American Journal of Hematology 27:163-168 (1988)

Prognostic Factors in Chronic Myelodysplastic Syndromes: A Multivariate Analysis in 107 Cases

Sonia Garcia, Miguel A. Sanz, Victoria Amigo, Pedro Colomina, M. Dolores Carrera, J. lgnacio Lorenzo, and Guillermo F. Sanz

Secci6n de Hematoiogia Cifnica, Servicio de Hematolog fa, Hospital La Fe, Valencia, Spain

A retrospective multivariate analysis of 37 clinical, biochemical, and hematological data was performed in 107 cases of primary myeiodispiastic syndromes (MDS) in order to recognize their prognostic significance. The most important individual variables, iso- lated in a previous univariate analysis, were placed In a multiple regression modeling procedure to identify major significant prognostic factors. Multivariate analysis tends to Identify prognostic variables containing significant independent predictive information. Characteristics were examined on both continuous and binary bases. The FAB ciassifi- cation was the first parameter entered In regression equations on both models, followed by platelet count, hemoglobin level, and circulating erythrobiasts in the binary model, and by hemoglobin level, systemic symptoms, platelet count, age, and dyserythro- poiesis in the continuous model. Our analysis confirms FAB ciasslfication as the main prognostic factor in MDS, supports the previously noted predictive value of platelet count, hemoglobin level, and age, and recognlses the importance of circulating ery- throblasts, systemic symptoms, and dyserythropoiesis as prognostic characteristics in MDS.

Key words: myeiodyspiasia, preleukemia

INTRODUCTION

The myelodysplastic syndromes (MDS) are a group of primary hematological disorders particularly common in the elderly, resulting from neoplastic transformation at the level of the pluripotent stem cell. MDS are associated with a high mortality that results from infections and/or bleeding caused by persistent cytopenias. However, some patients may remain symptom-free for prolonged periods of time [ 1 4 ] . In 1982, the French-American-British (FAB) Cooperative Group proposed a classification sys- tem for MDS [3], that recognised five clinical entities: refractory anemia (RA), refractory anemia with ring sid- eroblasts (RAS), chronic myelomonocytic leukemia (CMML), refractory anemia with excess of blasts (RAEB), and refractory anemia with excess of blasts “in transformation” (RAEB-t), with a variable incidence of morbidity and mortality. The prognostic value of the FAB classification and other individual features in MDS has been evaluated in some recent studies in order to

identify poor-prognosis patients [ 1,4-81, who might ben- efit from an early therapeutic approach. Unfortunately, most of these studies have been done on the basis of univariate analysis without taking into account correla- tion between variables. Multivariate analyses identify prognostic variables containing significant independent predictive information, and the prognostic models de- rived from these procedures offer higher predictive ca- pacity than that of isolated single variables contained in the multiple regression model. In this paper, we report the results of a multivariate survival analysis performed to evaluate the prognostic significance of multiple clinical and morphological characteristics at diagnosis, including the FAB classification, in a series of 107 patients with MDS.

Received for publication April 17, 1987; accepted October 8, 1987.

Address reprint requests to Miguel A. Sam, Secci6n de Hematologia Clinica, Servicio de Heniatologia, Hospital La Fe, Avda. Carnpanar 19, 46009 Valencia, Spain.

0 1988 Alan R. Liss, Inc.

164 Garcia et al

TABLE 1. Assessment of Dyshematopoiesis

Dyserythropoiesis Gigantism Multinuclearity Asynchrony of maturation Karyorrhexis Nuclear budding Basophilic stippling Howell-Jolly bodies Megaloblastic changes

Agranular neutrophils Abnormal granules Pelger-Huet anomaly Abnormal promyelocytes Dystrophic blasts

Dysthrombopoiesis Micromegakaryocytes Mononuclear megakaryocytes Megakaryocytes with

Dy sgranulopoiesis

multiple small nucleus

Marked: Two or more abnormalities in > 50% of the

Mild: Any abnormality in > 50% of the erythroblasts erythroblasts

Marked: Two or more abnormalities in > 50% of myeloid cells

Mild: Two or more abnormalities in 10-50% of myeloid cells or > 50% of agranular neutrophils

or 1&50% of agranular neutrophils

Marked: Any abnormality in > 50% of megakaryocytes Mild: Any abnormalilty in 10-50% of megakaryocytes

