Background: Polycythemia vera (PV) is a rare hematologic neoplasm with significant morbidities. Blood hyperviscosity reflected by elevated hematocrit (HCT) levels is believed to contribute to most disease symptoms and complications including thrombotic events (TE), which are among the most serious clinical manifestations of PV. Concurrent with ongoing Phase III trials in PV, this study assesses treatment patterns, TE event rates, and direct costs associated with PV in real-world practice settings. Preliminary results on incidence of TE events in a privately insured US population are presented here.
Methods: Data for this retrospective observational study were taken from the MarketScan database containing medical and drug utilization data for ~40 million unique individuals enrolled in employer-sponsored private health insurance plans in the United States. Patients with at least two claims of a PV diagnosis (International Classification of Diseases, 9th Clinical Modification [ICD-9-CM] code 238.4) at least 30 days apart, between July 1, 2008 and December 31, 2011, were selected for study inclusion. Using propensity score (PS) methods, we identified a non-PV control group matched with PV cases on demographics and comorbidities. PS matching is frequently used in observational studies to achieve balance on observed covariates between cases and controls. Incident TE events were identified using ICD-9 diagnosis codes and were assessed from the index date (first PV diagnosis) until the earliest of plan disenrollment or end of the database. Both fatal and nonfatal events were included, although a distinction could not be ascertained in these data. The TE event rate was defined as the number of unique patients with a new TE occurrence (i.e., incident cases) per 1000 person-years (PYs) of observation time. Unadjusted TE event rates and incident rate ratios (RR) expressing the increased frequency of events in PV cases relative to controls were estimated for overall (i.e., any), cerebral, arterial, and venous TE events. Kaplan-Meyer (KM) analyses were performed to estimate time to event for each of the event categories listed above.
Results: The cohort comprised 8,124 PV patients with mean age of 51 years; 71% of patients were male. Table 1 describes the quality of balance between cases and matched controls for selected variables that differed significantly before PS matching. As shown in Figure 1, TE event rates were significantly higher among PV cases compared to matched controls (overall: 14.3 vs. 4.9/1000 PYs [RR 2.90 (95% CI: 2.28-3.68)], cerebral: 3.5 vs. 1.3/1000 PYs [RR 2.66 (95% CI: 1.66-4.24)], arterial: 6.9 vs. 2.4/1000 PYs [RR 3.16 (95% CI: 2.05-4.87)], and venous: 4.7 vs. 1.5/1000 PYs [RR 2.94 (95% CI: 2.08-4.15)]). KM analyses showed that event probabilities over time were significantly higher among PV cases (Figure 2). Due to high censoring in both the groups, median time to TE event could not be calculated. However, among those with an event, median time to event was shorter (but not statistically significant) among PV cases compared to controls for overall (7.0 vs. 9.7 months [P=0.0817]), cerebral (8.5 vs. 13.1 months [P=0.0634]), and arterial (9.2 vs. 14.7 months [P=0.0768]), but slightly higher for venous (5.3 vs. 4.8 months [P=0.5583]).
Conclusion: Patients with PV have significantly higher rate of TE events compared with matched controls. This rate may be higher in patients with HCT greater than 45%. Further analyses in this ongoing study will assess patient level predictors of cardiovascular and TE events and their impact on overall economic burden of PV.
