Racial Differences in the Survival of Childhood BPrecursor Acute Lymphoblastic Leukemia: A Pediatric Oncology Group Study By Brad H. Pollock, Michael R. DeBaun, Bruce M. Camitta, Jonathan J. Shuster, Yaddanapudi Ravindranath, D. Jeanette Pullen, Vita J. Land, Donald H. Mahoney, Jr, Stephen J. Lauer, and Sharon B. Murphy Purpose: We conducted a historic cohort study to test the hypothesis that, after adjustment for biologic factors, African-American (AA) children and Spanish surname (SS) children with newly diagnosed B-precursor acute lymphoblastic leukemia had lower survival than did comparable white children. Patients and Methods: From 1981 to 1994, 4,061 white, 518 AA, and 507 SS children aged 1 to 20 years were treated on three successive Pediatric Oncology Group multicenter randomized clinical trials. Results: AA and SS patients were more likely to have adverse prognostic features at diagnosis and lower survival than were white patients. The 5-year cumulative survival rates were (probability ⴞ SE) 81.9% ⴞ 0.6%, 68.6% ⴞ 2.1%, and 74.9% ⴞ 2.0% for white, AA, and SS children, respectively. Adjusting for age, leukocyte count, sex, era of treatment, and leukemia blast cell ploidy, we found that AA children had a 42%

excess mortality rate compared with white children (proportional hazards ratio [PHR] ⴝ 1.42; 95% confidence interval [CI], 1.12 to 1.80), and SS children had a 33% excess mortality rate compared with white children (PHR ⴝ 1.33; 95% CI, 1.19 to 1.49). Conclusion: Clinical presentation, tumor biology, and deviations from prescribed therapy did not explain the differences in survival and event-free survival that we observed, although differences seem to be diminishing over time with improvements in therapy. The disparity in outcome for AA and SS children is most likely related to variations in chemotherapeutic response to therapy and not to compliance. Further improvements in outcome may require individualized dosing based on specific pharmacogenetic profiles, especially for AA and SS children. J Clin Oncol 18:813-823. © 2000 by American Society of Clinical Oncology.

CUTE LYMPHOBLASTIC leukemia (ALL) is the most common malignancy of childhood, accounting for one quarter of pediatric cancer. Subtypes of ALL are classified primarily by the immunophenotype of the leukemic blast cells. Approximately 84% of cases of ALL are classified as B-precursor ALL (clonal expansion of progenitors of B-cell lymphocytes), 14% are T-cell ALL, and 2% are B-cell ALL.1 There have been dramatic improvements in outcome for children with ALL in the past three decades, primarily because of the development of more effective combination chemotherapy regimens. Cure rates for Bprecursor ALL are currently approximately 70%.2 Earlier reports suggested that African-American (AA) children with leukemia have lower survival rates than do white children with leukemia.3 The racial disparity can be attributed not only to differences in the presenting features of disease,4,5 but to other factors as well.6,7 Although survival rates seem to be improving over time,2 there is little information available that relates this decrease to either change in presenting characteristics of patients or to improvements in therapy. Earlier studies have failed to include Spanish surname (SS) patients, failed to demonstrate significant SS outcome differences,8 or have not accounted for major biologic prognostic factors. For children who receive modern, multiagent, protocolbased chemotherapy, we hypothesized that outcomes for AA children compared with white children and SS children

compared with white children with newly diagnosed Bprecursor ALL are not equally shared, even after adjustment for known prognostic factors. We evaluated whether race is an independent prognostic factor for B-precursor ALL and described the relationship between race and survival rates over three successive Pediatric Oncology Group (POG) randomized clinical trials beginning in 1981.

A

From the University of Florida, and Pediatric Oncology Group Statistical Office, Gainesville, FL; Washington University, St Louis, MO; Midwest Children’s Cancer Center, and Children’s Hospital of Wisconsin, Medical College of Wisconsin, Milwaukee, WI; Children’s Hospital of Michigan, and Wayne State University, Detroit, MI; University of Mississippi School of Medicine, Jackson, MS; Children’s Memorial Hospital, and Pediatric Oncology Group Operations Office, Chicago, IL; Baylor College of Medicine, Houston, TX; and Emory University, Atlanta, GA. Submitted May 20, 1999; accepted October 25, 1999. Supported by research grants no. CA29139, CA30969, and CA37379 from the National Institutes of Health, Bethesda, MD. Address reprint requests to Brad H. Pollock, MPH, PhD, Pediatric Oncology Group Statistical Office 104 North, Main St, Suite 600, Gainesville, FL 32601-3330; email [email protected]. © 2000 by American Society of Clinical Oncology. 0732-183X/00/1804-813

