• Metformin use was associated with a significant decrease in risk of MPN, with the largest effect observed for the ET and PV subtypes.

  • A dose-response relationship supported the association with increasing treatment duration, in particular ≥5 years.

Abstract

Previous studies have suggested that metformin has beneficial effects beyond its glucose-lowering properties, particularly in terms of its potential as an antineoplastic and cancer-preventive agent. In this study, we aimed to investigate the association between metformin use and the risk of myeloproliferative neoplasms (MPN). We conducted a population-based case-control study using Danish registers. Cases with MPN diagnosed between 2010 and 2018 were identified, and metformin use before the MPN diagnosis was ascertained. We compared metformin use among cases with MPN and an age- and sex-matched control group from the Danish general population to estimate age- and sex-adjusted odds ratios (ORs) and fully adjusted ORs (aORs) for the association between metformin use and risk of MPN. The study population included 3816 cases and 19 080 controls. Overall, 7.0% of cases and 8.2% of controls were categorized as ever-users of metformin, resulting in an OR for MPN of 0.84 (95% confidence interval [CI], 0.73-0.96) and an aOR of 0.70 (95% CI, 0.61-0.81). Long-term metformin use (≥5 years) was more infrequent and comprised 1.1% of cases and 2.0% of controls, resulting in an OR of 0.57 (95% CI, 0.42-0.79) and an aOR of 0.45 (95% CI, 0.33-0.63). A dose-response relationship was observed when cumulative duration of treatment was analyzed, and this was consistent in stratified analyses of sex, age, and MPN subtypes. In conclusion, metformin use was associated with significantly lower odds of an MPN diagnosis, indicating its potential cancer-preventive effect. Given the retrospective design, causality cannot be inferred.

The classic Philadelphia chromosome–negative myeloproliferative neoplasms (MPN) are clonal neoplastic disorders that include the disease entities essential thrombocythemia (ET), polycythemia vera (PV), myelofibrosis (MF), and MPN unclassifiable (MPN-U). MPN are characterized by acquired somatic mutations, most commonly in the genes JAK2, CALR, or MPL, resulting in excessive proliferation of myeloid progenitors producing abnormal peripheral blood counts.1-3 Many patients experience constitutional symptoms due to a hyperinflammatory and hypermetabolic state affecting quality of life and well-being. Additionally, excessive formation of bone marrow (BM) fibrosis ultimately leads to disruption of normal BM function, with ineffective hematopoiesis leading to severe cytopenia and increased risk of leukemic transformation.4 The compensatory extramedullary hematopoiesis, mainly in the spleen, exacerbates constitutional symptoms and causes symptomatic organomegaly.

Metformin, an oral biguanide derivative used in the treatment of type 2 diabetes mellitus or metabolic syndrome, has become so widely used over the past decades that the drug is now among the most prescribed drugs in Denmark.5 The use of metformin has been associated with reduced solid tumor risk and mortality in meta-analyses,6 but results are inconsistent across studies and solid tumor subtypes. The cancer-preventive and clinical effects of metformin for hematological neoplasms, in particular MPN, are not well-investigated. However, preclinical studies have shown antileukemic activity of metformin in myeloid neoplasms by activating the AMP-activated protein kinase (AMPK)-mTOR pathway, downregulating JAK2/STAT signaling, and influencing mitochondrial activity, suggesting possible antineoplastic effects in myeloid neoplasms.7-9 Knowledge regarding potential pleiotropic and cancer-preventive effects of metformin is of great clinical and public health interest in a time with increased focus on drug repurposing for cancer treatment and prevention. The present study investigated the association between metformin use and the risk of MPN in the Danish population.

Study design and Danish nationwide health registers

This Danish population–based case-control study compared metformin use between patients diagnosed with MPN (cases) and a matched population from the Danish general population (controls) in the period between 2010 and 2018. Associations between MPN and metformin use, including duration of use, was estimated using odds ratios (ORs).

The following 5 nationwide health registers were merged by using the unique 10-digit personal identification number (CPR number) given to Danish citizens at birth or immigration10,11: the Danish Civil Registration System,10,11 the National Patient Register,12,13 the Danish National Chronic Myeloid Neoplasia Registry (DCMR),14 the Danish National Prescription Registry (DNPR),15,16 and the Danish Education Registers17 (the supplemental Material includes more detailed information).

Case and control selection

Cases with MPN were identified using the DCMR covering over 90% of MPN cases in Denmark since 2010.14 Cases had to be Danish residents with an MPN (ET, PV, MF, or MPN-U) diagnosis according to the World Health Organization (WHO) criteria from 2008 or later18,19 between 1 January 2010 and 31 December 2018. The index date was defined as the first registered date of the MPN diagnosis from the DCMR. Moreover, to ensure a medical history spanning a minimum of 10 years, all cases and controls were required to have been continuous residents for at least 10 consecutive years before the index date. Controls were selected by incidence density sampling in a 5:1 control-to-case ratio from the Danish general population matched on age, sex, and index date.20 Cases and controls were excluded if MPN or other cancers had been diagnosed before the index date (supplemental Material includes more detailed information, with the exception of nonmelanoma skin cancer and carcinomas in situ). The exclusion of previous cancers was aimed at preventing the inclusion of therapy-related MPN.

Metformin exposure assessment

Information on metformin exposure was retrieved from the DNPR (ATC code: A10BA02). Ever-use of metformin was defined as having redeemed at least 1 prescription of metformin before the index date, and never-use was defined as no filled prescription of metformin before the index date. Long-term use was defined as ≥5 years of cumulative treatment with metformin before the MPN diagnosis. The DNPR does not contain information on prescribed doses or indication for the treatment. Therefore, cumulative treatment duration was determined on the basis of the WHO-defined daily dose of metformin at 2000 mg, with an additional 25% of days to accommodate variations in prescription filling patterns and minor instances of noncompliance. The use of metformin in the 12 months before the index date was disregarded in the primary analyses to avoid reverse causation.

