A systematic review and meta‐analysis of the effectiveness of meal replacements for weight loss

Summary Meal replacements (MR) are generally not recommended in clinical guidelines for the management of obesity. The aim of this review is to provide an up‐to‐date systematic evaluation of the effect of weight loss interventions incorporating MR compared with alternative interventions on weight change at 1 year in adults with overweight or obesity. Six electronic databases were searched from inception to the end of August 2018 for randomized controlled trials comparing the effect of MR with interventions that did not include MR on weight at 1 year. We excluded studies using diets providing <3347 kJ/(800 kcal)/day and those which used total diet replacement (TDR) from this review. Risk of bias was assessed using the Cochrane risk of bias tool. Twenty‐three studies with 7884 adult participants were included. Six out of 23 studies were judged at low risk of bias across all domains, and 5/23 studies were judged at high risk of bias in at least one domain. Studies with similar intervention and comparators were grouped into five comparisons for analysis. Mean weight change at 1 year favoured the MR group relative to the control group in each comparison. In those comparisons where we conducted meta‐analysis, in people assigned to a diet incorporating MR, mean difference was −1.44 kg (−2.48 to −0.39 kg; I 2 = 38%) compared with alternative kinds of diets. In those assigned to a MR diet along with support, mean difference was −2.22 kg (−3.99 to −0.45, I 2 = 81%) compared with other diets with support and −3.87 kg (−7.34 to −0.40; I 2 = 60%) compared with other kinds of diet without support. In those assigned a MR diet with an enhanced level of support, mean difference was −6.13 kg (−7.35 to −4.91, I 2 = 19%) compared with alternative diets and regular support. Programmes incorporating meal replacements led to greater weight loss at 1 year than comparator weight loss programmes and should be considered as a valid option for management of overweight and obesity in community and health care settings.


| INTRODUCTION
Obesity is a condition of excess accumulation of body fat that causes premature mortality. It also causes substantial morbidity, including significantly increased risks of type 2 diabetes, cardiovascular disease, and several nonsmoking-related cancers, as well as physical impairments linked to excess weight such as breathlessness, joint problems, and back pain. 1,2 Nevertheless, the health risks of obesity are offset by weight loss, [3][4][5][6] with greater weight losses associated with greater health benefits. 7 The prevalence of excess weight is high and rising throughout the world; 8 thus, unless effective obesity treatment and prevention strategies are implemented, the increasing incidence of preventable diseases will place a growing burden on health systems and the global economy. 9 There is a pressing need to identify effective interventions to treat obesity that can be delivered at scale. Weight loss programmes that incorporate self-management strategies, such as the use of meal replacements (MR), may be a particularly useful approach, as they can be delivered within the community, potentially without support by health care professionals.
For the purposes of this review, we defined MR as discrete foods, food products, or drinks that are used to replace foods usually consumed at one or more meals with the intent to reduce daily energy intake for the purposes of achieving weight loss or weight maintenance following weight loss. MR programmes replace at least one meal per day and include at least one meal comprising conventional foods. This definition not only includes products purposefully marketed as MR for weight loss such as soups, shakes, and bars but also includes portion-controlled ready meals as well as discreet portions of readily available conventional foods such as breakfast cereal or rice. This review does not include very-low-energy diets (VLED) providing less than 3347 kJ (800 kcal)/day, or those that use low-energy total diet replacement (TDR), which use formula food products to replace all meals and snacks, but the impact of these types of programs on weight has been reviewed by others recently. 10,11 A previous systematic review and meta-analysis of interventions that incorporated MR for weight loss, using a similar definition as here, identified only six studies of MR programmes versus comparator. It is reported that the use of MR significantly increased short-term weight loss (3 months) compared with control. However, at 1 year, there was no statistically significant difference in weight change between the groups, though the point estimate favoured the intervention. 12 Uncertainty over the long-term effectiveness of MR may explain why many national guidelines do not reference MR, [13][14][15][16] or advise against their use 17 for the routine management of overweight and obesity.
However, in recent years, several further studies of the effects of MR with longer term follow-up have been conducted. We therefore set out to conduct an up-to-date systematic review of randomized controlled trials to examine the effectiveness of MR in people with overweight and obesity, compared with interventions that do not include MR, on weight loss at 1 year. The effects on cardiometabolic risk and any adverse effects were also investigated. We included interventions intended for the treatment of overweight or obesity and considered suitable for use in the community, without medical supervision.

