Obesity affects 13 % of adults worldwide and is a public health concern(1). In low- and middle-income countries, wealthy people are most affected. However, the reverse holds true in high-income countries(Reference Kim and von dem Knesebeck2). Addressing obesity in high-income countries is crucial since obesity disproportionately affects lower socio-economic groups and perpetuates health inequalities(Reference Bentley, Ormerod and Ruck3,Reference Templin, Tiago and Thomson4) .
Adult obesity is linked to chronic diseases such as cardiovascular diseases (CVD), diabetes, disability, cancers and raises mortality(Reference Stringhini, Carmeli and Jokela5,Reference Ameye and Swinnen6) . Chronic illness and weight discrimination limit employment and social prospects, resulting in a vicious cycle of health disparities and lost income(Reference Wimalawansa7). A modest one-kilogram weight loss can lower metabolic risk, delaying disease development(Reference Morris, Jebb and Oke8). Furthermore, treating obesity complications consumes 8·4 % of a country’s health budget(9). Considering these repercussions and secular trends, reducing obesity is promising in reducing the socio-economic and healthcare burden on society(Reference Peeters and Backholer10).
Low-income people are under-represented in obesity literature(Reference Harvey and Ogden11). It is hypothesised that the lack of economic and cultural capital hinders engagement in weight loss programmes(Reference Pampel, Krueger and Denney12). Besides, they may achieve poorer behavioural change outcomes following universal treatments according to the ‘Inverse Care Law’(Reference Watt13), leading to intervention-generated inequities(Reference White, Adams and Heywood14).
Behavioural strategies targeting diet and exercise are the cornerstones of affordable weight management(Reference Olateju, Ogwu and Owolabi15). Existing research on the effectiveness of adult obesity interventions covered strategies such as goal setting with self-monitoring(Reference Willmott, Pang and Rundle-Thiele16), group interventions(Reference Robertson, Avenell and Stewart17) and personalised feedback(Reference Sherrington, Newham and Bell18). However, lower socio-economic groups were not sampled; thus, findings may not be generalisable. Furthermore, there have been no previous comparisons between various interventions. Technology-based interventions employing webpages, mobile applications or tele-consults were also evaluated but findings were variable with little evidence of long-term efficacy(Reference Jahangiry and Farhangi19,Reference Hartmann-Boyce, Jebb and Fletcher20) .
To date, only one review focused on disadvantaged populations(Reference Hillier-Brown, Bambra and Cairns21); however, studies were outdated, featured middle-income countries and did not examine the impact on CVD risk factors. Moreover, the authors included observational studies with poor internal validity. Personalised lifestyle adjustments and community-based nutrition education were beneficial for short-term weight loss, but the authors did not specify strategies for commissioning in future trials. There were no objective pooled assessments available to determine the best interventions. As many trials have been published after this review, this review evaluates current evidence on the effectiveness of behavioural weight loss interventions for low-income populations based in high-income countries. We also examined the effects of weight loss interventions in attenuating CVD risk.
Methods
This review followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines(Reference Page, McKenzie and Bossuyt22). The review protocol was registered with PROSPERO (ID: CRD42022331776). No ethical approval was required as no individual patient-level data was sought.
Searches
We searched databases (PubMed, Web of Sciences, Cochrane Central Registry of Controlled Trials and EMBASE) randomised controlled trials (RCT), from November 2011 onwards to 1 February 2022 and re-ran on 1 May 2023. Additional references from previous reviews were screened to retrieve relevant studies. Only publications from November 2011 onwards were selected to provide an updated evidence base from the last review(Reference Hillier-Brown, Bambra and Cairns21).
Keywords and MeSH terms of two main concepts ‘weight loss interventions’ AND ‘low income’ were adapted for each database. The search was filtered to adults and RCT published in English. There were no limitations based on sex or sample size. Each study was checked against the World Bank Economic Classifications to confirm it was based in a high-income country. The full search strategy is presented in online Supplementary Materials Table S1.
