S efforts have been made to identify prognostic

Transcription

1 INITIAL FEATURES AND PROGNOSIS IN 363 CHILDREN WITH ACUTE LYMPHOCYTIC LEUKEMIA JOSEPH V. SIMONE, MD,* MANUEL S. VERZOSA, MD,+ AND JUDITH A. RUDY, RN~ The relationship of a variety of initial features and the outcome of therapy was analyzed for 363 children with acute lymphocytic leukemia (ALL). All had entered total therapy studies between 1962 and The standard for comparing outcome of therapy was whether patients with a given feature attained or exceeded the median duration of complete remission, hematologic remission or survival for the group. The results showed that, in general, the more massive or extensive the disease at diagnosis, the poorer the outcome. Factors associated with a significantly poorer prognosis included: initial leukocyte count above 100,000/mm3; spleen enlargement greater than 5 cm; me-, diastinal involvement and early CNS involvement. Children over 10 years old at diagnosis and Negro children also had a poor prognosis. From another viewpoint features were examined for patients who attained at least 3 years of continuous complete remission. This confirmed some earlier findings and, in addition, showed that children under 2 years of age at diagnosis or with hepatomegaly over 5 cm were less likely to attain this goal. With the exception of early CNS involvement, however, patients with excellent responses to therapy were found with each factor of poor prognosis. Two major factors were not analyzed because their relationship to prognosis is generally accepted: therapeutic differences and acute nonlymphocytic leukemia. Cancer 36: , INCE THE ADVENT OF EFFECTIVE THERAPY, S efforts have been made to identify prognostic factors in patients with cancer. Prognostication is important not only for the patient and his family but for accurate scientific evaluation of the efficacy of therapy. Over a relatively wide range of biological variation of both tumor and host, prognosis in cancer generally depends on the anatomical extent of the tumor, its innate sensitivity to therapy, and on the po- lrom the flematology Service, St. Jude Children s Research Hospital, Memphis, TN. * Member in Hematology-Oncology. Assistant Member in Hematology-Oncology. Research Assistant in Hematology-Oncology. Address for reprints: Joseph V. Simone, hli), St. Jude Children s Research Hospital, 332 North Lauderdale, P.O. Box 318, Xlemphis, TN , l his work was presented in part at the Keisenberg-Workshop, Prognostic Factors in Human Acute Leukemia, Ulm, October, Supported by Cancer Research Center Grant CA-08480, Research Project Grants CA and CA-130, and l raining <;rants C:A-(15176 and CA from the National Cancer Institute, by General Research Support Grant RR-,5584 lrom the National Institutes of Health, by Project <;rant C1-70 from the American Cancer Society, and by A1,SAC. Received for publication Ikcember 10, tency and selectivity of therapy. Useful sysi-ms have been devised for classification and prognostic stratification of some malignancies, particularly those appearing to arise in a single anatomical focus. Prognostic stratification has been more difficult in acute leukemia because it is disseminated at the time of diagnosis. Furthermore, it is only since effective therapy that results have spanned a sufficiently wide range t9 make prognosis of more I than academic interest. As therapy evolved, the response of patients with acute nonlymphocytic leukemia (ANLL) did not improve nearly as much as in patients with acute lymphocytic leukemia (ALL). Prognostic stratification of children with ALL, the single most common malignancy of childhood, has been possible to some degree.4+9v 4- l6 However, few reports have included sufficient numbers of patients who received uniform therapy to permit desirable precision. Furthermore, the standard prognostic determinant, duration of survival, is less appropriate today than duration of remission. 2 For this report the records of all children with ALL entering Total Therapy I-VII from 1962 to 1971 were analyzed for the presence and significance of prognostic factors and for interrelationships of the significant factors. 2099

