i. Search strategy for identification of studies providing information on clinical effectiveness

A search for relevant studies on clinical effectiveness will be conducted by means of electronic searches of key databases including MEDLINE, EMBASE, Science Citation Index and Biological Abstracts.

Searching for clinical information as contained in systematic reviews, meta-analyses or clinical trials. This will focus on the above key databases with the addition of the Cochrane library and specific trials registers. Published methods of searching specifically for systematic reviews and clinical trials as developed by the McMaster University Health Information Research Unit will be used. Specific concepts to be included in the literature searches will include terms relating to obesity (obesity.tw, obese, obesity (subject heading), obesity, morbid (subject heading)), and terms relating to agents (orlistat, sibutramine, rimonabant, anti-obesity agents (subject heading), Tetrahydrolipstatin (subject heading), sibutramine (subject heading), rimonabant (subject heading)). References will also be located through review of references for relevant articles and through citation search facilities via the Web of Science's Science Citation Index and Social Science Citation Index. Where systematic reviews already exist, these will be used to identify relevant studies and to inform subsequent analyses. In addition systematic searches of the Internet using various search engines will be used to identify unpublished materials and work in progress. Key authors and commercial organisations involved in the investigation of pharmaceutical agents will be contacted and asked for unpublished materials.

We will utilise a varied range of sources and search techniques to identify relevant literature. A comprehensive literature search will be undertaken in the major medical, health-related, science and health economic electronic bibliographic databases (i.e. CDSR, NHS DARE, NHS HTA, MEDLINE, EMBASE, CINAHL, Science Citation Index, PreMEDLINE, NHS EED, HEED, CENTRAL, Pascal, ASSIA, Social Care Online, Social Science Citation Index). In addition, various health service research and guideline producing bodies (e.g. SIGN, National Guidelines Clearinghouse, etc.) will be consulted via the internet and key organisations (e.g. National Obesity Forum) will be contacted. We will utilise the expertise within the group and consult with national and international experts in research and practice in obesity. Ongoing and recently completed research in the field will be identified through searching the National Research Register, ReFeR, Current Controlled Trials and its links, HSRProj and Index to Theses. Grey literature will be identified from searches of databases including Dissertation Abstracts and Inside Conferences. Finally, the reference lists of included studies will be examined for additional relevant references and, where appropriate, the citation facility in Web of Science will be used to search for specific papers and authors.

ii. Search strategy for identification of studies providing information on adverse effects

Supplementary searches will be conducted for data on adverse effects. No study restrictions will be utilised. Specific pharmacological databases will be used at this stage of the review. Reference will be made to published work on retrieval of adverse effects literature from the NHS Centre for Reviews and Dissemination.{Golder 2006, Golder 2006} We will also write to the manufacturers of these drugs to obtain any data on file.

iii. Search strategy for identification of studies providing information on adherence to treatment

Given that primary research suggests that lack of adherence to treatment is a major factor limiting the efficacy of anti-obesity drugs it would be valuable in answering the effectiveness-related questions to examine what we know on patients' perceptions of anti-obesity drugs. Ogden and Sidhu (2006) are among the first to examine specifically the qualitative experience of patients on obesity medication. {Ogden 2006} Other qualitative research on perceptions of obesity treatments will also be valuable. We therefore propose to conduct a tightly focused qualitative evidence synthesis using accepted methods of evidence interpretation and integration.{Pope, Mays and Popay 2007} This review will complement the effectiveness review and modelling work and provide added value by identifying the main variables that can impact on the anticipated effectiveness of anti-obesity medication.

iv. Search strategy for identifying economic evidence

In addition to the search strategies identified above systematic searches will take place of the specialist health economic data sources such as DARE, HTA Database (University of York), NHS EED and the Office for Health Economics HEED database. Economics filters used by the NHS CRD to populate the NHS EED database will be adapted to other databases.

a. Search strategy (example from EMBASE)

The search strategy will use the following terms:

1, obesity.tw, 2, obese.tw 3, obesity/4, obesity, morbid/5, or/1–4 6, orlistat 7, sibutramine 8, rimonabant 9, anti-obesity agents/10, Tetrahydrolipstatin/11, sibutramine/12, rimonabant/13, or/6–12 14, 5 and 13

Plus methodological filters as described above to locate high quality clinical effectiveness and cost-effectiveness studies. The results of the searches will be stored in a Reference Manager database.

b. Inclusion/exclusion criteria

• Types of studies: Randomised controlled trials (RCTs), incorporating any duration of therapy and any length of follow-up will be considered for inclusion in the review.
• Participants: i) RCTs recruiting adults (aged 18 years) defined as being overweight or obese. ii) RCTs recruiting adults wishing to maintain weight loss, having been previously overweight or obese. iii) Trials involving specific patient groups such as those with diabetes, hypertension or hyperlipidaemia will be included in the review, provided they meet the above criteria.
• Interventions: i) Evaluations of Orlistat, Sibutramine or Rimonabant used to treat overweight/obese patients or to maintain weight loss in previously overweight or obese patients. ii) Orlistat, Sibutramine or Rimonabant may be combined with other strategies such as dietary restriction or behavioural programmes. iii) Participants in control groups may receive placebo, an alternative anti-obesity pharmacological agent or an alternative anti-obesity intervention (e.g. based on dietary regimen, physical activity or behavioural modification).
• Studies recruiting people with eating disorders such as anorexia nervosa and bulimia nervosa will be excluded.
• In trials where overweight/obese participants were recruited as well as those with the above eating disorders, only those where results were presented separately for the overweight/obese participants will be included.

c. Outcomes

The primary outcome of the review will be an assessment of obesity/overweight status as measured by changes in body weight, fat content or fat distribution:

• Measures of weight change include absolute weight change and percentage weight change relative to baseline.
• Measures of fat content include BMI, ponderal index, skinfold thickness, fat-free mass, body percentage and fat change relative to baseline.
• Measures of fat distribution including changes in waist size, waist–hip ratio and girth–height ratio relative to baseline.

