Diagnostic Accuracy of Erythropoietin and Jakpot in Predicting JAK2-Postiive Erythrocytosis: A Retrospective Cohort Study

Introduction:

Erythrocytosis affects 0.5 – 4% of ambulatory patients and is a common reason for Hematology referral. Very few patients are ultimately diagnosed with a myeloproliferative neoplasm such as polycythemia vera. The JAKPOT score (Chin-Yee et al) incorporating platelet, neutrophil, and erythrocyte counts shows promise at identifying patients at low risk of JAK2-positive erythrocytosis, but has not been externally validated. Our aim was to compare the diagnostic accuracy of serum erythropoietin (EPO) with the JAKPOT score to identify patients with JAK2-positive erythrocytosis.

Methods:

Retrospective cohort study including all patients who had EPO and JAK2 molecular testing (V617F or exon 12) for undifferentiated erythrocytosis at a tertiary care centre in Hamilton, Canada between Dec. 2014 – Dec. 2022. Patients with hemoglobin < 125 g/L were excluded. Demographics, comorbidities, medications, and laboratory parameters were collected. Univariate analysis to detect predictors of JAK2 positivity was performed with Mann Whitney U-tests, two-tailed Welch t-tests and Chi-squared tests where appropriate. Multivariate logistic regression analysis was performed with a subset of the predictors. We determined the sensitivity (Sn), specificity (Sp), positive and negative likelihood ratios (+LR, -LR) of EPO, JAKPOT score, and a combination of serum EPO and JAKPOT (EPO-JAKPOT) to diagnose JAK2-positive erythrocytosis.

Results:

The initial cohort included 237 patients (74 female, mean age 57.7 years). Twenty-four (10%) patients were excluded because they were known to have donated blood or had been phlebotomized in the 3 months preceding laboratory investigation, leaving 213 patients for analysis.

The mean hemoglobin and hematocrit were 174 g/L and 0.52 L/L respectively. Forty patients (19%) had positive JAK2 molecular testing. Multivariate logistic regression showed that increasing age (p=0.001), increasing platelet count (p=0.006), decreasing serum ferritin (p=0.018) and low EPO (< 3.8 mU/mL) (p<0.001) were associated with JAK2-positive erythrocytosis, while the other components of the JAKPOT score, neutrophils (p=0.292) and erythrocytes (p=0.670), were not.

Low EPO had a Sn of 0.77 (95% CI, 0.62 – 0.87), Sp of 0.98 (95% CI, 0.94 – 0.99), +LR of 33 and -LR of 0.23 for the diagnosis of JAK2-positive erythrocytosis. A positive JAKPOT score had a Sn of 0.88 (95% CI, 0.73 – 0.94), Sp of 0.65 (95% CI, 0.57 – 0.72), +LR of 2.5 and -LR of 0.19 to diagnose JAK2-positive erythrocytosis. A JAKPOT score ≥ 1 or low EPO (EPO-JAKPOT+) had a Sn of 0.95 (95% CI, 0.83 – 0.98), Sp of 0.65 (95% CI, 0.58 – 0.72), +LR of 2.7 and -LR of 0.07. Restricting JAK2 testing to patients who were EPO-JAKPOT+ would have led to 55% fewer molecular tests in our cohort but would have missed 2 (5%) of JAK2+ patients.

Conclusions:

Individually, low EPO and the JAKPOT score had modest sensitivity for JAK2-positive erythrocytosis. Combining EPO and JAKPOT into the EPO-JAKPOT score increased the sensitivity to 95% and had a -LR of 0.07. While further validation is needed, this score has the potential to lessen molecular testing for erythrocytosis by identifying a population at low risk of disease.