2 2100 CANCER December 1975 Vol. 36 METHODS The treatment used in I-VII is outlined in Table 1 and has been reported in detail el~ewhere. -~* ~- ~ Two small studies performed parallel to the major studies are included in this report. Study V-A,5 which employed the same basic regimen as Study V, was designed for patients ineligible for the parent study because they had received previous therapy. Of 22 children, 16 had already relapsed on other therapy before entering the study. However, 6 had received only 1-8 weeks of initial remission induction therapy when referred to St. Jude and, after completion of this phase, subsequent therapy was identical to Study V. Therefore, the results of V and V-A are combined for analysis of durations of remission and survival. During a 6-month period in the 2-year span of Study VI, 11 children who were Negro and/or who had massive disease were entered in Study VI-A,4 a pilot study for patients at high risk for treatment failure. Therapy was similar to Study VI except that weekly asparaginase was added during remission induction and patients were not randomized, all receiving the 1 week of intensive intravenous chemotherapy and prophylactic craniospinal irradiation. From 1962 through 1971 a total of 363 children entered I -VII. Except for the six children in Study V-A, patients had received no prior therapy or less than 1 week of prednisone and vincristine. Definitions A diagnosis of acute lymphocytic leukemia (ALL) is made on the basis of excessive numbers of lymphoblasts and/or stem cells in an aspirated bone marrow specimen. In practice, this diagnosis includes all children with leukemia that is not characterized by Auer rods or by myelocytic, monocytic or histiocytic differentiation. Special cytochemical stains and biopsy specimens are sometimes helpful, but most often the cytological diagnosis is based on the interpretation of Wright-stained marrow smears by three experienced investigators. At this institution, ALL comprises 78% of all leukemia cases, acute nonly mphocy t ic varieties comprise I 1962 I v V* VI* v TABLE 1. Outline of Total Therapy of Childhood Acute Lymphocytic Leukemia Intensive Study Kemission chemotherapy Preventive Continuation & years induction phase CNS therapy chemotherapy Pred -VCK Pred - VCK Pred + VCR Pred + VCR Pred + VCK Pred + VCK + Dauno Pred + VCK None None 6-MP I.V. daily X 3 MTX I.V. daily X 3 Cycle1.V. X 1 As in 111 As in I11 As in Ill or none None * See text for details of companion Studles V-A and VI-A. 500 rads craniospinal 500 rads craniospinal 1200 rads craniospinal None 2400 rads cranial + I.T. MTX 2400 rads craniospinal or none 2400 rads cranial + I.T. MTX or 2400 rads craniospinal 6-MP P.O. daily VCK I.V. weekly Cyclo I.V. weekly 6-MP P.0. daily or IM I X I.M. weekly VCK I.V. every 2 weeks Cyclo I.V. every 2 weeks 6-MP P.O. daily M I X, Cyclo and VCK I.V. weekly Full or half dosage 6-MP P.O. daily MTX, Cyclo and VCK 1:V. weekly 6-MP P.0. daily MI X I.V. weekly Cyclo I.V. weekly Pred + VCK, 2-week course every 10 weeks 6-MP P.O. daily MI X P.O. weekly Cyclo P.0. weekly Pred + VCK, 2-week course every 12 weeks 6-MP P.O. daily M I X P.O. weekly Cyclo P.O. weekly &Pred + VCK, 2-week course every I2 weeks

3 No. 6 PROGNOSIS IN CHILDHOOD LEUKEMIA Simone et al % and chronic myelocytic leukemia 3%. Children with lymphoblastic or stem cell lymphosarcoma with marrow involvement at diagnosis are considered to have ALL and are included in these studies. Complete remission (CR) duration is the initial period free of all signs of ALL, whether hematologic, visceral or in the central nervous system (CNS). Hematologic remission (HR) duration is the initzal period free of signs of relapse in the blood or bone marrow. CNS leukemia is diagnosed when leukemia cells are found in a Wright-stained centrifugate of cerebrospinal fluid. For purposes of this report, patients who failed to attain complete remission in the initial 8 weeks of chemotherapy were dropped from the study and considered as having a zero duration of complete remission, even if they achieved remission with subsequent therapy. This appropriately reflects the poor prognosis of children with ALL who fail to respond to vincristine and prednisone. Therefore, patients who failed to attain complete remission or hematological remission are included for computation of median durations in each study. Features Analyzed A variety of features observed at the time of diagnosis before therapy were tabulated for all 363 patients entering these studies. Patients were stratified into two groups by sex and race, into three groups by age at diagnosis: less than 2 years, 2-10 years, and over 10 years. No patient was older than 20 years. Initial height and weight percentile for age were determined by standard growth grids and divided into three strata: less than the 25th percentile, 25th to 75th percentile and greater than the 75th percentile. Patients were stratified by their initial peripheral blood values: initial leukocyte count of less than 10,000/mm3, 10,000 to 49,999, 50,000 to 99,999 and 100,000 or above; hemoglobin concentrations of less than 6 g%, 6 to 8 g%, and greater than 8 g%; platelet count of less than 25,000/mm3, 25,000 to 49,999, 50,000 to 99,999 and 100,000 or above. Both liver and spleen enlargement in centimeters below the respective costal margins were divided into two strata: 0-5 cm and greater than 5 cm. Patients who had any lymph nodes at diagnosis greater than 2 cm in both diameters or who had enlarged lymph nodes other than