Abstract

Thymidine kinase (TK) activity is a marker of biological activity that allows the indolent and aggressive forms of chronic lymphocytic leukemia (CLL) to be distinguished. The aims of the study were to determine the relationship between TK activity and clinical status and prognosis, as well as to compare its activity with that of other prognostic and predictive factors. TK activity was measured in patient sera at the time of diagnosis using the DiviTum method, with the mean value being 439 Du/L. A correlation was found between TK activity and risk of disease progression (p=0.045). The optimal discriminative value of TK activity in the prediction of CLL progression was found to be 600 Du/L. TK activity significantly differed between the patients who achieved complete remission and those who only partially responded to therapy. In 93% of patients without any response to treatment and 18 out of 20 patients with progressive disease, TK activity over 600 Du/L was noted. In addition, all of the 10 patients with 17p13 deletion displayed TK activity of over 600 Du/L (p=0.0004). High TK activity also correlated with elevated levels of LDH (p=0.001) and β2‑microglobulin (p=0.03) in the study group. The results of the study indicated the importance of TK activity as a prognostic factor in patients with CLL.

Introduction

Chronic lymphocytic leukemia (CLL) is the most fre‑ quently diagnosed type of leukemia in Europe and North America with four new cases per year seen among 100,000 people [11,26]. Two clinical classifications were developed independently by Rai et al. and Binet et al. in the 1970s, and they still have significant prognostic value in clinical practice [3,22]. However, these classifi‑ cations are not relevant prognostic factors in patients with an early stage of the disease and do not predict the response to treatment. In recent years, a number of newer agents have been implemented, which are used to select patients with poorer prognosis and stage A disease according to Binet [23,24,27]. These include humoral fac‑ tors comprising the activity of thymidine kinase (TK) and concentrations of β2‑microglobulin in serum, the expression of CD38 on leukemic cells, as well as genetic markers, together with deletion 17p13 (del(17p)), TP53 mutation and the mutation status of IgVH [25]. Recently, an International Prognostic Index for CLL (CLL‑IPI) has been proposed, identifying four risk groups with signifi‑ cantly different five‑year survival. The CLL‑IPI takes into account the genetic risk factors (mutation status of IgVH, TP53 mutation and/or del(17p)), clinical stage, age, and β2 ‑microglobulin (β2 M) concentration, allowing the cli‑ nician to make a more accurate prognosis and plan an individualized treatment approach [12].

Although serum markers such as TK are not specific for CLL, they accurately reflect the activity and course of the disease. Thymidine kinase is involved in the synthe‑ sis of DNA strands and shows a significant correlation with cell proliferative activities in CLL. The results of several previous studies have highlighted the progno‑ stic significance of TK activity in patients with lympho‑ proliferative disorders [5,9,28]. In CLL, this correlates not only with the Rai clinical staging system but also with the biology of the disease, which allows for dif‑ ferentiation between its indolent and aggressive form [7,8,12,29,30]. It has also been shown that the activity of serum TK may be an independent predictive factor of disease duration without progression, and may com‑ plement the definition of indolent CLL in its early stage. A major problem in the wide use of TK as a prognostic factor for patients with CLL is the lack of a simple assay. The serum TK level is so low that the best method of measurement is to assess its activity [14].

Evaluation of enzyme level with the ELISA or dot blot methods is rarely applied in clinical practice [1]. Non‑radioactive measurement methods, which were developed in the 1990s, used a derivative of thymidine, azidothymidine, as a substrate, instead of a radioactively labeled base residue (TK Liaison probe). An alternative, non‑radioactive, method (DiviTum) uses as a substrate 5‑bromodeoxyuridine (BrdU), whose phosphorylation to monophosphate and then triphosphate (BrdUTP) is catalyzed by TK. BrdUTP is attached to a poly(A) oligo‑ nucleotide and then converted to a poly(A)/BrdU oligo‑ mer by reverse transcriptase, with the oligomer acting as a target for the antibody used in the probe. Both methods (TK Liaison and DiviTUM) are commercially available and have similar sensitivity [20], but neither has been widely used in patients with CLL.

The present study uses the DiviTUM method to assess the prognostic significance of TK in diagnosing CLL. The use of a DiviTum probe in determining TK activity was developed to avoid radioisotope determination. Yeast enzymes are used to induce phosphorylation in BrdU, a substrate for the studied enzyme. The DiviTum method is characterized by a higher sensitivity than the TK‑REA assay and enables more precise determination of the metabolic activity of the enzyme at higher activity.

