British journal
OJ
Haematology. 199 1, 77, 478-48 5
ADONIS
0007104891000772
Chronic lymphocytic leukaemia: prognostic value of lymphocyte morphological subtypes. A multivariate survival analysis in 146 patients TERESAVALLESPf. EMILIOMONTSERRAT*AND MIGUELA. S A N Z t Department of Haematology, Hospital General ‘Vall d’Hebron’, Barcelona, *Postgraduate School of Haematology ‘Farreras Valenti’, University of Barcelona, Hospital Clinic, Barcelona, and tllepartment of Haematology, Hospital ‘La Fe’, Valencia, Spain
Received 9 October 1990; accepted for publication 27 November 1990
Summary. Among other patient and disease characteristics. different morphological lymphocyte subtypes were analysed in 146 patients with chronic lymphocytic leukaemia (CLL)to establish their clinical significance and prognostic value. The univariate analysis selected, among other well-known variables. the following lymphocyte subtypes as significant in prognosis: prolymphocytes, granulated lymphocytes, cleaved lymphocytes and small-size lymphocytes. The presence of prolymphocytes and cleaved lymphocytes was correlated with a poor prognosis, whereas granular lymphocytes and small-size lymphocytes were related to a good prognosis. A multivariate regression analysis showed that, besides clinical
stages, haemoglobin level. WBC count, age. percentage of bone marrow erythroid cells, and sex. only prolymphocytes had independent prognostic significance. Prolymphocyte percentage correlated positively with characteristics expressing tumour mass such as WBC count, blood absolute lymphocyte count, serum lactate dehydrogenase level. number of enlarged lymph nodes, splenomegaly. and a high number of lymphocytes in bone marrow aspirate. Finally. a prolymphocyte threshold of 5 x 1 09/1was found to be useful not only to separate two different groups of patients in the whole series but also in Rai’s stages I1 and III+IV. and in Binet’s stages A and C.
The prognostic heterogeneity of chronic lymphocytic leukaemia (CLL) has prompted numerous studies to identify factors related to survival or disease progression. Clinical stages (Rai rt a/. 1975: Binet et al, 1981) are the most useful tools in CLL prognosis. Other reports demonstrate the prognostic importance of blood lymphocyte count (Baccarani et al. 1982; Rozman et a/. 1982). bone marrow histological patterns (Rozman et a / . 1984). cytogenetics (Han et a?. 1984: Juliusson, 1986), immunophenotype (Gordon et a/. 1983: Baldini et a/. 1985), lymphocyte doubling time in peripheral blood (Galton. 1966: Montserrat et al, 1986). lactate dehydrogenase levels (Lee et al, 1987). age, sex and response to treatment (Catovsky et a/, 1989). The variable morphology of lymphocytes in CLL has led to proposals to classify CLL in different subtypes: chronic lymphocytic leukaemia/prolymphocyticleukaemia (CLL/PL) (Melo et al. 1986;Bennett et al. 1989),large B-cell CLL (Orfao et al. 1988) and CLL of mixed cell type (Bennett et a/, 1989). On the other hand, such morphological heterogeneity has
been the basis for many studies looking at the clinical significance and prognostic value of lymphocyte size and morphology, with contradictory results. Most of these studies, however, have been focused on the prognostic significance of one specific morphological subtype, and multivariate analysis has been used in few reports. In the present report we analyse, in 146 CLL patients from a single unit, the clinical significance and prognostic value of different morphological lymphocyte subtypes in CLL by means of correlation and multivariate analyses. MATERIAL AND METHODS Patients. 1 58 consecutive patients, diagnosed and followed-up at the ‘Vall d’Hebron’ Haematology Department of Barcelona between 1976 and 1988. were initially included in this study. Twelve patients were excluded because of the presence of a Coombs positive haemolytic anaemia. Therefore, 146 were included in the final analysis. Seventeen patients had received previous treatment but they had active disease at the time of the study. Fifty patients were subsequently treated with regimens including alkylating agents.
Correspondence: Dr Teresa Vallespi. Department of Haematology, Hospital General ‘Vall d’Hebrcin’,08035 Barcelona, Spain.