Covariates and confounders

Data from the National Patient Register (diagnoses to assess comorbidity), the DNPR (prescriptions related to conditions or use of drug with suggested cancer-protective properties), and the Danish Education Registers (demographics and education) were used to adjust for the following prespecified potential confounders: (1) highest achieved educational level (primary school; high school; short/middle long education; long education) as a measure of socioeconomic status; (2) Charlson Comorbidity Index21 (CCI: 0; 1; ≥2); 3) use of drugs with suggested cancer-preventive effects22 including aspirin, nonsteroidal anti-inflammatory drugs, alendronate, immunosuppressants, and statins (≥2 previous prescriptions for all); and (4) markers of smoking (diagnosis of chronic obstructive pulmonary disease or use of bronchodilating inhalation agents containing an antimuscarinic component), markers of autoimmune diseases (AD) (diagnosis codes or prescription of immunomodulating drugs), and diagnoses related to excessive alcohol consumption (diagnose codes and prescription of disulfiram). Information to compute CCI was collected from the National Patient Register, using ICD-10 codes associated with the diagnoses of 19 chronic conditions. Detailed information regarding included ICD- 10 codes and definitions is provided in supplemental Table 1. In the primary analysis, data used for assessing confounding were excluded for the 12-months period leading up to the index date to avoid misclassification.

Statistical analysis

Categorical variables were reported as counts and percentages, and continuous variables were reported as medians with interquartile range ([IQR]; 25th to 75th percentile). For our primary analysis, we assessed the association between metformin use (exposure) and MPN (outcome) using conditional logistic regression to obtain sex- and age-adjusted ORs and a fully-adjusted model (aORs, adjusted for sex, age, calendar time, educational level, CCI, previous use of drugs with suggested cancer-preventive effects, history of alcohol-related diagnoses, overweight/obesity, chronic obstructive pulmonary disease, and AD) provided with 95% confidence intervals (CIs). Analyses stratified by subtype of MPN (MF, PV, ET, and MPN-U), sex (male and female), age group (<60 years; 60-75 years; >75 years), molecular subtype (JAK2- or CALR-mutated [see supplemental Table 2 for additional information]), and absence of AD were performed to investigate subgroup associations between long-term metformin use and MPN risk. Furthermore, for long-term use, analyses were repeated with single potential confounders to show the effects of single confounder adjustments and to analyze the contribution of each suggested confounder. Conducting a sensitivity analysis, we varied the lag time (ie, the disregarded time before index date when assessing exposure and confounders) for the analysis of long-term use by 6-month increments.

Statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc), Rstudio version 1.1.447 (Rstudio, Inc), and R version 3.6.1 (R Foundation for Statistical Computing). The study was approved by the institution responsible for the data (North Denmark Region, approval number: 2021-034) in accordance with the General Data Protection Regulation.

A total of 3816 cases and 19 080 matched controls were included (Table 1). The case cohort comprised 1306 (34.2%) patients with PV, 1307 (34.3%) with ET, 574 (15.0%) with MF, and 629 (16.5%) with MPN-U. Most cases had available information on molecular testing (96%), 2871 (78.4%) had a JAK2-V617F mutation, and 184 (5.0%) had a CALR mutation. Aspirin use was more common among cases (34.4% vs 26.4%), and cases had a higher CCI. Otherwise, baseline characteristics were balanced between groups (Table 1).

Table 1.

