Blood tests may predict early primary myelofibrosis in patients presenting with essential thrombocythemia Alessandra Carobbio, 1 Guido Finazzi, 1 Juergen Thiele, 2 Hans-Michael Kvasnicka, 3 Francesco Passamonti, 4 Elisa Rumi, 5 Marco Ruggeri, 6 Francesco Rodeghiero, 6 Maria Luigia Randi, 7 Irene Bertozzi, 7 Alessandro M. Vannucchi, 8 Elisabetta Antonioli, 8 Heinz Gisslinger, 9 Veronika Buxhofer-Ausch, 9 Naseema Gangat, 10 Alessandro Rambaldi, 1 Ayalew Tefferi, 10 and Tiziano Barbui 1 * According to World Health Organization (WHO)-defined criteria, patients presenting clinically as essential thrombocythemia (ET) may show early primary myelofibrosis (PMF) with accompanying thrombo- cythemia [1]. Previous clinicopathological studies revealed that labora- tory parameters like gender-matched hemoglobin (Hb), white blood cell (WBC) count, and particularly lactate dehydrogenase (LDH) values are significantly different in PMF [2]. By strictly applying the WHO cri- teria, our investigation was aimed to study sensitivity and specificity of these features in an exploratory cohort of 536 patients and to vali- date the results on an independently recruited series of 321 strictly corresponding patients. The discriminatory power of these parameters (Hb, WBC, and LDH) was tested by plotting their receiver operating characteristic curves. The best performance was found for LDH (areas under the curve, AUC 5 0.7059). WBC and Hb had superimposable curves, with AUC of 0.6279 and 0.6257, respectively. A diagnostic algo- rithm was generated by applying these parameters in a stepwise fash- ion. Nearly half of the patients could be correctly allocated to WHO- defined ET or early PMF in both cohorts investigated. It is important to note that this result does not substitute bone marrow morphology with hematological parameters, however, in clinical practice may alert physicians to get more suspicious of early PMF in a patient presum- ably presenting with ET . Following current criteria established by the World Health Organization (WHO), patients presenting with a clinical picture of essential thrombocy- themia (ET) can actually have an early primary myelofibrosis (PMF) [3]. Incidences may vary depending on the accuracy of applied diagnostic guidelines [4–6] and was recently reported in about 18% of cases of so- called ET [7]. Laboratory tests which are significantly different in early PMF as compared with histologically confirmed ET (WHO-ET) include decreased Hb, increased white blood cell (WBC) and lactate dehydrogen- ase (LDH) values. Although bone marrow biopsy represents the gold standard for differentiating WHO-ET from early PMF [8], in clinical prac- tice it might be very useful to know whether one or more of these base- line laboratory parameters may help to differentiate these two entities. To tackle this issue, we evaluated sensitivity (SE) and specificity (SP) of blood cells counts and LDH for the diagnosis of early PMF versus WHO- ET in an exploratory cohort of 536 histologically diagnosed patients aimed to propose an algorithm for the practical clinical use of these tests. This algorithm was then applied to an independently recruited validation cohort of 321 patients presenting with similar diagnosis of early PMF ver- sus WHO-ET. The starting point of this study was an international data- base of 1,071 patients with ET either confirmed by WHO criteria (891 cases) or revised to early PMF (180 cases) as detailed elsewhere [7]. From this database, 536 patients (50%) who had complete laboratory data measured at diagnosis were extracted and constituted the training set of our study. Clinical and laboratory characteristics of the 536 selected patients were not different from that of the initial population. To recognize early PMF, the salient parameters SE and SP have been evaluated by receiver oper- ating characteristic (ROC) curves (Fig. 1). The worst performance was registered for platelets (PLT) count: its areas under the curve (AUC) was only 0.5628, not significantly different from the reference value of 0.50 (P 5 0.154). Thresholds of Hb, WBC and LDH were searched to achieve at least 90% of SE or SP. Table I shows that Hb < 12 g/dL for women or <13 g/dL for men, or WBC >513 3 10 9 /L had higher SP (92 and 91%, respectively). High SP of these two tests is associated with few false positives (WHO-ET cases, incorrectly identified as early PMF), thus it is highly related to the presence of early PMF. Concerning high levels of LDH, we could not find an upper threshold with a SP >90% for diagnosis of early PMF and including at least 5% of patients. On the contrary, LDH < 200 mU/mL and WBC < 7 3 10 9 /L had good SE (91 and 94%, respectively). High SE produces few false negatives (early PMF cases not recognized and incorrectly classified as ET), thus it is highly related to the confirm of WHO-ET diagnosis. To optimize the use of this data in daily clinical practice, a step-by-step algorithm was proposed. By applying the criterium ‘‘anemia,’’ SP was 92% indicating that 46/536 (9%) of patients presenting anemia were correctly classified in 92% of cases as early PMF. In the 490 remaining patients, we applied two different WBC cut-offs: 97 (18%) and 43 (8%) patients were classified as WHO-ET or early PMF. Finally, in the remaining patients, LDH value (<200 mU/mL) classified as WHO-ET in another 13% of patients. Thus, following these steps, nearly half of patients (48%) could be classified as WHO-ET (by SE) or early PMF (by SP), assuming at each step a margin of error of about 10% (Fig. 2). For the remaining 50% of patients, laboratory results did not allow to suspect or exclude the presence of early PMF. To validate these results, the perform- ance of the same thresholds of laboratory parameters was tested in 321 patients classified as WHO-ET or early PMF (Cologne cohort). SP of ane- mia was 84%, WBC below 7 3 10 9 /L or above 13 3 10 9 /L had 91 and 81% of SE and SP, respectively. LDH values < 200 revealed a SE of 85%. By applying the above reported flow chart, the percentage of patients classified as WHO-ET or early PMF was almost similar (46%), even though the margin of error at each step was higher, likely due to smaller number of patients. In conclusion, while patients presenting clinically with ET can now be discriminated as true ET or early PMF by adopting the WHO 2008 criteria that require bone marrow histology, considering Hb, WBC, and LDH, this differentiation can be recognized in about 50% of patients with a good approximation. On the other hand, for a definitive proof, bone marrow his- tology remains still an integral part for final diagnosis. For this reason, it should be emphasized that results of our study do not substitute bone marrow biopsy with laboratory parameters to distinguish true ET from early PMF. Instead, it is to provide clinicians with laboratory parameters that should increase suspicion of early PMF in a patient with a working clinical diagnosis of ET. Methods Laboratory parameters considered in this study were those already found significantly different between early PMF and WHO-ET: Hb (g/dL), WBC (310 9 /L), PLT (310 9 /L), and LDH (mU/mL). The discriminatory ability of these parameters in correctly classifying patients in the early PMF or WHO- ET groups was initially tested by plotting their ROC curves and comparing the relative AUC with the value of 0.50 (which stands for the completely use- less application of the test) [9,10]. Three parameters with statistically signifi- cant discriminatory power were chosen (Hb, WBC, and LDH) and opportune thresholds searched to guarantee at least 90% of SE or SP [11]. Finally, a diagnostic algorithm was designed. The validation set of this analysis was constituted by 321 patients with WHO-ET (n 5 62) or early PMF (n 5 259) diagnosed by the same pathologist who confirmed the training set cohort and collected in the Institute for Pathology, University of Cologne, Germany. Letters V V C 2011 Wiley Periodicals, Inc. American Journal of Hematology 203 http://wileyonlinelibrary.com/cgi-bin/jhome/35105