| METHODS
A protocol for this review was published in advance and implemented without changes. 18

| Search strategy
MEDLINE, Embase, PsycINFO, CINAHL, Web of Science, and the Cochrane Controlled Register of Trials (CENTRAL) were searched from database inception to August 2018. We also screened references from systematic reviews identified through our search and requested papers from authors for those we were unable to obtain or for which we had abstracts only. The searches were not restricted by country or language.
The search strategy (MEDLINE) is included in Supplementary Table S1.
Studies were included if they recruited adults (≥18 years) with a body mass index (BMI) ≥25 kg/m 2 . Studies in pregnant women, people with eating disorders, those who had undergone bariatric surgery, or in those that included concomitant use of pharmacotherapies for the purposes of weight loss were excluded. We included randomized controlled trials of interventions incorporating the use of one or more MR daily, as part of a hypoenergetic diet intended for weight loss. We excluded interventions in which daily energy intake was restricted to <3347 kJ (800 kcal)/day, as these diets (VLEDs) require medical supervision and have been systematically reviewed previously. 10 Studies that used formula TDR were also excluded. To be included, studies must have had as a comparator either a weight loss intervention that did not include MR or offered no or minimal intervention. Studies which compared different types of MR but did not include a control arm were excluded. Studies were required to report participants' weight at least 1 year after enrolment.
Covidence (Cochrane) was used for abstract and full-text screening and data extraction. 19 Data were extracted on weight and adverse events at all reported time points as well as on fasting glucose; insulin; HbA1c; total, LDL, and HDL cholesterol; triglycerides; and systolic and diastolic blood pressure at 1 year. Studies were excluded if they did not present sufficient information to allow data extraction or quality assessment and if this information was not available in a trial protocol or provided to us on request by the authors. Initially, titles and abstracts were assessed for inclusion by two independent reviewers with disagreements resolved by discussion or by a third reviewer. Full papers of the included studies were obtained and independently assessed for inclusion by two reviewers, with disagreements resolved by discussion or by consultation with a third reviewer.

| Data extraction
Two reviewers independently used a prespecified and piloted data extraction form to characterize the population, intervention, control groups, and outcomes and to assess risk of bias. Disagreements were resolved by discussion or by a third reviewer. Authors were contacted for missing data and clarifications as required. Risk of bias was assessed using the Cochrane risk of bias tool, 20 consistent with the methods used in previous reviews of weight loss interventions. 21 If the outcome assessors were not blinded, we judged studies at low risk of detection bias where the outcome was objective (eg, body weight measured by study staff). Studies were judged at high risk of attrition bias if fewer than 50% of participants were measured at follow-up or if there was a difference of ≥20% drop-out rate between groups, for the primary outcome of weight at 1 year from baseline.
We categorized interventions including MR, and their comparators, to ensure that we compared effectiveness within meaningful groups. There were three types of active intervention. The first was stand-alone dietary advice which recommended the use of MR to replace one or more eating occasions with the remainder of the diet provided by conventional food (MR diet). The second included interventions which recommended the use of MR and also provided participants with additional support to lose weight (MR diet + support).
If the "support" offered was at a higher intensity than the support offered to the comparator group, we called this enhanced support, and this made up the third intervention group (MR diet + enhanced support). There were three types of comparators. The first comprised advice on dietary changes to aid weight loss that did not include the use of MR (diet only). The second was broader support for weight loss, which include advice on diet and other behavioural strategies, though the details may not be specified (diet + support). The third type of comparator intervention offered a nominal intervention for weight loss, either in the form of a printed leaflet or one-off educational lecture (minimal intervention).

