- Email: [email protected]
Morphometry For The Prognosis of Acute Lymphoblastic Leukemia in Childhood
Path. Res. Pract. 182,416-420 (1987)
Morphometry for the Prognosis of Acute Lymphoblastic Leukemia in Childhood * P. Tosi, P. Luzi, C. Miracco and R. Santopietro Institute of Pathological Anatomy, University of Siena, Italy
J. P. A. Baak Pathological Institute, Free University Hospital, Amsterdam, The Netherlands
C. Bernardini and A. Acquaviva Pediatric Clinic, University of Siena, Italy
SUMMARY The morphology of acute lymphoblastic leukemia is of prognostic significance, but due to the qualitative subjective nature of the assessment, disagreement between pathologists may be considerable. In order to investigate the adjuvant potential of morphometry, the present study was undertaken. Morphometry was applied in 21 children between 2 and 10 years of age with acute lymphoblastic leukemia (ALL) with a follow-up of at least 5 years. Of these, 9 patients died, 12 (57.1%) survived for at least five years (maximally 12 years). Morphometry appeared to have an important prognostic value, independent of the FAB classification currently in use. Combination of the nuclear-cell area ratio and the cell area gave a rather good discrimination, although these quantitative data form a morphological continuum. The positive results of this pilot study point to the necessity of evaluating the data on a larger material.
Introduction The prognosis of children with acute lymphoblastic leukemia (ALL) in the past decade has been (dramatically) improved. Today, 40% to 50% of the patients achieve complete remission and long-term disease-free survivaP, 4, 18. The choice of the different therapies available depends on the ALL subtypes. These subclassifications are based on various factors such as a) leukemic burden (ini..
* Based on a presentation held at the Xth European Congress
of Pathology, Athens, September 1985. 0344-0338/87/0182-0416$3.50/0
tial leukocyte count, hemoglobin level, platelet count, mediastinal mass, hepatosplenomegaly, lymphadenopathy and eNS leukemia) b) clinical factors (age at diagnosis, sex) and c) cytomorphologic subtypes (for review, see Poplack, 1982)20. Thus, the microscopical image is an essential factor for therapeutic decisions. In spite of this clinical importance, there is no uniformity among different pathologists when evaluating ALL6. As a result, a number of different suclassifications exists, based on morphological criteria (cell size, nuclear to cytoplasmic area ratio, number and size of nucleoli, characteristics of nuclear chromatin)5, 6, 13, enzyme and immunohistochemical features (cytoplasmic PAS reaction, © 1987 by Gustav Fischer Verlag, Stuttgart
Morphometry in Leukemia . 41 7
acid phosphatase and beta-glucuronidase activity?\ surface differentiation anti~ens for Band T-celllines 8, 9, and karyotypic abnormality . It is clear from multivariate statistical analysis that many of these prognostic characteristics are strongly correlated with each other22. Cytomorphology, however, seems to be of independent significant prognostic value l 5, 17, 19. For this reason every attempt to improve the reproducibility of morphologic criteria for ALL classification is important, and quantitative microscopy is especially promising. Many different quantitative studies have proven to be successful in diagnostic cyto- and histopathology (for an overview see Baak and Oort, 1983)3. In the area of malignant non-Hodgkin lymphomas, quantitation of nuclear and cytoplasmic features allowed the objective distinction between different subtypes 10, 26. DNA-ploidy as assessed by flow cytometry (FCM) II is reported to be a useful investigative method in acute leukemia. However, the equipment required for FCM is relatively expensive, and for large-scale retrospective studies the standard microscopical slides cannot be used. Static cytometry is a theoretical alternative for FCM, but at present tedious. In contrast, interactive morphometry, using a graphic tablet, has the advantage of objectivity, it is relatively fast and can be applied to standard specimens. It will be shown that with these inexpensive techniques it is possible to achieve objective cell typing in childhood ALL, thus providing morphological prognostic indicators of high reproducibility. Patients and Methods Patients
The patients described in this paper are a group of 21 children, 2-10 years of age, with acute lymphoblastic leukemia (ALL) diagnosed at the Pediatric Department of the University of Siena between 1972 and 1979. The minimum patient follow-up time was 5 years with a maximum of 12 years. The patients were subdivided into two groups according to survival: a) group 1: more than 5 years (long-term survivors), b )group 2: less than 5 years (short-term survivors). Lymphoblastic leukemia was diagnosed using air dried, Romanowsky-Giemsa stained bone-marrow aspirates. The bone marrow smears were prepared as follows: asmall drop of bone marrow was placed in the middle of a cover glass. A second cover glass was placed over the first so that they did not cover each other but so that the second cover glass was rotated 45 degrees on the first. The slides were handled by their free corners and pulled apart immediately after the drop of blood had spread between them. The slides were air dried and immediately stained. The diagnosis of ALL was established if blast cells lacked evidence of Auer rods and other criteria of myeloid or monocytic differentiation 2,5 . In the patients investigated, the percentage of blasts ranged from 40 to 100. All patients were untreated at the time of the diagnosis and analysis of the patients was performed for sex, age, mediastinal involvement, lymphadenopathy, splenomegaly, hepatomegaly, proportion of blast cells in the marrow, Hb concentration, platelet count, WBC count, and for sub typing according to the FrenchAmerican-British F ( AB) convention (L1, L2, L3). Because of the retrospective nature of the study, immunological studies were not complete in all patients, and therefore not further considered.
