2820480131_ftp jurnal cll

A New Prognostic Classification of Chronic Lymphocytic Leukemia Derived from a Multivariate Survival Analysis

J. L. BINET,* A. AUQUIER,t G. DIGHIERO,* C. CHASTANG,t H. PIGUET,* J. GOASGUEN,§ G. VAUGIER,Il G. POTRON,Y P. COLONA,# F. OBERLING,** M. THOMAS,tt G. TCHERNIA,** C. JACQUILLAT,§§ P. BOIVIN,""

C. LESTY,* M. T. DUAULT,* M. MONCONDUIT,* S. EELABBES,# AND F. GREMYt

Survivals of two series of CLL patients (99 from a retrospective series and 196 from a prospective series) were studied separately. The three main staging systems (Rai, Binet, Rundles) agreed well, but as far as survival is concerned, too many stages are defined. The authors performed a Cox multivariate analysis of survival in order to isolate important prognostic factors at diagnosis and to use them to build a simple three-stage classification. Thrombopenia and anemia appeared as the most important risk factors. Among the nonanemic and nonthrombopenic patients, the number of involved areas was clearly related to prognosis in the authors' two series. This study allowed the authors to propose a new classification in three prognostic groups. Group C: anemia (Hb < 10 g) and/or thrombopenia (platelets < 100,00O/mrn"); about 15% of the patients; median of 2 years. Group B: no anemia, no thrombopenia, three or more involved areas (counting as one each of the following: axiffary, cervical, inguinal, lymph nodes, whether unilateral or bilateral, spleen and liver); about 30% of patients; median of 7 years. Group A: no anemia, no thrombopenia, less than three involved areas; about 55% of patients; the survival of this group does not seem different from that of the French population of the same age and sex distribution. This three-stage classification only requires clinical examination and routine hemogram, has a good prognostic value which was confirmed on the series of Montserrat and Rozman (146 patients), and should therefore be helpful in planning new clinical trials.

Cancer 48:198-206, 1981.

HRONIC LYMPHOCYTIC L E U K E M I A (CLL) is a dis- C ease known to have a variable course; some patients die within one year after diagnosis while others live for longer than ten years. The numerous statistics published on CLL since Minot's and Isaacs'I9 initial publication have revealed a certain number of prognostic factors. The importance of sex, age, the degree of peripheral lymphocytosis, the degree of associated anemia or thrombocytopenia, cutaneous manifesta- tions, signs of inflammation and the onset of a sarcoma have all been stressed but it has not been possible to

From the Departments of Hematology, *Hdpital Pitie-Salpetriere, Paris; SHopital Henri Becquerel, Rouen; SHdpital Hotel Dieu, Rennes; 'Hdpital de la Source, Orleans; 7H6pital Robert Debre, Reims: #Hdpital Mustapha, Alger; **Hdpital Civil Strasbourg; f t Hdpital de Caen; $iHdpital Antoine Beclere, Clamart; 99Hdpital Saint Louis, Paris; ""Hdpital Beaujon, Clichy.

From the tINSERM U 88 "Methodologie Informatique et Statistique en Medecine," Hdpital Pitie-Salpetriere, Paris.

Address for reprints: Professor J . L. Binet, Department of Hematology, Chu Pitie-Salpetriere, 91, Boulevard de I'HGpital, 75634 Paris, Cedex 13, France.

The authors thank J . F. Boisvieux, J. Fermanian, and J . L. Golrnard for their review and comments, and also thank F. Dagonet for her assistance in drawing the figures.

Accepted for publication July 2, 1980.

determine the relative importance of each of these

dictable course of CLL has made indications for therapy particularly difficult. The work of Rai et (i1.25 has made a substantial contribution by grouping various factors. These authors were able to establish several stages with a prognostic significance. Binet et 0 1 . l postulated a clinical staging system similar to that of Rai et ul. which isolated pure splenic forms and associated anemia and/or thrombocytopenia in a single stage. Dighiero et ~ 1 . ~ emphasized the good prognosis of pure splenic forms of CLL, and postulated that the failure to recognize the good prognosis of these patients was probably due to the fact that they were often con- fused with other predominantly splenic hematologic disorders such as hairy cell leukemia and prolymphocytic leukemia." More recently Rundles and Moore"' proposed a staging system close to those of Rai and Binet, and Montserrat and Rozman'" tried to subdivide Stage I1 in the staging of Rai.