cervical or inguinal were identified and the outcome in both groups was compared individually to the patient population as a whole. The presence of mediastinal involvement (usually thy- mic) was determined by chest roentgenogram. This did not include young children with thymic shadows within the range of normal size for age. Early CNS involvement meant that leukemia cells were detected in the cerebrospinal fluid either at the time of diagnosis, at the time of initial marrow remission, or in Study VI, prior to randomization for preventive CNS therapy. In effect, this included patients with detectable CNS leukemia in the first 8 weeks after diagnosis in Study VI and in the first 4 weeks in all other studies. Method of Analysis Within each study group an analysis was made of the association of the above features with median durations of initial complete remission, initial hematologic remission and survival. Because of similar therapy and chronology, data from certain studies was combined yeilding the following analytic units: 1-111, IV, V + V-A, VI + VI-A, and VII. Patients who failed to respond to therapy or who died during the initial month of therapy were considered to have remission durations of zero. Remission was considered ended in patients who died, usually of infection, while still in initial remission. Since more than half the patients were still in remission or surviving in V-VII at the time of analyses (July 1973), medians were computed as of that time. Therefore, some patients who survive in initial remission are counted as not reaching the median. For this reason and because patients who failed to attain remission are included for computing the median, the medians in this report are understated and differ from reports of the individual studies. The principal method of analysis employed the following steps. First, the median durations of complete remission, hematological remission and survival were determined for each study. By definition, therefore, the proportion of patients attaining or exceeding the median is 0. Then, in patients with a particular feature, such as male sex or an initial leukocyte count above 100,000/mm3, the proportion attaining or exceeding the median is determined and compared to the whole group. If the proportions were about the same-0 -the feature had no prognostic value. If the proportion attaining the median was significantly less than 0, the feature had a negative association with prognosis. If the proportion attaining the median was significantly greater than 0, the feature had a positive association with prognosis. This method of analysis has two

4 2102 CANCER December 1975 Vol. 36 advantages. Since the actual months of remission and survival are not used, comparisons may be made between studies despite the 10-year span and the significant differences in actual median durations. When all studies showed no significant differences, results could be combined. Also, this system permitted the use of a relatively simple means of determining the statistical significance of differences-the 50% Probability Test which is a modification of the Binomial Test. As in other tests, one begins with the null hypothesis (no significant difference) and computes the probability (p value) that the observed difference could have occurred by chance alone. By convention a p value of 5% or less is considered sufficiently unlikely that the observed difference is statistically significant. After the significant factors were identified, their relationship to strata of the initial leukocyte count was analyzed and the statistical significance was determined by the Chi-squared Test. Finally, the initial features of patients in I-VI who were in continuous complete remission for at least 3 years were tabulated to determine whether the same prognostic associations app 1 ied. RESULTS Unless otherwise stated, the observed outcome in a subgroup of patients with a specific feature is compared to the expected outcome which is the outcome computed for each study group as a whole. Since the median durations are the standard of comparison, the proportion expected is 0 or 50%. Age and Sex In each study the outcome for patients 2-10 years of age was the same or better than for the group as a whole. This is reflected in the combined data for I-VII showing that 55 to 56% attained the medians. The outcome for patients over 10 years old was only slightly worse than expected in the earlier studies, but in Study VII, significantly fewer attained the median (Table 2). Consequently, the combined data shows that the outcome for patients over 10 years old at diagnosis is significantly worse than in patients 2-10 years old, and that the difference has increased with more modern and effective treatment. The outcome in children less than 2 years old at diagnosis was consistently worse than expected except