478
Lymphocyte Morphology in CLL 479 Dingnostic criteria. CLL diagnosis was based on the criteria recently recommended by the International Workshop on CLL ( 1989). In 2 3 cases with less than 1 5 x 1 0 y / l lymphocytes in peripheral blood the diagnosis was confirmed by cellsurface markers. Overall. the immunological phenotype was studied in 1 0 1 patients. Surface-membrane immunoglobulin, receptors for mouse erythrocytes and CD19. CD5. CDlO and FMC7 monoclonal antibodies were the most frequently used membrane markers. Lgmphocgte cell subtypes. The following morphological types were considered: small-size lymphocyte, large lymphocyte, prolymphocyte, cleaved lymphocyte. granular lymphocyte, and immunoblast. The morphological criteria employed to identify these subtypes were those recommended by Melo et a1 ( 1986),Orfao et a1 ( 1 988) and Bennett et nl(1989). Fig 1 shows examples of different lymphoid cell subtypes. Cell morphology was studied by light microscopy in peripheral blood films stained with May-Grunwald-Giemsa. A total of 200 lymphoid cells were counted on each slide by one observer (T.V.). The results were expressed as percentages. The identification of different lymphoid cells has proved to be reproducible when the peripheral blood smears are reviewed independently by two observers (Vallespi et al, 1987). O h r parameters. Patients were classified in the clinical stages defined by Rai et a1 ( 197 5) and Binet et a/ (19 8 1). Lymphocyte doubling time (LDT) is defined as the period needed to double the initial blood lymphocyte count, and was calculated as previously described (Montserrat et al. 1986). Statisticul methods. Twenty-nine patient and disease characteristics at diagnosis were analysed to establish their relationship with survival. Besides clinical stages and the number of different morphological subtypes of lymphocytes, as described above, the following variables were included: age, sex, first complaint (none, anaemia, haemorrhage. infection or systemic symptoms). hepatomegaly, splenomegaly. number of lymph node areas enlarged (considered as in the Binet et nl system, 1981), haemoglobin level, WBC count, blood lymphocyte count, neutrophil count, platelet number, as well as the percentage of lymphoid, myeloid and erythroid cells, and the proportion of megakaryocytes in bone marrow aspirate; serum lactate dehydrogenase level (LDH). serum immunoglobulin (IgA, IgG and IgM) levels, and lymphocyte doubling time. The Kaplan-Meier product limit method was used to estimate survival probability (Kaplan & Meier. 1958). All deaths. whether or not related to CLL. were considered as the endpoint of the follow-up. There were no patients lost to follow-up. Statistical comparisons between curves were based on log rank tests or. if applicable, the test for trend as recommended by Pet0 et a/ ( 1 977). Correlation studies were made by means of the Spearman method. In order to identify the most significant independent prognostic factors the proportional hazards regression method developed by Cox ( 1972) was used including characteristics selected as significant ( < 0 . 1 0 ) in the univariate analysis and those which in prior studies have shown prognostic significance. First complaint (1 14 cases). LDT ( 6 7 cases). LDH level ( 7 9 cases) and immunoglobulin levels ( 6 7 cases) were excluded from multivariate study in order to avoid making the study group
smaller. While the prognostic importance of bone marrow histological pattern has been appreciated ( R o m a n et a/. 1984). this feature was excluded from analysis because it was available only in more recent patients (91 cases). The forward stepwise regression procedure was stopped when the P value for entering a n additional factor was above 0.05. Variables were introduced in multivariate analysis according to two models, model A. where all characteristics were included, and model B. where clinical stages (Rai,Binet) were excluded. All analyses were performed by using DM, 3D. 1L and 2L programs from the Biomedical Data Package (BMDP)statistical library (Dixon. 1983) run on an IBM PC AT microcomputer (IBM. Danbury. Conn.). RESULTS Patient distribution according to the features studied is shown in Table I. The mean age was 6 5 . 8 years (range 4789). 9 4 were males and 52 females. Half of the patients presented no clinical symptoms at diagnosis. The median survival of the whole series was 4 9 months. At the time of the analysis 74 patients were still alive. Among patients with well-documented cause of death (41 cases). infection was the most frequent cause ( 5 1%). The proportion of different blood lymphoid cells in the whole series of patients is shown in Table 11. In most cases (87%) small-size lymphocyte was the predominant morphological subtype. Immunoblasts were found in only a few patients (16%). Univariate analysis of survival prognostic factors The results of the univariate analysis of survival prognostic factors are detailed in Tables I and 111. Four lymphocyte subtypes were correlated with survival: higher percentages of small lymphocytes and lymphocytes with azurophilic cytoplasmic granules were associated with a longer survival, whereas a higher proportion of cleaved lymphocytes or prolymphocytes were related to a shorter survival. Prolymphocyte cut-off points of both 5 x 1Oy/l (Fig 2) and 15 x 1Oy/I showed prognostic significance. Correlation analysis The correlation between different lymphoid types and the most important clinical and haematological characteristics is shown in Table IV. Prolymphocyte percentages correlated positively with clinical stages (Rai. Binet). WBC count. blood lymphocytosis. LDH level, number of lymph node areas enlarged, splenomegaly, and bone marrow lymphocyte infiltration. Small lymphocytes showed a negative correlation with Rai and Binet stages. LDH level, number of lymph node areas enlarged, splenomegaly, and bone marrow lymphocyte infiltration. Granular lymphocytes correlated negatively with clinical stages, WBC count, blood lymphocytosis, splenomegaly and bone marrow infiltration. and correlated positively with platelets number, and with a longer LDT. Multivariate analysis of survival prognostic factors The results of the multivariate analyses ofsurvival prognostic factors are shown in Tables V and VI. The best combination of
480
T. Vallespi, E. Montserrat and M.A. Sanz
Fig I . (a)Small lymphocytes: (b) large lymphocytes: (c)prolymphocyte: (d) cleaved lymphocyte: (e) immunobht: ( f ) granular lymphocyte (MayGriinwald-Giemsa. x 1000).
