Proposals and rationale for revision of the World Health Organization diagnostic criteria for polycythemia vera, essential thrombocythemia, and primary myelofibrosis: recommendations from an ad hoc international expert panel

Recent discoveries in the molecular pathogenesis of BCR-ABL–negative myeloproliferative disorders (MPDs) have raised the prospect of genetic classification and molecular diagnosis in these disorders.¹  The seminal reports in this regard were first published in early 2005 and described a somatic Janus kinase 2 (JAK2) mutation (JAK2617V>F; an exon 14 somatic 1849G>T mutation) in polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF)^2–5 ; mutational frequency is estimated at 95% for PV and 50% for both ET and PMF. Subsequently, JAK2617V>F was demonstrated in other myeloid neoplasms, albeit at a much lower mutational frequency.^6,7  Most recently, other MPD-associated JAK2 (exon 12 mutations)⁸  and MPL (MPL515W>L/K)⁹  mutations were described, the former in JAK2617V>F-negative PV⁸  and the latter in < 5% of patients with PMF or ET.^9,10  As such, a JAK2 mutation is present in virtually all patients with PV and thus constitutes a sensitive diagnostic marker for the disease. In addition, the incidence of JAK2617V>F in both ET and PMF is high enough to warrant diagnostic utility that is complementary to histologic impression

Prompted by these developments, members of the Steering Committee for the revision of the World Health Organization (WHO) Classification of Hematopoietic and Lymphoid Neoplasms approached key members of the International Working Group for Myelofibrosis Research and Treatment (IWG-MRT), the Myeloproliferative Diseases-Research Consortium (MPD-RC), and the European Collaboration on Low-Dose Aspirin in Polycythemia Vera (ECLAP) to enlist their cooperation in the development of revised diagnostic guidelines for the BCR-ABL-negative MPDs. E-conferences were initiated by 2 of the authors (A.T. and J.W.V.). The key participants are listed as coauthors of the current document and include clinical investigators, scientists, and pathologists, 2 of whom are members of the WHO Steering Committee (J.W.V. and J.T.). The resulting consensus document was also presented to members of the Clinical Advisory Committee for the revision of the WHO Classification of Myeloid Neoplasms (meeting March 10, 2007, Chicago IL, Clara D. Bloomfield, Chair), who endorsed the document and recommended its adoption by the WHO.

Table 1 outlines the current WHO criteria for the diagnosis of PV. As mentioned above, virtually all patients with overt (elevated hematocrit) or inapparent (hematocrit that does not exceed the upper reference range) PV carry JAK2617V>F or another functionally similar JAK2 mutation (exon 12 JAK2 mutations). Such mutations are absent in either secondary or spurious polycythemia. Therefore, most of the current WHO criteria, listed under either A-criteria or B-criteria (Table 1) are no longer necessary and can be replaced by adding “presence of JAK2617V>F or other functionally similar JAK2 mutation” as a major criterion (Table 2). Accordingly, we propose to replace the current list under the A-criteria with two “major” criteria: (1) laboratory evidence of increased hemoglobin, hematocrit, or red cell mass and (2) the presence of a JAK2 mutation (Table 2). Similarly, we propose to replace the current list of B-criteria with 3 biologically relevant “minor” criteria: (1) MPD-consistent bone marrow histology, (2) serum erythropoietin level below the reference range for normal, and (3) presence of endogenous erythroid colonies. Diagnosis of PV under the new proposed criteria will require either the presence of both major criteria and at least one minor criterion or the presence of the first major criterion and at least two minor criteria (Table 2).

In practice, meeting both of the newly proposed major criteria should capture more than 97% of cases with PV. However, to minimize consequences of false-positive molecular test results and further optimize diagnostic specificity, the additional presence of at least one of the 3 minor criteria will be required to make a diagnosis of PV (Table 2). The alternative diagnostic combination (first major criterion plus 2 minor criteria) should capture occasional cases of true PV that might be either negative for a known JAK2 mutation or associated with a mutation burden that is too low to be detected.¹¹  The requirement for two instead of one minor criteria in this instance reinforces the accuracy associated with each test and increases the comfort level to make the specific diagnosis in the context of a negative molecular test result.