in Study VII. However, the number in each study and the total for all studies are relatively small so the differences are not statistically significant. While it is possible that the improved outcome in Study VII signifies a modern trend, one cannot ignore the poor outcome in all other studies and the combined results. Of 33 patients in all studies combined, only onethird attained the median duration for CK. This difference is not statistically significant but may be clinically significant. A larger sample might clarify this point. The outcome in males and females was virtually identical in all studies with one exception (Table 2). In V + V-A, girls fared remarkably better with 17 of 19 attaining the median CK. This difference did not correlate with the presence of other factors and remains unexplained. TABLE 2. Kelationship of Age and Sex to Outcome in Childhood Acute Lymphocytic Leukemia Age-yrs VII >to I-VII < >10 Sex V + V-A Male Female I-VII Male Female * p ; + Proportion Attaining hledian Ihration Patients CK flu Survival p < 1% by The 50% Probability Test. CK = complete remission HK = hematologic remission, i9.27* S.so *.48.5i.35*.23* S * *,45,57

5 ~~ No. 6 Initial Height and Weight PROGNOSIS IN CHILDHOOD LEUKEMIA Simone et al These features were analyzed to determine whether children who were small or large for age might have a different prognosis, perhaps because of poor nutritional practices or chronic infection. As shown in Table 3, the outcome for patients with height gr weight above the 75th percentile was slightly worse than expected, but the difference was small and not significant. Race The outcome in Negro children was consistently worse than for white children in all studies (Table 4). The difference for all studies combined was statistically very significant. As we have reported previously, l4 and confirmed in correlations below, it is apparent that Negro children have more advanced disease at diagnosis. Initial Leukocyte Count The relatively poor prognosis of patients with high initial leukocyte counts has been reported by many investigators. The present data confirm this observation and reveal other pertinent information (Table 5). The totals for all studies show that the outcome in children with leukocyte counts below 10,000/mm3 was better than expected, nearly two-thirds attaining or exceeding the median for remission and survival. This was a consistent finding in each study as well as in the totals. With counts of 10,000 to 49,999/mm3, the outcome varied from study to study, but the combined data show that these patients did about as well as expected. In the 50,000 to 99,999/mm3 stratum, the outcome varied but usually was worse than expected, particularly in the duration of complete remission. With the exception of Study IV, the outcome with counts over 100,000/mm3 was poor and became even worse in later studies. The combined data show that between the chosen strata, there was an inverse relationship of initial leukocyte count to strata. The data also suggests that improvements in therapy have magnified these differences. Platelet Count and Hemoglobin Although the difference is not striking, the outcome was better in patients with initial platelet counts above 50,000/mm3 than for those below that figure. TLe difference was statistically significant for patients with platelet counts of 100,000/mm3 or above (Table 6) and was consistent through all studies. This may simply reflect the extent of leukemic infiltration of the Relationship of Initial Height and Weight to Outcome in Childhood Acute Lymphocytic Leukemia TABLE 3. Proportion Attaining Median Duration I-VI1 Patients CK HK Survival Height-percentile < so so.51 > Weight-percentile < > CK = complete remission HK = hematologic remission bone marrow or the platelet count may have independent prognostic value. There was no statistically significant relationship in any study or in the combined data of the initial hemoglobin level and outcome. Ironically, however, patients with hemoglobin levels above 8 g% fared somewhat worse than expected. Enlargement of Liver, Spleen and Lymph Nodes In all but the first study group, the outcome was worse than expected in patients with spleens palpated more than 5 cm below the costal margin at diagnosis. When the data from all studies was pooled, this difference was statistically significant (Table 7). The outcome in patients with liver enlargement greater than 5 cm was significantly worse than expected in Study VI, but no differences were observed in the other studies. Although the combined data show similar outcomes for both liver and spleen enlargement (Table 7), the difference in patients with liver enlargement is not statistically significant. In all studies combined, the outcome in patients with lymph nodes greater than 2 cm in both diameters or with enlarged lymph nodes other than cervical or inguinal was only slightly worse than expected (Table 7). Differences were found in individual studies. In Study V, one of four, and in Study VII, three of 14 patients with nodes greater than 2 cm attained the median for complete remission. However, differences within studies and in the combined data were not statistically significant. Mediastinal and Early CNS Involvement The outcome in patients with mediastinal involvement was consistently worse than expected. As shown in Table 8, only 8 of 37 patients at-