characteristics selected by means of the Cox proportional hazards regression method in model A (where all the variables were included) is shown in Table V. the variables being listed in the order entered by the forward stepwise modelling procedure. Rai’s clinical stage (the more advanced the stage, the shorter the survival). age (older patients with
worse prognosis), prolymphocyte percentages (the higher the proportion, the worse the prognosis) and WBC counts (related inversely to the prognosis) were the only characteristics selected. In model B (where clinical stages were excluded) the six variables selected were haemoglobin level (higher levels had a better prognosis), WBC counts. age,
Lymphocyte Morphology in CLL 481 Table 1. CLL: survival univariate analysis
Characteristic
Category
Clinical characteristics Age (years)
Sex
Systemic symptoms Lymphadenopathies (areas) Spleen enlargement Liver enlargement Haematological features in blood WBC ( X 10'/1) Lymphocytes ( x 10y/l)
I50 51-60 61-70 > 70 Male Female Yes No 0- 1
2-3 Yes No Yes No 125 26-50 > 50 I20 21-40
( x
10y/l)
Haemoglobin (g/dl)
Clinical stages Kai's stages
75
21 46 61
6
21 38
113 58 38
0-225 >225
52 27
25 20
0 I I1 Ill
C
45 23 51 10 14 45 74 24 73 45 24
9 9 31 9 12 9 40 21 22 26 21
0-6 >6
14 53
10 17
I+II
+
111 IV
A B
* NK: Not reached.
0.0002
41 69
0
Clinical evolution Lymphocyte doubling time (months)
124 99 46 41 53 46 42 154 91 43 39 82 40 68
42 22 1 36 29 46 19
N
Binet's stages
9 27 29 48 24 41 16 23 43 31 34 36 30
P value
71 50 7 58 70 97 29
5
Serum biochemical parameters LDH (U/l)
35 56 55 43 42 61 24 122 14
6
Median survival (months)
97
12.5 >2.5 < 10 ii.i-11 11.1-13 13.1-15 >15 I 50 5 1- 100 101-150 >150
Haematological features in bone marrow Erythroid cells (%) I10 11-25 >25 Myeloid cells ("/o) < 10 10 Megakaryocytes Normal Decreased Lymphocytes (%,I
14 27 57 47 94 52 58 56 74 64 47 90 51 86
No. dead
8 27 37 10 22 40 lh 56 14 5 22 21 10 5 7 17 40
>40
Neutrophils
No. of patients
6
39 57 27 6
8 30
124 60 39 124
0.5266 0 U l 16 0.0142 0.0007 0.0010
0~0002 0.0002
60
40 39 58 7 37 46 53 124 21 20 42 58
04200 5%
72
96
L.
120 144
MONTHS
* Percentages from all lymphoid cells.
Fig 2. Survival according to percentage of prolymphocytes in all patients ( I 5% prolymphocytes= 74 patients: > 5%= 65 patients).
prolymphocytes. percentage of bone marrow erythroid cells (higher percentage with longer survivals) and sex (women had a better prognosis) (Table VI).
Prognostic value of prolymphocytes The mean percentage of prolymphocytes for the whole series was 6.9 (SD 16.8).In the univariate analysisdifferent cut-off points of prolymphocytes demonstrated their prognostic value (Table I). The percentage of prolymphocytes in peripheral blood emerged as an independent prognostic factor of survival in both multivariate analyses. Moreover, a threshold of 5x1OY/1 prolymphocytes was found to be useful to separate two different groups ofpatients not only in the whole series but also in Rai's stages I1 and III+IV. and in Binet's stages A and C (Table VII, and Figs 3 and 4). Patients with high prolymphocyte percentages predominated in advanced
'0
24
48
72
96
Fig 3. Survival according to absolute number of prolymphocytes in Rai stage I1 patients ( I 5 x 1OY/1 prolymphocytes= 31 patients: > 5 x 1Oy/l prolymphocytes = 1 7 patients).
Table 111. CLL: survival univariate analysis. Prognostic value of morphological lymphoid cells subtypes.