The first major criterion, under the newly proposed criteria (Table 2), is notably different from that listed in the existing WHO criteria (Table 1). First, while retaining the hemoglobin and red cell mass threshold levels for PV diagnosis used in the current WHO criteria, the new first major criterion incorporates information on hematocrit to address potential discordance between hemoglobin level and hematocrit and accommodate the needs of medical centers that use hematocrit instead of hemoglobin level as their primary tool to evaluate polycythemia.¹²  We are very much aware of the controversies regarding which one of the 3 red cell parameters (hemoglobin level, hematocrit value, and measured red cell mass) is the most accurate in the determination of red cell volume. However, we do not believe that the issue will be resolved any time soon; thus, we opted to accommodate individual choice on the matter. The second modification of the first major criterion for PV diagnosis (Table 2) aspires to capture biologically true PV cases with a documented, sustained increase in their baseline hemoglobin level in excess of 20 g/L (2 g/dL), without reaching the diagnostic threshold. The particular measure helps bypass the need for red cell mass measurement in such cases and complements bone marrow histology in identifying early PV cases.

We recognize the possibility that the newly revised criteria could still miss occult PV. However, the identification of such cases might not be therapeutically relevant because (1) high-risk MPD patients receive similar drug therapy regardless of the specific MPD diagnosis and (2) the value of aggressive phlebotomy in aspirin-treated low-risk disease, within hematocrit ranges between 0.4 and 0.55 (40% and 55%), has been questioned by the European Collaboration on Low-Dose Aspirin in Polycythemia Vera (ECLAP) investigators.¹³  The newly proposed criteria will also allow identification of “atypical” or “early” PV that presents with the phenotype of “idiopathic erythrocytosis”¹⁴  because such cases have now been identified to display either exon 14 or exon 12 JAK2 mutations.^11,15  Some cases of idiopathic erythrocytosis remain molecularly undefined, but we do not believe that the situation warrants establishment of a separate set of diagnostic criteria. We are also aware of the possibility that certain cases of JAK2 mutation-positive “idiopathic” abdominal vein thrombosis could progress into overt PV but more information is needed in this regard before recognizing such cases in formal diagnostic criteria.¹⁶ 

Finally, hemoglobin level is expected to be lower in the presence of iron deficiency, which confounds the accurate interpretation of the hemoglobin/hematocrit level. However, for the purposes of research or clinical trials, a formal PV diagnosis requires demonstration of meeting the WHO criteria in terms ofhemoglobin/hematocrit level after iron replacement. Whether or not such a measure is clinically prudent is left to the discretion of the physician. In other words, in routine clinical practice, one should not be prevented from making a working diagnosis of PV, in the presence of iron deficiency, just because the WHO criteria are not met.

Table 3 outlines current WHO criteria for the diagnosis of ET. Many investigators have effectively argued that the use of 600 × 10⁹/L platelets as a threshold level compromises the detection of early-phase disease because the 95th percentile for normal platelet count, adjusted for gender and race, is below 400 × 10⁹/L.^17–21  Therefore, we start by proposing a change in the platelet threshold level required for ET diagnosis by lowering it to 450 × 10⁹/L from 600 × 10⁹/L (Table 4, criterion 1); a proposal supported by the current availability of a clonal marker (JAK2617V>F) that is present in 50% of ET patients. However, unlike the case with PV, the utility of mutation screening for JAK2617V>F for the diagnosis of ET or PMF is limited by suboptimal negative predictive value and lack of diagnostic specificity within the context of myeloid neoplasms.^7,22  Therefore, a bone marrow biopsy is still required to help with the differential diagnosis between JAK2617V>F-negative ET and reactive thrombocytosis and to differentiate ET from other chronic myeloid neoplasms, including cellular phase/prefibrotic PMF and myelodysplastic syndromes (MDS).²³  As such, we propose keeping bone marrow examination as one of the required criteria for ET diagnosis (Table 4, criterion 2).

In addition, we propose getting rid of separate categorization of negative exclusionary criteria and instead suggest the addition of two criteria (criteria 3 and 4), which serve in the same capacity, to the main list (Table 4). Criterion 3 requires the absence of WHO criteria that would otherwise classify a patient as having PV, PMF, chronic myelogenous leukemia (CML), or MDS. In this regard, it is important to exclude occult PV in the iron-deficient patient by a trial of iron replacement therapy. Similarly, careful bone marrow histologic assessment is necessary to exclude the possibility of cellular phase/prefibrotic PMF or MDS. Criterion 4 underlines the need to exclude reactive thrombocytosis either through the demonstration of JAK2617V>F or, in mutation-negative patients, by clinical assessment.^23,24  Accordingly, the diagnosis of ET would require fulfillment of all 4 main criteria listed in Table 4. However, it is possible that a JAK2617V>F-negative ET patient might have a concomitant condition known to be associated with reactive thrombocytosis. In this instance, diagnosis of ET can be made provided the first 3 criteria are met.