Journal of Clinical Oncology, Vol 18, No 4 (February), 2000: pp 813-823 Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

813

814

POLLOCK ET AL

PATIENTS AND METHODS

Study Subjects This historic cohort study included children with newly diagnosed B-precursor ALL who were enrolled onto phase III POG therapeutic protocols. POG is a consortium that represents approximately one half of all centers that treat childhood cancer in North America. The study population consisted of children treated on one of three randomized clinical trials (POG 8036, POG 8602, and POG 9005/6) for B-precursor ALL. Children younger than 1 year were excluded because of the inadequate number of patients for this group, the different biologic characteristics of their leukemia cells, and the fact that they received more intensive therapy on separate treatment protocols than did older children. Overall, more than 90% of eligible children with B-precursor ALL are enrolled onto POG therapeutic protocols at institutions that participate in POG ALL studies (unpublished data). This study excluded children who were treated on one of eight single-arm POG pilot protocols (N ⫽ 1,155) that were run concurrently with the phase III trials. The patient characteristics for those treated on pilot studies versus those treated on the randomized phase III trials (POG 8036, POG 8602, and POG 9005/6) were similar with respect to age, sex, leukocyte count, and DNA index (DI). Conclusions from analyses that combined patients treated on the pilot studies with those treated on the phase III studies were similar (data not presented). All patients were confirmed as having B-precursor ALL through the use of standardized central reference laboratory review procedures.9

Prognostic Factors Each patient’s racial classification was recorded at the time of diagnosis by the participating institutions according to the National Cancer Institute’s (NCI) coding scheme for the Clinical Therapeutics Evaluation Program of the Division of Cancer Treatment and Diagnosis. The categories included white (non-SS), AA (non-SS), SS, Native American, Asian, Hawaiian, Filipino, Indian subcontinent, and other. Analyses were limited to white, AA, and SS children because they accounted for more than 96% of the study population. B-precursor disease occurs less frequently in AA children.4,5 We restricted the analysis to children with B-precursor ALL who were 1 year or older at diagnosis, because completely separate, more intensive treatment protocols are used for infants with ALL. Biologic and clinical characteristics obtained at diagnosis included age, sex, WBC count, blast cell ploidy, platelet count, hemoglobin level, presence of the common ALL antigen (cALLA), and cytogenetic abnormalities. For presentation, we used the definition for poor risk based on a WBC count greater than 50,000 cells/mm3 and/or age ⱖ 10.00 years at diagnosis, which was developed as a consensus guideline for prognostic stratification.10 We analyzed leukemic cell cytogenetic data and the DI, a measure of leukemic blast cell ploidy, for patients registered on the two most recent POG treatment protocols (POG 8602 and POG 9005/6), for which cytogenetic assessment was centrally performed. DI was determined by flow cytometry, and the prognostic significance of DI ⱕ 1.16 was assessed.11 To characterize the variation in the institutions where patients were treated, we determined the average volume of new pediatric cancer patients seen annually at each institution and included this as an explanatory variable. For each patient, we created a summary variable that represented major deviations from the protocol regimen as determined by centralized review every 6 months.

Statistical Analysis Frequency tables were constructed to describe the relationship between race and categorical prognostic factors. The primary outcome measure assessed was survival, which was calculated as the interval of time from the date of diagnosis until the date of death from any cause or until the date of last contact. Separate analyses were performed comparing AA patients with white patients and SS patients with white patients, which were adjusted for various combinations of potential explanatory factors. Survival curves were generated using the method of Kaplan and Meier.12 Log-rank tests,13 stratified by NCI risk groups,10 were also used. Cox proportional hazards regression14 was used to estimate the proportional hazards ratio (PHR): this relative risk of mortality reflects the magnitude of the association between race and survival. The PHR was adjusted for other prognostic factors. We calculated 95% confidence intervals (CI) for the PHR estimates. Selected first-order interaction terms were included for age ⫻ race, sex ⫻ race, and age ⫻ sex. Multivariate analyses included the individual variables that are used to assign NCI risk groups. However, age was represented by two indicator variables (age ⬍ 3 years and age ⱕ 6 years), because very young and older children have poorer survival rates. To eliminate confounding that results from differences in access to salvage therapy, we also assessed event-free survival (EFS) as an outcome measure. EFS was calculated as the interval of time from the date of diagnosis until the date of first treatment failure (including failure to attain a complete response during the induction phase of the protocol, relapse, second malignancy, and death resulting from any cause) or until the date of last contact. Removal for bone marrow transplantation was counted as a censored event; only 0.64% of the study population was censored for bone marrow transplantation. Follow-up was through February 28, 1999.