All staging systems isolate a high-risk group of patients defined by anemia and/or thrombopenia. Among the remaining patients, who represent between 75% and 81% of the CLL cases depending on the classifi-

f a c ~ o r S ~ ~ ~ . . ~ . 1 0 . 1 ~ . l ~ , l S , 1 7 . 1 H,2 l , lZ ,24 ,2X The traditionally unpre-

0008-543X/81/0701/198/$1.00 0 American Cancer Society

198

No. 1 NEW PROGNOSTIC STAGING I N CLL . Binet et al. 199

cation, there is still a large variability in the course of the disease which is not well explained by the staging systems.

Given the importance of adapting therapy to the severity of the disease, we have found it necessary to tentatively define a new staging system aimed specifically at sorting out patients according to prognosis and comprising fewer stages. In the present study we analyzed two series of patients (a total of 295 subjects) with a survival model after that of COX,^ which allows one to take simultaneously into account the variables that may affect survival and to identify the important ones. We were led to define three groups corresponding to low-, intermediate- and high-risk patients. This conclusion was reached after preliminary comparisons of the main staging systems proposed in the literature.

Material and Methods

Put ie n t s

Two series of patients were analyzed in this study. The first series included patients followed in the

Hematology department of the Pitie Salpetriere Hos- pital in Paris. This retrospective series had already been studied in a previous paper.‘ Among the 129 patients studied before, but with four more years of follow-up, only in 99 was complete data available in all the variables considered in the statistical analysis. In Table 1, the 99 patients are characterized by several variables measured at time of diagnosis: age, sex, lymphocytosis, degree of bone marrow infiltration as measured by bone marrow aspiration, hemoglobin, platelet count, presence or absence of lymph node enlargement (cervical, axillary or inguinal), spleen and liver enlargement and polynuclear count. The distribution of follow-up time, with a maximum of 193 months, is also given in Table 1.

The second series includes patients from a coopera- tive study of the French Society of Hematology. This study started in May 1976 and therefore the longest follow-up time is 41 months. The 196 patients of this prospective study for whom complete data are available are also given in Table 1. We analyzed the two series separately because the follow-up times were very different and because the quality of the data of the prospective study was judged better.

Criteria f o r Diugnosis

In all patients the initial diagnosis was made on the basis of peripheral lymphocytosis (minimum of 4000/ mm3) combined with bone marrow lymphocyte infiltration (minimum of 40%).

TABLE 1. Characterization of the Two Series at Diagnosis

Retrospective Prospective series (99) series (196)

Sex Male

<60 years Mean age (years)

Cervical lymph nodes Normal

Axillary lymph nodes Normal

Inguinal lymph nodes Normal

Spleen Normal

Liyer Normal

Platelets 2 100,000/m3 Mean count (/mm’)

Hemoglobin 210 g Mean (g)

<50,000/mm3 Mean (/mm3)

33000/mmd Mean (/mmd)

160% Mean (%)

None Chlorambucil Other

(months)

Age

Lymphocytes

Neutrophils

Bone marrow infiltration

Prescribed treatment

Potential follow-up time

55%

30% 63.7

47%

58%

65%

64%

91%

94% 193,000

89% 13.0

80% 49,000

72% 13,478

44% 61

56% 32% 12%

s 6 0 (34%)

> 120 (27%) 61-120 (39%)

61%

31% 64.3

46%

46%

55%

60%

84%

90% 208,000

93% 12.8

76% 43,000

74% 4898

35% 66

37% 58%

5%

6 1 2 (33%) 13-24 (37%)

>24 (30%)

Poorly differentiated lymphocytic lymphomas, as well as rare leukemic forms of “well-differentiated lymphocytic lymphomas” with no initial bone marrow or peripheral blood involvement, were excluded from the study. Cases of so-called prolymphocytic leukemia which raise very different problems were also excluded.

Treatmenl

In the retrospective series, the vast majority of patients received either no treatment (56%) or chlorambucil (32%) at diagnosis. A few patients (12%) were initially treated in other ways (cortico-steroids, CVP, MOPP, splenic irradiation, splenectomy, etc.). In the prospective series, Binet Stage 0 and I1 patients were randomized between no treatment or levamizol (90 mg, 3 days every month). Stages I and I11 were randomized between chlorambucil alone (0.1 mg/kg, continuous regimen) and chlorambucil (at the same doses) combined with levamizol (same doses as above),

200 CANCER JUIJ? I 1981 Vol. 48

14.25

8.10

2.87

5.99

1.80

5 5 [D<O.