Materials and methods

Patient characteristics

The study group comprised 67 previously untreated CLL patients, 31 (53.7%) women and 36 (46.3%) men aged 28‑87 years. The median age was 71 years. Patient cha‑ racteristics are summarized in Tables 1 and 2. Diagno‑ stic criteria for CLL comply with the criteria proposed by Hallek et al. [6]. The diagnosis was based on the complete blood count and lymphocyte immunophenotyping test. Next, the Rai and Binet clinical stage as well as routine parameters in CLL, including LDH activity and β2 M con‑ centration, were determined. Therapy was administered in 27 (40.2%) patients, including nine patients (33.3%) treated according to the cyclophosphamide, cladribine (CC) regimen, seven (25.9%) according to the CC com‑ bined with rituximab (RCC) regimen, and seven were given chlorambucil in monotherapy. One patient rece‑ ived rituximab with methylprednisolone and one rituxi‑ mab with bendamustine. All investigations were carried out in accordance with the principles of the Declaration of Helsinki.

Evaluation of treatment efficacy

Evaluation of response to anti‑leukemic therapy was based on the International Workshop on Chronic Lym‑ phocytic Leukemia (IWCLL) criteria [6]: Physical exa‑ mination, imaging (X‑rays, ultrasound, computed tomography) examinations and assessment of the com‑ plete blood count as well as bone marrow were perfor‑ med. Progression‑free survival (PFS) was measured from the date of therapy initiation until disease progression or death.

Evaluation of chromosomal aberrations by the fluorescence in situ hybridization (FISH) test

Peripheral blood was taken prior to treatment. Chro‑ mosomal aberrations were analyzed by the FISH method according to generally accepted principles [15]. A standard panel of probes for the detection of 13q14 (D13S319), 11q22 (ATM), 17p13 (TP53) deletions, and tri‑ somy 12 (Cytocell, UK) was used.

Evaluation of thymidine kinase activity

Peripheral blood was taken prior to treatment. The TK activity in the serum was determined in the Genloxa laboratory in Puck, using the DiviTum manual kit (Bio‑ vica International AB, Uppsala, Sweden) in accordance with manufacturer’s instructions (www.Biovica.com) [2]. The activity was expressed in DiviTum/L (Du/L) units.

Statistical analysis

The Statistica for Windows program, version 9.0 (Stat‑ Soft, Inc., Tulsa, USA) was used for the statistical analy‑ sis. Differences were considered significant for values of p <0.05 in all assays. To verify the differences between the groups, the analysis of variance and Kruskal‑Wallis test were carried out. Logistic regression analysis based on univariate and multivariate models was applied to assess the relationship between the two‑state depen‑ dent variable and a group of independent variables. Receiver operating characteristic (ROC) curve analysis was combined with the determination of a continuous variable cut‑off point at which the optimal discrimi‑ nation properties were achieved, and this was used to dichotomize and construct the dependent variable for logistic regression analysis. A multivariate model was built stepwise, including variables significantly chan‑ ging the adjustment of univariate models. The analysis of survival along with the construction of Kaplan‑Meier curves was combined with the uni – and multivariate Cox regression models. The nonparametric Mann‑Whit‑ ney test was used to compare independent grouped variables. Pearson’s chi‑square test was used to analyze qualitative variables (categorical). The arithmetic mean (X), standard deviation (SD), median, 25th and 75th percentiles, and the range of values (minimum – maximum) were determined.

Results

Laboratory and clinical characteristics of patients quali‑ fied for the study are shown in Tables 1 and 2. The study group included 31 (46.3%) women and 36 (53.7%) men, aged 28‑87 years. At diagnosis, Rai stage 0 or 1 chro‑ nic lymphocytic leukemia was detected in half of the patients (49.25%), and Binet stage A CLL was found in 41 patients (61.2%). The median value of leukocytosis was 48.6 x 109 /L. The median values of hemoglobin concen‑ tration and platelet count were within the normal range: 12.3g/dL and 145×109 /L, respectively (Table 2).

Among the examined patients, 35 individuals (52.2%) did not require treatment for the entire period of observa‑ tion. Chemotherapy was necessary in 32 (47.7%) patients.