6 2104 C ANGER December Vol. 36 TABLE 4. Relationship of Race to Outcome in Childhood Acute Lymphocytic Leukemia Proportion Attaining Median 1)uration Race Patients CR HK Survival White Negro lo*. lo* IV White Negro 7.29.I4.I4 V + V-A White Negro V1 + VI-A White Negro *.17* VI I White Negro 10.lo*. 00'.I o* Total White Negro 42.26'.12'.14' * p ; + p < 1%; * p < 0.1% by The 50% Probability Test. CR = complete remission HR = hematologic remission tained the median for complete remission duration. This difference is both clinically and statistically significant. The outcome for patients with early CNS involvement was even worse; only 2 of 19 attained the median duration for complete or hematologic remission. Correlation of Significant Prognostic Factors It is possible that some of the significant prognostic factors identified above influence one an- other and, therefore, are not independent. A systematic correlation of factors would provide clues to their significance and interdependence. In Table 9, the correlation of initial leukocyte count and other factors is examined. For each stratum of leukocyte count, the proportion of patients possessing the listed features is compared to the expected proportion if there was no correlation with leukocyte count. The significance of any differences from the expected proportion was analyzed by the Chi-square test. TABLE 5. Relationship of Initial Leukocyte Count to Outcome in Childhood Acute Lymphocytic Leukemia Leukocytes Proportion Attaining Median Duration Survival ( 10S/mms) Patients CK HK < 10 loo+ IV <10 loo+ V + V-A <10 too+ VI + VI-A < 10 loo+ VII <I0 loo+ Total < 10 loo * p ; 'p < 1%; *p < 0.1% by The 50% Probability Test. CR = complete remission HR = hematologic remission '.65*.25.09*.65*.48.29*.20' ' *.62' ' ' *.63'.47.4 I.20*

7 No. 6 PROGNOSIS IN CHILDHOOD LEUKEMIA Simone et al TABLE 6. Relationship of Initial Platelet Count and Hemoglobin to Outcome in Childhood Acute Lymphocytic Leukemia Proportion Attaining Median Duration I-VII Patients CR HR Survival Plarelets ( 1O3/mm3) < J J * Hemoglobin (gm%) < > * p by The 50% Probability 'lest. CK = complete remission HK = hematologic remission As one might expect, the initial platelet count and leukocyte count were related inversely. While only 45% of all patients had leukocyte counts below 10,000/mm3, 64% of those with platelet counts above 100,000/mm3 were in this stratum. Also expected was the observation that patients with substantial splenomegaly or mediastinal involvement had higher initial leukocyte counts than the group as a whole. These factors appear to be related to the mass of tumor and dependent on the extent of infiltration. The distribution of patients over 10 years old was the same as for the population as a whole. Therefore, the poor prognosis in these patients appears to be independent of initial leukocyte count. TABLE 7. Relationship of Initial Enlargement of Liver, Spleen and Lymph Nodes to Outcome In Childhood Acute Lymphocytic Leukemia Proportion Attaining Median Duration I-VII Patients CK HK Survival Liver 0-5 cm >5 cm Spleen 0-5 cm >5 cm *.41*.43* Nodes >2 cm in both diameters Enlarged nodes other than cervical or inguinal * p by The 50% Probability Test. CR = complete remission HK = hematologic remission TABLE 8. Relationship of Mediastinal and Early CNS Involvement to Outcome in Childhood Acute Lymphocytic Leukemia Proportion Attaining Median Duration I-VII Patients CR HR Survival Mediastinum involved 37.22'.24'.32* CNS involved early 19.11'.I I'.16* * p ; ' p < 1% by The 50% Probability Test. CK = complete remission HK = hematologic remission Mediastinal involvement was determined by chest roentgenogram. Early CNS involvement means leukemia cells were detected in cerebrospinal fluid at the time of diagnosis, initial marrow remission or prior to randomization for preventive CNS therapy. As we have previously observed, a much greater than expected proportion of Negro children have high initial leukocyte counts. This explains in part why these children have a poor prognosis, but additional negative factors may also be responsible. Although slightly more patients with early CNS involvement had high leukocyte counts, the overall distribution was similar to the population as a whole. The lack of correlation with initial leukocyte count suggests