The article is also available on the Medical Observer website (may need registration) under the title, “What’s the evidence for metformin in the absence of diabetes?”.
Ahmed, a 25-year-old man living with bipolar affective disorder, saw me recently. His symptoms were well controlled with quetiapine (Seroquel), but I noticed that his psychiatrist had also prescribed metformin. Ahmed had neither diabetes nor “pre-diabetes”. I wondered at the evidence that supported this practice.
What is the effect of metformin on cardiovascular disease (CVD), in people without diabetes, on long-term antipsychotic medication?
What does the research evidence say?
Step 1: The Cochrane Library
The Cochrane Library does not have a systematic review on this question.
Step 2: TripDatabase
Next, I conducted a search using the TripDatabase PICO search tool (Participant: “antipsychotic”, Intervention: “metformin”, Comparator: blank, Outcomes: “cardiovascular disease”). It became obvious that no study that could directly answer the question had been published (or exists). Repeating the search with the “outcomes” field as blank demonstrated that the available literature mostly looked at weight. The most informative paper that I could find was a meta-analysis by Zheng et al. (2015) published in the Journal of Clinical Psychopharmacology .
I will use the systematic reviews critical appraisal sheet from the Centre for Evidence Based Medicine .
What PICO question does the systematic review ask?
In people receiving antipsychotic medication (Participants); what is the effect of metformin as prevention or treatment (Intervention); compared to placebo (Comparator); on weight gain and metabolic abnormalities (Outcome).
Is it clearly stated?
Yes. For instance, the review explicitly listed the primary and secondary outcomes (Table 1, p. 500) .
Is it unlikely that important studies were missed?
Yes. The authors searched multiple electronic databases, clinical study registries, and interestingly, also three major Chinese databases.
Were the criteria used to select articles for inclusion appropriate?
Probably. The authors only included randomised trials that reported at least one of the defined primary outcome measures (weight, BMI, fasting blood glucose, fasting insulin, total cholesterol, triglycerides).
Were the included studies sufficiently valid for the question asked?
Unclear. The authors formally assessed the risk of bias of the included studies using Jadad scales . Just over half of the included studies (11 of 21) could be considered “low quality” (Jadad scores 1-3).
Were the results similar between studies?
No, and complex. There was moderate to large heterogeneity in the results for many of the outcomes. In particular, there appeared to be meaningful differences between RCTs of Chinese and non-Chinese participants.
What were the results?
In non-Chinese participants, individuals who received metformin rather than placebo had:
- Lower weight gain/BMI, of a small-moderate to moderate effect size
- Lower total cholesterol and triglycerides, of a small to small-moderate effect size
There was large heterogeneity (inconsistency in results between studies) depending on the analysis and outcome. Trials of Chinese participants were associated with more favourable results from metformin.
Discussion and conclusion
The reviewed study provided a complex picture on the effect metformin has on a number of biometric measures that are CVD risk factors. There seems to be an effect on weight – metformin is associated with a small-moderate beneficial effect on the weight gain from antipsychotics. This effect appeared to be strongest when metformin was commenced in antipsychotic naïve patients. Curiously, low dose (≤ 750 mg/d) metformin was associated with more benefit than high dose .
At least one study indicated that metformin might be more effective than lifestyle intervention for “antipsychotic-induced weight gain” , but this was in a Chinese population. In this meta-analysis, studies in non-Chinese participants were associated with less benefit from metformin . Notably, studies from outside the field in the general population suggests that lifestyle measures are more effective than metformin for prevention .
Fundamentally, the meaningfulness of these changes in weight, a surrogate end point (see StatFacts), is quite unclear. Is the difference in weight gain clinically meaningful and does it actually change CVD risk? At present, there is no high quality evidence for our question – whether metformin improves CVD risk and outcomes in people taking antipsychotic medications. Although the practice is common and defensible, the lack of evidence should preclude it from being offered routinely.
Surrogate end points
A surrogate end point is “a laboratory measurement or physical sign that is used in therapeutic trials as a substitute for a clinically meaningful end point that is a direct measure of how a patient feels, functions, or survives” . The risk of over-reliance on surrogate end points is that it can lead to the adoption of therapies that are ultimately discovered to be useless or even harmful.
- Zheng W, Li XB, Tang YL, Xiang YQ, Wang CY, de Leon J. Metformin for Weight Gain and Metabolic Abnormalities Associated With Antipsychotic Treatment: Meta-Analysis of Randomized Placebo-Controlled Trials. J Clin Psychopharmacol 2015 Oct;35(5):499-509.
- Centre for Evidence Based Medicine. Systematic Review: Are the results of the review valid? Oxford: University of Oxford, 2005.
- Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996 Feb;17(1):1-12.
- Wu RR, Zhao JP, Jin H, et al. Lifestyle intervention and metformin for treatment of antipsychotic-induced weight gain: a randomized controlled trial. JAMA 2008 Jan 9;299(2):185-93.
- Mangahas T, Huang G, Neher J, Safranek S. Clinical Inquiry: Does metformin prevent diabetes in at-risk adults? J Fam Pract 2013 Aug;62(8):436-7.
- Temple R. Are surrogate markers adequate to assess cardiovascular disease drugs? JAMA 1999 Aug 25;282(8):790-5.