Eligibility
We considered studies to be eligible for inclusion only if they were RCT with concurrent controls, comprised adult participants (≥ 18 years) and evaluated behavioural interventions on overweight or obese individuals. Surgical and pharmacological treatments were beyond the scope of this meta-analysis. Trials were performed for ≥ 12 weeks (inclusion criterion used in Cochrane to avoid effect size exaggeration). The PICOS (Population, Interventions, Comparators, Outcomes, Study designs) method was used to identify studies in compliance with the eligibility criteria (Table 1).
Inclusion and exclusion criteria for eligibility of studies
Data extraction and synthesis
Data relating to source, demographics, interventions and results categorised into primary and secondary outcomes were extracted. Two independent reviewers (PL, YX) independently screened study titles, abstracts and full texts for studies to include. Reviewers (PL, YX) were blinded to each other’s decisions. Our screening resulted in 105 papers with a Cohen’s kappa of 0·80, indicating high inter-rater reliability between the two reviewers. Any disagreements that surfaced were discussed and resolved by a third reviewer (AW). No individual-level data were sought. Attrition was considered by extracting effects from intention-to-treat analyses. Attempts to obtain missing data from authors were futile.
Risk of bias assessment
Two authors (PL, YX) independently assessed the methodological quality of studies using the Cochrane Risk of Bias (RoB 2·0) tool(Reference Sterne, Savović and Page23). Each study was rated on five domains: selection (randomisation and allocation), performance (blinding of observers), attrition (incomplete outcome data), detection (blinding of outcome assessments) and reporting (selection of results). Disagreements were resolved by discussion or consultation with a third author. Studies were classified into: (i) low RoB, all criteria met; (B) some concerns, one or more of the criteria partly met; (C) high RoB, one or more criteria not met (online Supplementary Fig. S1(a)). High RoB studies were excluded from meta-analyses.
For each behavioural intervention strategy, the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) tool was used to assess the certainty of evidence, graded from low to high. Risk of bias, inconsistency (unexplained heterogeneity), indirectness (differences in population, intervention and outcome measures or indirect comparison), imprecision (uncertainty of results) and publication bias were considered(Reference Guyatt, Oxman and Vist24).
Statistical analyses
All statistical analyses were conducted in Review Manager Version 5.4.1 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Meta-analyses were conducted for overall effectiveness of behavioural strategies. For specific strategies identified during data extraction, we pooled studies with similar strategies for subgroup analyses. For studies with multi-component interventions, we also identified the key components to perform subgroup analyses for specific distinctive strategies. Only studies with low RoB or of some concerns were pooled.
For three-arm trials (two interventions with one control), intervention groups were combined using the in-built calculator in RevMan by combining the mean and effect size across both interventions to obtain the summary effect size. This is to generate a single pair-wise comparison with overcome unit-of-analysis errors(Reference Higgins, Thomas and Chandler25). Studies with multiple publications were reported singularly.
As outcomes were continuous variables measured on the same scale, means and standard deviations (s d) or confidence intervals (CIs) at baseline and end of intervention were extracted to calculate the pooled mean difference (MD) with its 95 % CI, using inverse variance method. Random-effect models were used as heterogeneity was assumed a priori due to the diversity of intervention components and comparator conditions. Fixed-effect models were not performed due to inter-study differences in participant demographics and settings.
Forest plots were utilised to identify effect estimates and CI. Study heterogeneity was assessed with Cochran’s Q tests and I-squared (I 2) statistics, where I 2 of ≤ 25 %, ≤ 50 % and ≤ 75 % correspond to small, moderate and large heterogeneity(Reference Higgins, Thompson and Deeks26). The difference in effect between groups was analysed at the 5 % significance level.
We inspected publication bias asymmetries using funnel plots when there were a minimum of 10 studies and ascertained the results by Egger’s regression test.
