TO THE EDITOR:
Thank you for your commentary1 relating to our paper entitled Metformin Use and Risk of Myeloproliferative Neoplasms - A Danish Population-Based Case-Control Study.2
Krecak et al highlights several intriguing aspects about the potential role of the metabolome in the pathogenesis of myeloproliferative neoplasms (MPNs) and associated comorbidities. They also point to possible implications for how novel treatment strategies could be investigated and developed. Although these important aspects cannot be addressed in a retrospective study, we greatly appreciate that our study has motivated Krecak et al to shed light on metabolomics in MPN research and motivate future prospective studies.
In the commentary, the authors propose to target the metabolome in MPN using drugs such as metformin. We agree that a number of preclinical observations, including studies conducted by Krecak et al,1 point to a possible role of statins and metformin in treating MPNs. These findings are supported by observational data including our recent registry-based studies.2,3 Importantly, confirmatory evidence from randomized clinical trials is not available at this moment. Although the open-label phase 2 FIBROMET3 trial4 did not show a disease-modifying effect of 24-month metformin use by reduction of bone marrow fibrosis in patients with primary myelofibrosis who were nondiabetic, some important observations were made. Intriguingly, reductions in proinflammatory cytokines and downregulation of the JAK-STAT activity were observed.
For now, in the absence of confirmatory evidence from clinical trials, drugs such as metformin and statins have no clinical role in the treatment of MPNs and MPN-associated clonal hematopoiesis of indeterminate potential. However, we endorse that the preclinical and observational findings warrant further research into the role of these medicines in MPNs. Going forward, we believe that clinical studies of metabolomics in MPNs should primarily focus on early-stage MPNs, including disease entities with low allele frequency of JAK2-V617F that do not meet diagnostic criteria for MPN. Hu et al5 recently reviewed the landscape of clinical trials focusing on targeting the metabolome using metformin in hematologic neoplasms. They found a number of trials mainly in multiple myeloma, non-Hodgkin lymphoma, myelodysplastic syndrome, or acute lymphoblastic leukemia. However, to our knowledge, no studies specifically addressing MPN and metabolome targeting have yet been initiated. In addition, the authors highlighted possible underdiagnosis of diabetes mellitus in patients with MPN due to artifactual lower Hemoglobin A1c (HbA1c) levels caused by increased glucose consumption, reduced red cell life span, and hydroxyurea-induced expansion of hemoglobin F. These speculations further support the importance of metabolome research in MPNs, not only to explore the disease-modifying potential but also to reduce the risk of MPN-associated complications such as arterial thrombosis.
In response to the request for additional data on the study, specifically regarding the role of metformin in nondiabetic patients, we suffer a notable limitation. Although the Danish registries offer high-quality and detailed data on many health care interactions, they lack information from the primary health care sector, such as the exact indication for metformin prescription. This gap is significant because most patients with type 2 diabetes mellitus are managed by their general practitioners, and only those requiring specialized hospital treatment are registered in our data set with an ICD-10 code or similar. As a result, we cannot ascertain the precise indication for metformin treatment in our study, which introduces a substantial risk of misclassification bias.
Keeping these limitations in mind, we have performed our analysis again using the same approach, methodology, and data as previously described,2 but this time with the exclusion of patients with registered markers of diabetes except metformin use. This was done using data from The National Patient Register6,7 and The Danish National Prescription Register,8,9 excluding markers of diabetes registered in contacts with the secondary health care sector (ie, hospitals), including ICD-8 codes (codes, 249.00-250.09) and ICD-10 codes (codes, E10.0-14.9), or prescription codes for antidiabetic medication using the Anatomical Therapeutic Chemical Classification codes (suffix A10, except metformin [A10BA02]).
As a result, 192 of 3816 cases (5.0%) and 970 of 19 080 controls (5.1%) from our original data set were excluded based on a registered marker of diabetes, leaving 3622 cases and 18 110 controls for analysis. Baseline characteristics are given in Table 1. Metformin use was, as expected, even rarer than in our first analyses, comprising 132 cases (3.6%) and 394 controls (2.2%) in the ever-user groups. Long-term use (≥5 years) was observed only in 5 cases (0.1%) and 15 controls (0.1%). This distribution resulted in a crude odds ratio (OR) of 1.71 (95% confidence interval [CI], 1.40-2.09) for the association between ever-use of metformin and MPN and adjusted OR of 1.63 (95% CI, 1.33-2.01) Table 2. Due to low numbers, we were not able to conduct the full analysis on dose-response relationship provided in our original analysis, however, when considering long-term users (≥5 years cumulative metformin use), the corresponding crude estimate for OR was 1.76 (95% CI, 0.64-4.90), and the adjusted OR was 1.63 (95% CI, 1.33-2.01) Table 2.