Among the prognostic variables, age, sex, WBC count and FAB subtype were used as identifiers for low, medium and high risk. On the basis of this classification the children were treated according to the protocols designed for each of these groups by the Italian Association for Leukemia (IAL)2. Using these criteria, two of the patients were at high risk (therapeutic protocol IAL 7602), the others were at medium risk (therapeutic protocol IAL 7401) (for details of the extensive cytostatic protocols, see reference 2). Central nervous system prophylaxis with cranial irradiation (2400 RAD) was performed in all the children. It is clear from Table 1 that these differences in classification and treatment did not interfere with the morphometric analysis or prognosis of the two subgroups studied. Quantitative studies
The morphometric analyses were all performed by one of the authors (CM). Romanowsky-Giemsa stained marrow smears, the same used for the original diagnosis of ALL, were used for the morphometrical study. A semiautomated interactive analyzer (IBAS, Kontron) was used, which consists of a magnetic tablet connected with a microcomputer and with a light emitting diode (LED). The smears were studied with a microscope (final enlargement 1250 x), equipped with a camera lucida. By means of the LED projected into the microscopic field under study, the boundaries of 100 blast cells as well as of their nuclei were outlined. Such a number was chosen because even after 60-70 measurements, mean, median and standard deviation values were quite constant. The observer selected the most cellular areas, after a thorough search of the whole slide. Repeated selection revealed that this procedure was reproducible. Once a particular area was selected, a random selection of the cells to be measured was followed, scanning neighbouring microscopical fields in a meandering way. All lymphoblastic cells in one field were measured, not in contact with any of the margins. The only further prerequisite for selection was that the cells and nuclei to be measured were intact and had clear profiles. All non-blast cells, if present, such as myeloid cells, erythrocytes, obvious plasma cells and megakaryocytes, were not measured. The following parameters were assessed: cell area, (nuclear/cell area ratio) x 1000, and nuclear regularity (form Ar). Form Ar is a size independent indicator of the regularity of a nuclear profile, calculated with the following formula: area :rc/4 . major diameter . minor diameter Major and minor diameters are calculated by the moment of inertia of an elliptical type structure. The ellipse has not the equivalent area of the traced structure, but the equivalent moment of inertia. Form Ar is = 1 for circle and ellipses, 0.999 to 0.990 for slightly irregular profiles, and < 0.990 for frankly irregular profiles. The intra- interobserver reproducibility of cell area, nuclear/cell area ratio, form Ar has been tested on all slides. The coefficient of variation of all quantitative features measured was less than 5%.
Statistics
Wi1coxon's test was used to assess significant differences between patients with a survival of more than 5 years and those with a survival of less than 5 years. As a level of significance, P < 0.05 was used. As the single variables (cell area, nuclear/cell area ratio, form Ar) showed some overlap, bivariate graphs of two features were also made.
418 . P. Tosi et al.
Results
Table 2. Median values of the variables
Data for patient identification and classification of the two survival groups investigated, according to age, sex, FAB subtyping, risk and therapy are shown in Table 1. The patients at medium and high risk are equally distributed over the two groups. The same is true for sex, FAB classification, and age. Therefore, the two groups can be compared with each other. Two of the 21 patients (9.5%) are at high risk (in agreement with the literature for this age group). Thus, the material can be regarded as representative. Results of measurements, expressed in terms of median values of cell area, of (nuclear/cell area ratio) X 1000, and of nuclear form Ar are presented in Table 2 and Fig. 1. Median values were chosen because the distribution of values of the parameters in single cases was not normal. The cell area is larger in the short-term survivors at a low level of significance, while the ratio nuclear/cell area is significantly lower in the same group, however, with a Table 1. Data of the 21 children with ALL used in the present study FAB O RiskOO
TherapyOO
11 11 11 11 11 12 11 L2 11 11 11 11
middle middle middle middle middle middle middle high middle middle middle middle
IAL IAL IAL IAL IAL IAL IAL IAL IAl; IAL IAL IAL
7401 7401 7401 7401 7401 7401 7401 7602 7401 7401 7401 7401
Short term survivors * * 1 11 93 M 61 M 11 2 L2 3 106 M 11 4 M 7 L2 5 M 92 11 6 31 M 11 7 54 F 11 8 40 M L2 9 104 F
middle middle middle high middle middle middle middle middle
IAL IAL IAL IAL IAL IAL IAL IAL IAL
7401 7401 7401 7602 7401 7401 7401 7401 7401
Sex Age months Long term survivors * 1 98 F F 2 96 28 3 M 4 50 F 5 50 M 6 113 M 7 39 M 8 79 M 24 9 M 10 45 M 11 27 F 12 25 M
mean ± SD
mean ± SD
56± 31.9
65± 35.3
° French-American-British Cooperative Group; according to I.A.L. (Italian Association against Leukemia); * survival of more than 5 years; * * survival of less than 5 years.