1.34

5.00

001 >

FIG. 1. Survival of the retrospective series by staging systems (99 patients). N = number of patients in stage; 0 = observed number of deaths; E = expected number of deaths: OiE = relative death rate; xp2 = logrank test statistic which has an approximate chi-square distribution with p degrees of freedom under the null hypothesis.

Stage IV patients were randomized between CVP and just supportive treatment.

Stuging S y s tcni ,$

Patients were allocated to one of five stages on the basis of clinical and hematologic findings, according to Binet's classification. In the prospective series the stages determined the treatments.

Stage 0: Peripheral and bone marrow lymphocytosis without lymph node enlargement or splenomegaly .

Srage I : Stage 0 with lymph node enlargement but without splenomegaly.

Stage 11: Stage 0 without lymph node enlargement but with palpable splenomegaly.

Stcige I l l : Stage 0 with lymph node enlargement and palpable splenomegaly.

Stcige IV: Stage 0 with either fairly severe anemia (hemoglobin < 10 g) or thrombocytopenia (platelets < 100,000/mm3).

In order to compare the different staging systems proposed, the same patients were sorted out along Rai's and Rundles' stagings which are summarized below.

Stages 0 and I in Rai's and Binet's systems are very similar. Rai's Stage I1 comprises patients with spleen or liver enlargement with or without lymph node enlargement. It includes in particular Binet's Stage I1 (pure splenic form). Stage I11 is defined by anemia (Hb < 11 g). Stage IV is defined by thrombopenia (platelets < 100,00O/mm") with or without anemia (Hb < 11 g).

Rundles' system defines Stage I as Stages 0 of Rai and Binet, Stage I1 is the association of Stages I and I1 of Binet, Stage 111 is comparable to Stage I11 of Binet, and Stage IV is close to Binet's Stage IV with the ad- dition of granulocytopenia (neutrophils < 1500/mm").

S t N t is tic ci I A n ci l y s i s

Survival curves were obtained by the product limit method of Kaplan and Meier.I6 Comparison of survival curves was based on the logrank test."3 Survival times were measured from the date of diagnosis and end- points were taken as deaths from all causes. We have attempted to explain the high variability in survival between patients by several variables measured at time of diagnosis. We used a multiple regression model developed by Cox6 for censored survival data which allows one to consider several variables simultaneously and to identify the variables that have an important effect on survival. This model allows one to consider continuous as well as categorical data. We used a binary code for the categorical variables that are all dichotomous and a logarithmic transformation of some of the continuous variables to avoid giving too much weight to extreme values. Model selection was based on an ascending stepwise procedure in which we stopped including variables when the maximum partial log-likelihood did not change significantly after inclusion of an extra variable. We defined a score for each individual as the component of the log-failure rate which does not change with time (i .e. zp in Cox's notation).

Results

In the first step, actuarial survival curves were con- structed according to the three classifications proposed in the literature. The survival curves for the two series are represented in Figures 1 and 2, up to 40 months for the prospective series (where the maximum follow-up time is 41 months) and up to 120 months for the retro-

No. 1 NEW PROGNOSTIC STAGING IN CLL . Binet et a1

spective series (with a maximum follow-up time of 193 months). As can be judged from the logrank test (x2 statistics displayed in Figs. 1 and 2) , the differences in survival between stages are highly significant within each series and each staging system. From the com- ponents of the xz statistics, it can be seen that these highly significant differences are mainly due to the bad prognosis of the anemic and thrombopenic patients. Indeed, in Binet's system, Stage IV, the median survival is 18 months for the prospective series of 196 patients and 24 months for the retrospective series of 99 patients (Figs. 1A and 2A). In Rai's system the median survival is 28 and 68 months respectively for Stage I11 patients (anemia only) and 14 and 15 months respectively for Stage IV patients (thrombopenia associated or not with anemia) (Figs. 1B and 2B). In Rundles' system, Stage IV, the median survival is 28 and 24 months respectively (Figs. 1C and 2C).