. | Cases . | Controls . |
---|---|---|
Total, N | 3622 | 18 110 |
Age, median (IQR), y | 68 (59-75) | 68 (59-75) |
Age group, n (%) | ||
<60 y | 956 (26.4) | 4 780 (26.4) |
60-75 y | 1665 (46.0) | 8 325 (46.0) |
>75 y | 1001 (27.6) | 5 005 (27.6) |
Male sex, n (%) | 1768 (48.8) | 8 840 (48.8) |
Metformin use before index date, n (%) | ||
Never-use | 3490 (96.4) | 17 716 (97.8) |
Ever-use | 132 (3.6) | 394 (2.2) |
Long-term use (≥5 y) | 5 (0.1) | 15 (0.1) |
Number of prescriptions for ever-users, n (%) | ||
1 | 13 (0.4) | 44 (0.2) |
2-4 | 27 (0.8) | 72 (0.4) |
≥5 | 92 (2.5) | 278 (1.5) |
Highest achieved education, n (%) | ||
Primary school | 1182 (32.6) | 5 808 (32.1) |
High school | 1514 (41.8) | 7 425 (41.0) |
Short/middle long education | 644 (17.8) | 3 295 (18.2) |
Long education | 211 (5.8) | 1 148 (6.3) |
Charlson comorbidity index, n (%) | ||
0 | 3178 (87.7) | 17 060 (94.2) |
1 | 295 (8.1) | 819 (4.5) |
≥2 | 149 (4.1) | 231 (1.3) |
Medical history, n (%) | ||
Alcohol related diagnoses | 243 (6.7) | 969 (5.4) |
Overweight and obesity-related diagnoses | 113 (3.1) | 560 (3.1) |
COPD | 320 (8.8) | 1 319 (7.3) |
Autoimmune disease | 304 (8.4) | 1 398 (7.7) |
Previous drug use, n (%) | ||
Aspirin | 1173 (32.4) | 4 202 (23.2) |
Other NSAIDs | 2980 (82.3) | 14 403 (79.5) |
Statins | 1175 (32.4) | 5 217 (28.8) |
Alendronate | 231 (6.4) | 1 050 (5.8) |
Immunosuppressants | 86 (2.4) | 449 (2.5) |
MPN subtype, n (%) | ||
Polycythemia vera | 1239 (34.2) | N/A |
Essential thrombocythemia | 1265 (34.9) | N/A |
Myelofibrosis | 525 (14.5) | N/A |
MPN-U | 593 (16.4) | N/A |
Molecular MPN subtype∗ , n (%) | ||
JAK2-V617F–mutated | 2729 (75.4) | N/A |
CALR-mutated | 179 (4.9) | N/A |
. | Cases . | Controls . |
---|---|---|
Total, N | 3622 | 18 110 |
Age, median (IQR), y | 68 (59-75) | 68 (59-75) |
Age group, n (%) | ||
<60 y | 956 (26.4) | 4 780 (26.4) |
60-75 y | 1665 (46.0) | 8 325 (46.0) |
>75 y | 1001 (27.6) | 5 005 (27.6) |
Male sex, n (%) | 1768 (48.8) | 8 840 (48.8) |
Metformin use before index date, n (%) | ||
Never-use | 3490 (96.4) | 17 716 (97.8) |
Ever-use | 132 (3.6) | 394 (2.2) |
Long-term use (≥5 y) | 5 (0.1) | 15 (0.1) |
Number of prescriptions for ever-users, n (%) | ||
1 | 13 (0.4) | 44 (0.2) |
2-4 | 27 (0.8) | 72 (0.4) |
≥5 | 92 (2.5) | 278 (1.5) |
Highest achieved education, n (%) | ||
Primary school | 1182 (32.6) | 5 808 (32.1) |
High school | 1514 (41.8) | 7 425 (41.0) |
Short/middle long education | 644 (17.8) | 3 295 (18.2) |
Long education | 211 (5.8) | 1 148 (6.3) |
Charlson comorbidity index, n (%) | ||
0 | 3178 (87.7) | 17 060 (94.2) |
1 | 295 (8.1) | 819 (4.5) |
≥2 | 149 (4.1) | 231 (1.3) |
Medical history, n (%) | ||
Alcohol related diagnoses | 243 (6.7) | 969 (5.4) |
Overweight and obesity-related diagnoses | 113 (3.1) | 560 (3.1) |
COPD | 320 (8.8) | 1 319 (7.3) |
Autoimmune disease | 304 (8.4) | 1 398 (7.7) |
Previous drug use, n (%) | ||
Aspirin | 1173 (32.4) | 4 202 (23.2) |
Other NSAIDs | 2980 (82.3) | 14 403 (79.5) |
Statins | 1175 (32.4) | 5 217 (28.8) |
Alendronate | 231 (6.4) | 1 050 (5.8) |
Immunosuppressants | 86 (2.4) | 449 (2.5) |
MPN subtype, n (%) | ||
Polycythemia vera | 1239 (34.2) | N/A |
Essential thrombocythemia | 1265 (34.9) | N/A |
Myelofibrosis | 525 (14.5) | N/A |
MPN-U | 593 (16.4) | N/A |
Molecular MPN subtype∗ , n (%) | ||
JAK2-V617F–mutated | 2729 (75.4) | N/A |
CALR-mutated | 179 (4.9) | N/A |
COPD, chronic obstructive pulmonary disease; IQR, interquartile range; MPN-U, MPN unclassifiable; N/A, not applicable; NNSAIDs, nonsteroidal anti-inflammatory drugs.
