Laboratory Correlates and Prognostic Significance of Granular Acute Lymphoblastic Leukemia in Children A Pediatric Oncology Group Study

Transcription

1 HEMATOPATHQLOGY AND LABORATORY HEMATOLOGY Original Article Laboratory Correlates and Prognostic Significance of Granular Acute Lymphoblastic Leukemia in Children A Pediatric Oncology Group Study LIZARDO CEREZO, M.D., JONATHAN J. SHUSTER, PH.D., D. JEANNETTE PULLEN, M.D., BRENDA BROCK, M.T., MICHAEL J. BOROWITZ, M.D., PH.D., JOHN M. FALLETTA, M.D., WILLIAM M. CRIST, M.D., AND DAVID R. HEAD, M.D. As part of a comprehensive prospective clinicopathologic study by the Pediatric Oncology Group (POG), 2,092 children with acute lymphoblastic leukemia (ALL) were evaluated by uniform morphologic, cytochemical, and immunologic methods to assess the frequency and implications of granular lymphoblasts. All cases were Sudan black or myeloperoxidase negative and met French-American-British (FAB) morphologic criteria for ALL. Granular ALL, characterized by the presence of more than 5% marrow blasts with at least three clearly defined azurophilic cytoplasmic granules, was identified in 56 of the 1,252 fully studied cases (4.5%). The frequency of granular features did not differ among early pre-b (4.3%), pre-b (3.6%), and T (5.8%) ALL; no cases were identified among the 12 patients with B ALL. Within the early pre-b/pre-b group, granular ALL was equally distributed between good- and poor-risk clinical groups but was more frequent among FAB L2 than FAB LI cases (12% vs. 2%; P < 0.001). Patients were treated with standard POG protocols for early pre-b/pre-b and T ALL. Complete remission (CR) rates were significantly lower for those with granular lymphoblasts, regardless of risk group, immunophenotype, or FAB type. Analysis of event-free survival (EFS) showed a significantly poorer outcome for granular early pre-b/pre-b cases with FAB L2 morphologic characteristics {P < 0.001) and for those classified as poor risk (P = 0.015). These findings suggest a relationship between granules and L2 morphologic characteristics in childhood ALL and indicate that the presence of granular lymphoblasts conveys a worse prognosis for certain subgroups of children with ALL. (Key words: ; Acute lymphoblastic leukemia; French-American-British morphologic criteria; Immunophenotype; Granular acute lymphoblastic leukemia) Am J Clin Pathol 1991;95: Acute lymphoblastic leukemia (ALL) with prominent scribed in several articles, all based on individual cases or azurophilic granulation of lymphoblasts has been de- small series.'" 5 These reports, representing a total of 14 cases in both adults and children, have emphasized the possibility that the presence of cytoplasmic granules may From the Orlando Regional Medical Center, Orlando, Florida; Pediatric Oncology Group Statistical Office and University of Florida, (AML). Little information is available concerning the im- lead to an incorrect diagnosis of acute myeloid leukemia Gainesville, Florida; University of Mississippi Medical Center, Jackson, munophenotypic and morphologic subtypes of ALL with Mississippi; Duke University Medical Center, Durham, North Carolina: and St. Jude Children's Research Hospital and the Departments of Pediatrics and Pathology, University of Tennessee, Memphis, Tennessee. cytologic granular lymphoblasts or the prognostic import of this finding. Received May 8, 1990; received revised manuscript and accepted for publication July 21, Supported in part by the following PHS Cooperative Agreement Grants awarded by the National Cancer Institute, DHHS: CA-29139, CA-15525, CA-03161, CA-29691, CA-20549, CA-28476, CA-33587, CA-28530, CA , CA-38859, CA-29293, CA-11233, CA-28383, CA-33625, CA , CA-41721, CA-28572, CA-25408, CA-4I188, CA-28439, CA , CA-28841, CA-41082, CA-15989, CA-03713, CA-05587, and by the American Lebanese Syrian Associated Charities (ALSAC). Address reprint requests to POG Operations Office: 4949 W. Pine Boulevard, Suite A, Second Floor, St. Louis, Missouri We present here the experience of the Pediatric Oncology Group (POG) with carefully characterized cases of granular ALL in children. Specifically, we examined the relationship of granular lymphoblasts to French-American-British (FAB) morphologic type, immunophenotype, risk-factor grouping, and treatment outcome. Assessment of both CR rates and EFS indicates that the presence of azurophilic granules has a negative impact on prognosis in childhood ALL. 526