Study cohort characteristics for cases and controls

CasesControls
Total, n 3816 19 080 
Age, median, y (IQR) 69 (59-76) 69 (59-76) 
Age group, n (%)   
<60 y, n (%) 981 (25.7) 4905 (25.7) 
60-75 y, n (%) 1755 (46.0) 8775 (46.0) 
>75 y, n (%) 1080 (28.3) 5400 (28.3) 
Male sex, n (%) 1875 (49.1) 9375 (49.1) 
Metformin use before index date, n (%)   
Never-use 3548 (93.0) 17 507 (91.8) 
Ever-use 268 (7.0) 1573 (8.2) 
Long-term use (≥5 y) 43 (1.1) 375 (2.0) 
No. of prescriptions for ever-users, n (%)   
15 (5.6) 98 (6.2) 
2-4 37 (13.8) 183 (11.6) 
≥5 216 (80.6) 1292 (82.1) 
Highest achieved education, n (%)   
Primary school 1277 (33.5) 6419 (33.6) 
High school 1576 (41.3) 7608 (39.9) 
Short/middle long education 659 (17.3) 3394 (17.8) 
Long education 216 (5.7) 1139 (6.0) 
CCI, n (%)   
3292 (86.3) 17 614 (92.3) 
340 (8.9) 1038 (5.4) 
≥2 184 (4.8) 428 (2.2) 
Medical history, n (%)   
Alcohol-related diagnoses 262 (6.9) 1104 (5.8) 
Overweight and obesity-related diagnoses 155 (4.1) 841 (4.4) 
Chronic obstructive pulmonary disease 348 (9.1) 1527 (8.0) 
Autoimmune disease 325 (8.5) 1611 (8.4) 
Previous drug use, n (%)   
Aspirin 1312 (34.4) 5034 (26.4) 
Other NSAIDs 3140 (82.3) 15 345 (80.4) 
Statins 1332 (34.9) 6389 (33.5) 
Alendronate 242 (6.3) 1114 (5.8) 
Immunosuppressants 95 (2.5) 492 (2.6) 
MPN subtype, n (%)   
PV 1306 (34.2) N/A 
ET 1307 (34.3) N/A 
MF 574 (15.0) N/A 
MPN-U 629 (16.5) N/A 
Molecular MPN subtype , n (%)   
JAK2-V617F–mutated 2871 (78.4) N/A 
CALR-mutated 184 (5.0) N/A 
CasesControls
Total, n 3816 19 080 
Age, median, y (IQR) 69 (59-76) 69 (59-76) 
Age group, n (%)   
<60 y, n (%) 981 (25.7) 4905 (25.7) 
60-75 y, n (%) 1755 (46.0) 8775 (46.0) 
>75 y, n (%) 1080 (28.3) 5400 (28.3) 
Male sex, n (%) 1875 (49.1) 9375 (49.1) 
Metformin use before index date, n (%)   
Never-use 3548 (93.0) 17 507 (91.8) 
Ever-use 268 (7.0) 1573 (8.2) 
Long-term use (≥5 y) 43 (1.1) 375 (2.0) 
No. of prescriptions for ever-users, n (%)   
15 (5.6) 98 (6.2) 
2-4 37 (13.8) 183 (11.6) 
≥5 216 (80.6) 1292 (82.1) 
Highest achieved education, n (%)   
Primary school 1277 (33.5) 6419 (33.6) 
High school 1576 (41.3) 7608 (39.9) 
Short/middle long education 659 (17.3) 3394 (17.8) 
Long education 216 (5.7) 1139 (6.0) 
CCI, n (%)   
3292 (86.3) 17 614 (92.3) 
340 (8.9) 1038 (5.4) 
≥2 184 (4.8) 428 (2.2) 
Medical history, n (%)   
Alcohol-related diagnoses 262 (6.9) 1104 (5.8) 
Overweight and obesity-related diagnoses 155 (4.1) 841 (4.4) 
Chronic obstructive pulmonary disease 348 (9.1) 1527 (8.0) 
Autoimmune disease 325 (8.5) 1611 (8.4) 
Previous drug use, n (%)   
Aspirin 1312 (34.4) 5034 (26.4) 
Other NSAIDs 3140 (82.3) 15 345 (80.4) 
Statins 1332 (34.9) 6389 (33.5) 
Alendronate 242 (6.3) 1114 (5.8) 
Immunosuppressants 95 (2.5) 492 (2.6) 
MPN subtype, n (%)   
PV 1306 (34.2) N/A 
ET 1307 (34.3) N/A 
MF 574 (15.0) N/A 
MPN-U 629 (16.5) N/A 
Molecular MPN subtype , n (%)   
JAK2-V617F–mutated 2871 (78.4) N/A 
CALR-mutated 184 (5.0) N/A 

IQR, interquartile range; NSAIDs, nonsteroidal anti-inflammatory drugs.

Among 3664 tested.

For the primary analysis, a total of 268 (7.0%) cases were categorized as ever-users of metformin as opposed to 1573 (8.2%) of controls, yielding an age- and sex-adjusted OR for MPN of 0.84 (95% CI, 0.73-0.96). Long-term use (≥5 years) was observed in 1.1% of cases and 2.0% of controls, resulting in an OR of 0.57 (95% CI, 0.42-0.79) (Table 2). When adjusting for potential confounders in the fully-adjusted model, the aORs for MPN were 0.70 (95% CI, 0.61-0.81) and 0.45 (95% CI, 0.33-0.63) for ever-users and long-term users, respectively (Table 2). When assessing the effect of cumulative duration of treatment, no clear relationship was found for the sex- and age-adjusted estimates. However, a dose-response relationship was present in the fully-adjusted model with aORs of 0.79 (95% CI, 0.62-1.00) for <1 year, 0.78 (95% CI, 0.64-0.95) for 1 to 4.99 years, 0.42 (95% CI, 0.29-0.61) for 5 to 9.99 years, and 0.56 (95% CI, 0.30-1.03) for >10 years (Table 2).

Table 2.

Association between exposure to metformin and risk of myeloproliferative neoplasm according to duration of metformin exposure

SubgroupCases, nControls, nOR (95% CI)Adjusted OR (95% CI)
Exposure     
Never-use 3548 17 507 1.0 (reference) 1.0 (reference) 
Ever-use 268 1573 0.84 (0.73-0.96) 0.70 (0.61-0.81) 
Long-term use (≥5 y) 43 375 0.57 (0.42-0.79) 0.45 (0.33-0.63) 
Cumulative treatment duration     
<1 y 86 465 0.91 (0.72-1.15) 0.79 (0.62-1.00) 
1-4.99 y 139 733 0.93 (0.77-1.12) 0.78 (0.64-0.95) 
5-9.99 y 31 287 0.53 (0.37-0.77) 0.42 (0.29-0.61) 
≥10 y 12 88 0.67 (0.37-1.23) 0.56 (0.30-1.03) 
SubgroupCases, nControls, nOR (95% CI)Adjusted OR (95% CI)
Exposure     
Never-use 3548 17 507 1.0 (reference) 1.0 (reference) 
Ever-use 268 1573 0.84 (0.73-0.96) 0.70 (0.61-0.81) 
Long-term use (≥5 y) 43 375 0.57 (0.42-0.79) 0.45 (0.33-0.63) 
Cumulative treatment duration     
<1 y 86 465 0.91 (0.72-1.15) 0.79 (0.62-1.00) 
1-4.99 y 139 733 0.93 (0.77-1.12) 0.78 (0.64-0.95) 
5-9.99 y 31 287 0.53 (0.37-0.77) 0.42 (0.29-0.61) 
≥10 y 12 88 0.67 (0.37-1.23) 0.56 (0.30-1.03) 

NSAIDs, nonsteroidal anti-inflammatory drugs.