Among 3481 tested.
Subgroup . | Cases, n . | Controls, n . | OR∗ (95% CI) . | Adjusted OR† (95% CI) . |
---|---|---|---|---|
Exposure | ||||
Never-use | 3490 | 17 716 | 1.0 (reference) | 1.0 (reference) |
Ever-use | 132 | 394 | 1.71 (1.40-2.09) | 1.63 (1.33-2.01) |
Long-term use (≥5 y) | 5 | 15 | 1.76 (0.64-4.90) | 1.58 (0.56-4.45) |
Subgroup . | Cases, n . | Controls, n . | OR∗ (95% CI) . | Adjusted OR† (95% CI) . |
---|---|---|---|---|
Exposure | ||||
Never-use | 3490 | 17 716 | 1.0 (reference) | 1.0 (reference) |
Ever-use | 132 | 394 | 1.71 (1.40-2.09) | 1.63 (1.33-2.01) |
Long-term use (≥5 y) | 5 | 15 | 1.76 (0.64-4.90) | 1.58 (0.56-4.45) |
NSAIDs, nonsteroidal anti-inflammatory drugs.
Adjusted for age, sex, and calendar time.
Adjusted for age, sex, and calendar time, in addition to (1) education level (primary school, high school, short/middle long education, and long education); (2) Charlson comorbidity index (0, 1, and ≥2); (3) previous use of aspirin, other NSAIDs, alendronate, immunosuppressants, and statins; and (4) previous history of alcohol related diagnoses, overweight and obesity-related diagnoses, chronic obstructive pulmonary disease, and autoimmune disease.
Although this analysis shows an association in the opposite direction of our first findings, that is, metformin use among individuals without markers of diabetes is associated with increased odds of MPN, we again underscore the aforementioned limitations. Because we cannot determine the exact indication for metformin use, we consider it very likely that patients using metformin in this analysis may be well-regulated diabetic patients without complications who are well managed by their general practitioner, leading to misclassification bias. Furthermore, there may be significant key differences between individuals using metformin who have contact with the secondary health care sector and those who do not, potentially leading to confounding. For example, “tolerable” diagnostic thresholds for elevated hematocrit or hemoglobin levels and general health behaviors might vary; that is, patients with healthier behaviors could be more likely to seek consultations with a hospital endocrinologist and thus more likely to undergo diagnostic workups for laboratory signs of MPN. If this is the case, these patients are excluded from this analysis.
Observational studies, including our published results on the cancer-protective effects of metformin and statins in MPN, are essential for generating hypotheses for clinically relevant prospective, randomized trials and studies elucidating disease biology. The literature contains examples in which randomized trials have failed to support hypotheses generated from such observational studies.
We strongly encourage colleagues to replicate and supplement this analysis on larger cohorts with more extensive data on diabetes. Despite the limitations of register-based studies, including the nebulous data on metformin use in nondiabetic patients, we extend our gratitude to Krecak et al for their insightful commentary. We hope our response brings some clarity and perspectives on the discussed issues and inspires further research.
Contribution: D.T.K., A.K.Ø., and A.S.R. conceived and designed the study, provided the study materials or patients, and collected and assembled the data; D.T.K., A.K.Ø., 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. reports consulting/advisory board fees from AbbVie, Atheneum, Immedica, and Astellas Pharma; and travel grants from Swedish Orphan Biovitrum AB. A.S.R reports consulting fees from AbbVie and Pfizer; and travel grant from Jazz Pharmaceuticals. The remaining authors declare no competing financial interests.
Correspondence: Anne Stidsholt Roug, Department of Hematology, Aarhus University Hospital, Palle Juul Jensens Blvd 99, Aarhus 8200, Denmark; email: [email protected].
References
Author notes
Per Danish law, the data cannot be shared directly; however, the data can be accessed through application. Additional information is available on request from the corresponding author, Anne Stidsholt Roug ([email protected]) or at www.rkkp.dk/in-english/ and www.dst.dk/en/TilSalg/Forskningsservice/Data.