Subgroup and sensitivity analyses
Subgroup analyses were undertaken for sex, intervention duration and BMI as independent moderating variables. Leave-one-out sensitivity analyses were employed to determine the impact of individual studies on the overall effect. If heterogeneity in the meta-analysis was moderate or high, outliers with non-overlapping 95 % CIs were removed. Additionally, for analyses with heterogeneity ≥ 50 %, study characteristics were explored for an explanation. As we aimed to demonstrate the effectiveness of weight loss interventions, high RoB studies were excluded in the meta-analysis as these adversely impacted the heterogeneity of the findings. To address any potential concerns regarding the exclusion of studies without sufficient justification, we have performed sensitivity analyses to demonstrate the robustness of our results (online Supplementary Materials).
Results
The search identified 755 unique publications. 195 duplicates were removed using EndNote (X9) software, and 455 were excluded following title and abstract screening. The remaining 105 papers were screened for full-text eligibility, leaving 14 papers. The screening process and reasons for exclusion are presented in Fig. 1.
PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart of the study selection process
Study characteristics
Characteristics of included studies are summarised in online Supplementary Table S2. The final selection of 14 trials included a total of 3618 adults, with a mean age of 40·2 years (sd 9·7) (range: 24·2–53·4 years), and BMI of 33·6 kg/m2 (sd 2·8) (range: 28·8–38·9 kg/m2). Studies (n 14) recruited samples from community settings (n 8), primary care settings (n 2), a mixture of community/primary care settings (n 3) or from work sites (n 1). All studies provided evidence for recruiting subjects in lower socio-economic groups.
Trials lasted from 16 weeks to 12 months, with three studies providing further follow-up from 22 weeks to 12 months(Reference Balk-Moller, Poulsen and Larsen27–Reference Rosas, Thiyagarajan and Goldstein29). Controls either received standard/usual care or were wait-listed.
Nine RCT were conducted in the USA(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Foley and Levine28,Reference Bennett, Steinberg and Askew30–Reference Van Name, Camp and Magenheimer36) , The remaining were conducted in Australia(Reference Hutchesson, Callister and Morgan37,Reference Lombard, Harrison and Kozica38) , Denmark(Reference Rosas, Thiyagarajan and Goldstein29), Scotland(Reference Dombrowski, McDonald and van der Pol39) and the UK(Reference McRobbie, Hajek and Peerbux40).
Eight RCT recruited female subjects(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Steinberg and Askew30–Reference Mayer, Vangeepuram and Fei33,Reference Van Name, Camp and Magenheimer36–Reference Lombard, Harrison and Kozica38) and five with mixed-gender subjects (females > 65 % of the population)(Reference Bennett, Foley and Levine28,Reference Rosas, Thiyagarajan and Goldstein29,Reference Phelan, Hagobian and Brannen34,Reference Samuel-Hodge, Garcia and Johnston35,Reference McRobbie, Hajek and Peerbux40) . One RCT recruited male subjects(Reference Dombrowski, McDonald and van der Pol39).
As studies integrated multi-component approaches, they were grouped into nine distinct strategies (CP = Calorie Prescription; DC = Dietary Counselling; FI = Financial Incentives; GS = Goal Setting; IF = Interactive Feedback; LC = Lifestyle Coaching; PA = Physical Activity; SM = Self-Monitoring; SS = Social support).
All studies had one intervention arm with the exception of two studies(Reference Bennett, Foley and Levine28,Reference Dombrowski, McDonald and van der Pol39) . In five RCT, intervention delivery required in-person attendance(Reference Bennett, Foley and Levine28,Reference Herring, Cruice and Bennett32,Reference Phelan, Hagobian and Brannen34,Reference Van Name, Camp and Magenheimer36,Reference McRobbie, Hajek and Peerbux40) . Nine studies involved the use of telephone calls, text messages, video conferences, web-based apps or a combination(Reference Balk-Moller, Poulsen and Larsen27,Reference Rosas, Thiyagarajan and Goldstein29–Reference Gilmore, Klempel and Martin31,Reference Mayer, Vangeepuram and Fei33,Reference Samuel-Hodge, Garcia and Johnston35,Reference Hutchesson, Callister and Morgan37–Reference Dombrowski, McDonald and van der Pol39) . Attrition varied considerably; from 4–36 % and 0–25 % in intervention and control groups, respectively.