Adjusted for age, sex, and calendar time.

Adjusted for in addition to the following: (1) education level (primary school; high school; short/middle long education; long education); (2) CCI (0; 1; ≥2); (3) previous use of aspirin or other NSAIDs; alendronate; immunosuppressants; and statins; and (4) history of alcohol-related diagnoses, overweight and obesity-related diagnoses, chronic obstructive pulmonary disease, and AD.

In stratified analyses including long-term users (≥5 years of cumulative treatment), the magnitude of the effect was observed across all age groups, with aORs of 0.47 (95% CI, 0.15-1.50) for <60 years, 0.45 (95% CI, 0.28-0.71) for 60 to 75 years, and 0.46 (95% CI, 0.27-0.78) for >75 years (Table 3). Furthermore, the protective effect of long-term metformin use was consistent according to sex, but stronger in male patients, with an aOR of 0.37 (95% CI, 0.23-0.60) compared with 0.56 (95% CI, 0.36-0.89) among female patients (Table 3). The protective effect of metformin was observed in all subtypes of MPN (PV: aOR, 0.45; [95% CI, 0.26-0.77]; ET: aOR, 0.33; [95% CI, 0.16-0.67]; MF: aOR, 0.65; [95% CI, 0.32-1.33]; MPN-U: aOR, 0.46; [95% CI, 0.20-1.06]) (Table 3). Additionally, when restricting the analysis to molecular subtypes, similar magnitude of protective effect was observed, with aORs of 0.47 (95% CI, 0.32-0.68) and 0.53 (95% CI, 0.11-2.50) for JAK2-V617F- and CALR-mutated MPN, respectively. Finally, when restricting the analysis to cases and controls without markers of AD, the effect was of the same magnitude as observed in the primary analysis, with an aOR of 0.46 (95% CI, 0.32-0.64) (Table 3).

Table 3.

Association between long-term exposure to metformin (≥5 years) and risk of myeloproliferative neoplasm according to subgroups

SubgroupOR (95% CI)Adjusted OR (95% CI)
Age group   
<60 y 0.96 (0.32-2.81) 0.47 (0.15-1.50) 
60-75 y 0.59 (0.38-0.91) 0.45 (0.28-0.71) 
>75 y 0.50 (0.30-0.85) 0.46 (0.27-0.78) 
Sex   
Male 0.46 (0.29-0.74) 0.37 (0.23-0.60) 
Female 0.71 (0.46-1.09) 0.56 (0.36-0.89) 
Disease subtype   
PV 0.57 (0.34-0.97) 0.45 (0.26-0.77) 
ET 0.39 (0.20-0.78) 0.33 (0.16-0.67) 
MF 0.89 (0.46-1.72) 0.65 (0.32-1.33) 
MPN-U 0.59 (0.27-1.30) 0.46 (0.20-1.06) 
Molecular subtype   
JAK2-V617F-mutated 0.61 (0.43-0.87) 0.47 (0.32-0.68) 
CALR-mutated 0.57 (0.13-2.48) 0.53 (0.11-2.50) 
No AD 0.59 (0.43-0.83) 0.46 (0.32-0.64) 
SubgroupOR (95% CI)Adjusted OR (95% CI)
Age group   
<60 y 0.96 (0.32-2.81) 0.47 (0.15-1.50) 
60-75 y 0.59 (0.38-0.91) 0.45 (0.28-0.71) 
>75 y 0.50 (0.30-0.85) 0.46 (0.27-0.78) 
Sex   
Male 0.46 (0.29-0.74) 0.37 (0.23-0.60) 
Female 0.71 (0.46-1.09) 0.56 (0.36-0.89) 
Disease subtype   
PV 0.57 (0.34-0.97) 0.45 (0.26-0.77) 
ET 0.39 (0.20-0.78) 0.33 (0.16-0.67) 
MF 0.89 (0.46-1.72) 0.65 (0.32-1.33) 
MPN-U 0.59 (0.27-1.30) 0.46 (0.20-1.06) 
Molecular subtype   
JAK2-V617F-mutated 0.61 (0.43-0.87) 0.47 (0.32-0.68) 
CALR-mutated 0.57 (0.13-2.48) 0.53 (0.11-2.50) 
No AD 0.59 (0.43-0.83) 0.46 (0.32-0.64) 

NSAIDs, nonsteroidal anti-inflammatory drugs.

Adjusted for age, sex, and calendar time.

Adjusted for in addition to the following: (1) education level (primary school; high school; short/middle long education; long education); (2) CCI (0; 1; ≥2); (3) previous use of aspirin or other NSAIDs, alendronate, immunosuppressants, or statins; and (4) history of alcohol-related diagnoses, overweight and obesity-related diagnoses, chronic obstructive pulmonary disease, and autoimmune disease

As a supplemental analysis, the effects of adjustment for single potential confounders were investigated, and estimates are given in supplemental Table 2. With the exception of adjustments for the previous use of aspirin and CCI, adjustment for each confounder had limited influence on the estimate. As a sensitivity analysis to assess the potential of reverse causation, we varied the lag time in 6-month increments, and no large differences were observed for our estimate (supplemental Table 3).

In this large Danish population–based case-control study, we showed an association between metformin use and decreased risk of MPN. The magnitude of this association showed a clear dose-response relationship with increasing cumulative treatment duration. Furthermore, this effect was numerically consistent in stratified analyses, implying effect in both male and female patients, different age strata, and MPN subtypes including molecular subtypes, although not all estimates reached statistical significance because of small numbers. Collectively, the results indicate that metformin may reduce the risk of MPN.

