Myelodysplastic syndromes (MDS): prognostic factors and scoring systems

Objective: To evaluate the score systems of Cassano and Sanz and suggest a new one. Design: Case series. Location: Teaching hospitais: EPM UNIFESP and Faculdade de Medicina de Botucatu. Participants: 59 patients diagnosed from 1979 to 1992.lntervention: Evaluation' of clinicai-laboratorial data. Measurement: Statistical comparison, uni and multivariate analysis and actuarial survival curves. Results: Cassano's system divided the patients into high and low risk (p=0.0966) while. Sanz's géive high, intermediate and lo\y risk (p=0.01 08). The univariate analysis showed hemoglobin, WBC count, E/M ratio, liver size and blast percentage in BM as statistically significant. The multivariate analysis showed blast percentage in BM (p=0.004) and Hb (p=0.050) as significant. Our system, considering the multivariate analysis data, divided the patients into high, intermediate and low risk (p=0.0038). Conclusions: Sanz's system was more functional than Cassano's, while ours showed predictive survival value and ease of use in clinicai practice.


INTRODUCTION
M yelOdYSPlasticsyndromes (MDS) are a complex group of hematological disordets characterized' by hypercellular bone marrow with dyshematopoiesis involving one or more celllineages and peripheral blood cytopenias that frequently transform into acute leukemia 1 ,2,3.Different authors have stressed that the MDS classification, proposed by the French-American-British (FAB) cooperative group in 1982 3 is only able to separate patients into two risk groups: refractory anemia (RA) plus refractory anemia with ringed sideroblasts (RARS) (low risk) and refractory anemia with excess of blasts (RAEB) plus refractory anemia with excess of blasts in transformation (RAEB-t) (high risk)4, 5, 6,7,8,9,10,11.
During the last 10 years there has been a growing interest in the analysis of variables of prognostic value in MDS, especially bacause of cases with unexpected clinicaI evolution or unclassified according to FAB. criteria l4 .Prognostic studies have recently been proposed using variables selected by univariate and multivariate regression analysis.Therefore, many scoring systems for predicting survival and leukemic transformation, as well as for selecting adequate therapeutic approaches for each individual case, have been proposed 9 ,11,15,16,17,18.The aims of this study were: 1) To evaluate the established scoring systems published by Cassano. 15

MATERIALS AND METHODS:
We studied 59 patients with MDS that were diagnosed at two different teaching hospitaIs in the state of São Paulo (Escola Paulista de MedicinalUNIFESP and Hospital da Faculdade de Medicina de Botucatu -Hospital da UNESP) from 1979 to 1992.
AlI patients were classified according to the estabilished scoring systems of Cassano l5 (Table 1) and Sanz 17 (Table 2).
ClinicaI features such as sex, race, age, interval between first symptoms and diagnosis, liver and spleen enlargement, and survival time were also examined.
In addition, cases were classified according to a scoring system using the significant variables of univariate and multivariate analysis.