Risk of bias
Three RCT (21 %) had high RoB(Reference Rosas, Thiyagarajan and Goldstein29,Reference Samuel-Hodge, Garcia and Johnston35,Reference Van Name, Camp and Magenheimer36) . Two studies failed to apply intention-to-treat analyses(Reference Rosas, Thiyagarajan and Goldstein29,Reference Van Name, Camp and Magenheimer36) ; in one, the attrition rate was > 50 %(Reference Rosas, Thiyagarajan and Goldstein29), and in another, controls received lifestyle advice from the centre’s nutritionist(Reference Van Name, Camp and Magenheimer36), leading to bias from deviations from intended interventions. Another failed to conceal allocation sequence, risking bias from randomisation(Reference Samuel-Hodge, Garcia and Johnston35).
Three RCT were low RoB (21 %)(Reference Gilmore, Klempel and Martin31,Reference Phelan, Hagobian and Brannen34,Reference Lombard, Harrison and Kozica38) . Approximately half (57 %)(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Foley and Levine28,Reference Bennett, Steinberg and Askew30,Reference Herring, Cruice and Bennett32,Reference Mayer, Vangeepuram and Fei33,Reference Hutchesson, Callister and Morgan37,Reference Dombrowski, McDonald and van der Pol39,Reference McRobbie, Hajek and Peerbux40) had some concerns due to imbalances in group sizes, inadequate descriptions of control treatments, unclear blinding of assessments and possibility of selective reporting (online Supplementary Fig. S1(b)). Due to the nature of the interventions, the lack of participant blinding did not affect RoB.
Primary outcomes
Weight
Studies reporting weight changes from baseline to intervention end were pooled and analysed. After excluding high RoB studies to prevent exaggeration of effect, 11 studies with 2562 participants remained(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Foley and Levine28,Reference Bennett, Steinberg and Askew30–Reference Phelan, Hagobian and Brannen34,Reference Hutchesson, Callister and Morgan37–Reference McRobbie, Hajek and Peerbux40) . Results favoured interventions, demonstrating significant reductions in body weight with moderate heterogeneity (MD: –1·56 kg (95 % CI –2·09, –1·03); P < 0·001, I 2 = 43 %) (Fig. 2). Leave-one-out sensitivity analyses were performed to identify heterogeneous studies. One outlier was removed(Reference Herring, Cruice and Bennett32) due to its large effect size and comparably strong heterogeneity contribution, but results remained robust although effect size was reduced slightly (MD: –1·27 kg (95 % CI –1·69, –0·86); P < 0·001 I 2 = 13 %) (online Supplementary Fig. S3).
Forest plot on mean difference in weight changes (kg) from baseline to the end of intervention (Overall effect from all 11 studies)
A slight symmetry was observed on the funnel plot, indicating the possibility of publication bias for the overall summary effect. However, Egger’s regression test (P = 0·32) ascertained that there was no publication bias.
Secondary outcomes
Glycaemic control
Two RCT (n 603)(Reference Bennett, Foley and Levine28,Reference Phelan, Hagobian and Brannen34) demonstrated improvements in HbA1c at end of interventions (MD: –0·05 % (95 % CI –0·10, –0·001); P = 0·05, I 2 = 0 %) (online Supplementary Fig. S4).
Cardiovascular risk
There were no differences in CVD risk factors such as waist circumference, systolic, diastolic blood pressure and LDL cholesterol (online Supplementary Fig. S5–S8).
Adverse events
No adverse events relating to interventions were documented. For individual intervention strategies and secondary outcomes, funnel plots were not conducted as there were less than ten studies to distinguish true asymmetries.
Overall effect of behavioural strategies
Intervention strategies were evaluated using GRADE criteria with their strength of effect size presented in Table 2. The full assessment is found in online Supplementary Materials Table S3. The forest plots are found in online Supplementary Fig. S9–S17.