To our knowledge, this study is the first to investigate this association specifically for MPN. However, a recent study collectively investigated hematological cancers including multiple myeloma, leukemia, and lymphoma among US veterans with diabetes.23 In this large retrospective cohort study using inverse probability of treatment weighting, metformin users aged 55 to 75 years had significantly lower risk for both incident solid and hematological cancers than sulfonylurea users with diabetes.23 The estimated hazard ratio (HR) for hematological malignancy was similar to that of solid cancers, with inverse probability of treatment weighting-adjusted HRs for hematological cancer of 0.68 (95% CI, 0.54-0.87) and 0.79 (95% CI, 0.63-0.98) for age of 55 to 64 and 65 to 75 years, respectively.23 In comparison with our study (and taking into account different age strata, control groups, and study designs), which found the most pronounced effect in older individuals aged >59 years (including those aged >75 years), Abdallah et al23 recorded the most pronounced effect in younger individuals and no effect in individuals aged ≥75 years.

Similarly, an observational study using the US veteran cohort investigated the effect of metformin use (≥4 years of treatment) on the progression to multiple myeloma from monoclonal gammopathy of undetermined significance. This study showed a significantly reduced risk of progression associated with metformin treatment, with an adjusted HR of 0.47 (95% CI, 0.25-0.87).24 Like multiple myeloma preceded by the premalignant disorder monoclonal gammopathy of undetermined significance, MPN develops from a stage characterized by clonal hematopoiesis of indeterminate potential (CHIP), and strong evidence suggests that CHIP-MPN are extremely underdiagnosed.25 A previous Danish study investigating the population prevalence of JAK2-V617F- and CALR-mutated clones surprisingly found a prevalence of 3.2%, with most being small clones with variant allele frequency <1%.25 Notably, only a minor fraction of these individuals had an MPN diagnosis, but ∼50% had abnormal laboratory findings. A CHIP stage of MPN with vague symptoms and chronic inflammation may also explain an observed increase in the utilization of health care resources before the diagnosis of an overt MPN.26 

Because the present study is merely observational and registry-based, we cannot assess the mechanism by which metformin might protect against the development of MPN. However, additionally supportive of a CHIP-MPN hypothesis, a recent phase 2 study failed to demonstrate an effect of metformin in reversing bone marrow fibrosis in patients with MF. Yet, the study showed significant downregulation of the JAK-STAT pathway and reduced levels of the proinflammatory cytokine interleukin-5.27 These results indicate that metformin does inhibit the central pathway in MPN in addition to dampening the chronic inflammation, which is considered a key factor for clonal expansion and evolution toward overt MPN.28,29 Consistent with this, we have previously shown that the use of statins, another agent assumed to possess antineoplastic and anti-inflammatory properties, also reduced the risk of developing MPN.30 Furthermore, a more recent study showed that statin treatment after the MPN diagnosis was associated with prolonged survival and reduced incidence of thrombosis.31 

Many other studies have investigated the association between metformin use and risk of solid cancers, and meta-analyses suggest a relative risk reduction at a magnitude of 30% to 55% for specific cancer sites including colorectal, liver, pancreatic, stomach, and esophageal cancer in addition to reduced risk of cancer-related death.6,32,33 Although observational studies provide a strong rationale for a cancer-protective effect of metformin, the efforts to explore this potential effect prospectively have been very limited. One reason may be the need for large numbers of study participants and long follow-up time to obtain adequate power to detect differences in cancer incidence. However, several clinical trials have been undertaken in populations at higher risk of developing malignancies. A phase 3 randomized study investigated the effect of metformin treatment on prevalence and number of colorectal adenoma or polyps among nondiabetic patients with recent polypectomy.34 This study showed that treatment with metformin resulted in a significantly lower risk of polyps and adenomas at the 1-year follow-up colonoscopy compared with placebo.34 Furthermore, among breast cancer patients treated with tamoxifen and at high risk of developing endometrial polyps, hyperplasia, and cancer, metformin appeared to inhibit and prevent tamoxifen-induced endometrial changes in a randomized clinical trial.35 Again, these studies suggest that metformin may excel in inhibiting premalignant states and hinder progression to overt disease.

Metformin either as a single agent or combined with cytotoxic, endocrine, or targeted therapy has also been investigated in a handful of solid cancers, but most of these studies have failed to demonstrate the clinical benefit of metformin.36 Latest, and most notably, the phase 3 randomized MA.32 trial, randomized over 3600 nondiabetic patients with high-risk nonmetastatic breast cancer to 850-mg metformin twice daily vs placebo for 5 years.37 The study did not meet the primary end point of improved invasive disease-free survival, and the investigators concluded that the addition of metformin for 5 years to standard-of-care treatment in this population did not show any beneficial clinical effects.37 Interestingly, in an explorative analysis of patients with human epidermal growth factor receptor 2–positive disease, the addition of metformin to standard-of-care was associated with longer invasive disease–free survival and overall survival for the C allele of the rs11212617 single-nucleotide variant. However, this should be interpreted with caution and is yet to be confirmed independently.37 

The present study has several strengths and limitations. Using the Danish registers ensures individual-level high-quality data and complete follow-up of every individual living in Denmark in the study period, allowing for detailed exposure and covariate measurements. Furthermore, using the DCMR to identify MPN cases ensures a high degree of diagnostic certainty, and the good coverage of the register of 90% in earlier periods and exceeding 90% in later periods entails limiting selection bias and high generalizability.14 Although the virtually similar proportions of patients with ET and PV in our cohort are surprising, this might reflect the well-known misclassification of JAK2-V617F–positive ET.38 Even if some patients may have been misclassified as having ET instead of PV, we still see a similar protective effect of metformin for both subtypes. Therefore, this potential misclassification has limited impact on our results. Furthermore, 16.5% of patients were classified as MPN-U, representing a clinically and molecularly heterogeneous group of patients including overlapping myelodysplastic and myeloproliferative syndromes. However, for these patients, the magnitude and direction of the association were similar to those observed in more established MPN entities.