| Analysis
Meta-analyses were conducted in Review Manager 5.2 (Cochrane) 22 to examine the difference between intervention and comparator groups for weight change at 1 year (primary outcome). Additionally, where available, we extracted interim weight change (~3 months, typically at the end of the main intervention) and weight change at 2 and 4 years after baseline. Further analyses were conducted to examine the changes in biomarkers of diabetes and cardiovascular risk at 1 year. Studies varied in whether and how they imputed data for participants lost to follow-up, and as the practices adopted for this can impact on the reported effect size, we recalculated weight change using the method of baseline observation carried forward (BOCF) to reduce spurious heterogeneity. 23 In studies that did not report baseline weight for the randomized participants, we assumed that there was no difference in baseline weight between those randomized and those who completed the study unless it was stated otherwise. In one study, it was unclear how many participants were lost to follow-up in each group, and we assumed equal drop-out for the purposes of the BOCF calculation. 24 A sensitivity analysis was conducted removing this study from the comparison.
Because the definitive numbers of participants achieving ≥5% or ≥10% weight loss were inconsistently reported in the included papers, we estimated these numbers based on the mean and standard deviation of the reported weight loss, assuming a normal distribution as a post hoc analysis.
All meta-analyses used a random effects model to account for differences in the intervention programmes and populations. Pooled results were calculated as mean differences in kilograms (kg) with 95% confidence intervals (CI) for continuous variables; weighted means were used to report individual group means. For the proportion of participants achieving ≥5% or ≥10% weight loss, pooled results are reported as odds ratio with 95% confidence intervals. The I 2 statistic was used to quantify statistical heterogeneity. 25 We used Cochrane guidance for interpretation of I 2 : 0% to 40% might not be important, 30% to 60% may represent moderate heterogeneity, 50% to 90% may represent substantial heterogeneity, and 75% to 100% considerable heterogeneity. 20 Where a study contributed more than one intervention arm to a meta-analysis, the control group was divided equally between interventions to avoid double counting in the pooled result.
In the preregistered protocol, we planned to use meta-regression to investigate whether characteristics of the MR interventions impacted the observed effect. However, there were few studies in each meta-analysis, and the data reporting the details of the programmes were limited such that we judged that the outcomes would not be robust, and this was omitted from our analysis.

| RESULTS
The literature search identified a total of 2924 papers. One hundred eighteen full papers were retrieved, and 24 papers reporting the results of 23 randomized controlled trials (RCTs) with 8253 participants were eligible for inclusion. The main reason studies were excluded is that they did not report body weight at 1-year follow-up or longer, and this information could not be obtained from authors.
For some studies, additional data were obtained from the authors. [26][27][28] The PRISMA flow chart is displayed in Figure 1

| Risk of bias
Summary risk of bias for the included studies is provided in Table 2, and justifications for any unclear of high judgments are provided in the supplementary material (Table S3). Six of the included studies were judged to be at low risk of bias across all domains. 33,38,41,44,45,48 Five studies were judged to be at high risk of bias in at least one domain, 24

| Weight change
Twenty-three studies were included in the primary analysis (weight change at 1 year) across the five comparisons, defined based on  (Table 4).

| MR diet vs diet only
Six studies compared a MR diet with diet only. 24

| MR diet + support vs diet + support
Ten studies (12 intervention arms and 10 control arms) were included in the MR diet + support versus diet + support comparison. 27 One study was omitted from the meta-analysis as weight data were not normally distributed, were reported in the paper as median and interquartile ranges, and we could not obtain further data. 45 In the nine stud-

| MR diet + support vs diet only
Two studies compared interventions involving a MR diet and support programme (n 95) with a diet-only comparator (n 95). 36,38 Mean weight change at 1 year was −5.3 kg for the MR diet + support and

| MR diet + enhanced support vs diet + support
Two studies tested a MR diet together with an enhanced support programme. 35

| MR diet + support vs minimal intervention
Three studies (four intervention arms) compared a MR diet plus support with a minimal intervention control. 39,41,44 Statistical heterogeneity was considerable (I 2 = 98%); therefore, we do not present pooled results here. In all three studies, participants in the intervention arm lost significantly more weight than those in the control arm at 1 year. In a study in China (n = 88), the mean difference was ( Figure 2). In the latter study, the mean difference in weight change   (Figures 4 and 5).
The other two studies did not report weight change beyond a year.

| Participants achieving >5% and >10% weight loss at 1 year
For each comparison, we calculated the odds ratio that participants would achieve a weight loss ≥5% and ≥10% from baseline at 1 year in the MR group compared with the control (

| Biochemical outcomes
There were very limited data on biochemical outcomes with only 12 studies reporting at least one biochemical outcome. 28,[32][33][34][38][39][40][41][42]44,46,47 Across all studies that reported outcomes, results consistently favoured meal replacement groups for HbA1 C . For the results for all other biochemical outcomes (glucose, insulin, lipids, and blood pressure), results were mixed and rarely reached statistical significance. Further detail can be found in Table 3 and Figures S1 to S9.