00
Cell area
(Nuclear/cell Form Ar area ratio) x 1000
Long term survivors * 1 2 3 4 5 6 7 8 9 10 11 12
63 39 30 33 59 70 36 36 33 21 37 46
900 920 850 910 869 900 910 900 940 850 920 890
0.973 0.992 0.985 0.988 0.992 0.984 0.992 0.994 0.995 0.998 0.995 0.992
mean ± standard deviation
41.9 ± 14.72
897 ± 27.8
0.990 ± 0.006
Short term survivors * * 1 55 2 36 3 85 4 45 5 102 6 50 7 51 30 8 9 58
851 830 880 870 840 860 880 800 830
0.995 0.993 0.994 0.994 0.993 0.992 0.984 0.990 0.996
mean ± standard deviation
849 ± 26.7
0.992 ± 0.003
56.9 ± 22.95
* survival of more than 5 years * * survival of less than 5 years large variation. The nuclear shape differs slightly between the two groups, but not significantly. Of the short-term survivors, 8 of the 9 cases have nuclei with such a feature, while 5 out of 12 cases of group 1 have nuclei with a more irregular profile (low values of form Ar, below 0.990). Thus the nuclear regularity descriptor (form Ar) is the least discriminant variable between the two groups. Bivariate graphs showed that using cell area and nuclear/cell area ratio a fairly good discrimination results between the two groups (Fig. 2). The long-term survivor misclassified (case 5 of group 1) (see Table 2) has values of cell area and nuclear/cell area ratio very close to the mean values of the other group.
Discussion Of the features investigated, the nuclear/cell area ratio is the best discriminator. However, by also using cell area, a slightly better discrimination can be obtained between survivors and non-survivors.
Morphometry in Leukemia . 419 (nuclear/cell area) ·1000
cell area ( ~m2) 120
•
100
940
0
920
00
.994
•
80
.990
0
0 0
••
••
0 40
d\,
•
00 0
20
000
880
0 60
900
•
.. • • • •
0 880
00
..
640
....•• •
ao
•
60
800
**
o
•
•
.988
0 0
•
.962
,-,
-'-.
0
-.
-----.,-p "'0
.978
•
•
820
0
•
100
•
(~m2)
00
.974
*
00 0
88
00
Cellular area
form Ar
0
NS
0
0 0
8 0
0
00
o long term survivors .
Nuclear areal Cellular area
• short term survivors
.810 NS · not significant
.800
.800
.000
Fig. 1. The variables in the different groups.
Fig. 2. Combination of nuclear/cell area ratio and cell area results in better discrimination of long-term (open circles) and short-term (closed circles) survivors.
A larger cell area and a lower nuclear/cell a,rea ratio are associated with short-term survival. Taking these two features into account, the nuclear shape adds little to the discrimination. The quantitative results of the present study are not new or even surprising. In fact, they agree quite well with the currently accepted classifications. Several different morphological classifications have been ~rofosed in the past in relation with survival. Mathe et al. 4. 1 distinguished five ALL types (microlymphoblastic, prolymphocytic, macrolymphoblastic, prolymphoblastic, and immunoblastic) based on cell diameter. Otherss, 13 suggested nucleolar prominence and cell size as important predictors of the prognosis. The correlation between prognosis and cell size could not be confirmed by others. Perhaps, this is due to a lack of reproducibility of the subjective assessmenf4• 27 • The quantitative results of the present study indicate the continuous nature of the significantly different features. There are no discrete subgroups, but the ALL favourable and unfavourable subtypes form a morphologic continuum. Many publications have shown that disagreement is high between pathologists -in these so-called continuous types of lesions3 , and even intra-observer inconsistencies exisfs.
The problem of reproducibility in ALL has also been considered by a French-American-British (FAB) cooperative working group, who first proposeds and then critically analyzed6 morphological criteria, expressed whenever possible in a semiquantitative way, for identifying three types of ALL (L1, L2, L3). The scoring system they used increased the concordance between the observers up to 84%6. The parameters considered were N/C ratio, number and prominence of nucleoli, nuclear membrane outline, cell size. L1 morphology, characterized by a smaller cell area, a higher N/C ratio, less prominent nucleoli, a regular nuclear profile, occasionally with clefting, were associated with the best prognosis. The L3 subclass, characterized by a larger cell area, a lower N/C ratio, more prominent nucleoli and a regular nuclear profile, has the worst prOf nosis1o. Other subtypes have been added by Miller et al. 6 between L1 and L2 on the basis of the relative predominance of Ll and L2 cells in an intermediate group. The existence of these different subclassifications of ALL iIJdicates that morphology is an essential prognostic factor, but also, that none of the available classifications for the prediction of prognosis is ideal. Again, the features used in the present study are not new; but the overriding advantage of the morphometric approach lies in its highly reproducible nature 12• 24.
. : p<0.05
• • : p
420 . P. Tosi et al.