that early CNS involvement is an independent prognostic factor not related to the total tumor mass. Since platelet count is related inversely to leukocyte count, one might conclude that early CNS leukemia and platelet count were independent. This hypothesis was examined, and as shown in Table 10, patients with early CNS leukemia have randomly distributed initial platelet counts. The last possibility examined in Table 9 was whether the absence of other significant factors was related to initial leukocyte count. The proportion of patients with none of the features had leukocyte counts slightly lower than the population as a whole, but the difference was not significant. Therefore, the absence of prognostic factors other than high leukocyte count had no independent prognostic value. Features of Patients with Long Remissions The standard used above for assessing prognosis was whether patients attained the median duration of remission and survival for the particular study group. Another means of determining prognostic factors is to examine the initial features of patients who have had excellent responses to therapy. Since all patients in

8 2106 CANCER December 1975 Vol. 36 TABLE 9. Correlation of Initial Leukocyte Count and Other Prognostic Factors in Childhood Acute Lymphocytic Leukemia Leukocyte Count ( 103/mm3) < Number of Patients Proportion of Patients I2 (Expected) Patients Observed Proportion p-value Platelets > 1O5/rnm < I% Spleen > 5 cm I5.14 < 0. I Y" Age > 1Oyrs Negroes I2.36 Mediastinal inv < 0.1% < 0. I?" CNS inv. early ll.26 None of above The Chi-squared Test was used to determine the probability (p) that the differences between the expected and observed proportions could have occurred by chance alone. I-VI have been under study for at least 3 years, the features were analyzed for patients having initial continuous complete remissions of 3 years or more. Patients who had even one CNS relapse are excluded, even though they may survive in continuous hematologic remission and no longer receive therapy. Also excluded were patients who died in remission, even though they never relapsed. In Table 11 the proportion of all 363 patients with a particular feature is compared to the proportion of the 81 patients in I-VI with that feature who attained complete remissions of 3 years or more. The statistical significance of any differences was estimated by appropriate tests. By this method of analysis, some factors retain the prognostic significance observed when median durations were used as the standard of outcome while others do not. For example, the proportion of children under 2 years old who attained 3 years of complete remission was significantly less than expected while the proportion over 10 years was about as expected. Although the proportion of Negroes attaining long remissions was less than the original proportion, surprisingly, the difference was not significant. This may indicate that Negro children do as well as white children with comparable tumor mass at diagnosis. The poor prognosis of children with initial TABLE 10. Correlation of Initial Platelet Count and Early CNS Leukemia in Childhood Acute Lymphocytic Leukemia Platelets (103/mm3) < Total Patients CNS leukocyte counts above 100,000/mm3, mediastinal involvement or early CNS involvement is confirmed by the significantly lower proportion attaining at least 3 years of complete remission. Early CNS leukemia is especially bad; not one patient attained the goal. Spleen size retained its prognostic significance and, whereas, in the earlier analysis liver size was not a statistically significant factor, it is significantly related to long remission. DISCUSSION The results reported here tend to confirm the principle that prognosis in children with ALL is related inversely to the initial mass and extent of leukemic infiltration. The only prognostic factor examined that cannot be explained in this way is age at diagnosis. The prognosis of children under 2 or over 10 years old is not as good as for those 2-10 years old." However, with the possible exception of early CNS involvement, the prognostic factors validly apply only to sizeable groups of patients because exceptional individuals are found in every category. For example, an 11-year-old Negro boy in Study V with an initial leukocyte count of 225,000/mm3 has been in initial complete remission for over 5 years and off all therapy for 2 years. The risk of giving grossly unconventional therapy and possibly losing the small but definite chance for a good result must be considered before devising new forms of therapy. The two most important prognostic factors in childhood leukemia have not been included in this report because their importance is generally accepted: cytologic type and therapy. Most would agree that children with the nonlymphocytic types of acute leukemia do poorly.