Characteristic
Category
~~
No. of patients
Patients dead
Median survival (months)
P value
~~
Small lymphocytes (%)
75
18 52 69
11 28 29
34 46 58
0.0319
Large lymphocytes (%)
5 10 11-20 > 20
75 34 30
42 13 13
44 69 49
0.9854
Cleaved lymphocytes 1%)
s 10 z10
117 22
55
53 34
0.023 0
13
74 65
29 39
68 42
0.0158
105 34
43 25
68 32
5
103 36
56 12
42 91
0.01 5 5
0
117 22
57 11
49 46
0.4329
Prolymphocytes (%)
5 5
>5 (x
1OY/I)
Lymphocytes with azurophilic cytoplasmic granules (%) lmmunoblasts
0
120 144 MONTHS
Lymphocyte Morphology in CLL
483
Table IV. Correlation between morphological lymphocyte subtype percentages (with prognostic value for survival) and other CIL parameters Small sire lymphocyte
Variable Clinical stage WBC ( x I O Y / l ) Lymph ( x 1(JY/1) Hb (g/dl) Platelets ( x loY/l) LDH W/U Lymph nodes (areas) Spleen (cm) BM infiltration ( X ) LDT
Negative*
Positive** Negative** Negative* Negative* Negative.
Prolymphocyte Positive* Positive; Positive* Negative* Positive** Positive** Positive* Positive*
Cleaved lymphocyte
Granular lymphocyte
Positive* -
Negative' Negative* Negative.
Negative** Positive** Positive** -
Negative**
P values less than 0.01:** P values less than 0.05. (-)
Positive* Negative* Negative* Positive*
Unrelated.
Table VI. Multivariate analysis: results from model B Order of entrance
55xlOQlL
Characteristic
Regression coefficient
Chisquare
-0.1 566
17.378 13.986 12.620 6.607 5.337 5.194
P value
1 50. 40.
24
48
Haemoglobin (g/dl) WBC ( x lOY/l) Age (years) Prolymphocytes (%) Erythroid cells in BM Sex
p.0.0045
-1
72
96
Table V. Multivariate analysis: results from model A
1 2
3 4
characteristic Rai's stage Age (years) Prolymphocyte (%) WBC ( x 10'/1)
Regression coefficient 0.6010 0.0650 0.0461 0.0026
5
14 8
12 8
20 3
0.0500
0-1
5
49
Binet's stages A
clinical stage (Kai, Binet), a n d presented more frequently high WBC a n d lymphocyte counts, low platelet counts, high LDH levels. high number of lymph node areas enlarged, splenomegaly. as well as a high degree of lymphocytic bone marrow intiltration (Table IV).
5 5
38
484
T. Vallespi, E. Montserrat and M . A. Sanz
DISCUSSION
In CLL the prognostic value of clinical stages (Rai et al, 1975; Binet et al. 1981). blood lymphocytosis (Baccarani et al, 1982: Rozman et al. 1982),bone marrow histology (Rozman ~ t a l1984)andLDT(Galton, . 1966: Montserratetal, 1986)is well established, and the value of some of these parameters has been corroborated in this study. However, the prognostic significance of different lymphocyte subtypes is still controversial. Although some attempts to correlate lymphocyte morphology and prognosis have been made, there are few studies analysing all the different lymphocyte subtypes by an appropriate statistical methodology (Melo et a!, 1987). This study was undertaken to address such an issue. Since the comparison between our results and those previously reported is difficult, not only because of the different methods used to evaluate lymphocyte size or morphology (Hemalog D. Coulter TF analyser, transmission electron microscope) but also because of the way the results are expressed, only studies performed by means of optical microscopy will be taken into account. In the univariate analysis the increased percentage of small size lymphocytes correlated with good prognosis. as reported by Dubner et a1 (1978). but in contrast to other studies (Knospe et al, 1977: Ghani & Krause, 1986). A percentage of more than 5% of large lymphocyte with cytoplasmic azurophilic granules was correlated with a better prognosis, a fact probably related to the non-leukaemic nature of this population. This finding is similar to that reported by Melo et a1 (1987). In our series large lymphocytes (‘benign atypical lymphocytes’), considered by some authors as related to a longer survival (Peterson et al, 1980), had no prognostic significance. Cleaved lymphocytes have been associated with a poor prognosis by some authors (Ralkiaer et al, 1983). but not by others (Ghani & Krause. 1986: Molica & Alberti, 1988). In our study the percentage ( > 10%)as well as absolute number ( > 10 x 1Oy/l) of cleaved lymphocytes influenced survival negatively. Such discrepancies can be due to the fact that a threshold of 5% cleaved lymphocytes, as used by some authors (Ghani & Krause. 1986: Molica & Alberti. 1988) to separate different risk groups, is too close to the normal value of such cells in peripheral blood. Indeed, by using this cut-off point we were unable to demonstrate its prognostic significance either. There is a more general agreement about the adverse influence of prolymphocytes in CLL prognosis, the only discrepancy being the number of cells most useful to separate different risk groups: 5% (Dubner et al, 1978). 10%(Economopoulos et al, 1982),and 15 x 10y/l (Melo et al, 1987). In our series both prolymphocytes percentages (5% and 10%) and an absolute number of 5 x 10y/l discriminate two different risk groups (P=O*O15.P=0.048 and P