Secondary outcomes of the review will be a) physiological changes occurring in association with changes in body weight/fat content/fat distribution such as changes in lipid profiles, glycaemic control among those with diabetes, and blood pressure, b) patient-related quality of life, c) information on adverse effects and d) costs.

d. Review methods

• References identified by the literature searches will be sifted in three stages. They will first be screened for relevance by title. The abstracts of those which are not excluded at this stage will then be read and finally, all manuscripts which seem to be potentially relevant will be obtained for a more detailed appraisal. Sifting will be undertaken by one reviewer, and to ensure consistency a sample of references will be checked by a second reviewer. All decisions will be coded and recorded in the Reference Manager database.
• Studies will be categorised according to the type of participant (see inclusion criteria). Data extraction will be undertaken by one reviewer, using customised data extraction forms, and checked by a second reviewer. Discrepancies will be discussed, and any which cannot be resolved will be referred for discussion to the study team. Data extraction will cover the design and conduct of trials, characteristics of participants and interventions, and outcomes.
• Quality checklists will be used to appraise each article included. The quality of randomised controlled trials will be assessed according to criteria based on those proposed by the NHS Centre for Reviews and Dissemination. Non-randomised forms of evidence of clinical effectiveness such as observational studies will be assessed using the Downs and Black checklist.{Tooth 2005} Attrition rates will be assessed and discussed as previous reviews have noted high attrition rates.
• Heterogeneity among the results will be explored with consideration given to the following: patient characteristics, study setting, patient selection, and outcome measures.
• Summary statistics will be derived for each study and a weighted average of the summary statistics will be computed across the studies. Statistical heterogeneity will be assessed using the I-squared measurement. The studies will be assessed clinically and methodologically to assess whether it is reasonable to meta-analyse the data. If so, the more conservative random-effects model will be used to account for small clinical and methodological variations between very similar high quality trials. Data from studies that score poorly on the quality assessment; or studies that are found to be statistically heterogeneous will not be combined. In these cases further investigation will be undertaken to identify factors that could potentially explain the heterogeneity. In addition, sensitivity analyses will be conducted to assess the impact of including these studies.

As no ‘head-to-head’ RCTs are expected (of the three drugs under consideration), a synthesis of the available evidence using indirect meta-analysis methods will be used {Caldwell et al 2005}. However, in elaborating the network to include other interventions (used either as a control intervention in pharmacological trials or as additional arms in such trials) Bayesian Mixed Treatment Comparison methods will almost certainly have to be used.{Salanti et al 2007;Lu & Ades, 2004} The analysis will incorporate both direct and indirect evidence to enable comparisons to be made between treatments, including not only estimation of all pair-wise comparisons, but also ranking of treatments in terms of clinical effectiveness. As part of the MTC analysis further issues will need to be addressed, including outcomes reported at multiple and different time points,{Lu et al 2007}, the fact that there will be heterogeneity in reporting, both in terms of outcome, e.g. BMI, weight change, hip-to-waist ratio {Nam 2007; Riley 2007}, change from baseline or otherwise {Abrams 2005}, extension of the network of evidence to include other comparators that have been evaluated in obese patients {Salanti 2007} and consistency of evidence {Lu & Ades 2006}. In addition there will be an assessment of publication bias (Sutton 2000) and exploration of whether clinical effectiveness varies with baseline obesity, e.g. BMI. The analysis will be done by the Department of Health Sciences at the University of Leicester, using the freely available software WinBUGS {Spiegelhalter, 2002}.

a. Search strategy

The search strategy will use the following terms: cost benefit, cost effectiveness, cost utility, cost consequences, cost minimisation, economic evaluation, quality of life, utility, incremental cost effectiveness analysis, incremental cost effectiveness ratio, net present value, incremental net benefit; combined with the search terms used in the effectiveness literature search strategy. Sensitive searching (e.g. economics [ec] as a floating subheading) will be used to pick up costs associated with the health conditions. The results of the searches will be stored in a Reference Manager database.

b. Inclusion criteria

English language papers reporting cost-effectiveness results in terms of cost per QALY or cost per life year gained for the three interventions Orlistat, Sibutramine or Rimonabant.

c. Screening strategy

All abstracts obtained by the computer search will be reviewed for relevance by the two economic analysts. Any disagreement will be resolved by discussion. All papers identified as relevant at the end of the abstract screening process will be obtained and entered into the quality assessment process. The results of the abstract screening will be recorded in the Reference Manager database, including the reason for excluding any paper from the quality assessment stage of the review.

Once papers selected for inclusion in the review have been obtained, a hand search of the reference lists will be undertaken to identify any potentially relevant papers not identified by the search of the literature databases. Any additional papers will be obtained and subjected to the abstract review process prior to inclusion or exclusion from the quality assessment process.

d. Quality assessment

Relevant studies will be critically appraised using the standard economic evaluation and modelling checklists.{Drummond, Eddy 1985} For papers reporting economic evaluations alongside clinical trials, the Drummond checklist will be supplemented with reference to the Good Practice Guidance produced by the ISPOR Task Force on Economic evaluations alongside clinical trials.{Weinstein 2003}

Additional searches will be conducted to identify evidence on quality of life (QoL) in obese individuals, natural history of weight gains, weight regain and relationships between weight changes and co-morbidities such as CHD and diabetes.