The potential clinical importance of the morphometric classification model results is obvious. It allows to discriminate different survival subgroups within one of the morphological FAB subtypes. Thus, the present data emphasize the concept that cytomorphology has a significant prognostic value, independent of other indicators of the behaviour of ALL As described above, the patient material can be regarded as representative, and the results are positive. Yet, a word of warning is indicated here. First, variations in tissue handling may have a disturbing influence. Although this is an implicit factor in morphological diagnosis, it is especially important for morphometrical features. Secondly, the number of patients is relatively small. The highly significant differences only partially set at rest, and investigation of a larger number of patients is mandatory. The number of patients did not allow the distinction of a set for learning and a set for testing. Thirdly, the selection of cells was reproducible between the present investigators, but it remains to be proved if this is the case for others, too. Finally, slight variations in measurements may occur, and further automatization is desired, if possible. References 1 Amadori S, Meloni G, Baccarini M, Haanen C, Willemre R, Corbelli G, Drenthe-Schonk A, Cardozo PL, Tura S, Mandelli F (1983) Long-term surviv,al in adolescent and adult acute lymphoblastic leukemia. Cancer 52: 30-34 2 A. I. L. (Italian Association Against Leukemia) Study group (1979) A cooperative study on the therapy of acute lymphoblastic leukemia. Results of Italian Association against leukemia. Haematologica 64: 119-147 3 Baak lPA, Oortl (1983) A manual of morphometry in diagnostic pathology. Springer-Verlag, Berlin-Heidelberg-New York 4 Basso G, Agostini C, Cocito MG, Pezzutto A, Destro R, Capuzzo F, Gazzola MV, Rimondo R, Zanesco L, Semenzato G (1984) Non-T, non-B childhood acute iymphoblastic leukemia. Correlation between cytochemical markers and first complete remission. Cancer 54: 981-985 5 Bennett 1M, Catovsky D, Daniel MT, Flandrin G, Galton DAG, Gralnick HR, Sultan C, French-American-British (FAB) Co-operative group (1976) Proposals for the classification of the acute leukaemias. Br 1 Haematol33: 451-458 6 Bennett 1M, Catovsky D, Daniel MT, Flandrin G, Galton DAG, Gralnick HR, Sultan C, The French-American-British (FAB) Co-operative Group (1981) The morphological classification of acute lymphoblastic leukemia: concordance among observers and clinical correlations. Br 1 Haematol 47: 553-561 7 Bloomfield CD, Rowley lD, Goldman AI, Lawler SD, Seeker Walker LM, Mitelman F (1983) Chromosomal abnormalities and their clinical significance in acute lymphoblastic leukemia. Cancer Res 43: 868-873 8 Bowman WP, Melvin SL, Aur R1A, Mauer AM (1981) A clinical perspective on cell markers in acute lymphocytic leukemia. Cancer Res 41: 4794-4801 9 Chessels 1M, Hardisty RM, Rapson NT (1977) Acute lymphoblastic leukemia in children: classification and prognosis. Lancet ii: 1307-1309
10 Crocher 1 (1984) Morphometric and related quantitative techniques in the study of lymphoid neoplasms. A review. 1 Pathol143: 69-80 11 Diamond LD, Nathawani BN, Rappaport H (1982) Flow cytometry in the diagnosis and classification of malignant lymphoma and leukemia. Cancer 50: 1122-1135 12 Lanham GR, Rivera G, Weiss K, Stass AS (1985) Comparison of morphology in ALL at presentation and relapse. Med Ped Oncol13: 1-3 13 Lee SL, Kopel Sand Glidwell 0 (1976) Cytomorphological determinants of prognosis in acute lymphoblastic leukemia of children. Semin Oncol3: 209-217 14 Mathe G, Pouillart P, Sterscu M, Arniel lL, Schwarzenberg L, Schneider M, Hayat M, De Vassal F, lasmin C, Lafleur M (1971) Subdivision of classical varieties of acute leukemia. Correlation with prognosis and cure expectancy. Europ 1 Clin BioI Res 16: 554-560 15 Mathe G, Belpomme D and Dantchev V (1975) Search for correlations between cytological types and therapeutic sensitivity of acute leukemias. Blood Cells 1: 37-49 16 Miller DR, Leikin S, Albo V (1979) Prognostic significance of lymphoblast morphology (FAB classification) in childhood leukemia (ALL). Proc Am Assoc Clin Oncol 20: 345-352 17 Miller DR, Leikin S, Albo V, Sather H, Hammond GD (1981) Prognostic importance of lymphoblast morphology (FAB classification) in childhood acute lymphoblastic leukemia. Br 1 Haematol 48: 199-206 18 Miller DR, Leikin S, Albo V, Sather H, Karon M, Hammond D (1983) Prognostic factors and therapy in acute lymphoblastic leukemia of childhood: CCG-141. A report from children cancer study group. Cancer 51: 1041-1049 19 Pantazopoulos N, Sinks LF (1974) Morphological criteria for prognostication of acute lymphoblastic leukemia. Br 1 Haematol 27: 25-30 20 Poplack DG (1982) Acute lymphoblastic leukemia and less frequently occuring leukemias in the young. In: Levin AS (Ed) Cancer in Young. Masson Publishing USA 21 Raney RB, Festa RS, Waldmann MTG, Manson D, Hann HWL (1979) The Periodic Acid-Schiff reaction and prognosis in children with acute lymphoblastic leukemia. Am 1 Hematol 6: 27-34 22 Robison L, Sather H, Coccia P, Nesbit M, Hammond G (1980) Assessment of the interrelationship of prognostic factors in childhood acute lymphoblastic leukemia. Am 1 Pediatr Hematol Oncol2: 5-13 23 Sen L, Borella L (1975) Clinical importance oflymphoblasts with T markers in childhood acute leukemia. N Eng! 1 Med 292: 828-832 24 Seshadri R, larvis LR, lamal 0, Skinner 1M (1985) A morphometric classification of acute lymphoblastic leukemia in children. Med Pediatr Oncol13: 214-220 25 Sherman AB, Koss LG, Adams SE (1984) Interobserver and intraobserver differences in the diagnosis of urothelial cells. Comparison with classification by computer. Anal Quant Cytol 6: 112-120 26 Tosi P, Leoncini L, Spina D, Del Vecchio MT (1984) Morphometric nuclear analysis of lymphoid cells in center cell lymphomas and in reactive germinal centers. Am 1 Pathol 117: 12-17 27 Viana MB, Maurer HS, Ferenc C (1980) Subclassification of acute lymphoblastic leukemia in children: analysis of the reproducibility of morphologic criteria and prognostic implications. Br 1 Haematol 44: 383-388
Received October 14, 19&6 . Accepted October 28, 1986
Key words: FAB-classification - ALL-Prognosis - Nuclear-cell area ratio Prof. Piero Tosi, Istituto Anatomia e Istologia Patologica, Universita degli Studi di Siena, Via delle Scotte, 53100 Siena, Italy
Morphometry for the Prognosis of Acute Lymphoblastic Leukemia in Childhood * P. Tosi, P. Luzi, C. Miracco and R. Santopietro Institute of Pathological Anatomy, University of Siena, Italy
J. P. A. Baak Pathological Institute, Free University Hospital, Amsterdam, The Netherlands
C. Bernardini and A. Acquaviva Pediatric Clinic, University of Siena, Italy
SUMMARY The morphology of acute lymphoblastic leukemia is of prognostic significance, but due to the qualitative subjective nature of the assessment, disagreement between pathologists may be considerable. In order to investigate the adjuvant potential of morphometry, the present study was undertaken. Morphometry was applied in 21 children between 2 and 10 years of age with acute lymphoblastic leukemia (ALL) with a follow-up of at least 5 years. Of these, 9 patients died, 12 (57.1%) survived for at least five years (maximally 12 years). Morphometry appeared to have an important prognostic value, independent of the FAB classification currently in use. Combination of the nuclear-cell area ratio and the cell area gave a rather good discrimination, although these quantitative data form a morphological continuum. The positive results of this pilot study point to the necessity of evaluating the data on a larger material.