In ET, bone marrow findings are remarkable for the presence of large, but mature-appearing, megakaryocytes with deeply lobulated and hyperlobulated nuclei that are most often dispersed throughout the biopsy sections, but sometimes also found in loose clusters.²⁵  Often the bone marrow is normally or only slightly hypercellular for the patient's age, and the increased trilineage proliferation (panmyelosis) that characterizes PV or the granulocytic proliferation and highly bizarre megakaryocytes that characterize the prefibrotic stage of PMF are not found in ET.²⁶  The presence of dyserythropoiesis, macrocytosis, monocytosis, pseudo Pelger-Huet anomaly, or other dysgranulopoietic changes of neutrophils and/or predominance of small megakaryocytes with monolobated nuclei suggest MDS rather than ET.^23,27  On the other hand, a spectrum of cytogenetic abnormalities, including del(20q), del(5q), and unbalanced whole-arm translocation between 1q and 7p have been reported in otherwise typical ET^28,29  and their presence, unless accompanied by histologic features of MDS, should not exclude the diagnosis of ET.³⁰  Nevertheless, we encourage obtaining cytogenetic studies at time of diagnostic bone marrow examination for ET and re-evaluation of the histologic impression in the case of unusual findings.

The current WHO criteria for PMF³¹  have separate tables for “prefibrotic” and “fibrotic” PMF (Tables 5,6). Under the newly proposed revised criteria (Table 7), 3 major and 4 minor criteria are enlisted. The first major criterion underscores histology as a critical diagnostic criterion for PMF. In this regard, there is little doubt about the existence of the histopathologic phenotype of PMF with minimal or no demonstrable excess in reticulin fibers, that is, the prefibrotic or “cellular phase.” Therefore, the demonstration of reticulin fibrosis, although characteristic, is not a required criterion for the diagnosis of PMF. Instead, the cardinal and therefore required features for PMF include increase in megakaryocyte growth associated with conspicuous abnormalities as well as granulocyte proliferation.²⁶  However, in the absence of reticulin fibrosis, meeting the first major criterion for PMF diagnosis requires, in addition to the aforementioned megakaryocyte changes, presence of bone marrow hypercellularity, granulocyte proliferation, and decrease in erythroid precursors.

The other two major criteria underline the need to exclude either myelofibrosis associated with another myeloid neoplasm (major criterion 2) or reactive bone marrow fibrosis (major criterion 3). The former requires the absence of meeting WHO criteria for another myeloid neoplasm and the latter the demonstration that either JAK2617V>F (or other clonal markers such as MPL515W>L/K) is present or bone marrow fibrosis is not attributed to a nonclonal process like (chronic) toxic myelopathies due to a variety of exogenous agents, lymphoid neoplasms such as hairy cell leukemia and lymphoma, or a metastatic malignancy.²⁴  The diagnosis of PMF according to the revised WHO criteria requires meeting all 3 major criteria and, to increase diagnostic accuracy, in the context of prefibrotic/cellular phase disease, 2 of 4 minor criteria (Table 7).

The distinction between PMF on one hand, and MDS with fibrosis, ET, or acute panmyelosis with myelofibrosis (in many cases a hyperfibrotic subtype of acute myeloid leukemia [AML] according to the WHO),^32,33  on the other, requires careful morphologic assessment.^26,34  PMF is characterized by megakaryocytes that are often found in sizable loose to tight clusters and range in size from small to large with an aberrant nuclear/cytoplasmic ratio and hyperchromatic, bulbous, or irregularly folded nuclei. Bare megakaryocytic nuclei are common.³⁵  In contrast, MDS is characterized by its classic dysmyelopoietic features. The dysplastic megakaryocytes seen in MDS are usually small and often have monolobated, hypolobated, or widely dispersed nuclei and are not usually found in large clusters (with the exception of some cases of fibrotic MDS) as are typical for MPD.³⁴  On the other hand, the presence of ringed sideroblasts, although seen frequently in MDS, may also be found in some cases of MPD, and therefore lacks diagnostic specificity.^36,37  Such contention is consistent with the demonstration of JAK2617V>F in a high proportion of cases with “refractory anemia with ringed sideroblasts and thrombocytosis (RARS-T).”^38–42  A proportion of these cases might “represent” true MPD cases (ET or PMF) with ringed sideroblasts.⁴³  Other cases, with more pronounced MDS-like abnormalities, such as dyserythropoiesis and anemia, might or might not display additional morphologic abnormalities of megakaryocytes that resemble those seen in MPD.^38–42 