9 No. 6 PROGNOSIS IN CHILDHOOD LEUKEMIA Simone et al TABLE 11. Comparison of Features at Diagnosis of 363 Patients with 81 Children in I-V1 Having Initial Complete Remission of at Least Three Years Features Age (yrs) < > 10 Sex Male Female Race White Negro Leukocyte Count ( 10a/mma < 10 loot Platelet Count ( 103/mm3) < loot Liver (cm) 0-5 >5 Spleen (cm) 0-5 > 5 Nodes < 2 cm Large Nodes (not cervical) of inguinal) Mediastinum Involved CNS Involved Early CK > 3 All patients Years Proportion Proportion p-value i ns. ns. ns. ns. < 590 Statistical significance was determined by the chi-squared Test, the zi Test or Poisson's Test, whichever was appropriate. The distinction from ALL usually is made easily, but a small proportion, less than 5%, are difficult to classify. These leukemias usually are characterized by primitive blast cells with no clearly distinguishing features by conventional or special stains. Occasionally, lymph node biopsy places it in the histiocytic or reticulum cell category. If opinion remains uncertain or divided after all appropriate diagnostic approaches have been exhausted, it is our practice to make a provisional diagnosis of ALL and to treat as such. Thereafter, unless convincing evidence to the contrary surfaces, the diagnosis remains ALL. It would be of inestimable value if these few cases could be classified with certainty. Little need be said about the relationship of therapy to prognosis in ALL. In our own studies the proportion of 5-year leukemia-free survivors increased from 0.17 in to 0.51 in Study V. Appropriately, investigators have concentrated their efforts on this prognostic factor because it is the one over which they have most control. It is noteworthy that despite major improvements in prognosis, the relative association of the factors has changed little. The outcome in patients with massive disease is better than it was 10 years ago, but they continue to do poorly, perhaps even worse, relative to all other patients. An observation with few exceptions in this report was the consistent agreement between complete remission, hematologic remission and survival. We have reported this relationship and demonstrated that complete remission duration is the earliest and most appropriate measure of response to therapy." Although correlations between certain initial features and prognosis were demonstrated in this study, exceptions were noted in every category. It is likely that other factors, hopefully more specific, will be identified so that better therapy may be devised. A promising approach is the identification of subpopulations of leukemic cells. A correlation of thymic enlargement and high leukocyte counts with blast cell surface features suggesting thymic origin was reported by Borella et al." It is possible that more extensive disease is due not to delayed diagnosis but to innately different tumor that has a biological advantage for spread or for impairing a specific host defense. REFERENCES 1. Aur, K. J. A,, Hustu, H. O., Verzosa, M. S., et al.: Comparison of two methods of preventing rentral nervous system leukemia. Blood 42: , Aur, R. J. A,, Simone, J. V., Hustu, H. L., et al.: A comparative study of central nervous system irradiation and intensive chemotherapy early in remission of childhood acute lymphocytic leukemia. Cancer 29: , Aur, R. J. A,, Simone, J., Hustu, H. O., et al.: Central

10 2108 CANCER December 1975 Vol. 36 nervous system therapy and combination chemotherapy of childhood lymphocytic leukemia. Blood 37: , Aur, K. J. A,, Simone, J. V., and Pratt, C. B.: Successful remission induction in children with acute lymphocytic leukemia at high risk for treatment failure. Cancer 27: , Aur, K. J. A,, Verzosa, M. S., Hustu, H. O., et al.: Response to combination therapy after relapse in childhood acute lymphocytic leukemia. Cancer 30: , Cutler, S. J., Axtell, L., and Heise, H.: Ten thousand cases of leukemia, J. Nut. Cancer Inst. 39: , Hardisty, K. M., and Till, M. M.: Acute leukemia, : Factors affecting prognosis. Arch. Dis. Child. 43:107-1 t5, hlodan, B., Virag, I., and Modan, M.: Survival in childhood malignancies: Assessment of the influence of age, sex and tumor type, with emphasis on long-term survivors. J. Nut. Cancer Inst. 43: , Pierce, hl. I., Borges, W. H., Heyn, K., et al.: Epidemiological factors and survival experience in 1770 children with acute leukemia treated by members of Children s Study Group A between 1957 and Cancer 23: , Pinkel, I).: Five-year follow-up of total therapy of childhood lymphocytic leukemia.3ama 216: , Pinkel, I]., Hernandez, K., Borella, L., et al.: Drug dosage and remission duration in childhood lymphocytic leukemia. Cuncer 27: , Simone, J. V., Aur, K. J. A,, Hustu, H. O., et al.: Total Therapy studies of acute lymphoeytic leukemia in children. Current results and prospects for cure. Cancer 30: , Simone, J. V., Holland, EZ, ard Johnson, W.: Fatalities during remission of childhood leukemia. Blood 39: , Walters, T., Bushore, M., and Simone, J.: Poor prognosis in Negro children with acute lymphocytic leukemia. Cancer 29: , Zuelzer, W. W.: Implications of long-term survival in acute stem cell leukemia of childhood treated with composite cyclic therapy.blood 24: , Zippin, C., Cutler, S. J., Reeves, W. J., Jr., et al.: Variation in survival among patients with acute lymphocytic leukemia. Blood , Borella, L., Sen, L., and Green, A. A,: Cell surface markers and clinical features in acute lymphmytic leukemia. Proc. Am. Assoc. Cancer Res. 15:122, 1974.