Introduction The prognosis of children with acute lymphoblastic leukemia (ALL) in the past decade has been (dramatically) improved. Today, 40% to 50% of the patients achieve complete remission and long-term disease-free survivaP, 4, 18. The choice of the different therapies available depends on the ALL subtypes. These subclassifications are based on various factors such as a) leukemic burden (ini..
* Based on a presentation held at the Xth European Congress
of Pathology, Athens, September 1985. 0344-0338/87/0182-0416$3.50/0
tial leukocyte count, hemoglobin level, platelet count, mediastinal mass, hepatosplenomegaly, lymphadenopathy and eNS leukemia) b) clinical factors (age at diagnosis, sex) and c) cytomorphologic subtypes (for review, see Poplack, 1982)20. Thus, the microscopical image is an essential factor for therapeutic decisions. In spite of this clinical importance, there is no uniformity among different pathologists when evaluating ALL6. As a result, a number of different suclassifications exists, based on morphological criteria (cell size, nuclear to cytoplasmic area ratio, number and size of nucleoli, characteristics of nuclear chromatin)5, 6, 13, enzyme and immunohistochemical features (cytoplasmic PAS reaction, © 1987 by Gustav Fischer Verlag, Stuttgart
Morphometry in Leukemia . 41 7
acid phosphatase and beta-glucuronidase activity?\ surface differentiation anti~ens for Band T-celllines 8, 9, and karyotypic abnormality . It is clear from multivariate statistical analysis that many of these prognostic characteristics are strongly correlated with each other22. Cytomorphology, however, seems to be of independent significant prognostic value l 5, 17, 19. For this reason every attempt to improve the reproducibility of morphologic criteria for ALL classification is important, and quantitative microscopy is especially promising. Many different quantitative studies have proven to be successful in diagnostic cyto- and histopathology (for an overview see Baak and Oort, 1983)3. In the area of malignant non-Hodgkin lymphomas, quantitation of nuclear and cytoplasmic features allowed the objective distinction between different subtypes 10, 26. DNA-ploidy as assessed by flow cytometry (FCM) II is reported to be a useful investigative method in acute leukemia. However, the equipment required for FCM is relatively expensive, and for large-scale retrospective studies the standard microscopical slides cannot be used. Static cytometry is a theoretical alternative for FCM, but at present tedious. In contrast, interactive morphometry, using a graphic tablet, has the advantage of objectivity, it is relatively fast and can be applied to standard specimens. It will be shown that with these inexpensive techniques it is possible to achieve objective cell typing in childhood ALL, thus providing morphological prognostic indicators of high reproducibility. Patients and Methods Patients
The patients described in this paper are a group of 21 children, 2-10 years of age, with acute lymphoblastic leukemia (ALL) diagnosed at the Pediatric Department of the University of Siena between 1972 and 1979. The minimum patient follow-up time was 5 years with a maximum of 12 years. The patients were subdivided into two groups according to survival: a) group 1: more than 5 years (long-term survivors), b )group 2: less than 5 years (short-term survivors). Lymphoblastic leukemia was diagnosed using air dried, Romanowsky-Giemsa stained bone-marrow aspirates. The bone marrow smears were prepared as follows: asmall drop of bone marrow was placed in the middle of a cover glass. A second cover glass was placed over the first so that they did not cover each other but so that the second cover glass was rotated 45 degrees on the first. The slides were handled by their free corners and pulled apart immediately after the drop of blood had spread between them. The slides were air dried and immediately stained. The diagnosis of ALL was established if blast cells lacked evidence of Auer rods and other criteria of myeloid or monocytic differentiation 2,5 . In the patients investigated, the percentage of blasts ranged from 40 to 100. All patients were untreated at the time of the diagnosis and analysis of the patients was performed for sex, age, mediastinal involvement, lymphadenopathy, splenomegaly, hepatomegaly, proportion of blast cells in the marrow, Hb concentration, platelet count, WBC count, and for sub typing according to the FrenchAmerican-British F ( AB) convention (L1, L2, L3). Because of the retrospective nature of the study, immunological studies were not complete in all patients, and therefore not further considered.
