A Simple Diagnostic Algorithm for the Prediction of JAK2 Mutation Status in Erythrocytosis Based on Complete Blood Count Data: A Multi-Center Retrospective Analysis

Background:
Screening for JAK2 mutations is instrumental in distinguishing polycythemia vera (PV) from non-neoplastic erythrocytosis. However, genetic test for all patients would place a burden on both medical facilities and healthcare economics. Although low erythropoietin (EPO) levels are a minor criterion in the 2016 WHO PV diagnostic criteria, nearly half of patients with non-neoplastic erythrocytosis also present with low EPO levels, complicating the differential diagnosis from PV. This study aims to identify predictors of JAK2 mutations using routine blood test data and develop a diagnostic algorithm for erythrocytosis.

Methods:
Patients with erythrocytosis (hemoglobin ≥16.0 g/dL for males and ≥15.0 g/dL for females) who underwent genetic testing at our institution were included in the training cohort. Laboratory data at the initial visit were collected. Multivariate analysis was employed to identify laboratory predictors of JAK2 mutations. Twelve parameters were analyzed: WBC count, neutrophil ratio, eosinophil ratio, RBC count, hemoglobin, hematocrit, MCV, RDW-CV, platelet count, platelet distribution width, LDH, and EPO. Receiver operating characteristic (ROC) analysis determined cutoff values that maximized the area under the curve (AUC), sensitivity, and specificity. Decision tree analysis utilized initial blood data as explanatory variables and JAK2 mutations as the response variable. The algorithm's efficacy was evaluated using a validation cohort comprising 188 patients from our hospital and three other facilities. Additionally, univariate regression analysis assessed the correlation between initial JAK2 allele burden and RDW-CV.

Results:
The training cohort consisted of 212 patients (153 males and 59 females; 56 JAK2 V617F-positive, one JAK2 exon12-mutated, and 155 JAK2-unmutated). Logistic regression analysis identified neutrophil ratio, RDW-CV, and platelet count as significant predictors of JAK2 mutations. ROC analysis yielded optimal cutoff values: neutrophil ratio 65.5% (sensitivity 92.6%, specificity 76.8%, AUC 0.923), platelet count 324 x 10⁹/L (sensitivity 92.7%, specificity 85.3%, AUC 0.926), and RDW-CV 14.7% (sensitivity 86.3%, specificity 90.5%, AUC 0.924). Decision tree analysis classified the patients with platelet count <360 x 10⁹/L and RDW-CV <15% as ‘likely JAK2-unmutated’ (122 of 123 patients in this group were truly JAK2-unmutated ), and the patients with platelet count ≥360 x 10⁹/L, neutrophil ratio ≥64%, and RDW-CV ≥15% as ‘likely JAK2 mutation-positive’ (32 of 33 patients in this group were truly JAK2-mutated). In the validation cohort (137 males and 51 females; 31 JAK2 V617F-positive and 157 JAK2-unmutated), 97.2% of ‘likely JAK2-unmutated’ was actually JAK2-unmutated, and 88.2% of ‘likely JAK2 mutation-positive’ was actually JAK2-mutated. Among the subgroup whose hemoglobin ≥16.5 g/dL for males and ≥16.0 g/dL for females (hemoglobin cutoff in the 2016 WHO PV diagnostic criteria) in the validation set (n=162), 97.5% of ‘likely JAK2-unmutated’ was actually JAK2-unmutated, and 91.6% of ‘likely JAK2 mutation-positive’ was actually JAK2-mutated. In 39 patients with JAK2 V617F allelic burden checked (median 38.42%, range 0.22-94%), a weak positive correlation (correlation coefficient 0.3752, p<0.05) was observed between initial JAK2 V617F allelic burden and RDW-CV.

Conclusion:
Platelet count, RDW-CV, and neutrophil ratio are effective predictors of JAK2 mutations. This simple algorithm can appropriately select patients with erythrocytosis for JAK2 genetic test, increase the pre-test probability of detecting JAK2 mutations, and potentially reduce the burden on patients and medical facilities as well as healthcare costs.