Treatment Regimens The schemas for the three treatment protocols are listed in Table 1. Dose modification criteria were determined before accrual. The regimens were primarily antimetabolite-based and have been described previously.15-18 Within each of the three successive protocols, patients were randomized up front within strata defined by age and WBC count. Assignment to either POG 9005 or POG 9006 was based on risk group, which was defined by age, WBC count, and tumor ploidy. These two protocols were run concurrently and shared a common treatment arm; thus, they are designated as POG 9005/6. White, AA, and SS patients were distributed relatively equally among treatment regimens within each protocol.

RESULTS

Study Population A total of 5,086 patients were accrued from June 3, 1981, through October 28, 1994. The ethnic composition was 4,061 (79.8%) white, 518 (10.2%) AA, and 507 (10.0%) SS. Patients were treated at 105 POG institutions. One institution accrued a total of 253 patients, 12 institutions each had between 100 and 200 patients, 24 institutions each provided between 50 and 99 patients, 60 institutions each had between two and 49 patients, and eight institutions each accrued a single patient. Patients were treated on one of

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

815

RACE DIFFERENCES IN SURVIVAL OF CHILDHOOD ALL Table 1.

Protocol Treatment Schemas

POG 8036

POG 8602

POG 9005

POG 9006

PRED, VCR

PRED, VCR, ASP

PRED, VCR, ASP

Consolidation

ASP, CTX, LDMTX

Continuation

PO MP/MTX/PRED or ⫹ IDM or ⫹ VCR/DOX, ARAC, 6TG/CTX TIT or IT MTX

IDMTX or IDMTX/ASP or IDMTX/ARAC PO MTX, PO MP, PRED, VCR

IDMTX or IDMTX/IVMP or dd PO MTX/IVMP IM MTX, PO MP

PRED, VCR, ASP, DNR IDMTX/IVMP or IDMTX/IVMP, ARAC/VM26, PRED/VCR/ASP/DNR/ARAC IM MTX, PO MP

TIT

IT MTX or TIT

Induction

CNS prophylaxis

TIT

Abbreviations: ARAC, cytarabine; ASP, asparaginase; CTX, cyclophosphamide; dd, divided dose; DNR, daunorubicin; DOX, doxorubicin; IDMTX, intermediate-dose methotrexate; IT, intrathecal; IM, intramuscular; IVMP, intravenous mercaptopurine; LDMTX, low-dose intravenous methotrexate; MTX, methotrexate; MP, mercaptopurine; PO, oral; PRED, prednisone; TG, thioguanine; TIT, triple intrathecal; VCR, vincristine; VM26, teniposide.

three consecutive phase III protocols: POG 8036 (June 1981 through January 1986), POG 8602 (February 1986 through January 1991), and POG 9005/6 (January 1991 through October 1994). Prognostic Factors by Race Prognostic factors by race are summarized in Table 2. AA patients had higher WBC counts compared with white patients (17.6% v 14.6%, respectively, with WBC ⱖ 50,000 cells/mm3; median WBC, 10,900 cells/mm3 v 8,200 cells/ mm3, respectively), were more likely to be cALLA-negative (13.0% v 10.5%, respectively), were more likely to present with CNS involvement at diagnosis (8.7% v 4.7%, respectively, with blasts and ⱖ five WBCs in the CSF), and were less likely to present with hyperdiploid blast cell DNA content (10.0% v 14.8%, respectively, with DI ⬎ 1.16). All of these characteristics reflect an increased likelihood of treatment failure. AA patients also had a slightly less favorable age distribution at diagnosis than did white patients (21% v 18% who were ⱖ 10 years of age at diagnosis). There were small differences in the distribution of the other clinical factors, including sex, platelet count, and lag-time (the interval of time between first symptom onset and diagnosis). Overall, adverse clinical prognostic features were more common for AA patients. To assess whether the size of the institution influenced outcome, we evaluated the average number of new patients seen annually. The AA patients were treated at institutions that had an average of 41 ⫾ 18 (mean ⫾ SD) new patients per year, compared with 43 ⫾ 24 new patients per year for institutions in which the white patients were treated.