Among the prognostic variables, age, sex, WBC count and FAB subtype were used as identifiers for low, medium and high risk. On the basis of this classification the children were treated according to the protocols designed for each of these groups by the Italian Association for Leukemia (IAL)2. Using these criteria, two of the patients were at high risk (therapeutic protocol IAL 7602), the others were at medium risk (therapeutic protocol IAL 7401) (for details of the extensive cytostatic protocols, see reference 2). Central nervous system prophylaxis with cranial irradiation (2400 RAD) was performed in all the children. It is clear from Table 1 that these differences in classification and treatment did not interfere with the morphometric analysis or prognosis of the two subgroups studied. Quantitative studies
The morphometric analyses were all performed by one of the authors (CM). Romanowsky-Giemsa stained marrow smears, the same used for the original diagnosis of ALL, were used for the morphometrical study. A semiautomated interactive analyzer (IBAS, Kontron) was used, which consists of a magnetic tablet connected with a microcomputer and with a light emitting diode (LED). The smears were studied with a microscope (final enlargement 1250 x), equipped with a camera lucida. By means of the LED projected into the microscopic field under study, the boundaries of 100 blast cells as well as of their nuclei were outlined. Such a number was chosen because even after 60-70 measurements, mean, median and standard deviation values were quite constant. The observer selected the most cellular areas, after a thorough search of the whole slide. Repeated selection revealed that this procedure was reproducible. Once a particular area was selected, a random selection of the cells to be measured was followed, scanning neighbouring microscopical fields in a meandering way. All lymphoblastic cells in one field were measured, not in contact with any of the margins. The only further prerequisite for selection was that the cells and nuclei to be measured were intact and had clear profiles. All non-blast cells, if present, such as myeloid cells, erythrocytes, obvious plasma cells and megakaryocytes, were not measured. The following parameters were assessed: cell area, (nuclear/cell area ratio) x 1000, and nuclear regularity (form Ar). Form Ar is a size independent indicator of the regularity of a nuclear profile, calculated with the following formula: area :rc/4 . major diameter . minor diameter Major and minor diameters are calculated by the moment of inertia of an elliptical type structure. The ellipse has not the equivalent area of the traced structure, but the equivalent moment of inertia. Form Ar is = 1 for circle and ellipses, 0.999 to 0.990 for slightly irregular profiles, and < 0.990 for frankly irregular profiles. The intra- interobserver reproducibility of cell area, nuclear/cell area ratio, form Ar has been tested on all slides. The coefficient of variation of all quantitative features measured was less than 5%.
Statistics
Wi1coxon's test was used to assess significant differences between patients with a survival of more than 5 years and those with a survival of less than 5 years. As a level of significance, P < 0.05 was used. As the single variables (cell area, nuclear/cell area ratio, form Ar) showed some overlap, bivariate graphs of two features were also made.
418 . P. Tosi et al.
Results
Table 2. Median values of the variables
Data for patient identification and classification of the two survival groups investigated, according to age, sex, FAB subtyping, risk and therapy are shown in Table 1. The patients at medium and high risk are equally distributed over the two groups. The same is true for sex, FAB classification, and age. Therefore, the two groups can be compared with each other. Two of the 21 patients (9.5%) are at high risk (in agreement with the literature for this age group). Thus, the material can be regarded as representative. Results of measurements, expressed in terms of median values of cell area, of (nuclear/cell area ratio) X 1000, and of nuclear form Ar are presented in Table 2 and Fig. 1. Median values were chosen because the distribution of values of the parameters in single cases was not normal. The cell area is larger in the short-term survivors at a low level of significance, while the ratio nuclear/cell area is significantly lower in the same group, however, with a Table 1. Data of the 21 children with ALL used in the present study FAB O RiskOO
TherapyOO
11 11 11 11 11 12 11 L2 11 11 11 11
middle middle middle middle middle middle middle high middle middle middle middle
IAL IAL IAL IAL IAL IAL IAL IAL IAl; IAL IAL IAL
7401 7401 7401 7401 7401 7401 7401 7602 7401 7401 7401 7401
Short term survivors * * 1 11 93 M 61 M 11 2 L2 3 106 M 11 4 M 7 L2 5 M 92 11 6 31 M 11 7 54 F 11 8 40 M L2 9 104 F
middle middle middle high middle middle middle middle middle
IAL IAL IAL IAL IAL IAL IAL IAL IAL
7401 7401 7401 7602 7401 7401 7401 7401 7401
Sex Age months Long term survivors * 1 98 F F 2 96 28 3 M 4 50 F 5 50 M 6 113 M 7 39 M 8 79 M 24 9 M 10 45 M 11 27 F 12 25 M
mean ± SD
mean ± SD
56± 31.9
65± 35.3
° French-American-British Cooperative Group; according to I.A.L. (Italian Association against Leukemia); * survival of more than 5 years; * * survival of less than 5 years.
00
Cell area
(Nuclear/cell Form Ar area ratio) x 1000
Long term survivors * 1 2 3 4 5 6 7 8 9 10 11 12
63 39 30 33 59 70 36 36 33 21 37 46
900 920 850 910 869 900 910 900 940 850 920 890
0.973 0.992 0.985 0.988 0.992 0.984 0.992 0.994 0.995 0.998 0.995 0.992
mean ± standard deviation
41.9 ± 14.72
897 ± 27.8
0.990 ± 0.006
Short term survivors * * 1 55 2 36 3 85 4 45 5 102 6 50 7 51 30 8 9 58
851 830 880 870 840 860 880 800 830
0.995 0.993 0.994 0.994 0.993 0.992 0.984 0.990 0.996
mean ± standard deviation
849 ± 26.7
0.992 ± 0.003
56.9 ± 22.95
* survival of more than 5 years * * survival of less than 5 years large variation. The nuclear shape differs slightly between the two groups, but not significantly. Of the short-term survivors, 8 of the 9 cases have nuclei with such a feature, while 5 out of 12 cases of group 1 have nuclei with a more irregular profile (low values of form Ar, below 0.990). Thus the nuclear regularity descriptor (form Ar) is the least discriminant variable between the two groups. Bivariate graphs showed that using cell area and nuclear/cell area ratio a fairly good discrimination results between the two groups (Fig. 2). The long-term survivor misclassified (case 5 of group 1) (see Table 2) has values of cell area and nuclear/cell area ratio very close to the mean values of the other group.
