgery; follow up or study duration; outcomes; outcome measurements; data analysis methods; details needed for risk of bias; author contact details; funding source. Data discrepancies were resolved through discussion. Three corresponding authors were contacted via email for further information and all responded. All analyses were performed on intention-to-treat data.

2.6 Assessment of heterogeneity

Clinical and methodological heterogeneity was assessed for diversity in participants, interventions, outcomes, study characteristics, and risk of bias for included studies to determine whether meta-analysis was appropriate. Statistical heterogeneity within each meta-analysis was assessed using the I2 statistic.15 Due to the small number of studies eligible for inclusion in this review the decision was made to include all studies in the meta-analysis regardless of the I2 statistic so long as the studies were similar with respect to clinical and methodological characteristics.

2.7 Assessment of publication bias

Published reports were compared against trial protocols to evaluate potential for publication bias. While assessment of publication bias was planned, there were insufficient studies (<10 studies) to construct a funnel plot or perform Egger's regression test.

2.8 Data synthesis

Data were analysed using Review Manager version 5.4.1 software from the Cochrane Collaboration.16 Differences between dichotomous outcomes are presented as odds ratios (ORs) and continuous outcomes are reported as mean difference (MD). 95% confidence intervals were calculated for both dichotomous and continuous outcomes. Random-effects meta-analysis was performed to pool outcomes from studies with similar characteristics. Due to the anticipated diversity of clinical and methodological characteristics of included studies, and because the influence of obesity appeared to vary, effect size was not assumed to be the same. Sensitivity analysis was not performed due to the small number of studies included. A priori, subgroup analysis was planned to compare participants who lost weight to those who did not, regardless of group allocation, however necessary data were not available. For continuous outcomes, pooled mean differences were calculated using the inverse-variance method. The Peto method was used for analysis of postoperative outcomes because it is the least biased and most powerful method for event rates below 1%.15 Data that could not be pooled are presented in tables with results summarised in text.

2.9 Assessing certainty in the findings

Two reviewers (NP, FG) applied the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach17 to evaluate the quality of evidence according to four levels (high, moderate, low, and very low) to quantify the degree of confidence in the reported results per outcome. A Summary of Findings (SoF) table was created using GRADEPro software (GRADEPro GDT 2015)18 (Table 2).

TABLE 2. GRADE summary of findings Weight-loss diets compared to usual care for patients with obesity awaiting elective surgery

Patient or population: patients with obesity awaiting elective surgery

Setting: community, outpatient setting (preoperatively), hospital inpatient (during surgery)

Intervention: weight-loss diets

Comparison: usual care

Outcomes Anticipated absolute effectsa (95% CI) Relative effect (95% CI) No. of participants (studies) Certainty of the evidence (GRADE) Comments Risk with usual care Risk with weight-loss diets Postoperative complications to 30 days post-surgery 78 per 1000 28 per 1000 (7–107) OR 0.34 (0.08–1.42) (p=0.14) 156 (3 RCTs)

⊕⊕◯◯

LOWb,c

The evidence suggests that weight-loss diets result in little to no difference in postoperative complications to 30 days post-surgery Preoperative weight change (%) The mean preoperative weight change (%) was −0.36% MD 6.07% lower (10.65 lower to 1.48 lower) (p=0.009) — 159 (3 RCTs)

⊕◯◯◯

VERY LOWc,d,e

Weight-loss diets may result in greater preoperative weight change (%) than usual care but the evidence is very uncertain Preoperative weight change (kg) The mean preoperative weight change (kg) was −0.58 kg MD 6.67 kg lower (12.09 lower to 1.26 lower) (p=0.02) — 159 (3 RCTs)

⊕◯◯◯

VERY LOWc,e,f

Weight-loss diets may reduce preoperative weight (kg) but the evidence is very uncertain Length of hospital stay post-surgery The mean length of hospital stay post-surgery was 41.9 h MD 3.72 h lower (10.76 lower to 3.32 higher) (p=0.30) — 156 (3 RCTs)