Protocol compliance was routinely assessed by the protocol principal investigator every 6 months. Major and minor protocol deviations were recorded. The assessment of these deviations integrates patient and treating-physician compliance. In addition, an overall evaluation of the patient’s ability to be assessed over the last 6 months is also recorded. One or more minor protocol deviations (a delay or reduction in dose) were recorded for 59% of AA patients compared with 65% of white patients, whereas an overall summary of the ability-to-be-assessed score19 was 95.4% for AA patients and 96.7% for white patients. There seemed to be little difference across these racial groups in how protocol treatment was administered. A greater proportion of SS children were accrued to POG 9005/6, the most recent study, than to the other POG protocols. This probably reflects the changes in institutional membership to the POG over time. SS children had similar WBC counts compared with white children (17.0% v 14.6%, respectively, with WBC ⱖ 50,000 cells/mm3; median WBC, 8,250 cells/mm3 v 8,200 cells/mm3, respectively). For SS children, 21.5% were ⱖ 10 years of age (v 17.7% for white children) and 8.6% were cALLA-negative (v 10.5% for white children). SS patients were treated at institutions with similar annual volumes of new patients compared with the treating institutions for white patients (39 ⫾ 19 [mean ⫾ SD] new patients per year v 41 ⫾ 18 new patients per year, respectively). The distribution of other prognostic factors was similar for white and SS children. Major protocol deviations were noted for 60% of the SS patients compared with 65% of the white patients. The

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

816

POLLOCK ET AL Table 2.

Distribution of Prognostic Factors by Race

White Prognostic Factor

Protocol POG 8036 ALINC 13 POG 8602 ALINC 14 POG 9005/6 ALINC 15 Sex Female Male Leukocyte count at diagnosis ⬍50,000 cells/mm3 ⱖ50,000 cells/mm3 Median, cells/mm3 Interquartile range, cells/mm3 Ploidy DI ⱕ1.16 ⬎1.16 Age at diagnosis ⬍10 years ⱖ10 years Mean ⫾ SD, years Median, years CNS WBC count at diagnosis ⬍5 cells/mm3 ⱖ5 cells/mm3 Platelet count at diagnosis ⬍100,000 platelets/mm3 ⱖ100,000 platelets/mm3 Median, platelets/mm3 Interquartile range, platelets/mm3 cALLa Positive Negative Lag-time Median, days Interquartile range, days

AA

%

Total No.

31.1 34.6 34.4

128 134 245

25.2 26.4 48.3

1,443 1,847 1,799

224 294

43.2 56.8

226 281

44.6 55.4

2,333 2,753

427 91

82.4 17.6

421 86

83.0 17.0

4,313 769

%

No.

1,154 1,534 1,376

28.4 37.8 33.8

161 179 178

1,883 2,178

46.4 55.6

3,465 592

85.4 14.6

8,200 4,000-25,000

%

10,900 4,600-32,000

3,463 601

85.2 14.8

466 52

3,345 719

82.3 17.7

410 108

1,098 54

95.3 4.7

147 14

845 307

73.3 26.7

110 51

5.4 ⫾ 4.2 4

49,000 22,000-108,000 1,224 144

89.5 10.5 17 8-31

SS No.

No.

422 85

83.2 16.8

4,351 738

398 109

78.5 21.5 5.9 ⴞ 4.2‡ 4

4,153 936

91.3§ 8.7

122 6

95.3 4.7

1,367 74

68.3 31.7

97 31

75.8 24.2

1,052 389

90.0* 10.0 79.2 20.8 6.0 ⴞ 4.2† 5

53,000 26,000-150,000 134 20

8,250 3,500-26,100

87.0 13.0 17 7-33

43,000 20,000-92,000 106 10

91.4 8.6

1,464 174

18 9-31

NOTE. P values ⬍ .05 are shown in bold for comparisons of white with AA children and white with SS children. Abbreviation: ALINC, acute leukemia in children. *P ⫽ .004. †P ⫽ .005. ‡P ⫽ .024. §P ⫽ .032. 㛳Interval between symptom onset and diagnosis.