Discussion Of the features investigated, the nuclear/cell area ratio is the best discriminator. However, by also using cell area, a slightly better discrimination can be obtained between survivors and non-survivors.
Morphometry in Leukemia . 419 (nuclear/cell area) ·1000
cell area ( ~m2) 120
•
100
940
0
920
00
.994
•
80
.990
0
0 0
••
••
0 40
d\,
•
00 0
20
000
880
0 60
900
•
.. • • • •
0 880
00
..
640
....•• •
ao
•
60
800
**
o
•
•
.988
0 0
•
.962
,-,
-'-.
0
-.
-----.,-p "'0
.978
•
•
820
0
•
100
•
(~m2)
00
.974
*
00 0
88
00
Cellular area
form Ar
0
NS
0
0 0
8 0
0
00
o long term survivors .
Nuclear areal Cellular area
• short term survivors
.810 NS · not significant
.800
.800
.000
Fig. 1. The variables in the different groups.
Fig. 2. Combination of nuclear/cell area ratio and cell area results in better discrimination of long-term (open circles) and short-term (closed circles) survivors.
A larger cell area and a lower nuclear/cell a,rea ratio are associated with short-term survival. Taking these two features into account, the nuclear shape adds little to the discrimination. The quantitative results of the present study are not new or even surprising. In fact, they agree quite well with the currently accepted classifications. Several different morphological classifications have been ~rofosed in the past in relation with survival. Mathe et al. 4. 1 distinguished five ALL types (microlymphoblastic, prolymphocytic, macrolymphoblastic, prolymphoblastic, and immunoblastic) based on cell diameter. Otherss, 13 suggested nucleolar prominence and cell size as important predictors of the prognosis. The correlation between prognosis and cell size could not be confirmed by others. Perhaps, this is due to a lack of reproducibility of the subjective assessmenf4• 27 • The quantitative results of the present study indicate the continuous nature of the significantly different features. There are no discrete subgroups, but the ALL favourable and unfavourable subtypes form a morphologic continuum. Many publications have shown that disagreement is high between pathologists -in these so-called continuous types of lesions3 , and even intra-observer inconsistencies exisfs.
The problem of reproducibility in ALL has also been considered by a French-American-British (FAB) cooperative working group, who first proposeds and then critically analyzed6 morphological criteria, expressed whenever possible in a semiquantitative way, for identifying three types of ALL (L1, L2, L3). The scoring system they used increased the concordance between the observers up to 84%6. The parameters considered were N/C ratio, number and prominence of nucleoli, nuclear membrane outline, cell size. L1 morphology, characterized by a smaller cell area, a higher N/C ratio, less prominent nucleoli, a regular nuclear profile, occasionally with clefting, were associated with the best prognosis. The L3 subclass, characterized by a larger cell area, a lower N/C ratio, more prominent nucleoli and a regular nuclear profile, has the worst prOf nosis1o. Other subtypes have been added by Miller et al. 6 between L1 and L2 on the basis of the relative predominance of Ll and L2 cells in an intermediate group. The existence of these different subclassifications of ALL iIJdicates that morphology is an essential prognostic factor, but also, that none of the available classifications for the prediction of prognosis is ideal. Again, the features used in the present study are not new; but the overriding advantage of the morphometric approach lies in its highly reproducible nature 12• 24.
. : p<0.05
• • : p
420 . P. Tosi et al.
