|Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meta-analysis of randomised clinical trials
|Stevens RJ, Ali R, Bankhead CR, Bethel MA, Cairns BJ, Camisasca RP, Crowe FL, Farmer AJ, Harrison S, Hirst JA, Home P, Kahn SE, McLellan JH, Perera R, Pluddemann A, Ramachandran A, Roberts NW, Rose PW, Schweizer A, Viberti G, Holman RR
The review concluded that there was no statistically significant beneficial effect of metformin on cancer or mortality outcomes compared with other drugs, placebo or usual care. The review was generally well conducted. The authors noted shortcomings in the quality of the evidence base, imprecision in the estimates and clinical differences between studies and appropriately concluded that more research was required.
To assess risk of cancer and all-cause mortality in adults treated with metformin.
MEDLINE, EMBASE and The Cochrane Library were searched up to April 2012 for full-text articles; limited search terms were reported. A further search was conducted of a clinical trials registry and the reference list of a relevant Cochrane systematic review.
Randomised controlled trials (RCTs) of at least 500 adult participants (with at least 100 participants randomised to metformin) that assessed metformin and with follow-up of at least one year were eligible for the review. Outcomes of interest were all-cause mortality, cancer incidence and cancer mortality. Participants were required to be free of pre-existing cancer. Trials were excluded if they had a crossover or pseudo randomised design, if most participants had a severely life threatening illness and if the effects of metformin alone could not be separated when metformin was used as a combination therapy.
Most participants in the included studies had type 2 diabetes and the rest were considered at risk of diabetes. Mean ages of participants ranged from 46 to 60 years. Approximately half of the participants were men. Metformin alone or metformin combined with saxagliptin or sulphonylurea was compared with placebo, usual care or any non-metformin oral antidiabetic therapy. Daily metformin dose ranged from 500mg to 2,550mg. Some studies had pretrial treatment with either diet or sulphonylurea. The authors documented treatments and comparators in separate tables available with the publication; studies included two to six treatment arms. The primary outcomes of interest in the included studies were diabetes-related outcomes such as incident diabetes or change in HbA1c. Follow-up ranged from one to 10.7 years. Studies were conducted in USA, UK, India or multiple centres worldwide.
Two independent reviewers selected studies for the review. Discrepancies were resolved by consultation with a third reviewer where necessary.
Assessment of study quality
Studies were assessed for quality using the Cochrane risk of bias tool with criteria that included random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors and risk of incomplete outcome data.
The authors did not report how many reviewers assessed studies for quality.
Data on the events and total person years for all-cause mortality and cancer incidence and mortality were extracted on an intention-to-treat basis. Relative risks (RR) were either extracted or approximated from other effect measures or estimated from the events and total person years data. Where data were not available for cancer incidence, relative risks for cancer mortality were extracted. In trials with multiple comparator arms, the placebo or usual care comparator arm was selected preferentially where available. Authors were contacted for additional data where necessary.
Where data on these outcomes were not available in the publication or could not be obtained from authors, information was extracted from publicly available sources where possible.
The authors did not state how many reviewers extracted data.
Methods of synthesis
Study results were synthesised in meta-analyses. Summary effect relative risks, together with 95% confidence intervals (CI), were calculated using a fixed-effect inverse variance model. Heterogeneity was quantified using the Ι² value. Random-effects models were used in sensitivity analyses.
Stratified analyses were undertaken according to comparator agent and level of blinding. Subgroup analyses were investigated trial duration (52 weeks to 366 days or >366 days). The authors reported that methods were used to take into account arms with zero events but they later reported that two trials with zero events were excluded from the analysis of cancer incidence. In studies with multiple comparative arms where placebo or usual care comparator arms were not selected preferentially, results for each comparison were combined in a single estimate by generalised least squares regression taking into account the correlation between the comparisons.
Post hoc sensitivity analyses were undertaken for cancer incidence, analyses of trials only of people with diabetes or where metformin was used as monotherapy, for all-cause mortality and exclusion of one trial that appeared to be an outlier.
Results of the review
Thirteen trials (23,729 participants, range 531 to 8,732) were eligible for the review. Five studies were open label (one used a blinded outcome assessment) and eight studies were double blind.