average ability-to-be-assessed score was 95.6% for SS patients compared with 96.7% for white patients. Univariate Survival Analysis The overall 5-year cumulative survival rate was higher in white patients compared with AA patients (81.9% ⫾ 0.6% [probability ⫾ SE] and 68.6% ⫾ 2.1%, respectively). Figure 1 shows the survival curves for white, AA, and SS children stratified by NCI risk group. For patients classified as good-risk, the 5-year survival was 87.6% ⫾ 0.7% for white patients and 76.6% ⫾ 2.5% for AA patients. For

poor-risk patients, the 5-year survival rate was 67.7% ⫾ 1.4% for white patients and 53.2% ⫾ 4.2% for AA patients. Adjusted for risk group, the overall survival for white patients compared with AA patients was significantly higher (stratified log-rank test, P ⬍ .0001). The 5-year EFS was also higher for white patients than for AA patients (Fig 2). The 5-year EFS rates were 74.3% ⫾ 0.9% (probability ⫾ SE) for white patients and 66.4% ⫾ 3.0% for AA patients in the good-risk group and 49.4% ⫾ 1.6% for white patients and 35.6% ⫾ 4.5% for AA patients in the poor-risk group (stratified log-rank test, P ⬍ .0001).

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

817

RACE DIFFERENCES IN SURVIVAL OF CHILDHOOD ALL

Fig 1. Kaplan-Meier estimates of survival are shown for 5,208 patients stratified by risk. Survival times are from the date of diagnosis. Log-rank tests compare white and AA children stratified by risk group and white and SS children stratified by risk group.

EFS closely paralleled survival, although the difference between them was slightly larger for good-risk white patients, perhaps reflecting increased salvage rates for white patients. As shown in Fig 1, SS children had a significantly lower survival than did white children: the 5-year survival rates were 81.9% ⫾ 0.6% (probability ⫾ SE) for white patients and 74.9% ⫾ 2.0% for SS patients (P ⬍ .0001). For good-risk white and SS patients, the 5-year survival rates were 87.6% ⫾ 0.7% and 81.8% ⫾ 2.4%, respectively. For poor-risk white and SS patients, the 5-year survival rates were 67.7% ⫾ 1.4% and 58.1% ⫾ 4.3%, respectively. Overall, white patients had significantly higher survival (stratified log-rank test, P ⬍ .0001). White children had higher EFS than did SS children (Fig 2). The 5-year EFS for good-risk patients was 74.3% ⫾ 0.9% (mean ⫾ SD) for white patients and 64.7% ⫾ 3.0% for SS children. For poor-risk white and SS children, the 5-year EFS rates were 49.4% ⫾ 1.6% and 41.9% ⫾ 4.3%, respectively (P ⬍ .0001).

Survival patterns over these treatment eras are shown in Fig 3. For white patients compared with AA patients, the 5-year survival rates were 75.0% ⫾ 1.3% (mean ⫾ SD) v 54.4% ⫾ 4.0% for POG 8036 (stratified by risk group, log-rank test; P ⬍ .0001), 84.4% ⫾ 0.9% v 71.1% ⫾ 3.4% for POG 8602 (P ⬍ .0001), and 87.0% ⫾ 1.1% v 81.0% ⫾ 4.0% for POG 9005/6 (P ⬍ .0001), respectively. Figure 4 shows the survival patterns for white compared with SS children in which the 5-year survival rates were 75.0% ⫾ 1.3% v 66.6% ⫾ 4.2% for POG 8036 (P ⫽ .03), 84.4% ⫾ 0.9% v 66.9% ⫾ 4.2% for POG 8602 (P ⬍ .0001), and 87.0% ⫾ 1.1% v 81.8% ⫾ 3.2% for POG 9005/6 (P ⫽ .017), respectively. AA children also experienced the greatest relative improvement in EFS. The 5-year EFS rates for white patients over the three successive protocols were 57.0% ⫾ 1.5%, 69.4% ⫾ 1.3%, and 71.2% ⫾ 2.2%. The 5-year EFS rates for AA patients were 40.4% ⫾ 4.1%, 63.1% ⫾ 3.8%, and 68.4% ⫾ 4.8%. The 5-year EFS rates for SS children were 50.5% ⫾ 4.7%, 51.9% ⫾ 4.6%, and 61.8% ⫾ 4.1%.

Survival Patterns Over Time

The relationship between race and survival, accounting for other explanatory prognostic factors, is presented in Table 3. When race was assessed alone, AA patients had an

During the three successive treatment eras, AA children experienced the greatest relative improvement in outcome.

Multivariate Survival Analysis: White/AA

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

818

POLLOCK ET AL

Fig 2. Kaplan-Meier estimates of EFS are shown by risk group. Survival times are from the date of diagnosis until the date of treatment failure (failure to attain a complete response during induction therapy, relapse, second malignancy, or death) or date of last contact.