The potential clinical importance of the morphometric classification model results is obvious. It allows to discriminate different survival subgroups within one of the morphological FAB subtypes. Thus, the present data emphasize the concept that cytomorphology has a significant prognostic value, independent of other indicators of the behaviour of ALL As described above, the patient material can be regarded as representative, and the results are positive. Yet, a word of warning is indicated here. First, variations in tissue handling may have a disturbing influence. Although this is an implicit factor in morphological diagnosis, it is especially important for morphometrical features. Secondly, the number of patients is relatively small. The highly significant differences only partially set at rest, and investigation of a larger number of patients is mandatory. The number of patients did not allow the distinction of a set for learning and a set for testing. Thirdly, the selection of cells was reproducible between the present investigators, but it remains to be proved if this is the case for others, too. Finally, slight variations in measurements may occur, and further automatization is desired, if possible. References 1 Amadori S, Meloni G, Baccarini M, Haanen C, Willemre R, Corbelli G, Drenthe-Schonk A, Cardozo PL, Tura S, Mandelli F (1983) Long-term surviv,al in adolescent and adult acute lymphoblastic leukemia. Cancer 52: 30-34 2 A. I. L. (Italian Association Against Leukemia) Study group (1979) A cooperative study on the therapy of acute lymphoblastic leukemia. Results of Italian Association against leukemia. Haematologica 64: 119-147 3 Baak lPA, Oortl (1983) A manual of morphometry in diagnostic pathology. Springer-Verlag, Berlin-Heidelberg-New York 4 Basso G, Agostini C, Cocito MG, Pezzutto A, Destro R, Capuzzo F, Gazzola MV, Rimondo R, Zanesco L, Semenzato G (1984) Non-T, non-B childhood acute iymphoblastic leukemia. Correlation between cytochemical markers and first complete remission. Cancer 54: 981-985 5 Bennett 1M, Catovsky D, Daniel MT, Flandrin G, Galton DAG, Gralnick HR, Sultan C, French-American-British (FAB) Co-operative group (1976) Proposals for the classification of the acute leukaemias. Br 1 Haematol33: 451-458 6 Bennett 1M, Catovsky D, Daniel MT, Flandrin G, Galton DAG, Gralnick HR, Sultan C, The French-American-British (FAB) Co-operative Group (1981) The morphological classification of acute lymphoblastic leukemia: concordance among observers and clinical correlations. Br 1 Haematol 47: 553-561 7 Bloomfield CD, Rowley lD, Goldman AI, Lawler SD, Seeker Walker LM, Mitelman F (1983) Chromosomal abnormalities and their clinical significance in acute lymphoblastic leukemia. Cancer Res 43: 868-873 8 Bowman WP, Melvin SL, Aur R1A, Mauer AM (1981) A clinical perspective on cell markers in acute lymphocytic leukemia. Cancer Res 41: 4794-4801 9 Chessels 1M, Hardisty RM, Rapson NT (1977) Acute lymphoblastic leukemia in children: classification and prognosis. Lancet ii: 1307-1309
10 Crocher 1 (1984) Morphometric and related quantitative techniques in the study of lymphoid neoplasms. A review. 1 Pathol143: 69-80 11 Diamond LD, Nathawani BN, Rappaport H (1982) Flow cytometry in the diagnosis and classification of malignant lymphoma and leukemia. Cancer 50: 1122-1135 12 Lanham GR, Rivera G, Weiss K, Stass AS (1985) Comparison of morphology in ALL at presentation and relapse. Med Ped Oncol13: 1-3 13 Lee SL, Kopel Sand Glidwell 0 (1976) Cytomorphological determinants of prognosis in acute lymphoblastic leukemia of children. Semin Oncol3: 209-217 14 Mathe G, Pouillart P, Sterscu M, Arniel lL, Schwarzenberg L, Schneider M, Hayat M, De Vassal F, lasmin C, Lafleur M (1971) Subdivision of classical varieties of acute leukemia. Correlation with prognosis and cure expectancy. Europ 1 Clin BioI Res 16: 554-560 15 Mathe G, Belpomme D and Dantchev V (1975) Search for correlations between cytological types and therapeutic sensitivity of acute leukemias. Blood Cells 1: 37-49 16 Miller DR, Leikin S, Albo V (1979) Prognostic significance of lymphoblast morphology (FAB classification) in childhood leukemia (ALL). Proc Am Assoc Clin Oncol 20: 345-352 17 Miller DR, Leikin S, Albo V, Sather H, Hammond GD (1981) Prognostic importance of lymphoblast morphology (FAB classification) in childhood acute lymphoblastic leukemia. Br 1 Haematol 48: 199-206 18 Miller DR, Leikin S, Albo V, Sather H, Karon M, Hammond D (1983) Prognostic factors and therapy in acute lymphoblastic leukemia of childhood: CCG-141. A report from children cancer study group. Cancer 51: 1041-1049 19 Pantazopoulos N, Sinks LF (1974) Morphological criteria for prognostication of acute lymphoblastic leukemia. Br 1 Haematol 27: 25-30 20 Poplack DG (1982) Acute lymphoblastic leukemia and less frequently occuring leukemias in the young. In: Levin AS (Ed) Cancer in Young. Masson Publishing USA 21 Raney RB, Festa RS, Waldmann MTG, Manson D, Hann HWL (1979) The Periodic Acid-Schiff reaction and prognosis in children with acute lymphoblastic leukemia. Am 1 Hematol 6: 27-34 22 Robison L, Sather H, Coccia P, Nesbit M, Hammond G (1980) Assessment of the interrelationship of prognostic factors in childhood acute lymphoblastic leukemia. Am 1 Pediatr Hematol Oncol2: 5-13 23 Sen L, Borella L (1975) Clinical importance oflymphoblasts with T markers in childhood acute leukemia. N Eng! 1 Med 292: 828-832 24 Seshadri R, larvis LR, lamal 0, Skinner 1M (1985) A morphometric classification of acute lymphoblastic leukemia in children. Med Pediatr Oncol13: 214-220 25 Sherman AB, Koss LG, Adams SE (1984) Interobserver and intraobserver differences in the diagnosis of urothelial cells. Comparison with classification by computer. Anal Quant Cytol 6: 112-120 26 Tosi P, Leoncini L, Spina D, Del Vecchio MT (1984) Morphometric nuclear analysis of lymphoid cells in center cell lymphomas and in reactive germinal centers. Am 1 Pathol 117: 12-17 27 Viana MB, Maurer HS, Ferenc C (1980) Subclassification of acute lymphoblastic leukemia in children: analysis of the reproducibility of morphologic criteria and prognostic implications. Br 1 Haematol 44: 383-388
Received October 14, 19&6 . Accepted October 28, 1986
Key words: FAB-classification - ALL-Prognosis - Nuclear-cell area ratio Prof. Piero Tosi, Istituto Anatomia e Istologia Patologica, Universita degli Studi di Siena, Via delle Scotte, 53100 Siena, Italy