MATERIALS AND METHODS ship on survival (univariate analysis). Variables Patients and Diagnostic Criteria considered for possible inclusion in the Cox regression

analysis were those for which there was some indication A total of 124 patients were diagnosed as having MDS of a significant association with survival in univariate

gested a possible association. The model was tested twice the present study 107 cases were evaluated. Borderline

by expressing the values in a continuous way (continuous cases of acute myeloblastic leukemia (AML) and patients with no bone marrow available for study were excluded. and by grouping into categories (binary model). Cases with dysmyelopoiesis secondary to vitamin B12 or The most discriminating cut-off level was employed in folic acid deficiency were not considered. None of the order to define categories of variables in the binary patients had previously received chemotherapy or radio- model. Statistical analyses were performed by using 1L

and 2L programs from the BMDP statistical package [ 111 therapy. Patients were studied for disease progression

running on an IBM PC AT microcomputer. and survival through May 1986. Patients were classified according to the FAB criteria

during the period from November 1971 to 1985* For analysis (P < .lo), or for which prior studies had sug-

[3]. Hematological examinations were performed using standard methods; bone marrow and peripheral blood smears were stained with May-Griinwald-Giemsa and reviewed by two observers. Marrow cell differential counts were performed on at least 500 cells. Dyshema- topoiesis was defined following the FAB [3] and Juneja criteria [2] (Table I).

Statistical Analysis The Kaplan-Meier product limit method was used to

estimate the probability of survival. Different curves were statistically compared by using the log-rank test or, if applicable, the test for trend, as recommended by Pet0 et a1 [9]. Further multivariate analysis, by means of a multiple regression model for censored survival data, developed by Cox [ 101, was employed in order to identify the most significant prognostic factors.

Thirty-seven patient and disease characteristics at di- agnosis were considered individually for their relation-

RESULTS

Sixty-one of 107 patients were men (57%) and 46 were women (43 %). The median age was 69 years (range 31- 92 years); only three patients (3%) were under age 50 years and 75% of the patients were above age 65 years. Table II shows the relationship of survival to age, sex, and 35 other patient and disease characteristics at diag- nosis. The median survival of the entire series was 11 months, with a range of 0.13 to 62 months. Twenty-one patients are still alive 10+ to 158+ months from diagnosis.

CI i nical Features Age and sex were not statistically related to the length

of survival. The presence of systemic symptoms (weak- ness, anorexia, and weight loss) was the only datum from the medical chart found to be significantly correlated with survival time (P = .004).

Prognostic Factors in Myelodysplastic Syndromes 165

TABLE II. Relationship Between Survival and Features at Diagnosis'

Median Features survival (all patients) Patients (months) P

Sex Males Females

< 70 > 70

No Yes

No Yes

No Yes

No Yes

No Yes

Age

Associated disease

Cardiopathy

Anemia symptoms

Hemorrhagic symptoms

Infection symptoms

Systemic symptoms (weakness, anorexia, and weight loss)

No Yes

No Yes

No Yes

No Yes

<7 >7

< 75 > 75

<2.5

Lymphadenopathy

Splenomegal y

Hepatomegaly

Hemoglobin gldl

Reticulccytes 109/L

Leucocytes 1 0 ~ 1 ~

>2.5

<2.5 >2.5

< 50 > 50

No Yes

<5% >5%

No Yes

Neutrophils 109/L

Platelets I O ~ / L

Circulating blasts

Circulating myeloid precursors

Circulating erythroblasts

107 61 (57) 46 (43)

53 (50) 54 (50)

80 (78) 22 (22)

92 (90) 10 (10)

14 (13) 92 (87)

82 (77) 24 (23)

95 (90) 11 (10)

89 (84) 17 (16)

95 (89) 11 (11)

92 (87) 14 (13)

84 (79) 22 (21)

23 (21) 84 (79)

65 (72) 25 (28)

15 (14) 92 (86)

57 (55) 50 (45)

25 (24) 79 (76)

89 (83) 18 (17)

94 (88) 13 (12)

14 (13) 93 (87)

11 8.8

17.8

9.7 11.9

10.7 11.4

10.8 4

9.4 10.7

10.7 6.8

10.4 9.6

11.6 4

9.2 24

10. I 10.7

9.4 10.7

5.9 11.7

14.8 7.7

5.7 10.8

10.7 11.2

3.3 14.9

11.6 4

11.5 5.1

2.5 11.7

NS

NS

NS

NS

NS

NS

NS

= . m 9

NS

NS

NS

=.021

NS

NS

NS

= .OOO16

= .0674

= ,0608

= .048

(continued)

TABLE II. Relatlonshlp Between Survlval and Features at Dlagnosls' (continued)

Median Features survival (all patients) Patients (months) P B.M. cellularity

Hypo Normal + hyper

B.M. promonocytes < 5 % > 5 %

< 50% >50%

Blasts I < 5 % > 5 %

Blasts I1 <5W > 5 %

No+mild Severe

Dy sgranulopoiesis No Mild +severe

Dysthrombopoiesis No Mild +severe

FAB categories RA RAS RAEB RAEB-t CMML RA+RAS RAEB+RAEB-t + CMML

Bun mg./dl. < 20 > 20

< 1.5 > 1.5

<4.5 >4.5

< 8 > 8

< I > I

< 55 > 55

<40 >40

LDH IU/L < 225 > 225

< 85 > 85

B.M. erythroblasts

Dyserythropoiesis

Creatinine mg./dl.