Among the remaining patients (nonanemic and nonthrombopenic), those with blood and bone marrow lymphocytosis only and the rare pure splenic forms of the disease seem to have the best prognosis. Long-term analysis (retrospective series) failed to show any dif- ference between patients with lymph node enlargement only and patients with lymph node enlargement associated with splenomegaly . These results indicate that the interest of these systems lies essentially in their ability to isolate a high-risk group.

We performed an analysis of the survival data aimed specifically at isolating important prognostic factors from which we derived a prognostic classification. We used a survival model after that of Cox in which several variables can be considered simultaneously. We selected all the variables listed in Table 1 for potential inclusion in the Cox model, except for bone marrow infiltration (which was missing data for 20% of the patients), neutrophils (which was missing data for 21% of the retrospective patients) and treatment.

The stepwise analysis on the prospective series of 196 patients selected thrombopenia and anemia as the most important prognostic factors, followed by splenomegaly. Scores that were closely related to in- stantaneous risks of death were calculated for each patient, based on the estimated parameters of the Cox model selected by the stepwise procedure and including thrombopenia, anemia and splenomegaly . A high score indicates an elevated risk of dying. Based on a histogram of these estimated scores drawn in Figure 3, we have divided the 196 patients in three groups of increasing risk. The separation between groups can be seen to be rather arbitrary since these scores appear continuous. The comparison of the survival curves of the three groups brings out the bad prognosis of the third group with largest scores (with

I 4

STAGE N* 0

20 1

E O / E

"k2-58.89(p41.001) I 4-

X

FIG. 2. Survival of the prospective series by staging systems (196 patients). N = number of patients in stage; 0 = observed number of deaths; E = expected number of deaths; O/E = relative death rate; xpe = logrank test statistic which has an approximate chi-square distribution with p degrees of freedom under the null hypothesis.

a relative death rate of 4.50 in the third group as compared with 0.43 in the first group and 0.83 in the second group; x2* = 36.94, P < 0.001). This is clearly confirmed in the retrospective series of 99 patients where scores have been calculated from the same estimated parameters and used to build three groups with the same boundaries (relative death rate equal to 5.77 in the third group as compared with 0.46 in the first group and 0.98 in the second group; x2' = 36.94, P < 0.001).

Since in the third group 19 of 26 patients in the prospective series and 9 of 11 patients in the retrospective series had anemia or thrombopenia, since the prognosis of this group is clearly worse than the others in the two series, and since the prognostic importance of anemia and thrombopenia was shown on other series

202

! I

! I

CANCER July I 1981 Vol. 48

-

I

FIG. 3 . Histogram of the scores of the 196 patients (Cox model). The scores, explained in Statistical Analysis, are equal, up to an additive constant, to the log failures rates.

of patients discussed in the literature, we thought the evidence was strong enough to justify putting anemic and thrombopenic patients aside and attempting a better prognostic analysis on the remaining patients. The effect of thrombopenia and anemia is indeed so overwhelming that it may hide more subtle effects for nonanemic and nonthrombopenic patients.

A decision had be taken concerning the threshold of anemia in the definition of the high-risk group, since Rai fixes it at 11 g and Binet at 10 g, even though it may be controversial to define a threshold in a continuous phenomenon. Among the eight patients from the prospective series with Hb levels between 10 and 11 g and platelet counts over 100,000/mm'3, none had died by 40 months, and among the five patients with the same characteristics from the retrospective series, one died at 40 months and another at 96 months. These results are not conclusive, since the number of patients is quite small, but they indicate that the threshold for Hb could be established at 10 rather than I 1 g.

After exclusion of anemic (Hb < 10 g) and thrombopenic patients (platelets < 100,000/mm"), 169 patients remained in the prospective series and 86 in the retrospective one. New Cox analyses on the two series allowed us to identify other prognostic factors. In the

TABLE 2. Staging Proposal

H b > l o g < 3 Enlarged areas* A

platelets a 100,000/mm3 3 3 Enlarged areas B

and/or (Any number of enlarged C

and

H b < l o g

platelets < 1W,000/rnm3 areas) ~ ~~

* Each of cervicul, oxilltrr-y , inguinrrl, area, (whether unilateral or bilateral), spleen and liver- count as one area; therefore the number of enlarged areas can take any value between 0 and 5 .

retrospective series, in which we observed 24 deaths among the 86 patients, inguinal lymph nodes and liver were selected as the most important risk factors. Fur- thermore, in this series, all patients with enlarged inguinal lymph nodes have enlarged cervical and axillary lymph nodes. In the prospective series, with only 15 deaths among the 169 patients because of the short follow-up time, spleen and axillary lymph nodes were selected. This importance of tumoral areas in the two series led us to consider the number of involved areas, as a prognostic factor, each of cervical, axillary, inguincrl areas (whether unilateral or bilateral), spleen and liver counting as one area, so that the number of enlarged areas runs from zero to five included.