OJ
Haematology. 199 1, 77, 478-48 5
ADONIS
0007104891000772
Chronic lymphocytic leukaemia: prognostic value of lymphocyte morphological subtypes. A multivariate survival analysis in 146 patients TERESAVALLESPf. EMILIOMONTSERRAT*AND MIGUELA. S A N Z t Department of Haematology, Hospital General ‘Vall d’Hebron’, Barcelona, *Postgraduate School of Haematology ‘Farreras Valenti’, University of Barcelona, Hospital Clinic, Barcelona, and tllepartment of Haematology, Hospital ‘La Fe’, Valencia, Spain
Received 9 October 1990; accepted for publication 27 November 1990
Summary. Among other patient and disease characteristics. different morphological lymphocyte subtypes were analysed in 146 patients with chronic lymphocytic leukaemia (CLL)to establish their clinical significance and prognostic value. The univariate analysis selected, among other well-known variables. the following lymphocyte subtypes as significant in prognosis: prolymphocytes, granulated lymphocytes, cleaved lymphocytes and small-size lymphocytes. The presence of prolymphocytes and cleaved lymphocytes was correlated with a poor prognosis, whereas granular lymphocytes and small-size lymphocytes were related to a good prognosis. A multivariate regression analysis showed that, besides clinical
stages, haemoglobin level. WBC count, age. percentage of bone marrow erythroid cells, and sex. only prolymphocytes had independent prognostic significance. Prolymphocyte percentage correlated positively with characteristics expressing tumour mass such as WBC count, blood absolute lymphocyte count, serum lactate dehydrogenase level. number of enlarged lymph nodes, splenomegaly. and a high number of lymphocytes in bone marrow aspirate. Finally. a prolymphocyte threshold of 5 x 1 09/1was found to be useful not only to separate two different groups of patients in the whole series but also in Rai’s stages I1 and III+IV. and in Binet’s stages A and C.
The prognostic heterogeneity of chronic lymphocytic leukaemia (CLL) has prompted numerous studies to identify factors related to survival or disease progression. Clinical stages (Rai rt a/. 1975: Binet et al, 1981) are the most useful tools in CLL prognosis. Other reports demonstrate the prognostic importance of blood lymphocyte count (Baccarani et al. 1982; Rozman et a/. 1982). bone marrow histological patterns (Rozman et a / . 1984). cytogenetics (Han et a?. 1984: Juliusson, 1986), immunophenotype (Gordon et a/. 1983: Baldini et a/. 1985), lymphocyte doubling time in peripheral blood (Galton. 1966: Montserrat et al, 1986). lactate dehydrogenase levels (Lee et al, 1987). age, sex and response to treatment (Catovsky et a/, 1989). The variable morphology of lymphocytes in CLL has led to proposals to classify CLL in different subtypes: chronic lymphocytic leukaemia/prolymphocyticleukaemia (CLL/PL) (Melo et al. 1986;Bennett et al. 1989),large B-cell CLL (Orfao et al. 1988) and CLL of mixed cell type (Bennett et a/, 1989). On the other hand, such morphological heterogeneity has
been the basis for many studies looking at the clinical significance and prognostic value of lymphocyte size and morphology, with contradictory results. Most of these studies, however, have been focused on the prognostic significance of one specific morphological subtype, and multivariate analysis has been used in few reports. In the present report we analyse, in 146 CLL patients from a single unit, the clinical significance and prognostic value of different morphological lymphocyte subtypes in CLL by means of correlation and multivariate analyses. MATERIAL AND METHODS Patients. 1 58 consecutive patients, diagnosed and followed-up at the ‘Vall d’Hebron’ Haematology Department of Barcelona between 1976 and 1988. were initially included in this study. Twelve patients were excluded because of the presence of a Coombs positive haemolytic anaemia. Therefore, 146 were included in the final analysis. Seventeen patients had received previous treatment but they had active disease at the time of the study. Fifty patients were subsequently treated with regimens including alkylating agents.
Correspondence: Dr Teresa Vallespi. Department of Haematology, Hospital General ‘Vall d’Hebrcin’,08035 Barcelona, Spain.