For the present study, we also recognize limitations related to the registry-based retrospective design. Firstly, patients receiving metformin may differ substantially from patients who did not, leading to potential unmeasured bias in our estimates. Studies on drug–cancer relationships have the potential to suffer from healthy user bias. Namely, patients using metformin may be more aware about their health and thus improving their health habits, leading to a lower risk of developing cancer, including MPN.39 These factors and habits are extremely difficult to measure and adjust for, and despite efforts to adjust for several conditions known to influence metformin prescription (exposure) and the development of MPN (outcome), the data quality for confounder assessment remains a limitation in the present study. We cannot exclude the presence of residual confounding related to lifestyle, including smoking, overweight, and dietary habits. Particularly, obesity is an important confounding factor due to the association between obesity-related inflammation and clonal expansion of CHIP-related clones.40 To adjust for overweight and obesity, we used diagnostic codes from the National Patient Register and classified 4.1% and 4.4% as overweight or obese among cases and controls, respectively. We acknowledge that this proportion is much lower than expected from a recent Danish study with data from 2021 showing that 18.4% of adults were considered obese.41 This discrepancy may originate in the registration procedure of primary and secondary diagnostic codes in our health care system, where the primary cause of contact is prioritized. A previous study evaluating ICD-10 codes for overweight and obesity found a coverage of only ∼11%, but with a very satisfactory positive predictive value for registered patients.42 Furthermore, increased medical surveillance and attention around the prescription of a new drug could also affect our results given that more regular laboratory tests could detect abnormal blood counts, leading to further investigations. To assess this bias, we conducted a sensitivity analysis by varying the lag time for exposure and confounder assessment around the diagnosis, which did not alter our findings significantly.

Moreover, metformin users could also differ in other risk factors associated with decreased or increased risk of cancer (eg, socioeconomic factors or lifestyle) compared with individuals using other types of antidiabetic medications or no antidiabetic medication. The presence of type 2 diabetes mellitus has been associated with increased risk of subtypes of MPN in 1 study,43 whereas another found no association44; however, for this study, if an association exists, it would bias the estimate against no association between metformin use and risk of MPN.

In summary, this study showed significant reduction in risk of MPN among metformin users and a dose-response relationship, with long-term use of metformin having more protective effects. However, overall, given the rarity of MPN, the absolute risk reduction is small and primary prophylaxis is not feasible. Given the retrospective design, causality cannot be established from the present study, and further studies are needed. We believe that our findings merit future well-designed prospective studies on a high-risk population, that is, patients with low level JAK2-V617F positive clones not meeting the current WHO criteria for MPN.

The authors express profound gratitude to everyone who has participated in forming and collecting data for the National Chronic Myeloid Neoplasia Registry, the Danish Clinical Quality Program, and the National Clinical Registries for providing and governing the data.

This study was supported by a grant from the Danish Cancer Society to D.T.K. and A.S.R. (grant number R327-A18949) and to T.C.E.-G. (grant number R274-A17146).

Contribution: D.T.K., A.K.Ø., L.H.K.J., M.T.S., T.C.E.-G., and A.S.R. conceived and designed the study; D.T.K., A.K.Ø., M.T.S., L.H.H., J.S., M.H.H., A.P.V., M.B., H.C.H., and A.S.R. provided the study materials or patients; D.T.K., A.K.Ø., and A.S.R. collected and assembled the data; D.T.K., A.K.Ø., L.H.K.J., M.T.S., T.C.E.-G., and A.S.R. analyzed and interpreted the data; and all authors wrote the manuscript, approved the final version of the manuscript, and are accountable for all aspects of the work.

Conflict-of-interest disclosure: D.T.K. serves as a member on consulting/advisory boards for AbbVie, Atheneum, and Astellas Pharma. H.C.H. is the chairman of the scientific committee for the INCA 33989-101 study (mCALR-101), Incyte; is a member on the advisory boards of AOP Orphan and Incyte; and is part of the scientific committee of Incyte. A.S.R. received consulting fees from AbbVie and Pfizer and a travel grant from Jazz Pharmaceuticals. The remaining authors declare no competing financial interests.

The current affiliation for L.H.K.J. is Novo Nordisk.

Correspondence: Anne Stidsholt Roug, Department of Hematology, Aarhus University Hospital, Palle Juul Jensens Blvd 99, 8200 Aarhus N, Denmark; email: annrou@rm.dk.