2 CEREZO ET AL. Granular Lymphoblasts in ALL 527 Morphologic/Cytochemical Studies METHODS Patients Between May 1981 and January 1986, 2,092 children with newly diagnosed ALL were registered on POG protocol #8035 (Laboratory Subclassification in Acute Lymphocytic Leukemia of Childhood). Of these children, 1,252 had complete studies of both morphologic characteristics and immunologic phenotype (early pre-b, preb, B, or T); 1,020 early pre-b/pre-b cases and 223 T cases were eligible for statistical analysis ofresponse to induction therapy and 1,021 early pre-b/pre-b and 226 T cases for statistical analysis of EFS. The cutoff for outcome analysis was November 14, The diagnosis of ALL was made by examination of peripheral blood and bone marrow at the individual POG member hospitals, with the use of published FAB criteria and myeloperoxidase or Sudan black and nonspecific esterase stains.6,7 Wright-Giemsa-stained smears were reexamined by two hematopathologists (L.C. and D.R.H.) for confirmation of diagnosis, FAB subclassification, and evaluation for the presence of granular lymphoblasts. Granular lymphoblasts were defined as blasts containing three or more clearly defined cytoplasmic azurophilic granules, each 0.5 nm or greater in diameter (Fig. 1). Cases containing 5% or more such blasts in marrow smears were considered positive for granular lymphoblasts. FIG. 1. Bone marrow. Two cases of acute lymphoblastic leukemia with numerous granular lymphoblasts. Wright-Giemsa stain. Case 1: A (X 1,200). B(X 1,200). Case 2: C (X 1,200). D (XI,500). Vol. 95 No. 4

3 528 HEMATOPATHOLOGY AND LABORATORY HEMATOLOGY Article Immunophenotyping Viable cells for each case were forwarded to central reference laboratories at Duke University (M.J.B.) and the University of Alabama at Birmingham (W.M.C.) for immunophenotyping by previously published methods. 8 " 10 Briefly, pre-b ALL was defined by the presence of :10% of blasts with cytoplasmic fluorescence for n heavy chain without light chain or surface immunoglobulin (slg); T ALL by blasts reacting s40% above controls with pan-t-cell antisera; B ALL by the expression of slg on ^ 10% of the blasts; and early pre-b ALL as all remaining cases. 8 " 10 At the time of this study, myeloid antibodies were not used routinely by POG for immunophenotyping. Risk Group All cases of early pre-b and pre-b ALL were assigned to clinical risk groups with the use of published POG criteria." Briefly, this determination is based on age, white blood cell (WBC) count, and the presence or absence of bulky disease, defined as lymphadenopathy, splenomegaly, hepatomegaly, or a mediastinal mass. (Cases of T or B ALL were considered to be poor risk and were not included in these determinations.) The good-risk group included patients aged years with WBC counts less than 100,000 and no bulky disease and those aged and years old with WBC counts less than 10,000 and no bulky disease. All other patients were considered poor risk. Therapy Treatment was administered according to separate protocols for patients with early pre-b and pre-b ALL (POG study 8036, ALinC-13, a comparison of complex multidrug continuation therapies after a standard vincristine-prednisone induction regimen), 12 T ALL (POG study 7837, a modified LSA 2 L 2 treatment regimen), 13 or B ALL. There were insufficient cases of B ALL to permit comparison of outcome and clinical characteristics. Statistical Considerations The major end point of this study was EFS, defined as the time from achieving a complete response to the earliest event (relapse, death, or last contact). Cases in which a complete response was not achieved were considered failures at time zero. A secondary end point was the CR induction rate. Estimates of two-year and four-year EFS were calculated by the method of Kaplan-Meier 14 with standard errors of Peto and associates. 15 Comparisons >f EFS survival were made with the log-rank test. 16 Comparisons involving binomial events, such as response, were made with the use of the traditional chi-square or, where expected numbers were insufficient, by the exact unconditional chisquare statistic. 