We found that the number of involved areas, al- though not as good as the Cox score based on inguinal lymph nodes and liver in the retrospective series, was a better prognostic factor than the Cox score in the prospective series (this could be established by comparisons of maximum partial log-likeli- hoods). Similarly, the number of involved areas was not as good as the optimal Cox score in the prospective series based on spleen and axillary lymph nodes, but was better than the same Cox score calculated for the retrospective series. Therefore, the number of involved areas'appears as a good prognostic summary in the two series by comparison with the optimal Cox scores. A comparison of survival curves according to the number of involved areas confirms its prognostic importance (logtank test for trend equal to 12.17 in the retrospective series, P < 0.001, and to 4.38 in the prospective series, P < 0.05). The simplicity of this prognostic factor requiring only clinical examination led us to consider it as a basis for a classification of the nonanemic and nonthrombopenic patients. We sorted out patients in two groups: Group A including those with less than three enlarged areas and Group B with three or more enlarged areas. As expected, the two groups have well-separated survival curves in the two series.

These two sets of analyses, first on the complete series and second on the nonanemic and nonthrombopenic patients only, lead us to propose a new three- stage prognostic classification for CLL patients (Table 2j. Anemic and/or thrombopenic patients constitute Group C. The remaining patients are divided into two groups according to the number of involved areas defined above ( < 3 , 33). The survival curves of the5e three groups are clearly distinct in both series (see Figs. 4 and 5 for the prospective and retrospective series respectively). The median is not reached at 10 years for Group A, is 7 years for Group B and 2 years

No. 1

FIG. 4. Survival of the retrospective series according to the new staging proposal. N = number of patients in stage; 0 = observed number of deaths; E = expected number of deaths; O/E = relative death rate. xZ2 = logrank test statistic which has an approximate chi-square distribution with 2 degrees of freedom under the null hypothesis.

FIG. 5 . Survival of the prospective series according to the new staging proposal. N = number of patients in stage; 0 = observed number of deaths; E = expected number of deaths; O/E = relative death rate; x12 = logrank test statistic which has an approximate chi-square distribution with 2 degrees of freedom under the null hypothesis.

NEW PROGNOSTIC STAGING I N CLL . Binet et a / . 203

204 CANCER July I 1981 Vol. 48

TABLE 3. Characterization of the Staging Proposal on the Two Series

Retrospective series Prospective series

A (57) B (29) c (13) A (102) B (67) C (27)

Sex

Age

Male

<60 years Mean (years)

Cervical lymph nodes normal

Axillary lymph nodes normal

Inguinal lymph nodes normal

Spleen normal

Liver normal

Enlarged areas < 3

Platelets z 100,000imrn" Mean count (imm")

Hemoglobin 210 g Mean (g)

Lymphocytosis <50,00O/mm:' Mean (imm'l)

Neutrophils 23000 Mean ( m d )

<60% Mean (%)

Bone marrow infiltration

44% 66% 77% 61% 61% 59%

28% 65

77%

91%

100%

81%

96%

100%

34% 62

0%

0%

3%

45%

86%

0%

31% 65

23%

3 8%

46%

31%

77%

46%

27% 65

80%

78%

89%

81%

97%

100%

36% 61

4%

6%

9%

42%

70%

0%:.