1.
Kralovics
R
,
Passamonti
F
,
Buser
AS
, et al
.
A gain-of-function mutation of JAK2 in myeloproliferative disorders
.
N Engl J Med
.
2005
;
352
(
17
):
1779
-
1790
.
2.
Nangalia
J
,
Massie
CE
,
Baxter
EJ
, et al
.
Somatic CALR mutations in myeloproliferative neoplasms with nonmutated JAK2
.
N Engl J Med
.
2013
;
369
(
25
):
2391
-
2405
.
3.
Pikman
Y
,
Lee
BH
,
Mercher
T
, et al
.
MPLW515L is a novel somatic activating mutation in myelofibrosis with myeloid metaplasia
.
PLoS Med
.
2006
;
3
(
7
):
e270
-
e1151
.
4.
Zahr
AA
,
Salama
ME
,
Carreau
N
, et al
.
Bone marrow fibrosis in myelofibrosis: pathogenesis, prognosis and targeted strategies
.
Haematologica
.
2016
;
101
(
6
):
660
-
671
.
5.
Christensen
DH
,
Rungby
J
,
Thomsen
RW
.
Nationwide trends in glucose-lowering drug use, Denmark, 1999-2014
.
Clin Epidemiol
.
2016
;
8
:
381
-
387
.
6.
Gandini
S
,
Puntoni
M
,
Heckman-Stoddard
BM
, et al
.
Metformin and cancer risk and mortality: a systematic review and meta-analysis taking into account biases and confounders
.
Cancer Prev Res
.
2014
;
7
(
9
):
867
-
885
.
7.
Biondani
G
,
Peyron
JF
.
Metformin, an anti-diabetic drug to target leukemia
.
Front Endocrinol
.
2018
;
9
(
10
). 446.
8.
MacHado-Neto
JA
,
Fenerich
BA
,
Scopim-Ribeiro
R
, et al
.
Metformin exerts multitarget antileukemia activity in JAK2V617F-positive myeloproliferative neoplasms
.
Cell Death Dis
.
2018
;
9
(
3
):
311
.
9.
Kawashima
I
,
Kirito
K
.
Metformin inhibits JAK2V617F activity in MPN cells by activating AMPK and PP2A complexes containing the B56α subunit
.
Exp Hematol
.
2016
;
44
(
12
):
1156
-
1165.e4
.
10.
Pedersen
CB
.
The Danish civil registration system
.
Scand J Public Health
.
2011
;
39
(
suppl 7
):
22
-
25
.
11.
Pedersen
CB
,
Gøtzsche
H
,
Møller
JO
,
Mortensen
PB
.
The Danish civil registration system. A cohort of eight million persons
.
Dan Med Bull
.
2006
;
53
(
4
):
441
-
449
.
12.
Lynge
E
,
Sandegaard
JL
,
Rebolj
M
.
The Danish national patient register
.
Scand J Public Health
.
2011
;
39
(
suppl 7
):
30
-
33
.
13.
Schmidt
M
,
Schmidt
SAJ
,
Sandegaard
JL
,
Ehrenstein
V
,
Pedersen
L
,
Sørensen
HT
.
The Danish national patient registry: a review of content, data quality, and research potential
.
Clin Epidemiol
.
2015
;
7
(
17
):
449
-
490
.
14.
Bak
M
,
Ibfelt
EH
,
Stauffer Larsen
T
, et al
.
The Danish national chronic myeloid neoplasia registry
.
Clin Epidemiol
.
2016
;
8
(
8
):
567
-
572
.
15.
Pottegård
A
,
Schmidt
SAJ
,
Wallach-Kildemoes
H
,
Sørensen
HT
,
Hallas
J
,
Schmidt
M
.
Data resource profile: the Danish national prescription registry
.
Int J Epidemiol
.
2017
;
46
(
3
). 798-798f.
16.
Wallach Kildemoes
H
,
Toft Sørensen
H
,
Hallas
J
.
The Danish national prescription registry
.
Scand J Public Health
.
2011
;
39
(
suppl 7
):
38
-
41
.
17.
Jensen
VM
,
Rasmussen
AW
.
Danish education registers
.
Scand J Public Health
.
2011
;
39
(
suppl 7
):
91
-
94
.
18.
Swerdlow
SH
,
Campo
E
,
Harris
NL
, et al
. WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. 4th ed.
WHO Press
;
2008
.
19.
Swerdlow
SH
,
Campo
E
,
Pileri
SA
, et al
.
The 2016 revision of the World Health Organization classification of lymphoid neoplasms
.
Blood
.
2016
;
127
(
20
):
2375
-
2390
.
20.
Lash
TL
,
VenderWeele
TJ
,
Haneuse
S
,
Rothman
KJ
. Modern Epidemiology. 4th ed.
Wolters Kluwer
;
2021
.
21.
Charlson
ME
,
Pompei
P
,
Ales
KL
,
MacKenzie
CR
.
A new method of classifying prognostic comorbidity in longitudinal studies: development and validation
.
J Chronic Dis
.
1987
;
40
(
5
):
373
-
383
.
22.
Friis
S
,
Kesminiene
A
,
Espina
C
,
Auvinen
A
,
Straif
K
,
Schüz
J
.
European code against cancer 4th edition: medical exposures, including hormone therapy, and cancer
.
Cancer Epidemiol
.
2015
;
39
(
suppl 1
):
S107
-
S119
.
23.
Abdallah
M
,
Dumontier
C
,
Orkaby
A
, et al
.
Metformin use and incidence of cancer: a large retrospective cohort study of U.S. veterans
.
Blood
.
2022
;
140
(
suppl 1
):
10991
-
10992
.
24.
Chang
S-H
,
Luo
S
,
O'Brian
KK
, et al
.
Association between metformin use and progression of monoclonal gammopathy of undetermined significance to multiple myeloma in US veterans with diabetes mellitus: a population-based retrospective cohort study
.
Lancet Haematol
.
2015
;
2
(
1
):
e30
-
e36
.
25.
Cordua
S
,
Kjaer
L
,
Skov
V
,
Pallisgaard
N
,
Hasselbalch
HC
,
Ellervik
C
.
Prevalence and phenotypes of JAK2 V617F and calreticulin mutations in a Danish general population