Statistica/ Ana/ysis
Actuarial survival probability "curves were plotted according to the method of Kaplan and Meyer 19 • Different curves were compared statistically using the Cox-Mantel (log rank) or the generalized Wilcoxon test 20 • For univariate analysis, the cut-offlevel of each quantitative variable was established based on data in the literature.In some cases, the cut-off was established by trial and error, until "p" values were found close to 5%.For qualitative variables, the different categories were compared to each other.After prognostic features were selected by univariate analysis," multivariate analysis was performed according to Cox's modeFI.Variableswhich remained significant were included in the equation, the relative risk for each patient was estimated, and ~hepopulation was divided into three risk groups: low, intermediate and high.
The patients were then divided into 3 risk groups: low (RR 0.04 -0.15), intermediate (RR 0.18 -0.33) and high (RR 0.36 -2.57).The Wilcoxon test showed a siginificant difference in survival when low versus high (p=0.0346)and low versus intermediate risk groups (Figure 1) were compared.
Considering the significant variables of univariate and multivariate analysis a new scoring system was elaborated:  Using univariate analysis, liver size was a significant variable.Patients with an enlarged liver had a survi vaI of 19.30 months versus 84.70 months in those with unpalpable liver (p=0.0070).This was probably associated with more aggressive FAB groups and with the liver infiltration seen in more "aggressive" FAB subgroups (RAEB, RAEB-t and CMMoL).
Hemoglobin values were analyzed in the same ~anner as in Sanz'sl7 work, comparing the survival curves ofthree groups: as proposed by Sanz et aI.(1989): < 8 g%, 8-10 g% and > 10 g%, but was not significant.When the patients were divided into two groups, with Hb 6 g/dl (median survival of 24.40 months) and with Hb > 6 g/dl (median survival of 49.60 months), univariate and multivariate analysis showed statistically significant difference.These results confirm that hemoglobin leveIs are an important prognostic indicator, agreeing with other au thors4, 9,13, 17,22,23,24,25,26.Univariate analysis showed that WBC count was significant (p=0.0214)when the survivaI curves of groups with WBC counts of2.0 x 10 9 1 1(median survivaI of 16.10 months) and > 2.0 x 10 9 I I (median survivaI of 46.20 months) were compared (p=0.0214),confirming the importance of peripheraI cytopenias as observed by others 17.
In contrast to the Iiterature, in this study platelet counts did not appear to be a significant prognostic factor in this population.
Bone marrow cellularity was not significant, but the .'.median survival in the group with normocellularity was higher (84.70 months) than that of the hypercellular and Some authors have stressed that BM hypercellularity is an indicator of poor prognosis Ii,22,27.BM qualitative variables such as dyserythropoiesis, dysgranulopoiesis and dysmetakaryiocytopoiesis were not statistically significant.
BM fibrosis was present to some extent in 40% of patients, but was not significant.
The presence of ALIPs in BM biopsies has been demonstrated I0,25,28, 29,30,31 as a significant prognostic factor, but not in the present population.
Comparisons of erythro/myeloid ratio were significant (p=0.0456) in univariate analysis when comparing the survival curves of groups with RE/M <0.40 versus >0.40, with median survival of 19.60 and 84.70 months respectively (p=0.0456).This was aIs o observed by Cassano l5 , when comparing the survival curves of groups with RE/M <0.53 versus >0.53.The significance of this variable may be related to the increase in myeloid lineage and BM blast cells.
The BM blast cells percentage was the most significant prognostic factor in seen in this population, as already stressed by many other authors4,9, 10,13,23,24,26.Statistical significance was seen in univariate and multivariate analysis (p=0.0025 and p=0.0040, respectively), when the following survival curves were compared: • <5% BM blast cells: median survival of 84.70 monthsnths.• 5 -10% BM blast cells: median survival of 35.30 months.• >10% BM blast cells: median survival of 7.20 months.BM blast cell percentage, combined with the cytogenetics abnormalities, are considered to be the most important prognostic factors for survival ofMDS patients.
Scoring system "A" of Cassano et aI (1990) divided the patients into two risk groups: score <5 (low risk) with survival >50% in the period analyzed, and score >5 (high risk) with survival of 31.10 months, but was not statistically significant despite the difference in survival time.This is very interesting scoring because it included BM biopsy and qualitative data (dysmegakaryopoiesis, fibrosis and presence of ALIPs).This population showed a long survival rate in the high risk group (31.10 months), in concordance with the original work.Using Sanz's scoring system (Sanz et aI, 1989), patients were divided into three risk groups: score 0-1, with median survival of 84.70 months; score 2-3, with median survival of 16.10 months; score 4-5 months, with median survival of 7.20 months.The comparison of alI survi vaI curves was significant using the Wilcoxon test (p=0.0108)and also using Cox Mantel: 0-1 versus 2-3 (p=0.0027),0-1 versus 4-5 (p=0.0027) and 2-3 versus 4-5 (p=0.0080).Sanz's system seems to be more appropriate and more easily applicable for clinicaI practice than Cassano's.
The population could be separated into three risk groups using Cox's relative risk rnodel: 0.004-0.15-low risk, with >50% survival during the period studied; 0.18-0.33-intermediate risk, with >50% survival during the period studied; and 0.36-2.57-high risk, with survival of 16.10 months.The difference among the survival curves was significant (p=O.O165).
Using our proposed scoring systern, which includes variables derived from univariate analysis, ie WBC count and RE/M, patients were divided into three risk groups: O (low) with survival above 80% during the study period; 1 (intermediate) with median survival of 31.10 months and 2+3 (high) with median survival of 12.80 months (p=0.0038).
We conclude that the new scoring system presented here is easier to apply than Sanz's and Cassano's because it includes variables that are easily accessible to clinicians.
Cytogenetics abnormalities have been considered an important prognostic factor for survival of MDS patients.However, as it is not yet a test available to all patients, we consider that future studies in our country should include cytogenetic analysis. .