P h o s p h o ~ ~ mg . /dl .

Uric ac. mg./dl.

Bilirubin mg./dl.

SGPT IU/L

SGOT IU/L

Alkaline phosphatase IU/L

13 (11) 85 (87)

98 (92) 09 (8)

87 (81) 20 (19)

47 (44) 60 (56)

88 (82) 19 (18)

89 (83) 18 (17)

35 (33) 71 (67)

38 (37) 64 (63)

13 (12) 18 (17) 36 (33) 20 (19) 20 (19) 31 (29) 76 (71)

58 (56) 45 (44)

73 (79) 19 (21)

77 (92) 7 (8)

83 (86) 13 (14)

83 (82) 18 (18)

83 (91) 8 (9)

86 (86) 14 (14)

42 (47) 47 (53)

64 (61) 41 (39)

4 11.2

10.8 8.3

11.3 5.5

20.2 6.3

11.6 3. I

11.6 3.8

17.1 7.8

8.5 11.6

23 31 8 4 9.4

27.8 8

14.9 8.4

11 5.5

9.2 17

10.8 9.9

11.2 6.3

10.4 11.7

10.7 6.3

10.1 9.2

13.8 7.2

= .03

NS

NS

= .00028

=.0139

= .05

= ,0348

NS

NS

NS

= . m 5

NS

NS

NS

NS

NS

NS

NS

NS

NS

*NS denotes nonstatistical significance (P > .lo).

Prognostic Factors in Myelodysplastic Syndromes 167

Peripheral Blood and Bone Marrow Cytologic Features

The peripheral blood parameters that had a significant adverse effect on survival were hemoglobin level <7 g/ dl. (P = .021), circulating erythroblasts (P = .048), and a thrombocyte count of <50 x 109/L (P = .OOOl). The presence of blasts and immature myeloid precursors in peripheral blood also adversely affected survival, al- though without reaching statistical significance (.05 < P c .lo).

The features of the bone marrow smears that were demonstrated to have an effect on survival were cellular- ity (P = .03), presence of >5% of blasts I (P = .OOO2), > 5 % of blasts II (P = .013), dyserythropoiesis (P = 0.05), and dysgranulopoiesis (P = 0.03).

The FAB categories showed a strong statistical signifi- cance difference on survival. We found an important difference in survival between RA + RAS group (median survival 28 mo) and RAEB+RAEB-t+CMML group (median survival 8 mo) (P = .00004).

Biochemical and Other Laboratory Variables Several biochemical parameters were analysed for their

prognostic significance. Survival for patients with high levels of serum alkaline phosphatase, BUN, serum aspar- tate aminotransferase (SGOT), and phosphorus were considerably shorter than for those who did not suffer from these abnormalities. There was no relationship be- tween the level of creatinine, uric acid, bilirubin, serum alanine aminotransferase (SGPT), and lactic dehydroge- nase (LD).

Multivariate Analysis The best combinations of patient and disease character-

istics selected by means of the Cox regression propor- tional hazards method are given in Table ILI, with variables listed in the order entered by the forward step-

TABLE 111. Order of the Variables Entering the Regression and Level of Significance'

Binary Model P Continuous model P

FAB classification ,001 FAB classification .001 Platelet count .001 Hemoglobin level .011 Hemoglobin level .017 Systemic symptoms .005 Circulating erythroblasts ,049 Platelet count ,018

Age ,016 Dyserythropoiesis ,047

*Regression equations: 1 n[h(t)/ho(t)] = 1. I4(FAB - 1.7) -0.76 (platelets- 1.7) -0.62 (hemoglobin- 1.8) + 0.68 (erythroblasts- 1 . 1 ) . In [h(t)/ho(t)] = 0.27 (FAB-3.1) -0.19 (hemoglobin-8.4) + 0.96 (systemic symptoms- 1.2) -0.003 (platelets-156) + 0.04 (age-69) + 0.3 (dyserythropoiesis- 1.6).

wise modeling procedure. In all instances, the FAB clas- sification was the first parameter to enter the regression models, entering also in different order the hemoglobin level and platelet count. However, the presence of circu- lating erythroblasts also contributed to a better definition of the binary model, whilst in the continuous model three other variables provided significant additional informa- tion: systemic symptoms, age, and dyserythropoiesis.

DISCUSSION

Over the past few years numerous investigators have been engaged in improving the knowledge of prognostic factors in several heterogeneous hematological diseases. However, the prognostic factors in MDS have not been sufficiently and appropriately studied. Most analyses have been carried out under univariate procedures, and no account has been taken of correlation between covariates. The purpose of this paper is to perform a multivariate analysis by means of a proportional hazard model [lo], in order to identify the prognostic variables that show significant independent predictive value on survival.