First‑line treatment resulted in a complete remission (CR) in eight patients, partial remission (PR) in 10, and stable disease (SD) in four. Ten patients did not respond to the applied therapy. Progression of the disease in the treated group occurred in 19 out of 32 patients.

Serum TK activity was measured in the peripheral blood of all patients. The median TK activity was 439 Du/L. The ROC curve analysis showed that the TK activity obtained an optimal discriminative ability to predict disease pro‑ gression at the value > 600 Du/L. This value was deter‑ mined as the optimal cut‑off point in subsequent analysis.

The relationship between TK activity and response to treatment was shown by the defined multi‑state varia‑ ble taking into account the patients who achieved CR, PR, SD and NR (no remission) (p = 0.0018) (Fig. 1). The TK activity was compared in particular categories of the variable defining the treatment outcome (Table 3). Significant differences in the baseline TK level were noted between the patients who achieved CR and those who did not (p = 0.0018). In all patients who achieved PR following treatment, the TK activity was above the usual cut‑off value (600 Du/L). TK activity above 600 Du/L was found in 92.8% of the patients with no response to treat‑ ment. The progression‑free survival (PFS) analysis reve‑ aled significant differences in patients in whom the TK level was ≥600 Du/l at diagnosis, as compared to PFS in those with the TK level <600 Du/l (p = 0.045) (Fig. 2).

A significant correlation between serum TK activity and severity of disease was detected at diagnosis, according to Rai (p = 0.0001) and Binet (p = 0.006) staging (Fig. 3). However, no significant correlation was found between TK activity and age of patients.

Significantly higher TK activity was observed in patients with lymphadenopathy than in patients without this disorder (p = 0.013) (Fig. 4). No significant correlation was observed between TK activity and presence of sple‑ nomegaly (p = 0.05) (Fig. 4). Patients with autoimmune hemolytic anemia (AIHA) showed significantly higher TK activity than those without (p=0.005).

The patients with del(17p) were found to have a signi‑ ficantly higher TK level than the group of patients without del(17p) (p=0.0004) (Fig. 5). In all patients with del(17p), the enzyme activity exceeded 600 Du/L. No significant difference in TK activity was observed between the patients with del(11q), those with del(13q) and those with trisomy 12. The TK activity in patients with β2 M> 2.2 was significantly higher than in patients with β2 M <2.2 (p = 0.03) (Fig. 5).

Discussion

Our findings indicate that high serum TK activity cor‑ relates with a poorer response to treatment. In a study of the TK activity in 188 patients with progressive or advanced CLL treated with fludarabine, Di Raimondo et al. [4] found significantly greater TK activity in 92% of patients, and similarly to our study, correlated with tumor mass parameters (lymphocytosis, bone marrow cellularity). In addition, the authors noted that TK acti‑ vity may be a parameter predicting the response to tre‑ atment, which is consistent with our observations. It has been demonstrated that 83% of patients with serum TK activity <10 U/L obtained partial or complete remission after the treatment with fludarabine, whereas only 45% of patients with TK levels >10 U/L have been found to have a satisfactory response to treatment [4].

Other studies have shown that TK activity has indepen‑ dent prognostic value in patients with CLL, and particularly in patients with less advanced disease (Binet A) [7]. Hallek et al. [8] suggest that in these patients, TK activity in serum can be a particularly valuable prognostic mar‑ ker. In the Binet A‑stage patients at high risk and with TK activity> 7.1 U/L, the mean PFS time was found to be about eight months, while the PFS time was 49 months for Binet A patients at low risk, with an enzyme level <7.1 U/L [8]. Other studies have demonstrated that only three parameters can be used as independent prognostic fac‑ tors for PFS, including TK above 7.1 U/L, the presence of lymphadenopathy and lymphocytosis above 75×109 /l [13].

Matthews et al. [18] compared the activity of serum TK with other parameters associated with a worse prognosis, including the absence of gene mutation for IgVH, adverse chromosomal aberrations, increased expression of zeta chain‑associated protein of 70 kDa kinase (ZAP‑70) and CD38 expression [18]. It was found that TK activity was significantly higher in patients without IgVH mutations and, as in our study, with del(17p); however, in contrast to our study, they also reported a correlation between TK activity and trisomy 12 and del (11q). Konoplev et al. pointed out that, like CD38 and ZAP 70, TK activity may be an intermediate exponent of the presence or absence of IgVH mutations [14]. The correlation between TK acti‑ vity in plasma and the presence of IgVH mutation has also been observed by other researchers. By determining the activity of LDH, IgVH mutation and CD38 expression, it was found that only increased TK activity is an independent factor associated with mutation absence [18,30].