37% 65

22%

22%

37%

26%

66%

30%

100% 224.700

100% 206,900

54% 134.000

100% 217.700

100% 190.600

30% 107,000

100% 13.5

100% 13.2

1 6% 8.9

100% 14.0

100% 13.0

48% 9.8

87% 26,400

73% 57,000

61% 130,400

89% 21,300

72% 49,100

37% 1 1 1,200

71% 4505

79% 33.838

58% 4167

78% 43 90

64% 4250

81% 8410

58% 51.8

27% 69.3

1 870 79.7

54% 54.2

17% 75.0

21% 78.2

for Group C on the retrospective series. A similar trend is observed at 40 months on the prospective series (92% survivors for Group A, 62% for Group B and median survival at 18 months for Group C). The three groups, which represent about 55% of the patients for A, 30% for B and 15% for C, are characterized in Table 3, for the two series.

(Figs. 1 and 2). Nevertheless they all present two dif- ficulties. First, from a prognostic standpoint it appears that too many stages are defined-at most three different prognostic patterns come out; moreover, sorting out patients in as many as five stages adds further complications in the conduct and evaluation of clinical trials, Second, the prognostic of about 55% of the patients (Stages I and I1 from Rai, I and I11 from Binet) is still difficult to assess. The goal of this study was to perform a prognostic analysis of survival of CLL patients with statistical methods that allow simultaneous analysis of several variables and to use the results of this analysis to define a prognostic classification with fewer stages in order to facilitate clinical trials. To study the relationship between survival of CLL patients and factors measured at time of diagnosis, we used a survival model due to Cox6 in which the simultaneous effect of several potential prognostic variables can be investigated. Based on this analysis we were able to define three groups of increasing risk according to thrombopenia, anemia, lymph nodes, hepatic and splenic enlargement (Table 2). The prog-

Discussion

Staging has achieved widespread acceptance in many types of cancer, such as Hodgkin's disease where it is used as a basis for allocating treatment. Even though it may be controversial to attempt to divide the disease in stages since we are probably dealing with a continuous phenomenon, a somehow arbitrary separation in stages is nowadays necessary to improve compre- hension of the disease and optimize therapeutic deci- sions and clinical trials.

The comparison of the three main classifications of CLL proposed in the in our two series of patients showed good agreement between them

No. 1 NEW PROGNOSTIC STAGING I N CLL . Binet et al. 205

nostic of these three groups is clearly different in the two series studied (Figs. 4 and 5).

The overwhelming importance of thrombopenia and anemia revealed by the Cox analysis justified setting anemic and thrombopenic patients aside in order to attempt a better prognostic analysis of the remaining patients. Cox analyses on these patients selected inguinal lymph nodes and liver as the most important prognostic factors in one series and spleen and axillary lymph nodes in the other. The number of involved areas was found to be, globally in the two series, as good a prognostic summary as either the Cox score based on inguinal lymph nodes and liver or the Cox score based on spleen and axillary lymph nodes. The simplicity of the number of involved areas led us to prefer it to the Cox scores for classifying nonanemic and nonthrombopenic patients.

Our study suggests that the threshold for hemoglobin should be placed at 10 g, and not 11 g as proposed by Rai et al.25 even though the evidence is not strong enough because of small numbers of patients with hemoglobin between 10 and 11 g. It should be emphasized, however, that while it is necessary to fix a threshold in order to define a classification, setting a threshold is arbitrary in a continuous phenomenon. Moreover, the threshold should probably not be unique since hemoglobin is well known to vary with sex.

As opposed to Rundles and Moore,’6 we could not find any prognostic significance of the neutrophil count. Neutrophil count as determined in most laboratories is not reliable in the forms associated with high WBC. They may be increased in these forms and it is well known that high lymphocytosis is statistically associated with the advanced forms of the disease.

After exclusion of anemic and thrombopenic patients, our classification is based only on the number of involved areas. This seems to be, at least in our two series of patients, as reliable as the more com- plicated scoring system recently proposed by Mont- serrat and Rozman ,2” whose prognostic significance was confirmed in our series. Furthermore, these authors were also able to confirm the prognostic value of our staging system (Montserrat and Rozman, per- sonal communication).

A three-stage classification could have been con- structed either from Binet’s staging system by grouping Stages 0 and 11, Stages I and I11 and Stage IV, or almost equivalently from Rai’s staging system, as sug- gested by Rai et a l . ,25 by grouping Stage 0, Stages I and I1 and Stages I11 and IV (Figs. 1 and 2). The comparison which follows between the above grouping of five Binet stages in three and the new proposal indicates the interest of the latter.