Effect estimate with 95 % CI based on individual intervention function, with GRADE assessment for the certainty of the evidence, relating to weight changes (kg)
All nine strategies were beneficial for weight reduction but only two yielded effect sizes of > 2 kg. Financial incentives in the form of cash vouchers or material goods resulted in the greatest weight reduction (GRADE certainty: moderate)(Reference Bennett, Steinberg and Askew30–Reference Herring, Cruice and Bennett32,Reference Dombrowski, McDonald and van der Pol39) , followed by interactive feedback given through mobile applications, text messages or voice responses (GRADE certainty: moderate)(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Steinberg and Askew30,Reference Gilmore, Klempel and Martin31,Reference Mayer, Vangeepuram and Fei33) .
Three studies incorporated calorie prescriptions(Reference Balk-Moller, Poulsen and Larsen27,Reference Gilmore, Klempel and Martin31,Reference Hutchesson, Callister and Morgan37) . One trial limited calories to 1800 kilocalories daily,(Reference Gilmore, Klempel and Martin31) whereas others maintained a daily deficit of 200–600 kilocalories(Reference Balk-Moller, Poulsen and Larsen27,Reference Hutchesson, Callister and Morgan37) . Five studies(Reference Balk-Moller, Poulsen and Larsen27,Reference Mayer, Vangeepuram and Fei33,Reference Phelan, Hagobian and Brannen34,Reference Hutchesson, Callister and Morgan37,Reference McRobbie, Hajek and Peerbux40) involved dietitians while three studies(Reference Bennett, Foley and Levine28,Reference Bennett, Steinberg and Askew30,Reference Lombard, Harrison and Kozica38) recruited community or peer leaders to give comparable effects through regular check-ins with advices on general healthy eating and physical activity information on a weekly to monthly basis (GRADE certainty: low).
Eight studies incorporated physical activity(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Foley and Levine28,Reference Gilmore, Klempel and Martin31–Reference Phelan, Hagobian and Brannen34,Reference Dombrowski, McDonald and van der Pol39,Reference McRobbie, Hajek and Peerbux40) . Physical activity directives ranged from the use of weights, exercise CDs, group classes, increasing daily steps to at least thirty minutes of physical activity almost daily. Six studies incorporated goal setting where participants made action plans(Reference Bennett, Foley and Levine28,Reference Gilmore, Klempel and Martin31,Reference Mayer, Vangeepuram and Fei33,Reference Phelan, Hagobian and Brannen34,Reference Hutchesson, Callister and Morgan37,Reference McRobbie, Hajek and Peerbux40) , while seven studies included regular weighs and disbursement of calorie counters, food diaries and fitness trackers as part of self-monitoring(Reference Balk-Moller, Poulsen and Larsen27,Reference Bennett, Steinberg and Askew30,Reference Herring, Cruice and Bennett32,Reference Mayer, Vangeepuram and Fei33,Reference Hutchesson, Callister and Morgan37–Reference Dombrowski, McDonald and van der Pol39) .
Seven studies incorporated elements of social support(Reference Bennett, Foley and Levine28,Reference Bennett, Steinberg and Askew30,Reference Herring, Cruice and Bennett32,Reference Phelan, Hagobian and Brannen34,Reference Hutchesson, Callister and Morgan37,Reference Lombard, Harrison and Kozica38,Reference McRobbie, Hajek and Peerbux40) . Group sessions for workouts, cooking demonstrations and nutrition classes were most common. Two hosted private social media groups on Facebook and Instagram(Reference Bennett, Steinberg and Askew30,Reference Hutchesson, Callister and Morgan37) . Compared with trials without social support (MD: –1·83 kg; (95 % CI –2·29, –1·07); P < 0·001, I 2 = 0 %), GRADE certainty: moderate), trials incorporating social support demonstrated less weight reduction with moderate heterogeneity (I 2 = 56 %, P = 0·04).