No differences were found between TK activity and patient age in the present study. Szantho et al. [29] compared the results of TK activity in different age groups of patients with CLL and in healthy volunteers and proposed that the range of enzyme activity should be changed depending on patient age. In both healthy volunteers and patients with leukemia, the enzyme activity was higher among younger people as compared to the group of elderly patients.

Attention should be drawn to the prognostic model pro‑ posed recently by Pflug et al. [21], in which TK activity was used as one of the prognostic parameters. The rese‑ archers prospectively analyzed 23 prognostic markers in 1,948 patients from three randomized studies conducted by the German CLL Study Group. The multivariate Cox regression model was used to assess the impact of TK and seven other independent prognostic factors on ove‑ rall survival: age, sex, general condition according to the Eastern Cooperative Oncology Group (ECOG), del(17p), del(11q), the state of IgVH mutation, β2 M concentration and TK activity in serum. Letestu et al. [16] demonstra‑ ted that a relatively simple evaluation of the TK level may be one of the key parameters classifying Binet stage A patients into high‑risk or low‑risk groups. Those authors suggest that early treatment may be possible in some of the patients from the stage A low‑risk group, who are at a greater risk of progression. This model allows us to reduce the number of ambulatory visits for patients at low risk and, on the other hand, to monitor more frequently patients at high risk of progression [21].

To conclude, our findings confirm the value of serum TK as a prognostic factor. The method used in the study is relatively simple and is available for the majority of hematological centers.

Acknowledgements

We thank Anna Wiśniewska and Edward Lowczowski from the Medical University of Lodz for editorial assi‑ stance. The study was supported in part by grant fun‑ ding from the Medical University of Lodz.