By definition all Stage 0 and I1 patients belong to

Group A, and Stage IV patients to Group C. Among the 43 Stage I and I11 patients from the retrospective series, 14 were allocated to Group A and the remaining 29 patients formed Group B. Only 3 of the 14 A patients died as compared with 14 out of the 29 B patients. This is not significant by the logrank test for comparing survival curves (xz = 3.22, 0.05 < P < O.lO) , but this is probably due to the small numbers involved. It is remarkable that the same tendency is observed in the prospective series, in which more patients are involved, but the follow-up time does not exceed 41 months (among the 105 Stage I and I11 patients, 2 out of the 38 allocated to A died, whereas 10 out of the 67 allocated to B died, x2 = 2.15, 0.10 < P < 0.20).

Moreover it seems that the inclusion of these Stage I and I11 patients in Group A does not lead to a lower survival curve than that of Binet Stage 0. Finally, the survival of Group A patients does not seem to be different from that of a sample of French people of the same age and sex distribution with a five-year survival rate of 0.87. So, this staging proposal may be able to disclose some patients from Binet Stages I and 111, whose prognosis does not seem to be different from Stage 0. On the other hand, it seems that the relatively bad prognosis of Group B patients is due, at least in part, to the inclusion of what is known in the literature as “tumoral forms” of CLL. In our study we tried to assess the importance of the size of the lymph nodes by dividing them arbitrarily in two groups-more than 4 cm and less than 4 cm-and splenic size by dividing it as above or below the umbilical line. Even though these parameters may give an account of tumoral mass, we have not enough patients with lymph nodes greater than 4 cm and spleen below the umbilical line to assess its importance. By accumulating more patients, we hope to be able to address this question in the future.

The proposed staging system is related to factors other than the ones used to build it, as can be seen in Table 3: Lymphocytosis count and degree of bone marrow infiltration increase as we go from Group A to Group C. On the other hand, age, sex, and polynuclear counts do not seem to be related to the staging system. We have not considered in the study treatment or other variables such as cytologic aspect and size from cells,’ kineticsg and immunologic markers (percentage of B and T cell^)^,^,^^ and lymphangiographic pattern,I3 which were missing for most patients. Some of these variables, such as size of cells, have been shown to have some prognostic significance, but this seems to be correlated to the stages of the disease.2 Even though responsiveness to chlorambucil has been shown to be related to prognosis, this factor was not considered in this study, which was limited to variables measured at initial diagnosis.

206 CANCER July 1 1981 Vol. 48

The fact that 44% patients from the retrospective series and 63% patients from the prospective series were treated after diagnosis might constitute a major bias in this study. Indeed, the treatment may modify the natural history of the disease and change the prognostic importance of some variables. This major pitfall is very difficult to overcome, since it is not possible to have a large series of patients representing all stages of CLL and left untreated. Nevertheless, most treated patients in our series received chlorambucil, and, to our knowledge, the effectiveness of this therapy remains unproved.'.15

In conclusion, a new staging system, with fewer stages, than those described before, is proposed. This system appears to be simple, requiring only clinical examination and routine hematologic analysis, and has a good prognostic value. We hope that this classification will be helpful for planning new clinical trials on this long-term disease, where progress in therapy is still needed.

REFERENCES

1 . Binet JL, Le Porrier M, Dighiero G, ei a / . A clinical staging system for chronic lymphocytic leukemia. Prognostic significance. Cuncer 1977; 40:855-865.

2. Binet JL, Vaugier G, Dighiero G, D'Athis Ph, Charron D. lnvestigation of a new parameter in chronic lymphocytic leukemia: The percentage of large peripheral lymphocytes determined by the Hemalog D. Prognostic significance. Am J Med 1977: 63:683-688.

3. Boggs DR, Sofferman SA, Wintrobe MM, Cartwright GE. Factors influencing the duration of survival of patients with chronic lymphocytic leukemia. Ant J Med 1966; 40:243-254.

4. Brouet JC, Flandrin G, Sasportes M, Preud'homme JL, Seligmann M. Chronic lymphocytic leukemia of T-cell origin. Lancet 1975: 2890.

S. Coeur P, Gentilhomme 0, Morel P, Germain D. Valeur pronostique de I'hemogramme et du myelogramme initial dans la LLC (Etude statistique de 310 observations par la methode actu- arielle). Nouv Rev Fr Hemuiol (Blood Cells) 1970; 10:566-575.