Subgroup analyses
Subgroup analyses were performed for sex ((online Supplementary Fig. S18), intervention duration online Supplementary Fig. S19) and BMI (online Supplementary Fig. S20) and are presented in Table 3. Females were observed to lose significantly greater weight than males despite similar interventions. Longer interventions lasting 12 months or more were more effective for weight loss, while the extent of weight loss was directly proportional based on the severity of obesity. No further subgroup analyses were undertaken for intervention type as the number of studies is small.
Subgroup analyses based on sex, intervention duration and BMI, relating to weight changes (kg)
Discussion
This is the first review to use meta-analyses to examine behavioural strategies for weight reduction among adults belonging to lower socio-economic groups. Nine unique strategies were identified. Financial incentives produced the greatest weight reductions with moderate certainty of evidence.
Unlike the previous review conducted by Hllier et al, which focused on interventions that specifically targeted reducing socio-economic inequalities(Reference Hillier-Brown, Bambra and Cairns21) our review aimed to examine the direct effects of interventions to adults belonging to lower socio-economic groups. Comparing both studies, it was evident that both individual and community weight loss programmes were effective, especially among women, considering the under-representation of men in these studies. Additionally, the regional concentration in the former review prompted us to broaden our search to include studies conducted outside of the USA in our current study.
Overall, results favoured interventions, and the use of behavioural strategies demonstrated a modest, but significant weight loss over controls. Despite the relatively small effect (1·0–2·5 kg), findings are consistent with a recent meta-analysis in predominantly high-income countries where interventions experienced 2·3 kg greater weight loss(Reference Madigan, Graham and Sturgiss41). Similarly, another meta-analysis showed that a 1 kg weight reduction reduced blood pressure(Reference Neter, Stam and Kok42), while other trials have utilised a 2 kg weight loss as a clinically meaningful endpoint since it lowers the risk of developing diabetes(Reference Stevens, Obarzanek and Cook43,Reference Clark, Keith and Weiner44) .
Despite reductions in HbA1c, other CVD risk factors such as waist circumference, blood pressure and serum lipids showed no improvements. Since many studies did not report these attributes, the relationship between interventions and CVD risk could not be properly evaluated. However, a secondary analysis of subjects participating in the National Health and Nutrition Examination Survey found that low income was associated with CVD risk, underscoring the need to improve weight outcomes in this population(Reference Seligman, Laraia and Kushel45).
Predictably, there was heterogeneity across the studies identified. Although the meta-analyses provided strong evidence for the overall effectiveness of behavioural strategies, effect sizes varied between strategies. As a consequence, independent investigations were conducted for each strategy through subgroup analyses while exploring differences in sex, intervention duration and BMI classifications.
Compared with a mixed-gender sample, female-targeted interventions produced twice as much weight loss. A recent meta-analysis found no significant differences between gender-targeted studies for weight loss; however, the subjects were restricted to young adults which limited the generalisability of our sample(Reference Sharkey, Whatnall and Hutchesson46). Another older review found that men lost significantly more weight than women when strategies were restricted to diet and exercise(Reference Williams, Wood and Collins47). As majority of our cases were females and only one study recruited male subjects(Reference Dombrowski, McDonald and van der Pol39), the true effect of sex differences could not be determined.
Compared with overweight participants, Class II obese subjects lost twice as much weight whereas Class I obese subjects lost 1·5 times as much weight. Although moderate heterogeneity (I 2 = 40 %, P = 0·15; I 2 = 51 %, P = 0·11) was observed, results were insignificant. Findings are consistent with another meta-analysis where comparable subjects (mean BMI range 25–40 kg/m2) found that severely obese participants shed more weight than their less obese counterparts(Reference Barte, Veldwijk and Teixeira48).
Interestingly, participants lost more weight in trials that did not integrate social support. These findings should, however, be interpreted with caution as this subgroup showed greatest heterogeneity when compared with others (I 2 = 56 %, P = 0·04). Additionally, it was noted that over the past decade, technology-based interventions have been progressively trialled. With increasing popularity, it is important to ensure equity across socio-economic classes as affordability of computers and smartphones may be a problem(Reference Ashrafian, Toma and Harling49).