References

• 1. Alegre M.M., Weyant M.J., Bennett D.T., Yu J.A., Ramsden M.K.,Elnaggar A., Robison R.A., O’Neill K.L.: Serum detection of thymidinekinase 1 as a means of early detection of lung cancer. AnticancerRes., 2014; 34: 2145‑2151
  Google Scholar
• 2. Bacovsky J., Myslivecek M., Minarik J., Scudla V., Pika T., Zaple‑talova J., Petrova P., Bartkova M., Adam T., Gronowitz S.J.: Analy‑sis of thymidine kinase serum levels by novel method DiviTum inmultiple myeloma and monoclonal gammopathy of undeterminedsignificance – comparison with imaging methods 99mTc‑MIBI scin‑tigraphy and 18F‑FDG PET/CT. Biomed. Pap. Med. Fac. Univ. PalackyOlomouc Czech Repub., 2015; 159: 135‑138
  Google Scholar
• 3. Binet J.L., Auquier A., Dighiero G., Chastang C., Piguet H., Goas‑guen J., Vaugier G., Potron G., Colona P., Oberling F., Thomas M., Tch‑ernia G., Jacquillat C., Boivin P., Lesty C., Duault M.T., MonconduitM., Belabbes S., Gremy F.: A new prognostic classification of chroniclymphocytic leukemia derived from a multivariate survival analysis.Cancer, 1981; 48: 198‑206
  Google Scholar
• 4. Di Raimondo F., Giustolisi R., Lerner S., Cacciola E., O’Brien S., Kan‑tarjian H., Keating M.J.: Retrospective study of the prognostic roleof serum thymidine kinase level in CLL patients with active diseasetreated with fludarabine. Ann. Oncol., 2001; 12: 621‑625
  Google Scholar
• 5. Gronowitz J.S., Hagberg H., Källander C.F., Simonsson B.: Theuse of serum deoxythymidine kinase as a prognostic marker, andin the monitoring of patients with non‑Hodgkin’s lymphoma. Br. J.Cancer, 1983; 47: 487‑495
  Google Scholar
• 6. Hallek M., Cheson B.D., Catovsky D., Caligaris‑Cappio F., Dighie‑ro G., Döhner H., Hillmen P., Keating M.J., Montserrat E., Rai K.R.,Kipps T.J., International Workshop on Chronic Lymphocytic Leu‑kemia: Guidelines for the diagnosis and treatment of chronic lym‑phocytic leukemia: a report from the international Workshop onChronic Lymphocytic Leukemia updating the National Cancer In‑stitute‑Working Group 1996 guidelines. Blood, 2008; 111: 5446‑5456
  Google Scholar
• 7. Hallek M., Langenmayer I., Nerl C., Knauf W., Dietzfelbinger H.,Adorf D., Ostwald M., Busch R., Kuhn‑Hallek I., Thiel E., EmmerichB.: Elevated serum thymidine kinase levels identify a subgroup athigh risk of disease progression in early, nonsmoldering chroniclymphocytic leukemia. Blood, 1999; 93: 1732‑1737
  Google Scholar
• 8. Hallek M., Wanders L., Ostwald M., Busch R., SenekowitschR., Stern S., Schick H.D., Kuhn‑Hallek I., Emmerich B.: Serumβ2‑microglobulin and serum thymidine kinase are independent pre‑dictors of progression‑free survival in chronic lymphocytic leukemiaand immunocytoma. Leuk. Lymphoma, 1996; 22: 439‑447
  Google Scholar
• 9. Hallek M., Wanders L., Strohmeyer S., Emmerich B.: Thymidinekinase: a tumor marker with prognostic value for non‑Hodgkin’slymphoma and a broad range of potential clinical applications. Ann.Hematol., 1992; 65: 1‑5
  Google Scholar
• 10. Hamblin T.: Chronic lymphocytic leukaemia: one disease or two?Ann. Hematol., 2002; 81: 299‑303
  Google Scholar
• 11. Hayat M.J., Howlader N., Reichman M.E., Edwards B.K.: Cancerstatistics, trends, and multiple primary cancer analyses from theReferencesSurveillance, Epidemiology, and End Results (SEER) Program. On‑cologist, 2007; 12: 20‑37
  Google Scholar
• 12. International CLL‑IPI working group: An international progno‑stic index for patients with chronic lymphocytic leukaemia (CLL‑IPI):a meta‑analysis of individual patient data. Lancet Oncol., 2016; 17:779‑790
  Google Scholar
• 13. Källander C.F., Simonsson B., Hagberg H., Gronowitz J.S.: Serumdeoxythymidine kinase gives prognostic information in chroniclymphocytic leukemia. Cancer, 1984; 54: 2450‑2455
  Google Scholar
• 14. Konoplev S.N., Fritsche H.A., O’Brien S., Wierda W.G., KeatingM.J., Gornet T.G., St Romain S., Wang X., Inamdar K., Johnson M.R.,Medeiros L.J., Bueso‑Ramos C.E.: High serum thymidine kinase 1 le‑vel predicts poorer survival in patients with chronic lymphocyticleukemia. Am. J. Clin. Pathol., 2010; 134: 472‑477
  Google Scholar
