Diabetes mellitus, metformin, and the risk of MPN

TO THE EDITOR:

With great interest, we have read the recent article by Kristensen et al¹ that reported that metformin use may be associated with a lower risk of chronic myeloproliferative neoplasms (MPNs), including essential thrombocythemia (ET), polycythemia vera (PV), and myelofibrosis. These interesting results may be clinically relevant but could also reflect specific biology of glucose metabolism in patients with MPNs.

Even though metformin use has been associated with a lower risk of different solid and hematologic cancers in prior studies,² possibly due to dampening of cancer-related chronic inflammation, other pathophysiological mechanisms in patients with MPNs may be responsible as well. Preclinical study in transgenic mice models expressing Janus Kinase 2 (JAK2) has demonstrated increased energy requirements and metabolic reprogramming that subsequently led to hypoglycemia and adipose tissue atrophy without signs of insulin resistance. Hypoglycemia in these mice models correlated with hyperactive erythropoiesis due to increased consumption of glucose in erythroid cells. Interestingly, high glucose diet did not ameliorate hypoglycemia but rather fueled erythrocytosis and splenomegaly, whereas intermittent fasting attenuated the MPN phenotype. In addition, activation of mutant JAK2 signaling altered the expression of genes involved in the regulation of metabolic pathways, causing enhanced glycolysis and increased oxidative phosphorylation. More importantly, therapeutic targeting of PFKFB3, an early rate-limiting enzyme in glycolysis, induced apoptotic cell death and reduced the growth of human JAK2-expressing cell lines.³ These cell-autonomous metabolic alterations in MPNs could also be targeted by metformin, which is a potent inhibitor of glycolysis and oxidative phosphorylation.⁴ This may provide an explanation for potential clinical benefits of metformin with respect to MPN occurrence and survival, but a prospective randomized trial is needed.

A similar effect was observed for statins in which the disruption of lipid shafts in the MPN-cell membrane leads to attenuated JAK2 signaling.⁵ In fact, patients with MPNs often present with hypocholesterolemia, which is thought to occur due to the increased consumption of lipids by the strongly proliferating hematopoietic cells.⁶ Retrospective studies have shown that the use of statins could be associated with lower incidence of MPNs,⁷ lower phlebotomy requirements in PV,⁸ lower thrombotic risk, and prolonged survival in both ET and PV⁹ but not lower symptom burden.¹⁰ Context-dependent JAK2 signaling modulates a multitude of metabolic processes, and its inhibition has a potential for both beneficial and undesirable metabolic consequences in various tissues.¹¹ Ruxolitinib, a JAK1/JAK2 inhibitor, was shown to revert metabolic perturbations in patients with MPNs who often recover from hypolipoproteinemia, gain weight,^12-14 and can also improve their muscle mass upon exposure to the drug.¹⁵ Importantly, the results from the COMFORT-1 randomized clinical trial have shown that the risk of hyperlipidemia with ruxolitinib use is minimal, with most ruxolitinib-treated patients with myelofibrosis not exceeding total cholesterol and low-density lipoprotein (LDL) levels above 240 mg/dL (6.2 mmol/L) and 160 mg/dL (4.1 mmol/L), respectively.¹⁴ However, it remains unclear whether and to what extent reverting hypolipoproteinemia in MPNs may add to higher cardiovascular risk, especially when considering that “safe” total cholesterol and LDL levels in patients with MPNs remain undefined. Notably, one retrospective multicenter study including patients with ET and PV has suggested that optimal LDL levels for the prediction of thrombotic events may be <70 mg/dL (1.8 mmol/L),¹⁶ which is similar to general population that is considered to be at high cardiovascular risk.¹⁷ Nevertheless, validation on larger number of patients with MPNs is needed to confirm these observations.

