Metformin and cancer in type 2 diabetes: a systematic review and comprehensive bias evaluation.

Existing Reviews

Please note, new claims can take a short while to show up.

No claims yet.

Annotation Summary

Term Occurence Count Dictionary
diabetes mellitus 1 endocrinologydiseases
metformin 72 endocrinologydiseasesdrugs
type 2 diabetes mellitus 1 endocrinologydiseases

Graph of close proximity drug and disease terms (within 200 characters).

Note: If this graph is empty, then there are no terms that meet the proximity constraint.

Review

Having read the paper, please pick a pair of statements from the paper to indicate that a drug and disease are related.

Select Drug Character Offset Drug Term Instance
metformin 689 12/2016AbstractAbstractBackground: Existing observational studies provide conflicting evidence for the causal effect of metformin use on cancer risk in patients with type-2 diabetes, and there are concerns about bias affecting a number
metformin 914 studies.Methods: MEDLINE was used to identify observational studies investigating the association between metformin and overall or site-specific cancer in people with type-2 diabetes. A systematic data extraction and
metformin 1377 rated as unlikely, low, medium or high.Results: Of 46 studies identified, 21 assessed the effect of metformin on all cancer. Reported relative risks ranged from 0.23 to 1.22, with 12/21 reporting a statistically
metformin 1734 low or unlikely risk of bias in all domains. Two of these had results consistent with no effect of metformin ; one observed a moderate protective effect overall, but presented further analyses that the authors
metformin 2093 baseline adjustment and 35 from possible time-dependent confounding.Conclusions: Observational studies on metformin and cancer varied in design, and the majority were at risk of a range of biases. The studies least likely
metformin 2267 range of biases. The studies least likely to be affected by bias did not support a causal effect of metformin on cancer risk.IntroductionResearch exists to suggest type 2 diabetes mellitus (T2DM) may be a risk
metformin 2578 affect this risk.[3] Multiple observational studies have reported an apparent protective effect of metformin , a common first-line therapy for T2DM, against incidence of any cancer.[5] However, a number of potential
metformin 2810 identified within some of these studies.[10] There are limited data from clinical trials comparing metformin with other treatments, though one meta-analysis of adverse events from trials did not find any association
metformin 2935 treatments, though one meta-analysis of adverse events from trials did not find any association between metformin use and cancer occurrence.[11].Particular difficulties arise for observational studies in this context
metformin 3071 occurrence.[11].Particular difficulties arise for observational studies in this context because treatment with metformin for T2DM changes through time (is ‘time varying’), and is influenced by disease severity. This means
metformin 3236 and is influenced by disease severity. This means that disease severity may be a confounder between metformin use and cancer, but will also be on the causal pathway since metformin is prescribed in order to control
metformin 3307 may be a confounder between metformin use and cancer, but will also be on the causal pathway since metformin is prescribed in order to control disease severity. For example, glycated haemoglobin (HbA1c), a measure
metformin 3502 (HbA1c), a measure of long-term blood glucose control, and body mass index (BMI) are predictive of metformin use according to well-defined treatment guidelines for T2DM,[12] but use of metformin will likely influence
metformin 3588 predictive of metformin use according to well-defined treatment guidelines for T2DM,[12] but use of metformin will likely influence future HbA1c and BMI. There is also evidence that both BMI[13] and HbA1c[14] affect
metformin 4150 by prior treatment’ (Box 1).Reviews to date have examined existing evidence for the link between metformin use and cancer; however, some were not comprehensive[10] and others have not systematically evaluated
metformin 4423 this study was to summarize existing observational studies investigating possible associations between metformin use and cancer risk in patients with T2DM, and to systematically examine the research design and analysis
