Metformin in the management of antipsychotic-induced weight gain in adults with psychosis: development of the first evidence-based guideline using GRADE methodology

Background Adjunctive metformin is the most well-studied intervention in the pharmacological management of antipsychotic-induced weight gain (AIWG). Although a relatively unaddressed area, among guidelines recommending consideration of metformin, prescribing information that would facilitate its applied use by clinicians, for example, provision of a dose titration schedule is absent. Moreover, recommendations differ regarding metformin’s place in the hierarchy of management options. Both represent significant barriers to the applied, evidence-based use of metformin for this indication. Objective To produce a guideline solely dedicated to the optimised use of metformin in AIWG management, using internationally endorsed guideline methodology. Methods A list of guideline key health questions (KHQs) was produced. It was agreed that individual recommendations would be ‘adopted or adapted’ from current guidelines and/or developed de novo, in the case of unanswered questions. A systematic literature review (2008–2020) was undertaken to identify published guidelines and supporting (or more recent) research evidence. Quality appraisal was undertaken using the Appraisal of Guidelines Research and Evaluation II tool, A Measurement Tool to Assess Systematic Reviews (AMSTAR) assessment,and the Cochrane Risk of Bias 2 tool, where appropriate. Assessment of evidence certainty and recommendation development was undertaken using Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology. Findings We confirmed that no published guideline—of appropriate quality, solely dedicated to the use of metformin to manage AIWG was available. Recommendations located within other guidelines inadequately addressed our KHQs. Conclusion All 11 recommendations and 7 supporting good practice developed here were formulated de novo. Clinical implications These recommendations build on the number and quality of recommendations in this area, and facilitate the optimised use of metformin when managing AIWG.

For both the guideline and database search, several searches were undertaken using different combinations of MeSH terms and free text words, as outlined in Table 1. The search was carried out between December 2019 -January 2020.

Search number
Terms used 1 Antipsychotic OR Psychosis OR Schizophren* OR "Severe Mental Illness" AND Pharmac* OR Metformin OR Treatment OR Intervent* AND management AND "metabolic side-effects" OR "metabolic disturbance" OR "physical health"

2
Search 1 but substitute antipsychotic with psychotropic OR Neuroleptic OR Aripiprazole OR misulpride OR Chlorpromazine OR Clozapine OR Fluphenazine OR Haloperidol OR Olanzapine OR Risperidone OR Paliperidone OR Quetiapine OR Zuclopenthixol 3 Search 1 + 2 but substitute Metabolic side-effects/metabolic disturbances with "weight-gain" OR Weight For the database search, a search date between 1/1/2008 -1/1/20 was applied to all searches following a preliminary review of the area identifying that the majority of research was published during this interval. This was complemented by a manual search of included guidelines' reference lists.
3. An advanced Google search, as guideline developers are increasingly publishing their guidelines on the Web for availability.
Google search criteria are outlined in Table 2. A search date of 1/1/2008 -1/1/2020 was applied and English language restriction set. The first 200 hits were screened. The search was carried out between December 2019 -January 2020.

Database
Search strategy Google 1. Advanced search: "psychosis" AND this exact word or phrase "management in adults" + "guideline".
2. Advanced search: antipsychotic-induced weight gain AND this exact word or phrase "management in adults" + "guideline".

A.2 Inclusion and exclusion criteria
Guidelines were included if they: 1. Were written in English; 2. Contained a focused section on pharmacological management of antipsychotic-induced weight gain (AIWG); 3. Were published between 1/1/2008 -1/1/2020; 4. Specified the target population as adults ≥18 with a psychotic illness, other than in the context of Bipolar Affective Disorder (BPAD); 5. Are evidence-based guidelines i.e. those that include a report on systematic literature searches and contain explicit links between individual recommendations and supporting evidence and; 6. Were developed for use within countries with a relevant healthcare system for the Irish context.
Guidelines were excluded if they: 1. Didn't address any of our key health questions (KHQs); 2. Were developed in a healthcare system not applicable to the Irish setting; 3. Were written by only a single author;

4.
Didn't contain a documented systematic evidence compilation and review process and; 5. Were guidelines based on obesity management in the general population only.
All guidelines were initially identified by their title from the three sources by one reviewer (IF). Two reviewers (IF and EC) then completed the second screen of all guidelines and article abstracts to identify those for full review. Differences were resolved via discussion. Data abstraction was undertaken by one reviewer (IF), but was then checked by a second reviewer (EC).
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