It should also be pointed out that, in contrast to other metabolic comorbidities, the incidence of diabetes mellitus (DM) in MPNs seems to be quite low (4.7%-16.9%), and it significantly varied among different retrospective studies.¹⁸ This is puzzling, especially when considering that MPNs typically occur in older patients, who are burdened with cardiovascular and metabolic diseases. There are several possible explanations for this. First, artifactual hypoglycemia may be frequent in patients with MPNs due to leukocyte-driven glucose consumption.¹⁹ Second, short red blood cell life span and different cytoreductive treatments may lead to lower glycated hemoglobin (HbA1c) levels; moreover, patients with MPNs most frequently receive hydroxyurea which drives the expansion of HbF, potentially causing falsely lower HbA1c levels. Actually, HbA1c levels in patients with ET and PV do not seem to correlate with erythrocyte count, hematocrit, or Hb levels.²⁰ Therefore, it is possible that an unknown proportion of patients with MPNs and DM are diagnostically “missed” in everyday practice. This observation may be clinically relevant, especially when considering the potential impact of DM on thrombotic risk in the general population. However, the exact effect of DM presence on thrombotic risk in MPNs remains elusive, with multiple studies reporting no significant association.^18,21-23 As MPNs are rare disorders with survival measured in decades, the expected number of cardiovascular events during follow-up is often low. This makes randomized clinical trials for elucidating the potential effect of DM on thrombotic risk in MPNs unlikely. For this reason, clinical research focusing on the prediction of thrombotic events in MPNs usually relies on analyses of retrospective data sets, in which DM presence is coded only when registered as such in the medical documentation or when included patients use different antidiabetic drugs. This can lead to situations such as patients with chronic heart failure using sodium glucose cotransporter 2 inhibitors (SGLT2-is), even though they might not have DM, being coded as such. Large time spans during which patients with MPNs and DM were enrolled in such registries also inevitably introduce significant biases and different confounders. For example, diagnostic criteria for DM have evolved throughout the years, and over long periods of time, patients were exposed to different antidiabetic and other cardioprotective medications, with the real possibility of better outcomes in contemporary patients. The latter is even more important when considering the recent treatment paradigm shift in the management of DM in which more focus is placed on cardiovascular protection and less on HbA1c control, especially in the older population.²⁴ 

We have recently conducted a multicenter international retrospective study¹⁸ with an aim to elucidate the optimal HbA1c levels in MPNs and observed that patients with ET and PV with baseline HbA1c ≥7.2% may be having higher thrombotic risk; this effect was independent of JAK2 mutation and was evident in both low- and high-risk patients. Moreover, there were significant interactions between advanced age, higher blood counts, higher HbA1c, and higher thrombotic risk, suggesting that uncontrolled DM may act synergistically with the MPN clone and drive cardiovascular disease development.^21,22 Nevertheless, additional validation of these findings in larger date sets is needed.

Further research is also needed to unravel which antidiabetic drug classes (or their combinations) may represent the best treatment approaches for the management of DM in MPNs. In the general population, favorable cardioprotective effects were most evident with SGLT2-i and glucagon-like peptide 1 agonists.^24,25 Antidiabetic drugs from the class of SGLT2-i may promote erythrocytosis and are able to reveal patients with “masked” MPN,²⁶ which is contrary to the above mentioned negative associations between MPN and DM. These medications may also cause increased phlebotomy requirements in established patients with MPNs, and there are preliminary signals from smaller case series suggesting a higher thrombotic risk in MPNs with SGLT2-i use,²⁷ although it is unclear whether this risk was intrinsic to these patients irrespectively of drug exposure. On the contrary, this drug class may contribute to blood plasma depletion and deinflammation and thus may help to reduce the thrombotic risk in MPNs through additional novel mechanisms.^28-32 Due to small number of patients with MPNs and DM included in retrospective studies, data regarding the efficacy of other frequently used antidiabetic drugs, that is, metformin or glucagon-like peptide 1 agonists in patients with MPNs, are still lacking. Therefore, elucidating the potency of different antidiabetic compounds with respect to glycemic control and cardiovascular risk in MPNs is of utmost importance. International collaborations including larger number of patients with MPNs and DM would be a good way to start. Finally, continuous glucose monitoring and assessment of glycated proteins may be a reasonable novel approach for DM diagnosis and treatment in MPNs, because these particular methods are independent of patient's blood counts.³³ However, further research is needed.

We congratulate Kristensen et al¹ on providing the results that substantially add to the pool of knowledge on targeting metabolic comorbidities and the risk of MPNs. Indeed, targeting the metabolome may represent a novel and an exciting approach for the treatment of MPNs. Considering that indications for metformin use may vary among different individuals (DM, prediabetes, obesity, polycystic ovaries, nonalcoholic fatty liver disease, and prevention of long COVID-19, etc) it would be of great interest if authors could report what proportion of MPN and control patients had DM in this study and whether metformin use contributed independently of DM to the lower risk of MPNs. This approach may potentially balance the risk of having lower incidence of DM if being diagnosed with MPN (or vice versa) and could help us to better understand whether DM presence and/or metformin use were independently associated with a lower risk of MPNs in the current data set.

Contribution: All authors wrote, drafted, and approved the final version of the commentary.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ivan Krecak, Department of Internal Medicine, General Hospital of Šibenik-Knin County, Stjepana Radica 83, Šibenik 22000, Croatia; email: krecak.ivan@gmail.com.