metformin 5939 Reviews and meta-analysis were not included. Studies were required to present a measure of effect of metformin on risk of cancer incidence (either all cancer or site-specific) in patients with T2DM, with age adjustment
metformin 6265 excluded.During an initial title and abstract screen, reviews, meta analyses and editorial pieces that looked at metformin and cancer were retained so that reference lists could be checked. Additionally, papers that appeared
metformin 8272 (14)10 (42)13 (28) Prostate3 (14)8 (33)11 (24) Stomach1 (5)4 (17)5 (11)Definition of exposure to metformin for primary estimate Any exposure14 (64)8 (33)22 (48) Any exposure but minimum time/number of prescriptions
metformin 8466 time/number of prescriptions needed1 (5)2 (8)3 (7) Total exposure (number of prescriptions/time on metformin )6 (27)4 (17)10 (22) Monotherapy1 (5)8 (33)9 (20) Randomization0 (0)1 (4)1 (2) Combination therapy
metformin 9223 (2) Rosiglitazone0 (0)1 (4)1 (2) Sulphonylurea2 (9)9 (38)11 (24) Any other OAD3 (14)4 (17)7 (15) No metformin (combination of diet and other OADs)17 (77)7 (29)24 (52)bNew users of OADs Yes3 (14)7 (29)10 (22) No17
metformin 11098 and analysis methods were unlikely to induce a systematic difference between risk of cancer between metformin users and non-users. Low risk meant that there was small possibility of bias but the potential magnitude
metformin 13556 for. If studies omitted a particular confounder because they found it did not alter the estimate of metformin on cancer risk in a multivariable model, then they were not deemed to be at risk of bias due to its
metformin 14884 exploratory analysis designed to investigate whether between-study heterogeneity in the observed effect of metformin could be explained by bias and other study level factors, a random effects meta-regression was performed.
metformin 15646 Characteristics included were comparator exposure [diet only, other oral antidiabetic drugs (OADs), less metformin and no metformin (diet and other OADs combined)], bias in exposure definition, bias in outcome definition,
metformin 15663 included were comparator exposure [diet only, other oral antidiabetic drugs (OADs), less metformin and no metformin (diet and other OADs combined)], bias in exposure definition, bias in outcome definition, bias from
metformin 19862 collected data from a specific cancer or diabetes clinic.A total of 22 studies (46%) defined exposure to metformin as any exposure, without considering overall duration. Three further studies refined this definition
metformin 20126 prescriptions before an individual was considered exposed. Nine studies (20%) looked at monotherapy with metformin and 10 studies (22%) used total exposure to enable dose-response analyses. The remaining two studies
metformin 20247 studies (22%) used total exposure to enable dose-response analyses. The remaining two studies looked at metformin in combination with specific OADs, with a comparator group that allowed the estimation of the effect
metformin 20366 combination with specific OADs, with a comparator group that allowed the estimation of the effect of just metformin . The most frequently used comparator group was no metformin, used in 24 studies (52%). Use of sulphonylurea
metformin 20426 allowed the estimation of the effect of just metformin. The most frequently used comparator group was no metformin , used in 24 studies (52%). Use of sulphonylurea [another popular first-line oral agent; 11 studies (24%)]
metformin 20624 11 studies (24%)] was also a common comparator.There were 116 estimates presented for the effect of metformin on risk of cancer when considering separate estimates for different cancer sites. A total of 21 studies
metformin 21045 lung (12 studies) and prostate (11 studies). Other sites had less than 10 estimates each.Effect of metformin on cancer riskFigure 2 displays the study estimates and 95% confidence intervals (CIs) for relative
metformin 21202 estimates and 95% confidence intervals (CIs) for relative risk [odds ratio (OR) or hazard ratio (HR)] of metformin use on incidence of all cancer. Estimates and 95% CIs for the four most commonly studied site-specific
metformin 21459 2.Estimated relative risk (odds ratio or hazard ratio) with 95% CI for the 21 studies examining use of metformin and risk of all cancers, and corresponding assessment of bias according to pre-specified criteria. a Represents
metformin 21646 pre-specified criteria. a Represents the hazard ratio for cancer risk per one extra prescription of metformin .Figure 3.Estimated relative risk (odds ratio or hazard ratio) with 95% CI for 4 most commonly studied