17 Because of the relative rarity of cases positive for granular lymphoblasts, multivariate methods were not considered appropriate. Prevalence RESULTS Fifty-six of the 1,252 fully studied cases of ALL (4.5%) were positive for the presence of granular lymphoblasts. Table 1 characterizes these patients by risk group, immunophenotype, and FAB subtype. Cases were similarly distributed among immunophenotypic groups (4.3% of early pre-b, 3.6% of pre-b, and 5.8% of T cases). There were no granular cases among the 12 patients with B ALL, who were excluded from additional analysis. Granular lymphoblasts were identified significantly more often among cases with FAB L2 than LI morphologic characteristics (12% vs. 2%;P< 0.001; see Table 1). This finding held for early pre-b, pre-b, and T cases and for each combination of risk group and immunophenotype with the exception of poor-risk pre-b ALL. Table 2 provides data on age, WBC count, sex, and common acute lymphoblastic leukemia antigen (CALLA) blast reactivity for patient groups defined by immunophenotype and the presence or absence of granular lymphoblasts. Morphology Other than showing the presence of granules in blasts, all granular cases studied were morphologically typical FAB LI or L2 ALL. Chromatin was finely granular to slightly clumped; there was no evidence of myeloid differentiation; Auer rods were absent; nuclear shape varied; and the spectrum of nucleoli was typical of that generally observed in series of ALL. The granules were azurophilic, rather than the deep purple granules of promyelocytes or TABLE 1. PREVALENCE OF GRANULAR LYMPHOBLASTS BY IMMUNOPHENOTYPE, FAB, AND RISK GROUP FAB Groups Immunophenotype/ P Value Risk Group LI L2 Li (LI vs. L2) Early pre-b Good risk 7/342(2.0%) 7/93(7.5%) 0/ Poor risk 9/283(3.2%) 12/86(14.0%) 0/ Pre-B Good risk 1/76 (1.3%) 3/21(14.3%) 0/ Poor risk 3/99 (3.0%) 1/21(4.8%) 0.87 T 4/164(2.4%) 9/58(15.5%) 0/4 <0.001 A.J.C.P. il 1991

4 CEREZO ET AL. 529 Granular Lymphoblasts in ALL TABLE 2. PATIENT CHARACTERISTICS BY IMMUNOPHENOTYPE AND PRESENCE OR ABSENCE OF GRANULAR LYMPHOBLASTS TABLE 4. EVENT-FREE SURVIVAL FOR CASES WITH AND WITHOUT GRANULAR LYMPHOBLASTS BY IMMUNOPHENOTYPE AND RISK GROUP Early Pre-B/Pre-B T Without With Median age (year) Median WBC count (X10 9 /L) Female CALLA negative (J5) Without (n = 983) % 10% With (n = 43) % 14% Without (n = 213) % 65% With (n = 13) % 82% the differentiated granules of maturing myeloid cells. varied in shape, with round, oval, crescentic, or doughnut forms. The larger granules were frequently surrounded by a halo (Fig. 1). Complete Remission Children with granular ALL had significantly lower CR rates on the POG protocols for both early pre-b/pre-b ALL (P < 0.001) and T ALL (P = 0.013). Among patients with early pre-b or pre-b ALL, the differences were significant for both FAB L1 and L2 subtypes and both goodand poor-risk groups (Table 3). Although patients with the L2 subtype overall had a lower CR rate than patients with the LI subtype (L2 = 91%, LI = 98%) (P < 0.001), the difference in CR rate based on granular blasts was actually greater (81 % with vs. 97% without) (P = < 0.001). Event-free Survival Among early pre-b/pre-b cases, therapy failed significantly more often in those with granular lymphoblasts (P = 0.02). Analysis by risk group suggests that this difference was attributable to the higher failure rate in the poor-risk category (Table 4). Within the early pre-b/pre-b group, the failure rate among patients with LI morphologic characteristics did not differ based on the presence of granular lymphoblasts (P = 0.75); however, L2 cases with granules had a significant increase in failure rate (P n Fail Expected* n Fail Expected* P Value good risk poor risk Early pre-b/pre-b total T * Expected value is calculated by the log-rank test under the null hypothesis of target population equivalent survival curves. The P value for early pre-b/pre-b total is adjusted for risk group. Five patients are excluded because of unknown event-free survival. Fail indicates relapse, death, or failure to achieve complete response. < 0.001) (Table 5). There were no significant differences in EFS for LI versus L2 overall or for LI without granules versus L2 without granules (P = 0.13 and 0.45, respectively). Only among granular cases were L2 morphologic characteristics associated with a significantly higher failure rate than those with LI features (P < 0.025). There were more failures than expected among T ALL cases with granular lymphoblasts, but the difference was not statistically significant (see Table 4). Kaplan-Meier estimates of two- and four-year EFS are given in Table 6 by immunophenotype, risk group, and FAB subtype. DISCUSSION Granular lymphoblasts are a relatively uncommon morphologic finding in childhood ALL. In our series, the frequency of this finding was 4.5%, somewhat lower than the 7.5% reported in a smaller series of children treated at a single institution. 5 The use of both Giemsa and Wright-Giemsa stains in that series may explain this discrepancy, because Giemsa staining enhances granules in immature cells; we used only the Wright-Giemsa stain. We found no significant difference in the frequency of granular ALL by immunophenotype, whereas the previous series detected this feature only in common ALL. The reasons for this disparity are unclear but may relate TABLE 3. COMPLETE REMISSION RATE VERSUS PRESENCE OF GRANULAR LYMPHOBLASTS IN ALL BY IMMUNOPHENOTYPE, FAB, AND RISK GROUP* Early Pre-B/Pre-B FAB LI FABL2 Good Risk Poor Risk Total T Without granules With granules With versus without 759/774 (98%) 18/20 (90%) P = /198(93%) 17/23 (74%) P = /515(99%) 16/18 (89%) P = /462 (95%) 19/25 (76%) P = /977 (97%) 35/43 (81%) P< /211 (93%) 8/12 (67%) /> = * Six early pre-b/pre-b patients with unknown response are excluded. In addition, four L3 patients are excluded from the FAB analysis. Three T patients were excluded for unknown response. Vol. 95 No. 4

5 530 HEMATOPATHOLOGY AND LABORATORY HEMATOLOGY Original Article to the larger size of our series (1,252 vs. 66 fully studied patients). We found that granular lymphoblasts were associated with lower CR rates in both early pre-b/pre-b and T ALL and in all combinations of risk groups and FAB subtypes analyzed; the CR rate was particularly poor in granular L2 ALL. Further, in early pre-b/pre-b ALL, granular lymphoblasts had a significant negative impact on EFS in the poor-risk group and FAB L2 cases. Granular lymphoblasts were more common in FAB L2 cases, a relationship that held true when cases were further subdivided by risk group and immunophenotype. This could simply reflect the frequently increased volume of cytoplasm in L2 blasts. However, the greater effect of granules than L2 morphologic characteristics on CR rate, the greater effect of granules than L2 morphologic characteristics on EFS, the absence of a significant effect of L2 morphologic characteristics on EFS except in granular cases, and the particularly bad prognosis of granular L2 cases suggest a more complex relationship. These findings further suggest that a major reason for the poor prognosis of L2 ALL may actually be the high incidence of granular cases in L2 ALL. The biologic implications of granules in lymphoblasts are unclear. None of our cases showed evidence of myeloid differentiation. All were Sudan black or myeloperoxidase negative, were negative for myeloid (cytoplasmic) nonspecific esterase (NSE) (some cases of ALL are NSE positive with either a Golgi pattern or [in granular ALL] a positive stain in granules), demonstrated the usual spectrum of ALL morphologic characteristics, lacked Auer rods, and had a frequency of CALLA positivity similar to that of nongranular cases. On Romanovsky staining the granules were azurophilic rather than the deep purple of primary granules or the color of any of the three secondary granules in granulocytes. Immunophenotyping clearly indicated a lymphoid origin of these blasts, although the possibility that some represent mixed lineage acute leukemia cannot be excluded. Prior studies 2,5 have demonstrated the presence of acid phosphatase, alpha- TABLE 5. EVENT-FREE SURVIVAL FOR CASES WITH AND WITHOUT GRANULAR LYMPHOBLASTS BY FAB MORPHOLOGIC CHARACTERISTICS (early pre-b/pre-b) LI L2 Total* n Without Fail Expected n With Fail Expected P Value 0.75 < * Adjusted for LI versus L2. Nine patients are excluded for either unknown event-free survival or L3 FAB morphologic characteristics. Fail indicates relapse, death, or failure