In:
Blood, American Society of Hematology, Vol. 110, No. 11 ( 2007-11-16), p. 681-681
Abstract:
Established risk factors for thrombosis in Essential Thrombocythemia (ET) including age and previous vascular events have been incorporated into algorithms for risk assessment in clinical trials. Our interest is now to refine this risk stratification by adding, to this established predictive model, leukocytosis, found to be a new risk factor for these events. For this purpose we used the C statistic estimate that defines, from a Cox multivariable model, the probability of concordance between leukocytosis and events among comparable patients during the time. C statistic values range from 0.5 (usefulness of the test) to 1 (perfect discriminatory test) and allow to evaluate the specificity and sensitivity of the test (leukocytosis) in analogy of the area under the curve (AUC) used for assessing the accuracy of the diagnostic tests. This analysis provides an assessment of the incremental role of leukocytosis, in addition to conventional risk factors, for discriminating ET patients with or without thrombosis. Finally, by a receiver operating characteristic (ROC) curve, we looked for the best cut-off of leukocytes to stratify patients into risk categories. During follow-up (median 4.5 years) 657 ET (PVSG and WHO diagnostic criteria) patients, 212 males, 445 females, median age 52 years (range 8–93), seen in two academic Italian institutions, had 72 major thrombotic events (28 venous, 44 arterial). Cox proportional hazard model was performed to analyse the thrombotic risk, adjusted for the following baseline variables: centre, sex, standard risk factors (age ≥ 60 years and/or prior thrombosis), hemoglobin, hematocrit, platelet, leukocytes and JAK2V617F allele burden. Results confirmed that age, prior events and leukocytosis are independent risk factors for thrombosis. Interestingly, a gradient between white blood cell (WBC) number and venous and arterial events was documented [Reference category: WBC 〈 7.2 x109/L, WBC 7.2 − 8.7x109/L: RR=2.4, WBC 8.7 − 10.4x109/L: RR=2.7, WBC 〉 10.4 x109/L: RR=3.0, all p-values 〈 0.05]. On the contrary, no significant (p 〉 0.05) association was found either for JAK2V617F allele burden [Reference category: JAK2V617F 0%, JAK2V617F 1–25%: RR=1.2, JAK2V617F 26–50%: RR=1.5, JAK2V617F 〉 50%: RR=1.8] and for the other laboratory parameters. No centre-confounding effect was found. C statistics were then calculated on two Cox models for the prediction of major thrombosis in the follow-up of individual patients. The first model, including age over 60 years and/or prior thrombosis, showed a C statistic of 0.63. In the second one, by adding leukocytes at diagnosis, the C statistic was significantly improved (0.67). The best leukocyte cut-off values for predicting the events (ROC curve) ranged from 9.0 to 9.5 (x109/L), which corresponded to the highest sensitivity and specificity rates. In conclusion, we confirmed in this large retrospective cohort of ET patients that leukocytosis is an independent risk-factor for thrombosis. Moreover, we demonstrated by C statistic that leukocytosis has an incremental prognostic role in addition to conventional risk factors and found the best leukocytes cut-off able to discriminate between the group that will have thrombosis and the group that will not. These findings constitute a solid background to stratify patients in future clinical trials in ET.
Type of Medium:
Online Resource
ISSN:
0006-4971
,
1528-0020
DOI:
10.1182/blood.V110.11.681.681
Language:
English
Publisher:
American Society of Hematology
Publication Date:
2007
detail.hit.zdb_id:
1468538-3
detail.hit.zdb_id:
80069-7
Permalink
