Why molecular response in polycythemia vera remains an elusive endpoint

Polycythemia vera is a rare, chronic myeloproliferative neoplasm characterized by an overproduction of red blood cells, and often white blood cells and platelets. This uncontrolled proliferation increases the risk of thrombotic events, bleeding, and progression to myelofibrosis or acute myeloid leukemia. Current management strategies focus on controlling hematocrit, alleviating symptoms, and reducing thrombotic risk, typically involving phlebotomy and cytoreductive agents.¹

Hydroxyurea (HU) has long served as a first-line cytoreductive therapy, but its use is associated with side effects and a subset of patients develop resistance or intolerance. Newer agents, including interferons and ruxolitinib, have emerged, offering alternatives for patients who do not respond adequately to or tolerate HU. But the direct comparative efficacy of these newer agents against each other, or against HU in specific patient subgroups, remains largely unaddressed by direct head-to-head trials.¹

The Attempt to Synthesize Evidence

Walden and colleagues conducted a targeted literature review and feasibility assessment to determine if a network meta-analysis (NMA) could compare ropeginterferon alfa-2b-njft against peginterferon alfa-2a or ruxolitinib, using hydroxyurea as a common comparator.¹ The team screened 193 PubMed records and 460 conference abstracts published between January 2014 and May 2024. They ultimately included 40 records, representing evidence from 11 randomized controlled trials and 10 observational studies.¹

The investigators focused on several key endpoints: complete hematologic response (CHR), molecular response (MR), allele burden, event-free survival, and safety. A network meta-analysis requires a connected evidence network, where treatments are linked through common comparators, allowing for indirect comparisons. Of the 40 included studies, only 20 formed connected evidence networks for the endpoints of interest. This initial reduction already highlighted a fragmented evidence base.¹

The primary goal was to assess the homogeneity of patient populations, treatment regimens, and endpoint definitions across these studies, which is critical for a robust NMA. A network meta-analysis relies on the assumption that studies are sufficiently similar in design and patient characteristics that any differences in outcomes can be attributed to the treatments themselves, rather than confounding factors. Without this homogeneity, indirect comparisons become unreliable and potentially misleading.¹

Why the Numbers Did Not Add Up

The feasibility assessment revealed substantial heterogeneity across the included studies, ultimately precluding a robust network meta-analysis. Patient populations varied significantly, encompassing newly diagnosed patients, high-risk and low-risk cohorts, and patients who were either hydroxyurea-refractory or -intolerant. This variation in baseline characteristics means that the observed treatment effects might not be comparable across studies, as the underlying disease biology and patient responsiveness could differ substantially between these groups.¹

Definitions of complete hematologic response also differed, with some requiring the absence of disease-related symptoms in addition to hematologic parameters, while others did not. This inconsistency in outcome definitions makes it difficult to pool data or compare results directly. A CHR definition that includes symptom resolution, for example, sets a higher bar than one focused solely on blood counts, making a direct comparison of CHR rates between studies with different definitions problematic.¹

Molecular response thresholds were inconsistent as well. Molecular response, often measured by the reduction in JAK2V617F allele burden, is an increasingly important endpoint in PV, reflecting a deeper disease control beyond hematologic parameters. But without standardized thresholds for what constitutes a molecular response (e.g., a 25% reduction, a 50% reduction, or complete molecular remission), comparing the efficacy of different treatments on this endpoint becomes impossible. This is a recurring issue in molecular testing in other conditions, where consensus on thresholds is still evolving.¹

Follow-up durations varied considerably across the studies. Some trials had relatively short follow-up periods, while others extended for several years. This discrepancy is particularly problematic for a chronic disease like PV, where long-term outcomes, such as event-free survival and progression-free survival, are critical. Short follow-up periods may miss late-onset adverse events or long-term benefits, making it difficult to draw definitive conclusions about the comparative long-term efficacy and safety of treatments.¹

The definition of standard of care also presented a challenge. In some studies, standard of care comprised almost exclusive use of hydroxyurea, while in others, it involved mixed regimens that could include phlebotomy, aspirin, or other cytoreductive agents. This variability in the comparator arm introduces further confounding, as the baseline treatment context against which new therapies are evaluated is not consistent. This makes it difficult to isolate the specific effect of the investigational drug.¹

The Broader Implications of Heterogeneity

The findings from Walden and colleagues highlight a fundamental challenge in generating robust comparative evidence for rare conditions like polycythemia vera. The lack of standardized clinical trial designs and outcome definitions across studies severely limits the ability to synthesize existing data, even with advanced statistical methods like network meta-analysis. This problem extends beyond PV, affecting many rare diseases where patient populations are small and trial designs often adapt to practical constraints.¹

The inability to perform a reliable NMA means that clinicians are left without a clear hierarchy of evidence for choosing between ropeginterferon alfa-2b, peginterferon alfa-2a, ruxolitinib, and hydroxyurea in various patient subgroups. Each drug has its own profile, and individual trial results, while informative, do not provide the head-to-head comparisons that would offer definitive guidance. This forces clinicians to make treatment decisions based on a more fragmented understanding of the evidence, often relying on expert opinion and institutional experience. For a comprehensive overview of haematological conditions, the Oxford Handbook of Clinical Haematology remains a valuable resource.

The study highlights the importance of prospective, harmonized research efforts. Future clinical trials in PV, and indeed in other rare diseases, must adopt standardized patient selection criteria, consistent endpoint definitions, and comparable follow-up durations to enable meaningful comparisons. Without such standardization, the field will continue to struggle with an evidence base that is rich in individual studies but poor in synthesizable, comparative data. This issue is particularly relevant when considering the role of minimal residual disease (MRD) status, which is gaining traction as a prognostic marker but lacks consistent definitions across trials.¹

The authors concluded that an NMA for PV treatments was not feasible due to significant clinical and methodological heterogeneity across studies. This included differences in patient characteristics, treatments, outcome definitions, and follow-up times. This is not a failure of the analytical method itself, but rather a reflection of the underlying data quality and consistency. The problem lies not in the inability to perform the statistical calculations, but in the inability to draw valid clinical inferences from such disparate inputs.¹

The open-label design of many PV trials is an obvious caveat, introducing potential bias in symptom reporting and investigator assessment. But even beyond blinding, the fundamental differences in what was measured and in whom it was measured proved insurmountable. The trial was not powered to detect differences in molecular response across these varied populations, and that gap matters for a disease where deeper remission is increasingly sought.¹