The burden of adult obesity has been studied extensively, and pressure is mounting to halt the growth of obesity in lower socio-economic groups. This is evident with population strategies, such as the USA’s Health Equity Resource Toolkit(50) and Australia’s National Obesity Strategy 2022–2032(51), highlighting the need to focus on lower socio-economic populations. According to a policy paper, the healthcare system could save £105 million over 5 years if everyone who is overweight or obese, lost 2·5 kg each(52). Our review builds on this knowledge to form a suite of behavioural strategies to address obesity. However, it is necessary to emphasise that weight reductions are not causal to improving health inequalities; socio-economic determinants of health and the living environment must be considered in addition to weight loss(Reference Borys, Richard and Ruault Du Plessis53).
Implications for research and practice
Longer trials demonstrated weight loss advantages over shorter trials. However, it was not possible to study the fidelity of weight maintenance as few studies recorded follow-up data and the longest follow-up was only one year(Reference Bennett, Foley and Levine28). As weight regain susceptibility rises 36 weeks post-intervention(Reference Machado, Guimarães and Bocardi54), it is important to consider obesity’s chronicity and offer extended support. Notably, although financial incentives resulted in the greatest weight loss of all nine interventions, recent literature suggests that offering incentives for intermediate behavioural improvements, rather than weight reduction per se, may be more beneficial for sustaining weight loss(Reference Jay, Orstad and Wali55). Overall, this review’s favourable findings support the use of behavioural strategies to reduce obesity but future studies should examine their cost-effectiveness.
Strengths and limitations
This review uses strong methodological criteria to synthesise updated, high-quality evidence. To ensure the relevance of the interventions studied and their alignment with technological advancements and digital platforms, we focused exclusively on publications from the last decade. This decision was influenced by the findings of Hilier et al(Reference Hillier-Brown, Bambra and Cairns21), where the interventions primarily consisted of face-to-face nutrition or exercise programmes, except for a single study. In our present study, 64 % of interventions employed tele/video or application-based approaches to engage with participants. This shift towards technology-enabled methods highlights the evolving landscape of intervention strategies. As obesity may be a stigmatising, sensitive issue, only summary data from ethics-approved research were used. All weights and blood results were measured to limit reporting bias. Most intervention strategies (7/9) were of low heterogeneity, giving robust evidence for the strength of effect
Only 21 % of publications were assessed as low RoB. Hence, only studies of low RoB or with some concerns were meta-analysed, while the GRADE tool allows for a pragmatic interpretation of the certainty of evidence. Lack of translation resources limited investigations to English-language publications. Most studies (64 %) were from the USA, thereby limiting the generalisability of findings owing to differences in healthcare systems in other nations. Males were under-represented, and there was inadequate reporting on ethnic diversity. Future reviews should re-evaluate these statistics to identify which targeted interventions are most appropriate. Lastly, this paper could not ascertain if technological interventions were superior to traditional in-person delivery methods due to high heterogeneity.
Conclusion
Behavioural strategies have been shown to be effective in supporting modest, but significant weight loss among individuals with lower socio-economic status living in high-income nations. Of the nine distinct strategies, providing financial incentives resulted in the greatest weight reductions. Few studies explored the relationship between these strategies and CVD risk factors; hence, the impact remains unclear. Although this study advances our understanding of the effectiveness of various intervention strategies, future research should sample a larger population that is reflective of the population’s diversity. Comparative analyses should also be conducted on the cost-effectiveness of these strategies to determine the economic credentials of interventions.
Acknowledgements
There are no competing interests or funding to be declared.
Priscilla Li, Yingxiao Huang, and Alvin Wong contributed to acquisition, analysis, or interpretation of the data. Priscilla Li drafted the manuscript. Alvin Wong and Yingxiao Huang critically revised the manuscript. All three authors contributed to conception/design of the research. All authors agree to be fully accountable for ensuring the integrity and accuracy of the work. All authors have read and approved the final manuscript.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114523001940