• 15. Kotkowska A., Wawrzyniak E., Blonski J.Z., Robak T., Koryc‑ka‑Wolowiec A.: Chromosomal aberrations in chronic lymphocy‑tic leukemia detected by conventional cytogenetics with DSP30 asa single agent: comparison with FISH. Leuk. Res., 2011; 35: 1032‑1038
  Google Scholar
• 16. Lejcko J., Jungerova J., Topolcan O., Koza V.: Thymidine kinase atmalignant lymphogranuloma and Non‑Hodgkin Lymphoma. Prakt.Lek., 1992; 72: 171‑173
  Google Scholar
• 17. Letestu R., Lévy V., Eclache V., Baran‑Marszak F., Vaur D., Na‑guib D., Schischmanoff O., Katsahian S., Nguyen‑Khac F., Davi F.,Merle‑Béral H., Troussard X., Ajchenbaum‑Cymbalista F.: Prognosisof Binet stage A chronic lymphocytic leukemia patients: the strengthof routine parameters. Blood, 2010; 116: 4588‑4590
  Google Scholar
• 18. Magnac C., Porcher R., Davi F., Nataf J., Payelle‑Brogard B., TangR.P., Oppezzo P., Lévy V., Dighiero G., Ajchenbaum‑Cymbalista F.: Pre‑dictive value of serum thymidine kinase level for Ig‑V mutationalstatus in B‑CLL. Leukemia, 2003; 17: 133‑137
  Google Scholar
• 19. Matthews C., Catherwood M.A., Morris T.C., Kettle P.J., DrakeM.B., Gilmore W.S., Alexander H.D.: Serum TK levels in CLL identifyBinet stage A patients within biologically defined prognostic sub‑groups most likely to undergo disease progression. Eur. J. Haema‑tol., 2006; 77: 309‑317
  Google Scholar
• 20. Nisman B., Allweis T., Kadouri L., Mali B., Hamburger T., BarasM., Gronowitz S., Peretz T.: Comparison of diagnostic and prog‑nostic performance of two assays measuring thymidine kinase 1activity in serum of breast cancer patients. Clin. Chem. Lab. Med.,2013; 51: 439‑447
  Google Scholar
• 21. Pflug N., Bahlo J., Shanafelt T.D., Eichhorst B.F., Bergmann M.A.,Elter T., Bauer K., Malchau G., Rabe K.G., Stilgenbauer S., Döhner H.,Jäger U., Eckart M.J., Hopfinger G., Busch R., Fink A.M., WendtnerC.M., Fischer K., Kay N.E., Hallek M.: Development of a comprehen‑sive prognostic index for patients with chronic lymphocytic leuke‑mia. Blood, 2014; 124: 49‑62
  Google Scholar
• 22. Rai K.R., Sawitsky A., Cronkite E.P., Chanana A.D., Levy R.N.,Pasternack B.S.: Clinical staging of chronic lymphocytic leukemia.Blood, 1975; 46: 219‑234
  Google Scholar
• 23. Robak T., Hus I., Błoński J., Giannopoulos K., Jamroziak K., RolińskiJ., Smolewski P., Wołowiec D.: Rekomendacje diagnostyczne i terapeu‑tyczne dla przewlekłej białaczki limfocytowej w 2014 r. – raport GrupyRoboczej PTHiT oraz PALG – CLL. Acta Haematol. Pol., 2014; 45: 221‑239
  Google Scholar
• 24. Robak T., Jamroziak K., Robak P.: Current and emerging treat‑ments for chronic lymphocytic leukaemia. Drugs, 2009; 69: 2415‑2449
  Google Scholar
• 25. Sagatys E.M., Zhang L.: Clinical and laboratory prognostic indica‑tors in chronic lymphocytic leukemia. Cancer Control, 2012; 19: 18‑25
  Google Scholar
• 26. SEER. Chronic Lymphocytic Leukemia. Stat Fact Sheets. http://seer.cancer.gov/statfacts/html/clyl.html (02.07.2011)
  Google Scholar
• 27. Stelmach P., Błoński J.Z., Robak T.: Czynniki prognostyczne w prze‑wlekłej białaczce limfocytowej. Onkol. Prakt. Klin., 2014; 10: 322‑329
  Google Scholar
• 28. Stelmach P., Robak T.: Pathogenesis, prophylaxis and treatmentof infections in patients with chronic lymphocytic leukemia. PostępyHig. Med. Dośw., 2013; 67: 560‑568
  Google Scholar
• 29. Suki S., Swan F. Jr., Tucker S., Fritsche H.A., Redman J.R.,Rodriguez M.A., McLaughlin P., Romaguera J., Hagemeister F.B.,Velasquez W.S., Sarris A.H., Younes A., Cabanillas F.: Risk classi‑fication for large cell lymphoma using lactate dehydrogenase,beta‑2 microglobulin, and thymidine kinase. Leuk. Lymphoma,1995; 18: 87‑92
  Google Scholar
• 30. Szánthó E., Bhattoa H.P., Csobán M., Antal‑Szalmás P., ÚjfalusiA., Kappelmayer J., Hevessy Z.: Serum thymidine kinase activity:analytical performance, age‑related reference ranges and valida‑tion in chronic lymphocytic leukemia. PLoS One, 2014; 9: e91647
  Google Scholar
• 31. Xu W., Cao X., Miao K.R., Qiao C., Wu Y.J., Liu Q., Fan L., LiJ.Y.: Serum thymidine kinase 1 concentration in Chinese patientswith chronic lymphocytic leukemia and its correlation with otherprognostic factors. Int. J. Hematol., 2009; 90: 205‑211
  Google Scholar

Full text