478
Lymphocyte Morphology in CLL 479 Dingnostic criteria. CLL diagnosis was based on the criteria recently recommended by the International Workshop on CLL ( 1989). In 2 3 cases with less than 1 5 x 1 0 y / l lymphocytes in peripheral blood the diagnosis was confirmed by cellsurface markers. Overall. the immunological phenotype was studied in 1 0 1 patients. Surface-membrane immunoglobulin, receptors for mouse erythrocytes and CD19. CD5. CDlO and FMC7 monoclonal antibodies were the most frequently used membrane markers. Lgmphocgte cell subtypes. The following morphological types were considered: small-size lymphocyte, large lymphocyte, prolymphocyte, cleaved lymphocyte. granular lymphocyte, and immunoblast. The morphological criteria employed to identify these subtypes were those recommended by Melo et a1 ( 1986),Orfao et a1 ( 1 988) and Bennett et nl(1989). Fig 1 shows examples of different lymphoid cell subtypes. Cell morphology was studied by light microscopy in peripheral blood films stained with May-Grunwald-Giemsa. A total of 200 lymphoid cells were counted on each slide by one observer (T.V.). The results were expressed as percentages. The identification of different lymphoid cells has proved to be reproducible when the peripheral blood smears are reviewed independently by two observers (Vallespi et al, 1987). O h r parameters. Patients were classified in the clinical stages defined by Rai et a1 ( 197 5) and Binet et a/ (19 8 1). Lymphocyte doubling time (LDT) is defined as the period needed to double the initial blood lymphocyte count, and was calculated as previously described (Montserrat et al. 1986). Statisticul methods. Twenty-nine patient and disease characteristics at diagnosis were analysed to establish their relationship with survival. Besides clinical stages and the number of different morphological subtypes of lymphocytes, as described above, the following variables were included: age, sex, first complaint (none, anaemia, haemorrhage. infection or systemic symptoms). hepatomegaly, splenomegaly. number of lymph node areas enlarged (considered as in the Binet et nl system, 1981), haemoglobin level, WBC count, blood lymphocyte count, neutrophil count, platelet number, as well as the percentage of lymphoid, myeloid and erythroid cells, and the proportion of megakaryocytes in bone marrow aspirate; serum lactate dehydrogenase level (LDH). serum immunoglobulin (IgA, IgG and IgM) levels, and lymphocyte doubling time. The Kaplan-Meier product limit method was used to estimate survival probability (Kaplan & Meier. 1958). All deaths. whether or not related to CLL. were considered as the endpoint of the follow-up. There were no patients lost to follow-up. Statistical comparisons between curves were based on log rank tests or. if applicable, the test for trend as recommended by Pet0 et a/ ( 1 977). Correlation studies were made by means of the Spearman method. In order to identify the most significant independent prognostic factors the proportional hazards regression method developed by Cox ( 1972) was used including characteristics selected as significant ( < 0 . 1 0 ) in the univariate analysis and those which in prior studies have shown prognostic significance. First complaint (1 14 cases). LDT ( 6 7 cases). LDH level ( 7 9 cases) and immunoglobulin levels ( 6 7 cases) were excluded from multivariate study in order to avoid making the study group
smaller. While the prognostic importance of bone marrow histological pattern has been appreciated ( R o m a n et a/. 1984). this feature was excluded from analysis because it was available only in more recent patients (91 cases). The forward stepwise regression procedure was stopped when the P value for entering a n additional factor was above 0.05. Variables were introduced in multivariate analysis according to two models, model A. where all characteristics were included, and model B. where clinical stages (Rai,Binet) were excluded. All analyses were performed by using DM, 3D. 1L and 2L programs from the Biomedical Data Package (BMDP)statistical library (Dixon. 1983) run on an IBM PC AT microcomputer (IBM. Danbury. Conn.). RESULTS Patient distribution according to the features studied is shown in Table I. The mean age was 6 5 . 8 years (range 4789). 9 4 were males and 52 females. Half of the patients presented no clinical symptoms at diagnosis. The median survival of the whole series was 4 9 months. At the time of the analysis 74 patients were still alive. Among patients with well-documented cause of death (41 cases). infection was the most frequent cause ( 5 1%). The proportion of different blood lymphoid cells in the whole series of patients is shown in Table 11. In most cases (87%) small-size lymphocyte was the predominant morphological subtype. Immunoblasts were found in only a few patients (16%). Univariate analysis of survival prognostic factors The results of the univariate analysis of survival prognostic factors are detailed in Tables I and 111. Four lymphocyte subtypes were correlated with survival: higher percentages of small lymphocytes and lymphocytes with azurophilic cytoplasmic granules were associated with a longer survival, whereas a higher proportion of cleaved lymphocytes or prolymphocytes were related to a shorter survival. Prolymphocyte cut-off points of both 5 x 1Oy/l (Fig 2) and 15 x 1Oy/I showed prognostic significance. Correlation analysis The correlation between different lymphoid types and the most important clinical and haematological characteristics is shown in Table IV. Prolymphocyte percentages correlated positively with clinical stages (Rai. Binet). WBC count. blood lymphocytosis. LDH level, number of lymph node areas enlarged, splenomegaly, and bone marrow lymphocyte infiltration. Small lymphocytes showed a negative correlation with Rai and Binet stages. LDH level, number of lymph node areas enlarged, splenomegaly, and bone marrow lymphocyte infiltration. Granular lymphocytes correlated negatively with clinical stages, WBC count, blood lymphocytosis, splenomegaly and bone marrow infiltration. and correlated positively with platelets number, and with a longer LDT. Multivariate analysis of survival prognostic factors The results of the multivariate analyses ofsurvival prognostic factors are shown in Tables V and VI. The best combination of
480
T. Vallespi, E. Montserrat and M.A. Sanz
Fig I . (a)Small lymphocytes: (b) large lymphocytes: (c)prolymphocyte: (d) cleaved lymphocyte: (e) immunobht: ( f ) granular lymphocyte (MayGriinwald-Giemsa. x 1000).