6. Cox DR. Regression models and life tables. J R Siui Soc 1972; (B)34: 187-220.

7. Dighiero G, Charron D, Debre P, ri d. Identification of a pure splenic forme of chronic lymphocytic leukemia. Br J Haemui 1979; 41:169- 176.

8. Dighiero G, Follezou JY, Roisin JP, Ternynck T, Binet JL. Comparison of normal and CLL lymphocyte surface Ig determinants using peroxidase-labeled antibodies. 11. Quantification of light chain determinants in atypical lymphoid leukemia. Blood 1976; 48:559-566.

9. Dighiero G, Vaugier G , Charron D, D-Athis Ph, Binet JL.

Variations in lymphocyte counts four hours after administration of hydrocortisone in patients with chronic lymphocytic leukemia. Blood 1977; 49:7 19- 728.

10. Ezdinli EZ, Stutzman L. Chlorambucil therapy for lymphomas and lymphocytic leukemia. J A M A 1965; 191: 100- 106.

11. Galton DAG, Goldman JM, Wiltshaw E, Catovsky D, Henry K, Goldenberg GJ. Prolymphocytic leukemia. Br J Haematol 1974; 27:7-23.

12. Gray JL, Jacob A, Block M. Bone marrow and peripheral blood lymphocytosis in the prognosis of chronic lymphocytic leukemia. Cancer 1974; 33: 1169- 1178.

13. Grellet J , Curet P, Laemmel MG, Cazale P, Dighiero G, Binet JL. Correlation between lymphographic grouping and clinical stages in chronic lymphocytic leukemia. A m J Roentgenol 1979; 133: 797-803.

14. Hansen MM, Chronic lymphocytic leukemia clinical studies based on 189 cases followed for a long time. Scand J Haenzuiol (Suppl) 1973; 18:3-286.

15. Houdard M. Traitement et evolution de 93 malades atteints de leucemie lymphoi'de chronique. Theis, Faculte de Medecine Paris VII, 1973.

16. Kaplan E, Meier P. Non parametric estimation from incom- plete observations. J Am Slat Assoc 1958; 53:457-481.

17. Leave11 BS. Chronic lymphocytic leukemia. A study of the incidence and factors influencing the duration of life. Am J Med Sci 1938; 196:329-340.

18. Lortholary P, Boiron M, Ripault M, Levy JP, Manus A, Bernard J. Leucemie lymphoi'de chronique secondairement as- sociee a une reticulopathie maligne, syndrome de Richter. Nouv Rev Fr Hematol (B lood Cells) 1964; 4:621-645.

19. Minot GB, Isaacs R. Lymphatic leukemia: age incidence, duration and benefit derived from irradiation. Boston Med Surg

20. Montserrat E, Rozman C. Sub-classification of stage I1 chronic lymphocytic leukemia with prognostic implications. Lancet 1979: 2:854.

21. Osgood EE. Treatment of chronic leukemias. J Nucl Med 1964; 5:139-153.

22. Pascucci LM. Chronic leukemia: A statistical study of symptoms, duration of life and prognosis. Rndiology 1942; 39: 75-80.

23. Peto R, Pike MC, Armitage P, ei al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. 11. Analysis and examples. Br J Cuncer 1977; 35:l-39.

24. Phillips EA, Kempin S, Passe S, Mike V, Clarkson B. Prog- nostic factors in chronic lymphocytic leukemia and their implications for therapy. Clin Haerncitol 1977; 6:203-222.

25. Rai KR, Sawitsky A, Cronkite EP, Chanana AD, Levy RN, Pasternack BS. Clinical staging of chronic lymphocytic leukemia. Blood 1975; 46:219-234.

26. Rundles RW, Moore JO. Chronic lymphocytic leukemia. Cancer 1978; 42:941-945.

27. Ternynck T, Dighiero G, Follezou J , Binet JL.. Comparison of normal and CLL lymphocyte surface 1G determinants using peroxidase-labeled antibodies. 1. Detection and quantitation of light chain determinants. Blood 1974; 43:789-795.

28. Zippin C, Cutler SJ, Reeves WJ, Lum D. Survival in chronic lymphocytic leukemia. Blood 1973: 42:367-376.

J 1924; 191:l-9.

2820480131_ftp jurnal cll

Documents