Cancer incidence (11 trials with 398 cancers during 51,681 person years): For metformin compared to any comparator, the relative risk of any cancer was not statistically significant (1.02, 95% CI 0.82 to 1.26; Ι²=24%). In subgroup analyses, relative risk of any cancer was not statistically significant (0.98, 95% CI 0.77 to 1.23; Ι²=13%; six trial arms) when metformin was compared with active comparators. Relative risk of any cancer was not statistically significant (1.36, 95% CI 0.74 to 2.49) when metformin was compared with placebo or usual care (number of trials not clear). Sensitivity analysis with a random-effects model, subgroup analysis according to duration of the trial and post hoc analyses restricted to people with diabetes or where metformin was used as monotherapy did not markedly change the findings. Results were not influenced by differences in study quality.
All-cause mortality (13 trials with 566 recorded deaths during 66,447 person years): For metformin compared to any comparator, relative risk of all-cause mortality was 0.99 (95% CI 0.83 to 1.17; Ι²=23%). In subgroup analyses, relative risk of all-cause mortality was 1.05 (95% CI 0.84 to 1.33; Ι²=0%; seven trials) when metformin was compared with active comparators. Relative risk of all-cause mortality was 0.91 (95% CI 0.7 to 1.18; Ι²=51%; six trial arms) when metformin was compared with placebo or usual care. Subgroup analyses according to trial duration did not change the results markedly. The authors reported that a post hoc analysis with the exclusion of one trial reduced all-cause mortality when metformin was compared with usual care or placebo (RR 0.67, 95% CI 0.49 to 0.93; Ι²=0%). There was no evidence of a significant difference when metformin was compared with any comparator (RR 0.91, 95% CI 0.75 to 1.10).
There was no statistically significant beneficial effect of metformin on cancer or mortality outcomes but the evidence base was limited.
The review addressed a clear research question supported by appropriate inclusion criteria. Relevant sources were searched and attempts made to find unpublished studies which minimised the chance of publication bias. Appropriate methods were used to select studies but the authors did not state how many reviewers performed quality assessment or data extraction so reviewer error and bias could not be ruled out. A valid tool was used to assess studies for risk of bias; some studies were not blinded and the authors acknowledged that this could have resulted in bias.
The included studies had a wide range of comparators but analyses were stratified to assess differential effects. Arms in some trials included other glucose-lowering drugs so it was difficult to separate out the effects of metformin alone compared with control. Synthesis of studies and assessment of heterogeneity was appropriate. Appropriate sensitivity and subgroup analyses were undertaken to assess the robustness of results to variations in the included studies. As the outcomes of interest in the review were not the primary outcomes in the original studies there was variability in the measurement of cancer outcomes and lack of adjudicated endpoints that potentially influenced the accuracy of the data. The authors documented compliance and other treatment details in tables available with the publication. There were insufficient studies to assess effects on individual cancer endpoints or to assess long-term outcomes. Confidence intervals for many of the findings were wide and this made it difficult to determine the reliability of the results.
The review was generally well conducted. The authors noted the shortcomings in the quality of the evidence base and clinical differences between studies. They concluded that although the review found no beneficial effect of metformin on cancer or mortality, more research was required to further explore this. This conclusion seems appropriate.
Implications of the review for practice and research
Practice: The authors did not state any implications for practice.
Research: The authors stated that two ongoing trials were investigating whether metformin lowered cancer risk.
Stevens RJ, Ali R, Bankhead CR, Bethel MA, Cairns BJ, Camisasca RP, Crowe FL, Farmer AJ, Harrison S, Hirst JA, Home P, Kahn SE, McLellan JH, Perera R, Pluddemann A, Ramachandran A, Roberts NW, Rose PW, Schweizer A, Viberti G, Holman RR. Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meta-analysis of randomised clinical trials. Diabetologia 2012; 55(10): 2593-2603
Other publications of related interest
Stevens RJ, Ali R, Bankhead M, Bethel MA, Cairns BJ, Camisasca RP et al. Erratum to: Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meat-analysis of randomised clinical trials. Diabetologia (2012) 55: 3399-3400.
Subject indexing assigned by NLM
Adult; Aged; Diabetes Complications /complications; Diabetes Mellitus /drug therapy; Female; Follow-Up Studies; Humans; Hypoglycemic Agents /therapeutic use; Male; Metformin /therapeutic use; Middle Aged; Neoplasms /epidemiology /mortality; Randomized Controlled Trials as Topic; Risk Factors; Survival Rate
Date bibliographic record published
Date abstract record published
This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.