characteristics selected by means of the Cox proportional hazards regression method in model A (where all the variables were included) is shown in Table V. the variables being listed in the order entered by the forward stepwise modelling procedure. Rai’s clinical stage (the more advanced the stage, the shorter the survival). age (older patients with
worse prognosis), prolymphocyte percentages (the higher the proportion, the worse the prognosis) and WBC counts (related inversely to the prognosis) were the only characteristics selected. In model B (where clinical stages were excluded) the six variables selected were haemoglobin level (higher levels had a better prognosis), WBC counts. age,
Lymphocyte Morphology in CLL 481 Table 1. CLL: survival univariate analysis
Characteristic
Category
Clinical characteristics Age (years)
Sex
Systemic symptoms Lymphadenopathies (areas) Spleen enlargement Liver enlargement Haematological features in blood WBC ( X 10'/1) Lymphocytes ( x 10y/l)
I50 51-60 61-70 > 70 Male Female Yes No 0- 1
2-3 Yes No Yes No 125 26-50 > 50 I20 21-40
( x
10y/l)
Haemoglobin (g/dl)
Clinical stages Kai's stages
75
21 46 61
6
21 38
113 58 38
0-225 >225
52 27
25 20
0 I I1 Ill
C
45 23 51 10 14 45 74 24 73 45 24
9 9 31 9 12 9 40 21 22 26 21
0-6 >6
14 53
10 17
I+II
+
111 IV
A B
* NK: Not reached.
0.0002
41 69
0
Clinical evolution Lymphocyte doubling time (months)
124 99 46 41 53 46 42 154 91 43 39 82 40 68
42 22 1 36 29 46 19
N
Binet's stages
9 27 29 48 24 41 16 23 43 31 34 36 30
P value
71 50 7 58 70 97 29
5
Serum biochemical parameters LDH (U/l)
35 56 55 43 42 61 24 122 14
6
Median survival (months)
97
12.5 >2.5 < 10 ii.i-11 11.1-13 13.1-15 >15 I 50 5 1- 100 101-150 >150
Haematological features in bone marrow Erythroid cells (%) I10 11-25 >25 Myeloid cells ("/o) < 10 10 Megakaryocytes Normal Decreased Lymphocytes (%,I
14 27 57 47 94 52 58 56 74 64 47 90 51 86
No. dead
8 27 37 10 22 40 lh 56 14 5 22 21 10 5 7 17 40
>40
Neutrophils
No. of patients
6
39 57 27 6
8 30
124 60 39 124
0.5266 0 U l 16 0.0142 0.0007 0.0010
0~0002 0.0002
60
40 39 58 7 37 46 53 124 21 20 42 58
04200 5%
72
96
L.
120 144
MONTHS
* Percentages from all lymphoid cells.
Fig 2. Survival according to percentage of prolymphocytes in all patients ( I 5% prolymphocytes= 74 patients: > 5%= 65 patients).
prolymphocytes. percentage of bone marrow erythroid cells (higher percentage with longer survivals) and sex (women had a better prognosis) (Table VI).
Prognostic value of prolymphocytes The mean percentage of prolymphocytes for the whole series was 6.9 (SD 16.8).In the univariate analysisdifferent cut-off points of prolymphocytes demonstrated their prognostic value (Table I). The percentage of prolymphocytes in peripheral blood emerged as an independent prognostic factor of survival in both multivariate analyses. Moreover, a threshold of 5x1OY/1 prolymphocytes was found to be useful to separate two different groups ofpatients not only in the whole series but also in Rai's stages I1 and III+IV. and in Binet's stages A and C (Table VII, and Figs 3 and 4). Patients with high prolymphocyte percentages predominated in advanced
'0
24
48
72
96
Fig 3. Survival according to absolute number of prolymphocytes in Rai stage I1 patients ( I 5 x 1OY/1 prolymphocytes= 31 patients: > 5 x 1Oy/l prolymphocytes = 1 7 patients).
Table 111. CLL: survival univariate analysis. Prognostic value of morphological lymphoid cells subtypes.