Once again, the FAB classification was found to be the most useful prognostic variable in all regression models obtained from our MDS series. This finding seems to have almost universal agreement in previous studies [4,6,8,9]. However, in our series only two groups could be clearly separated in terms of survival curves: RA +RAS constituted a good prognostic group, whilst RAEB +WEB-t +CMML provided poor prediction with similar median survival rates.

Apart from the FAB classification, the independent prognostic value of reported variables such as peripheral blood findings, multilineage defects, and other disease characteristics must be questioned because of controver- sial results and inadequate statistical approaches. When correlation between covariates was considered by means of a multivariate analysis, Foucar found that the FAB subtype and weight loss were the most significant of 72 variables tested; no other covariates gave sufficient ad- ditional prognostic information to justify their inclusion in the model [ 121. However, as well as the FAB classifi- cation, we have found a significant independent predic- tive value for the hemoglobin level and platelet count in both continuous and binary models. The influence of one or two cytopenias on survival had been previously re- ported directly [1,5] or indirectly [4,7]. The other vari- ables that contributed to a better definition of each model had not been previously noted as influencing survival in MDS, except for age [6,13,14] and systemic symptoms, which included weight loss in the Foucar report [ 121.

To summarize, our study of prognostic factors in MDS confirmed the FAB classification as the most important

168 Garcia et al

prognostic variable. The predictive value of the variables platelet count, hemoglobin level, and age, included in

prognostic factors in 193 patients. Cancer 52:83-90, 1983. 6. Tricot G , Vlietinck R. Booeaerts MA. Hendrickx B. De Wolf- -

our regression formula, had been previously recognised. However, this is not the case regarding three other sig- nificant variables (systemic symptoms, dyserythro- poiesis, and circulating erythroblasts) whose predictive capacity had not been identified in other studies.

7.

8.

REFERENCES

I .

2.

3.

4.

5 .

Weisdorf DJ, Oken MM, Johnson GJ, Rydell RE: Chronic my- elodysplastic syndrome: Short survival with or without evolution to acute leukaemia. Br J Haematol 55:691-700. 1983. Juneja Sk, Imbert M, Jouault H, Scoazec JY, Sigaux F, Sultan C: Haematological features of primary myelodysplastic syn- dromes (PMDS) at initial presentation. A study of 118 cases. J Clin Pathol 36: 1129- 1135, 1983. Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DAG, Gralnick HR, Sultan C: Proposals for the classification of the myelodysplastic syndromes. Br J Haematol 51: 189-199, 1982. Mutii GJ, Stevens JR, Oscier DG, Hamblin TJ, Machin D: 13. Myelodysplastic syndromes: A scoring *system with prognostic significance. Br J Haematol59:425-433, 1985. Coiffer B, Adeleine P, Viala JJ, Bryon PA, Fiere D, Gentil- homme 0, Vuvan H: Dysmelopoietic syndromes. A search for

9.

10.

11.

12.

14.

Peeters C, Van den Berghe i, Verwilghen RL: Prognostic factors in the myelodysplastic syndromes: Importance of initial data on peripheral blood counts, bone marrow cytology, trephine biopsy and chromosomal analysis. Br J Haematol60:19-32, 1985. Varela BL, Chuang C, Woll JE, Bennett JM: Modification in the classification of the primary myelodysplastic syndromes: The addition of a scoring system. Hematol Oncol3:55-63, 1985. Vallespi MT, Torrabadella M, Julia A, Irriguible D, Jaen A, Acebedo G, Triginer J: Myelodysplastic syndromes: A study of 101 cases according to the FAB classification. Br J Haematol

Pet0 R, Pike MC, Armitage P: Design and analysis of random- ized clinical trials requiring prolonged observation of each pa- tient. Br J Cancer 35: 1-39, 1977. Cox DR: Regression models and life tables. J R Statist SOC B

Dixon WJ: “BMDP Statistical Software.” Berkeley: University of California Press. 1983. Foucar K, Langdon RM, Armitage JO, Olson DB, Carroll TJ: Myelodysplastic syndromes: A clinical and pathologic analysis of 109 cases. Cancer 56553-561, 1985. Izrael V, Jacquillat C , Chastang C, Weil M, de Heaulme M, Boiron M, Bernard J: Donnees nouvelles sur les leucemies oli- goblastiques. Nouv Presse Med 497-952, 1975. Joseph AS, Cinokoati KI, Hunt L, Gary CG: Natural history of smouldering leukaemia. Br J Cancer 46: 160-166, 1977.

61183-92, 1985.

34: 187-220, 1972.