⊕⊕◯◯

LOWb,c

The evidence suggests that weight-loss diets result in little to no difference in length of hospital stay post-surgery Note: GRADE Working Group grades of evidence: High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect. Abbreviations: CI, confidence interval; OR, odds ratio; MD, mean difference. a The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). b Downgraded by one level due to high risk of bias in one study due to deviation from intended intervention and missing outcome data and some concerns in the selective reporting of outcomes in the remaining studies. c Downgraded by one level because the sample size does not meet the Optimal Information Size (OIS). d Downgraded by one level due to high risk of bias in one study due to deviation from intended intervention and missing outcome data. e Downgraded by one level due to inconsistency (high I2). f Downgraded by one level due to high risk of bias in the largest study due to selective reporting of outcomes and high risk of bias in one study due to deviation from intended intervention and missing outcome data. 3 RESULTS 3.1 Study inclusion

A total of 15 547 references were retrieved through searches of electronic databases with another 568 references from other sources. After excluding duplicates and irrelevant articles from titles and abstracts, 16 articles remained (Figure 1). Thirteen articles were excluded (Table S1) because three19-21 included patients with BMI less than 30 kg/m2, one commenced the weight-loss intervention in the postoperative period,22 one involved the intervention crossing over to the postoperative period,23 two were incomplete,24, 25 two were unpublished with no data available for inclusion,26, 27 two were retrospective studies,28, 29 one was an editorial commentary30 and one was an RCT with a weight-loss diet in both the control and intervention groups.31 Three RCTs met the eligibility criteria.32-34 No prospective non-RCTs were eligible for inclusion. No additional studies were identified by reference searching, yielding three articles that satisfied inclusion criteria and provided quantitative data for analysis.

PRISMA flow diagram

3.2 Methodological quality

Risk of bias assessments was completed for three outcomes: postoperative complications, preoperative weight change and length of stay (LOS) (Figure 2). One study demonstrated some concerns across all outcomes due to selective reporting of outcomes.32 Similarly, one study had some concerns for postoperative complications and LOS, and high risk of bias for preoperative weight change due to selective reporting of outcomes.34 One study demonstrated high risk of bias across all outcomes due to deviation from the intended intervention, missing outcome data and selective reporting of outcomes for preoperative weight change.33

Risk of bias assessments for postoperative complications, preoperative weight change and length of hospital stay

Regarding the quality of the dietary interventions, no studies clearly stated a dietitian was involved in designing the intervention. Whilst all studies assessed dietary compliance with various methods, none utilised a validated tool to assess participant dietary compliance.

3.3 Characteristics of included studies

The three studies included were RCTs published in English. Two studies involved participants undergoing general surgery32, 33 and one examined participants undergoing total knee arthroplasty for knee osteoarthritis.34 A total of 173 participants were recruited with 82 and 87 participants randomised to control and intervention groups, respectively, aged from 21 to 85 years, with a baseline BMI greater than 30 kg/m2. Primary outcomes differed between studies; however, each study recorded postoperative complications to different time points and weight change from baseline. Characteristics of the included studies are presented in Tables 3 and 4. Each of the three studies implemented a weight loss diet involving very-low-calorie diets (VLCD) via meal replacement formulas, with differing calorie restrictions.

TABLE 3. Characteristics of included studies Study Burnand et al.32 Hollis et al.33 Liljensøe et al.34 Study design RCT RCT RCT Country United Kingdom Australia Denmark Journal International Hepato-Pancreato-Biliary Association Inc. Nutrition & Dietetics Scandinavian Journal of Surgery Age 21–69 years Mean: 51.6 ± 13.1 years 46–85 years Gender (n)

Male (4)

Female (42)

Male (17)

Female (29)

Male (22)

Female (54)

Diagnosis (n)

Biliary colic (40)

Cholecystitis (4)

Obstructive jaundice (2)

Not specified Knee osteoarthritis Operation type (n) Laparoscopic cholecystectomy

Laparoscopic cholecystectomy (27)

Umbilical hernia repair (11)

Ventral hernia repair (5)

Inguinal hernia repair (3)