Characteristic
Category
~~
No. of patients
Patients dead
Median survival (months)
P value
~~
Small lymphocytes (%)
75
18 52 69
11 28 29
34 46 58
0.0319
Large lymphocytes (%)
5 10 11-20 > 20
75 34 30
42 13 13
44 69 49
0.9854
Cleaved lymphocytes 1%)
s 10 z10
117 22
55
53 34
0.023 0
13
74 65
29 39
68 42
0.0158
105 34
43 25
68 32
5
103 36
56 12
42 91
0.01 5 5
0
117 22
57 11
49 46
0.4329
Prolymphocytes (%)
5 5
>5 (x
1OY/I)
Lymphocytes with azurophilic cytoplasmic granules (%) lmmunoblasts
0
120 144 MONTHS
Lymphocyte Morphology in CLL
483
Table IV. Correlation between morphological lymphocyte subtype percentages (with prognostic value for survival) and other CIL parameters Small sire lymphocyte
Variable Clinical stage WBC ( x I O Y / l ) Lymph ( x 1(JY/1) Hb (g/dl) Platelets ( x loY/l) LDH W/U Lymph nodes (areas) Spleen (cm) BM infiltration ( X ) LDT
Negative*
Positive** Negative** Negative* Negative* Negative.
Prolymphocyte Positive* Positive; Positive* Negative* Positive** Positive** Positive* Positive*
Cleaved lymphocyte
Granular lymphocyte
Positive* -
Negative' Negative* Negative.
Negative** Positive** Positive** -
Negative**
P values less than 0.01:** P values less than 0.05. (-)
Positive* Negative* Negative* Positive*
Unrelated.
Table VI. Multivariate analysis: results from model B Order of entrance
55xlOQlL
Characteristic
Regression coefficient
Chisquare
-0.1 566
17.378 13.986 12.620 6.607 5.337 5.194
P value
1 50. 40.
24
48
Haemoglobin (g/dl) WBC ( x lOY/l) Age (years) Prolymphocytes (%) Erythroid cells in BM Sex
p.0.0045
-1
72
96
Table V. Multivariate analysis: results from model A
1 2
3 4
characteristic Rai's stage Age (years) Prolymphocyte (%) WBC ( x 10'/1)
Regression coefficient 0.6010 0.0650 0.0461 0.0026
5
14 8
12 8
20 3
0.0500
0-1
5
49
Binet's stages A
clinical stage (Kai, Binet), a n d presented more frequently high WBC a n d lymphocyte counts, low platelet counts, high LDH levels. high number of lymph node areas enlarged, splenomegaly. as well as a high degree of lymphocytic bone marrow intiltration (Table IV).
5 5
38
484
T. Vallespi, E. Montserrat and M . A. Sanz
DISCUSSION
In CLL the prognostic value of clinical stages (Rai et al, 1975; Binet et al. 1981). blood lymphocytosis (Baccarani et al, 1982: Rozman et al. 1982),bone marrow histology (Rozman ~ t a l1984)andLDT(Galton, . 1966: Montserratetal, 1986)is well established, and the value of some of these parameters has been corroborated in this study. However, the prognostic significance of different lymphocyte subtypes is still controversial. Although some attempts to correlate lymphocyte morphology and prognosis have been made, there are few studies analysing all the different lymphocyte subtypes by an appropriate statistical methodology (Melo et a!, 1987). This study was undertaken to address such an issue. Since the comparison between our results and those previously reported is difficult, not only because of the different methods used to evaluate lymphocyte size or morphology (Hemalog D. Coulter TF analyser, transmission electron microscope) but also because of the way the results are expressed, only studies performed by means of optical microscopy will be taken into account. In the univariate analysis the increased percentage of small size lymphocytes correlated with good prognosis. as reported by Dubner et a1 (1978). but in contrast to other studies (Knospe et al, 1977: Ghani & Krause, 1986). A percentage of more than 5% of large lymphocyte with cytoplasmic azurophilic granules was correlated with a better prognosis, a fact probably related to the non-leukaemic nature of this population. This finding is similar to that reported by Melo et a1 (1987). In our series large lymphocytes (‘benign atypical lymphocytes’), considered by some authors as related to a longer survival (Peterson et al, 1980), had no prognostic significance. Cleaved lymphocytes have been associated with a poor prognosis by some authors (Ralkiaer et al, 1983). but not by others (Ghani & Krause. 1986: Molica & Alberti, 1988). In our study the percentage ( > 10%)as well as absolute number ( > 10 x 1Oy/l) of cleaved lymphocytes influenced survival negatively. Such discrepancies can be due to the fact that a threshold of 5% cleaved lymphocytes, as used by some authors (Ghani & Krause. 1986: Molica & Alberti. 1988) to separate different risk groups, is too close to the normal value of such cells in peripheral blood. Indeed, by using this cut-off point we were unable to demonstrate its prognostic significance either. There is a more general agreement about the adverse influence of prolymphocytes in CLL prognosis, the only discrepancy being the number of cells most useful to separate different risk groups: 5% (Dubner et al, 1978). 10%(Economopoulos et al, 1982),and 15 x 10y/l (Melo et al, 1987). In our series both prolymphocytes percentages (5% and 10%) and an absolute number of 5 x 10y/l discriminate two different risk groups (P=O*O15.P=0.048 and P