metformin 21934 triangle, Cohort studies by filled circles.For all cancer, 18/21 studies estimated a protective effect of metformin , with 12/16 having upper confidence limits below 1. The magnitude of the effect estimates ranged from
metformin 22511 missing data[43] where it was rated unknown. Three of these studies saw no evidence of an effect of metformin . One study estimated a modest protective effect of long-term use (> 60 months) in comparison with short-term
metformin 23055 estimates in Figure 2. Of the 12 studies that estimated a statistically significant protective effect of metformin , eleven had at least medium risk of bias in at least two domains; nine had medium or high risk of bias
metformin 23818 effect on log risk ratioP-value*Estimate 95% CI for effect on log risk ratioP valueComparator groupNo metformin 0 (ref)0 (ref)0 (ref)0 (ref)0 (ref)Diet only−1.16 (−2.41, 0.10)d0.217Less metformin0.14 (−0.28,
metformin 23905 valueComparator groupNo metformin0 (ref)0 (ref)0 (ref)0 (ref)0 (ref)Diet only−1.16 (−2.41, 0.10)d0.217Less metformin 0.14 (−0.28, 0.55)−0.04 (−0.43 , 0.34)d0.3860.05 (−0.34, 0.44)d0.107−0.37 (−0.81, 0.07)d0.625−1.66
metformin 25337 expected change in the log risk ratio (either HR or OR, depending on analysis method) for the effect of metformin on cancer, for each study level predictor. For example, a study of metformin and lung cancer, in which
metformin 25414 method) for the effect of metformin on cancer, for each study level predictor. For example, a study of metformin and lung cancer, in which there is high risk of bias from exposure definition, is estimated to have
metformin 26387 other antidiabetic treatments were accounted for in the analysis. These studies considered exposure to metformin as fixed from baseline [‘intention to treat’ (ITT) principle], and had confounders measured immediately
metformin 27395 therefore the potential for these studies to have adjusted for factors on the causal pathway between metformin and cancer was high. For the cohort studies, most used BMI and HBA1c measurements at or close to the
metformin 28539 different time windows to measure exposure, meaning the overall chance of seeing individuals exposed to metformin is systematically different between the cases and controls. Bias was most often introduced into cohort
metformin 30638 and exposure definition. The model estimated that using a comparator group of diet, as opposed to no metformin , made metformin appear more protective, whereas using other OADs or less metformin as a reference group
metformin 30654 definition. The model estimated that using a comparator group of diet, as opposed to no metformin, made metformin appear more protective, whereas using other OADs or less metformin as a reference group made metformin
metformin 30721 as opposed to no metformin, made metformin appear more protective, whereas using other OADs or less metformin as a reference group made metformin appear less protective. However, this model was estimated to still
metformin 30757 metformin appear more protective, whereas using other OADs or less metformin as a reference group made metformin appear less protective. However, this model was estimated to still have 85% residual variation due to
metformin 31084 models for colorectal, lung and breast cancers, using other OADs as the comparator was estimated to make metformin appear more protective.The strongest predictor of heterogeneity for studies of lung cancer was risk
metformin 31336 present, was estimated to reduce the log risk ratio by 0.44, 95% CI (0.17, 0.72) P = 0.007, making metformin appear more protective. For breast cancer, the strongest predictor was use of an incident user cohort,
metformin 31460 protective. For breast cancer, the strongest predictor was use of an incident user cohort, which made metformin look less protective. This predictor was also identified for studies of lung and pancreatic cancer,
metformin 31946 colorectal cancer, the strongest predictor was biased exposure definition, which was estimated to make metformin appear more protective.DiscussionThe 46 studies examined in this review did not provide consistent evidence
metformin 32098 studies examined in this review did not provide consistent evidence to support a protective effect of metformin on risk of cancer. Two of three studies with low or unlikely risk of bias for all categories had estimates