Total knee arthroplasty (TKR) Baseline BMI (kg/m2), mean (SD)

Intervention: 34 (3.44)

Control: 33.6 (3.35)

Intervention: 40.3 (6.0)

Control: 40.7 (5.9)

Intervention: 31.6 95% CI (30.6–32.6)

Control: 31.2 95% CI (29.8–32.6)

Trial size at baseline 46 50 77 Trial size at intervention 46 46 76 Intervention Very-low-calorie diet Very-low-calorie diet Low energy liquid diet Dietitian involved No (advice only) Yes Yes Control Standard care. Dietitian available for advice only Standard care Standard care Study duration 2 weeks preoperatively

8 weeks preoperatively

30 days postoperatively

8 weeks preoperatively

12 months postoperatively

Primary outcome Intra-operative time Feasibility of implementing a preoperative very-low-calorie diet weight programme for patients with obesity awaiting general elective surgery.

Body weight (kg)

Short-form 36 subscale Physical Component Score (PCS)

Secondary outcome

Weight change

Post-op complications

Length of stay (hours)

Day-case rates

Perceived difficulty of procedure by surgeon

Surgical complications

Weight (kg)

Waist circumference (cm)

Muscle mass (kg)

Fat mass (kg)

Quality of life (Impact of Weight on Quality of Life-lite questionnaire)

Intra-operative time (min)

Length of stay (days)

Short-form 36 subscale Mental Component Score (MCS)

Knee Injury and Osteoarthritis Outcome Score (KOOS)

6-Minute walk test

Fat mass (lean and bone)

Bone mineral density

Lipids

Length of stay (days)

Intra-operative time

Wound secretions at day 0

Blood pressure and heart rate

TABLE 4. Characteristics of weight-loss diets from included studies Study Burnand et al.32 Hollis et al.33 Liljensøe et al.34 Intervention Very-low-calorie diet Very-low-calorie diet Low-energy liquid diet Calorie restriction Total caloric intake of 800 calories/day 700–800 calories/day with ≥0.75 g/kg adjusted body weight protein

Preoperative phase: 810 calories/day

Postoperative maintenance phase: 1200 calories/day

Intervention details Participants were given a diet sheet to follow

Optifast meal replacement shakes

The consumption of three to four Optifast shakes mixed on water with an additional ≥two cups (non-starch) vegetable or salad, at least 2 L of energy free fluids, and one teaspoon of vegetable oil were recommended daily for 8 weeks

Preoperative phase: Formula diet (Cambridge Weight Plan®, Northants, UK) consisting of ready-to-use meal, bars, and sachets to mix with water or skimmed milk (7.5 dL a day) to make shakes, soups, or porridge, consumed four times a day

Nutritional education: weekly group sessions of 1.5 h led by an experienced dietitian for 8 weeks

Maintenance phase: Regular meals combined with one formula diet serving per day. Eight group sessions led by the study dietitian

Duration 2 weeks preoperatively 8 weeks preoperatively 8 weeks preoperatively and 12 months postoperatively Was a dietitian involved in administering the diet? No. Dietitians were not directly involved but were available to provide advice as needed Yes Yes Was a dietitian involved in the designing of the dietary intervention? Yes. A member of the research team was a dietitian Yes. The lead researcher is a senior dietitian No information Was there monitoring of adherence to the dietary intervention? All patients were asked to complete a detailed diary survey for the 2 weeks prior to surgery Adherence to the very-low-calorie diet programme was evaluated using the presence of urinary ketones in ≥50% of fortnightly samples collected by the clinic nurse over the 8-week period No information Was a validated tool utilised to measure dietary adherence? No information No No information 3.4 Review findings

The details for all primary and secondary outcomes are described in Tables S2 and S3. Meta-analysis was undertaken where data were available as the included studies were similar with respect to participant and weight-loss intervention characteristics and study design.

3.4.1 Postoperative complications

The primary outcome, postoperative complications to 90 days, was reported in one study with a follow-up period of 1-year post-surgery.34 Data for postoperative complications to 30 days was available from three studies with 156 participants. Random-effects meta-analysis did not show a sta