to achieve complete response. TABLE 6. KAPLAN-MEIER ESTIMATES OF EVENT-FREE SURVIVAL (%) ± STANDARD ERROR (%) total LI L2 good risk poor risk T Two Years Four Years Without With Without With 73 ±2 74 ±2 69 ±3 84 ±2 60 ±2 54 ±4 58 ± 8 75 ± ± ± ± ± ±2 58 ±5 70 ±3 46 ± ± ± ± ± ± 18 naphthyl acetate esterase, and alpha-naphthyl butyrate esterase activity in granules in some cases of ALL, suggesting that they may be analogous to lysosomal granules found in stimulated normal lymphocytes. One can only speculate on the observed relationship between the presence of granules and poorer treatment response, a relationship that crosses immunophenotypic, risk, and morphologic groups. Perhaps the granules do represent lysosomes and indicate an activated phenotype among lymphoblasts, regardless of morphologic (FAB) category or immunophenotype. The practical implications of granular lymphoblasts in childhood ALL are threefold: (1) their presence appears to convey a worse prognosis; (2) their high incidence in L2 cases suggests a causal relationship between L2 morphologic characteristics and granules and may explain, in part, the worse prognosis associated with L2 cases in some studies; and (3) the presence of granules may cause diagnostic confusion with AML. The latter situation is easily resolved by careful morphologic examination of Romanovsky-stained material, routine use of cytochemical techniques in evaluating acute leukemia, and use of immunophenotyping studies to corroborate morphologic findings. Acknowledgments. The authors thank Kim Nalewaja and Peggy Vandiveer for typing this manuscript and Christy Wright for editorial assistance. REFERENCES 1. Brunning RD, Parkin J, Dick F, et al. Unusual inclusions occurring in the blasts of four patients with acute leukemia and Down's syndrome. Blood 1974;44: Grogan TM, Insalaco SJ, Savage RA, Vail ML. Acute lymphocytic leukemia with prominent azurophilic granulation and punctate acidic nonspecific esterase and phosphatase activity. Am J Clin Pathol 1981;75: Komiyama A, Ogawa M, Eurenius K, et al. Unusual cytoplasmic inclusions in blast cells in acute leukemia. Arch Pathol Lab Med 1976;100: A.J.C.P. April 1991

6 CEREZO ET AL. 531 Granular Lymphoblasts in ALL 4. Simpkins H, Shoaf F, Katz J. An acute granular lymphoid leukemia with unusual cytochemistry and immunologic phenotype. Hum Pathol 1987;18: Stein P, Peiper S, Butler D, Melvin S, Williams D, Stass S. Granular acute lymphoblastic leukemia. Am J Clin Pathol 1983;79: Bennett JM, Catovsky D, Daniel M-T, et al. Proposals for the classification of the acute leukaemias. Br J Haematol 1976;33: Bennett JM, Catovsky D, Daniel M-T, et al. The morphologic classification of acute lymphoblastic leukaemia. Concordance among observers and clinical correlations. Br J Haematol 1981 ;47: Anderson JK, Metzgar RS. Detection and partial characterization of human T and B lymphocyte membrane antigens with antisera to HSB and SB cell lines. J Immunol 1978;120: Pullen DJ, Falletta JM, Crist WM, et al. Southwest Oncology Group experience with immunological phenotyping in acute lymphocytic leukemia of childhood. Cancer Res 1981;41: Vogler LB, Crist WM, Bockman DE, et al. Pre-B-cell leukemia a new phenotype of childhood lymphoblastic leukemia. N Engl J Med 1978;298: van Eys J, Berry D, Crist W, et al. Treatment intensity and outcome for children with acute lymphocytic leukemia of standard risk. Cancer 1989;63: Crist W, Boyett J, Jackson J, et al. Prognostic importance of the pre-b-cell immunophenotype and other presenting features in B- lineage childhood acute lymphoblastic leukemia: a Pediatric Oncology Group Study. Blood 1989;74: Shuster JJ, Falletta JM, Pullen DJ, et al. Prognostic factors in childhood T-cell acute lymphoblastic leukemia: a Pediatric Oncology Group study. Blood 1990;75: Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. Journal of the American Statistical Association 1958;53: Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. II. Analysis and examples. Br J Cancer 1977;35: Peto R, Peto J. Asymptotically efficient rank invariant test producers. Journal of the Royal Statistical Society, Series A 1972;135: Shuster J: Exact B and Conf: exact unconditional procedures for binomial data. American Statistician 1988,42:234. Vol. 95 No. 4