.
Blood
.
2019
;
134
(
5
):
469
-
479
.
26.
Christensen
SF
,
Svingel
LS
,
Kjaersgaard
A
, et al
.
Healthcare resource utilization in patients with myeloproliferative neoplasms: a Danish nationwide matched cohort study
.
Eur J Haematol
.
2022
;
109
(
5
):
526
-
541
.
27.
Campos
PDM
,
Pagnano
KB
,
Mancuso
RI
, et al
.
Final results of the fibromet trial: an open label phase II study to evaluate metformin effects on bone marrow fibrosis and disease progression in primary myelofibrosis patients
.
Blood
.
2021
;
138
(
suppl 1
):
2584
.
28.
Hasselbalch
HC
,
Silver
RT
.
New perspectives of interferon-alpha2 and inflammation in treating Philadelphia-negative chronic myeloproliferative neoplasms
.
Hemasphere
.
2021
;
5
(
12
):
e645
.
29.
Hasselbalch
HC
,
Elvers
M
,
Schafer
AI
.
The pathobiology of thrombosis, microvascular disease, and hemorrhage in the myeloproliferative neoplasms
.
Blood
.
2021
;
137
(
16
):
2152
-
2160
.
30.
Tuyet Kristensen
D
,
Kisbye Øvlisen
A
,
Hjort Kyneb Jakobsen
L
, et al
.
Use of statins and risk of myeloproliferative neoplasms: a Danish nationwide case-control study
.
Blood Adv
.
2023
;
7
(
14
):
3450
-
3457
.
31.
Podoltsev
NA
,
Wang
R
,
Shallis
RM
, et al
.
Statin use, survival and incidence of thrombosis among older patients with polycythemia vera and essential thrombocythemia
.
Cancer Med
.
2023
;
12
(
18
):
18889
-
18900
.
32.
Franciosi
M
,
Lucisano
G
,
Lapice
E
,
Strippoli
GFM
,
Pellegrini
F
,
Nicolucci
A
.
Metformin therapy and risk of cancer in patients with type 2 diabetes: systematic review
.
PLoS One
.
2013
;
8
(
8
):
e71583
.
33.
Aljofan
M
,
Riethmacher
D
.
Anticancer activity of metformin: a systematic review of the literature
.
Future Sci OA
.
2019
;
5
(
8
). FSO410.
34.
Higurashi
T
,
Hosono
K
,
Takahashi
H
, et al
.
Metformin for chemoprevention of metachronous colorectal adenoma or polyps in post-polypectomy patients without diabetes: a multicentre double-blind, placebo-controlled, randomised phase 3 trial
.
Lancet Oncol
.
2016
;
17
(
4
):
475
-
483
.
35.
Davis
SR
,
Robinson
PJ
,
Jane
F
, et al
.
The benefits of adding metformin to tamoxifen to protect the endometrium—a randomized placebo-controlled trial
.
Clin Endocrinol
.
2018
;
89
(
5
):
605
-
612
.
36.
Lord
SR
,
Harris
AL
.
Is it still worth pursuing the repurposing of metformin as a cancer therapeutic?
.
Br J Cancer
.
2023
;
128
(
6
):
958
-
966
.
37.
Goodwin
PJ
,
Chen
BE
,
Gelmon
KA
, et al
.
Effect of metformin vs placebo on invasive disease-free survival in patients with breast cancer: the MA.32 randomized clinical trial
.
JAMA
.
2022
;
327
(
20
):
1963
-
1973
.
38.
Hasselbalch
HC
.
Time for revival of the red blood cell count and red cell mass in the differential diagnosis between essential thrombocythemia and polycythemia vera?
.
Haematologica
.
2019
;
104
(
11
):
2119
-
2125
.
39.
Lund
JL
,
Richardson
DB
,
Stürmer
T
.
The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application
.
Curr Epidemiol Rep
.
2015
;
2
(
4
):
221
-
228
.
40.
Pasupuleti
SK
,
Ramdas
B
,
Burns
SS
, et al
.
Obesity-induced inflammation exacerbates clonal hematopoiesis
.
J Clin Invest
.
2023
;
133
(
11
):
e163968
.
41.
Schramm
S
,
Sørensen
TIA
,
Davidsen
M
,
Tolstrup
JS
.
Changes in adult obesity prevalence in Denmark, 1987-2021: age–period–cohort analysis of nationally representative data
.
Eur J Public Health
.
2023
;
33
(
3
):
463
-
467
.
42.
Gribsholt
SB
,
Pedersen
L
,
Richelsen
B
,
Thomsen
RW
.
Validity of ICD-10 diagnoses of overweight and obesity in Danish hospitals
.
Clin Epidemiol
.
2019
;
11
:
845
-
854
.
43.
Leal
AD
,
Thompson
CA
,
Wang
AH
, et al
.
Anthropometric, medical history and lifestyle risk factors for myeloproliferative neoplasms in the Iowa Women’s Health Study cohort
.
Int J Cancer
.
2014
;
134
(
7
):
1741
-
1750
.
44.
Duncombe
AS
,
Anderson
LA
,
James
G
, et al
.
Modifiable lifestyle and medical risk factors associated with myeloproliferative neoplasms
.
Hemasphere
.
2020
;
4
(
1
):
e327
.

Author notes

D.T.K. and A.K.Ø. contributed equally to this study.

Per Danish law, the data cannot be shared directly; however, they can be accessed through application.

Additional information is available on request from the corresponding author, Anne Stidsholt Roug (annrou@rm.dk) or at www.rkkp.dk/in-english/ and www.dst.dk/en/TilSalg/Forskningsservice/Data.

The full-text version of this article contains a data supplement.

Supplemental data