| Adverse events
Only two studies reported information on adverse events (AE). 26,41 Neither of these studies gave details on the total number of AEs reported by intervention arm, but the Look Ahead Research Group 2007 reported that there were no between-group differences in the frequency of hypoglycaemia, fractures, amputations, congestive heart failure, or occurrence of gallstones among 5145 participants. 50 Flechtner-Mors et al reported that there were no adverse events that could be attributed to the intervention. 26 We also considered nutritional deficiencies to be an AE. Five studies assessed diet quality during the interventions, and these studies reported that diet quality was improved in those randomized to MR. 29   This review expands on the indicative findings of a previous review. 12 By including more studies, and reducing the confidence interval around the point estimate, the present findings demonstrate that the weight loss at 1 year in interventions that include MR products is statistically significantly greater than in interventions that do not use these products. Other recent reviews have included weight loss interventions incorporating meal replacements. 49,51 but these reviews have considered the use of meal replacements alongside other weight loss interventions, and have therefore been unable to determine the specific effect of meal replacements per se. TDR have also been reviewed in various populations. 10,11,52 Although TDR products are like MR in that they are designed to replace usual food-based meals, TDR products are formulated to be used as the sole source of nutrition, that is, to replace all food in the diet. MR aims to replace one or more usual meals, but the instructions for use suggest eating a food-based meal at least once a day to ensure adequate intake of nutrients and vitamins. MR can be purchased over the counter and used without guidance from a health professional, whereas weight loss programmes incorporating TDR are advised to be undertaken under supervision. Given these differences between MR and TDR programmes, it is important to consider their effectiveness independently.
The key finding of this review is that participants assigned to a MR diet compared with a diet only approach (akin to a self-directed weight loss attempt) lose an additional 1.44 kg at 1 year, and this difference appears to be maintained up to 4 years. Likewise, many people join behavioural weight loss programmes to provide support to lose weight. The results of the present study suggest that all other things being equal, incorporating meal replacement products into such behavioural weight loss programmes enhances their effect (reflected in the MR diet + support versus diet + support comparison), with a 2.22 kg greater weight loss at 1 year in the MR group. Although these differences may appear to be modest, currently around 70% of the UK population with a BMI of 30 kg/m 2 are trying to lose weight, 53 and a sustained weight loss of even 1 kg will bring substantial benefits to public health. 54,55 Moreover, since the comparative interventions also lead to weight loss, by and large incorporating MR into weight loss programmes increases the proportion of participants who achieve ≥5% and ≥10% weight loss at 1 year.
Meal replacement products are consumer food products (as opposed to medical devices or drugs), widely available to be purchased over the counter without prescription, but at present are infrequently used by those attempting weight loss in their weight loss attempts. 56 Advice and guidance provided by clinicians to encourage their use could deliver benefits at minimal cost to health care providers. Few studies included in the review reported adverse events, including the impact on the nutritional quality of the data. Two studies found no evidence of adverse events arising from the use of these products, 26,41 and five studies report that they may improve the overall nutritional quality of the diet. 5,6,18,19,35 The greater weight loss observed in programmes incorporating MR suggests that this approach makes it easier to adhere to a reduced energy diet. However, there has been little detailed study of eating behaviours. It is possible that these fixed-energy, portion-controlled, or prepacked foods contain less energy than most self-selected meals and snacks, or that the structure and external control associated with their use facilitate adherence. Further work could usefully examine the behavioural processes that facilitate weight loss which would be of use when designing effective weight loss programmes to enhance effectiveness and reach.
These findings provide important new evidence to inform clinical guidelines. At present in many countries, clinical recommendations for the treatment of obesity advise that individuals attempting weight loss should be advised to aim for an energy deficit of 2092 to 4184 kJ/day (500-1000 kcal), but they do not suggest that MR could be used to help individuals achieve this deficit nor do they recommend MR as an effective dietary strategy for weight loss. 13,15 Guidelines in some countries do refer to meal replacement programmes. In the United States, the guidelines state that the strength of evidence on the longer term effect on weight of MR is low. 57 Singaporean guidelines note that meal replacements may induce greater acute reductions in weight but are not advised for the long-term management of overweight and obesity. 17 Australian guidelines do not recommend meal replacements for weight loss beyond their use as part of a VLED. 16 In conclusion, this review provides evidence that MR is an effective intervention for the treatment of overweight and obesity at a year, especially when used together with a support programme. MR could be recommended for inclusion in weight management programmes offered by professionals or as part of a self-management strategy for people with overweight or obesity.