B.2 Inclusion and exclusion criteria
Evidence was included if: i. Was based on Randomised Controlled Trials (RCTs), Cochrane or systematic reviews, or meta-analyses; ii. Was published between 1/1/2008 -1/1/2020; iii. Was available in full text; iv. Was published in a peer-reviewed journal; v. Was published in English; vi. The cohort were adults ≥18 with a psychotic illness, other than in the context of bipolar affective disorder; vii. The intervention was the use of metformin alone in the treatment of AIWG; viii. The comparator was placebo, usual care or non-pharmacological methods of treating AIWG and; ix. Anthropometric measurements associated with the interventions(s) were reported as the primary outcome.
Evidence was excluded if: i. The intervention was used to attenuate weight gain from other psychotropics, apart from antipsychotics; ii. The applied use of metformin was solely in the prevention of AIWG; iii. Was based on low levels of evidence including case reports, editorials and commentaries and; iv. Primary outcomes reported weren't associated with anthropometric measurements.
Title review was undertaken by the primary author. Studies that clearly weren't relevant were eliminated. Abstracts of remaining studies were then eliminated by two independent reviewers. Effort was made to contact authors for clarification, where needed. Where uncertainty remained, third party consultation was available. Data extraction was primarily performed by the primary author using a pre-designed data extraction form.     * Adverse event data was not synthesized quantitatively for several reasons. This included evidence of incomplete and selective outcome reporting in some studies meaning that data were incomplete. As highlighted in 8/10 studies in the De Silva et al., meta-analysis did report adverse event data, however in only 6/8 cases numerical data (total study numbers: placebo n =223, metformin n =227) was reported that could be considered for aggregation. Furthermore, a range of side effects were reported with very small number of events per subgrouping e.g. within GI side effects, some studies reported on nausea, others on diarrhoea only. In other cases a range of seemingly unrelated side effects were reported including somnolence and extrapyramidal s/e, which haven't been related to metformin treatment previously. One study only gave adverse event data where the rate of that adverse event that affected >5% of the overall sample. We received no response from study authors for greater details where the study report did not provide numerical data, or where the report stated that there was no significant difference in adverse events, but did not report the adverse effects measured or number of associated events. Due to missing data, the sample size of available studies where data was available to combine, and smaller number of events within these subcategories of side effectseven amongst s/e that are known to be common to metformin i.e. gastrointestinal side effects, we decided that pooling these results was unlikely to yield clinically meaningful results due to low power and precision of any estimate drawn, and that certainty in collated estimates would be very low. We agreed quantitative aggregation wouldn't yield any further benefits over narrative synthesis and thus, wouldn't influence our application of the evidence base specific to this outcome and resultant recommendations. ** Results for this study were assessed separately to the meta-analysis by De Silva et al. 1 The guideline development group chose to do this primairly because the population under assessment in the RCT by Chiu et al.,were those treated with clozapine i.e. participants were those with treatment-resistant schizophrenia. The mean duration of illness was 25-28 years across treatment groups and thus previous antipsychotic exposure was very likely significant, although not specified in the study report. 2 As the effects of metformin in this group compared to those who have a much lesser burden of antipsychotic exposure are likely to be different, as highlighted in the De Silva et al., sub-group analysis, 1 a decision was made not to collate these results. Furthermore, doses used were significantly lower than those applied across studies included in the De Silva et al., meta-analysis. 1,2 As the study report by Chiu et al., did not provide any measure of variance of effect alongside point estimates, 2 this represented another barrier to aggregation of study results. Requests for such measures from the study authors by the guideline development group went unanswered. a. = Rated serious as there was an unclear risk of bias for blinding in four studies and an unclear risk of bias related to drop-out rates in one. Despite anthropometric outcomes being objective, it was felt that if participants, research personnel and/or outcome assessors were aware of individual(s) assignment, this may lead to an alteration in health behaviours and reduce certainty associated with estimates. b = Marked as serious due to high risk of bias in 2/10 studiesdue to no reporting on this outcome i.e. selective outcome reporting. In 2/8 that did, this data was incomplete i.e. evidence of incomplete outcome reporting. c = Although data wasn't pooled, results appeared inconsistent. This was worsened by incomplete and selective outcome reporting in minority of studies. d = Downgraded as no confidence intervals were reported for any of the outcomes assessed.  There was no statistically significant difference between any type of adverse event seen in the two groups. Numerically higher rates of nausea was seen in the metformin treated groups. There were five serious adverse events in the trial that led to withdrawal. All were exacerbation of psychosis.
⨁⨁⨁◯ MODERATE IMPORTANT CI: Confidence interval; MD: Mean difference a. Rated down by one level as the included cohort were FEP participants who had gained >10% of their baseline bodyweight and in all groups' total exposure to antipsychotics was < 1 year. Thus, this cohort are only representative of some of the SU that will be affected by guideline recommendations i.e. generalisability reduced.

PICO Key Health Questions Addressed:
KHQ 3 -Should metformin + non-pharmacological vs. non-pharmacological treatment alone be used in the management of AWIG in adults with psychosis? KHQ 5 -Where metformin is identified as being effective in a particular cohort, what dose of metformin should be used? KHQ 6 -Where metformin is identified as being effective in a particular cohort, for how long should metformin be used? KHQ 7 -Where metformin is being used for the management of AIWG vs. usual care or placebo what are the potential harms associated with its use in adults with psychosis? Rates of adverse gastrointestinal events were 12.5% in the metformin + lifestyle group and 15.6% in the lifestyle + placebo group, with no significant difference between groups (P<0.88). No difference in dropouts between groups due to adverse GI events. Notably in this study the maximum dose of metformin used was 750mg/day. This dose combined with the small numbers in each group may have led to the lack of significance between groups, as this is a common significant side effect of metformin in groups with bigger numbers.

CI: Confidence interval; MD: Mean difference
Footnotes: a = Rated down by one level as the included cohort were FEP participants who had gained >10% of their baseline bodyweight and in all groups' total exposure to antipsychotics was < 1 year. Thus, this cohort are only representative of some of the SU that will be affected by guideline recommendations i.e. generalisability reduced.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) (1) Evidence of high risk of bias due to in selective outcome reporting in 1/5 studies.
(2) Unclear risk of bias in 2/5 studies across RoB outcomes related to sequence generation for randomization, allocation sequence concealment and blinding of participants and personnel.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)