Careful bone marrow morphologic examination is crucial in distinguishing ET from prefibrotic PMF, an important detail that was overlooked by the original Polycythemia Vera Study Group (PVSG) criteria,^44,45  despite its potential confounding effect on both survival and rate of myelofibrotic transformation.⁴⁶  Such distinction takes into account the degree of bone marrow cellularity (marked hypercellularity in cellular-phase PMF), the presence of prominent granulocyte proliferation with left-shifted forms (typical in PMF but absent in ET), and in particular megakaryocyte morphology (with the aforementioned nuclear features in PMF and giant, mature-appearing megakaryocytes with deeply lobulated and hyperlobulated nuclei in ET).¹⁷  Nevertheless, to reinforce diagnostic accuracy, the revised WHO diagnostic criteria for PMF require the presence of at least 2 of 4 PMF-characteristic peripheral blood or clinical features: leukoerythroblastosis, increased serum lactate dehydrogenase level, anemia, and palpable splenomegaly.

Patients with acute (malignant) myelofibrosis or panmyelosis with myelofibrosis usually present with severe constitutional symptoms, pancytopenia, and mild or no splenomegaly and feature an increase in blood and/or bone marrow blast count that will partially approach or fulfill the required threshold for AML diagnosis.³³  In this context, it has been shown that the latter is a heterogeneous entity that should be differentiated from AML³³  and MDS with accompanying myelofibrosis and a low blast count.^34,47  Nevertheless, in all cases with myelofibrosis, blasts may be difficult to accurately estimate, and CD34 assessment by immunohistochemistry may be invaluable in such cases.⁴⁸  Finally, the presence of either del(13)(q11∼13q14∼22) or der(6)t(1,6)(q21∼23;p21.3) is strongly suggestive but not diagnostic for PMF.^49,50 

The proposed revisions to the diagnostic guidelines require the time-honored integration of clinical, laboratory, and pathologic findings. However, now the laboratory data include evaluation for a molecular defect, JAK2617V>F, that when present clearly identifies the process as neoplastic and thus eliminates the need for additional tests previously performed to exclude reactive myeloproliferation caused by other diseases. The revised criteria also take advantage of the recently appreciated contribution of specific morphologic abnormalities to the diagnosis of the MPDs. The histologic differences among the entities outlined here are recognized by experienced hematopathologists, and there is ongoing investigation to establish even more precise quantitative definitions similar to those established for other myeloid neoplasms.⁵¹ 

The objective of establishing diagnostic criteria is neither to capture all biologically true cases of disease nor guarantee 100% diagnostic specificity. For example, recent communications have disclosed the presence of JAK2617V>F in “idiopathic” abdominal vein thrombosis in patients who otherwise do not fulfill conventional diagnostic criteria for either PV or ET.¹⁶  Some of these patients subsequently developed erythrocytosis and/or thrombocytosis, whereas others appear to have stable counts for many years. Regardless, such cases represent unclassified MPDs and should not be forced into a specific diagnosis, for research purposes. We provide this example to emphasize the point that not fulfilling a given set of published criteria does not prevent one from making a specific working diagnosis that is based on sound clinical judgment.

Contribution: A.T., J.T., A.O., and J.W.V. prepared the initial draft of the proposals for revision of the WHO diagnostic criteria for polycythemia vera, essential thrombocythemia, and primary myelofibrosis. A.T. wrote the paper. J.T., A.O., and J.W.V. participated in the writing of the paper. All of the authors participated in the discussion and have reviewed and approved the current version of the manuscript.

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

J.T. and J.W.V. are members of the Steering Committee for Revisions of the WHO Classification of Haematopoietic and Lymphoid Neoplasms.

Correspondence: Ayalew Tefferi, Mayo Clinic, 200 First Street SW, Rochester, MN 55905; e-mail:tefferi.ayalew@mayo.edu.