metformin 32244 studies with low or unlikely risk of bias for all categories had estimates consistent with no effect of metformin . The third study had an estimate consistent with a moderate protective effect; however this study included
metformin 32414 protective effect; however this study included many analyses, and also reported that when comparing metformin exposure with other classes of oral antidiabetics, the risk of cancer did not differ between drugs.
metformin 34056 those studies with the highest risk of bias overall.Figure 4.Estimates of relative risk of cancer from metformin use, ordered by risk of bias from exposure assessment only (left) and by overall risk of bias (right).
metformin 34694 possible to eliminate all subjectivity from this process.Figure 5A represents the total causal effect of metformin use on cancer risk that we wish to estimate in a simple example where we assume HbA1c is the only time-dependent
metformin 35214 measurement is taken any time during follow-up, which may result in ‘adjusting out’ any effect of metformin that is mediated through HbA1c. In Figure 5C, because treatment may change after baseline, the single
metformin 36164 When applied and analysed carefully, it will give an unbiased estimate of the effect of initiating metformin compared with initiating (as an example) sulphonylurea on development of cancer. However, this is not
metformin 36346 cancer. However, this is not necessarily equivalent to estimating causal pharmacological effect of metformin use on cancer incidence and may be inappropriate if the comparator in question may itself affect risk
metformin 36835 meaningful impact of time-dependent confounders affected by previous treatment on the estimated effect of metformin on cancer risk.Figure 5.Directed Acyclic Graphs (DAGs) to represent estimated causal pathways for A)
metformin 38107 the estimate of effect under this approach.In order to estimate the causal pharmacological effect of metformin on risk of cancer, the ideal would be to emulate a randomized controlled trial where patients are randomized
metformin 38236 the ideal would be to emulate a randomized controlled trial where patients are randomized to either metformin or diet only. This would involve comparison of patients initiating metformin with those controlling
metformin 38313 are randomized to either metformin or diet only. This would involve comparison of patients initiating metformin with those controlling their disease by diet only, and correctly adjusting for disease severity at time
metformin 40299 systematically identified and assessed the existing literature on the pharmacoepidemiological question of metformin use and cancer risk. The search identified a large number of studies from varying countries and journals,
metformin 42895 interpreted cautiously.Overall, the existing literature provides inconsistent answers to the question of metformin use and cancer risk in type 2 diabetes. Variation in design of studies and the potential for many kinds
metformin 43376 effects that have been observed are a result of immortal time bias and other issues relating to how metformin use is defined. Studies without such biases tend to have estimates closer to the null, and whereas an
metformin 43498 defined. Studies without such biases tend to have estimates closer to the null, and whereas an effect of metformin use on risk of subsequent cancer in patients with type 2 diabetes cannot be excluded, the previously
metformin 44182 conflict of interest to report.Key MessagesMany existing observational studies investigating the effect of metformin use on cancer incidence in patients with type 2 diabetes have risk of bias.No studies to date have used
metformin 44539 be affected by previous treatment.Studies at lowest risk of bias do not support the hypothesis that metformin is protective against cancer.Previously reported large protective associations are unlikely to be causal.Supplementary
Select Disease Character Offset Disease Term Instance
diabetes mellitus 2338 not support a causal effect of metformin on cancer risk.IntroductionResearch exists to suggest type 2 diabetes mellitus (T2DM) may be a risk factor for cancer,[1],[2] and observational studies have suggested that diabetic
type 2 diabetes mellitus 2331 did not support a causal effect of metformin on cancer risk.IntroductionResearch exists to suggest type 2 diabetes mellitus (T2DM) may be a risk factor for cancer,[1],[2] and observational studies have suggested that diabetic

You must be authorized to submit a review.