80% increased mortality rate compared with white patients (PHR ⫽ 1.80; 95% CI, 1.53 to 2.11). When age, sex, WBC count, and era of treatment were included in the analysis, the association between AA race and poor survival was slightly attenuated but remained strong (PHR ⫽ 1.68; 95% CI, 1.43 to 1.97). None of the interaction terms reached statistical significance at the P ⫽ .10 level. Cytogenetic data were centrally collected for the two most recent treatment protocols. Blast cell ploidy, as measured by DI ⱕ 1.16, was independently and strongly associated with decreased survival (PHR ⫽ 2.72; 95% CI, 1.99 to 3.72). Although the inclusion of ploidy and other covariates further attenuated the PHR for race, there remained a 42% excess mortality rate for AA patients compared with white patients (PHR ⫽ 1.42; 95% CI, 1.12 to 1.80) after adjustment for age, WBC count, DI, sex, and era of treatment. The incremental significance of other explanatory variables was examined by first forcing age, WBC count, leukemia blast cell ploidy, and era of treatment into a regression model and then determining the incremental significance of other single variables (Table 4). For AA patients compared with white patients, deviation from

prescribed protocol treatment was independently associated with survival. So too were the presence of the t(9;22) and t(4;11) translocations as well as the presence of trisomy 21 and simultaneous trisomies 4 and 10, although the occurrence of these cytogenetic abnormalities was rare. In no instance did the inclusion of any one of these variables significantly attenuate the association between race and survival; ie, the lower limit of the 95% CI remained greater than the null value of 1.0 for each model. Multivariate Survival Analysis: White/SS The multivariate comparison of white with SS patients paralleled the comparison of white with AA patients (Table 3). The PHR for SS patients, considered without adjustment for other explanatory variables, was 1.23 (95% CI, 1.13 to 1.34). When prognostic factors other than ploidy were added, the PHR was unchanged (PHR ⫽ 1.26; 95% CI, 1.15 to 1.37). Finally, adding ploidy to the model slightly changed the PHR to 1.33 (95% CI, 1.19 to 1.49). For SS children compared with white children, protocol deviation, t(9;22), t(4;11), and t(1;19) translocations, trisomy 21, and simultaneous trisomies 4 and 10 were independent prognostic factors. Again, none of these factors significantly de-

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

819

RACE DIFFERENCES IN SURVIVAL OF CHILDHOOD ALL

Fig 3. Comparing white and AA children, Kaplan-Meier estimates of overall survival are shown for the three consecutive protocols (POG 8036, POG 8602, and POG 9005/6). Racial differences were significant for each protocol, although the magnitude of the difference decreased for the most recent protocols.

creased the magnitude of association between race and survival. DISCUSSION

In this cohort, AA and SS children with B-precursor ALL presented with more adverse prognostic features at diagnosis compared with white children. AA and SS children had significantly higher mortality than did white children. Even after adjustment for differences in presenting features, the mortality rate was 42% higher for AA children compared with white children and 37% higher for SS children compared with white children. In earlier investigations, racial differences in ALL survival were largely attributed to differences in presenting features.4,5 More modern ALL therapy has increasingly used presenting features for triage for risk-based treatment. Over the past decade within the POG, assignment to a particular regimen has relied on the identification of prognostic factors such as age, WBC count, immunophenotype, blast cell ploidy, and presence of specific cytogenetic markers. In our cohort, none of these prognostic biologic markers, either alone or in combination, explained the strong association between race and outcome. Pui et al2 examined survival differences for white and AA children. For children with ALL treated at St Jude from

1962 to 1983, it was reported that AA patients had a statistically significant 89%-increased mortality compared with white patients, although a major limitation of this study was the lack of adjustment for major biologic risk factors other than age and leukocyte count at diagnosis. This disparity in survival all but disappeared for children treated from 1984 to 1992, for which period the excess mortality rate was 4%. Our study included only children treated on modern antimetabolite-based regimens during an era in which leukemia outcomes had already dramatically improved.20 Our analysis was also restricted to children with B-precursor ALL and, thus, was not confounded by differences in the incidence of immunophenotypes such as T-cell ALL or B-cell ALL for children of different racial groups. We accounted for a full complement of known leukemia prognostic factors as well as measures of protocol compliance. The disparity in survival and EFS that we observed could not be explained by difference in clinical presentation, tumor biology, or measures of compliance. Our results are consistent with a recent report showing inferior survival rates for SS and AA children.21 For POG clinical trials, the survival differences between AA patients and white patients have been dramatically reduced over the three successive eras of treatment. AA patients have gained the greatest incremental improvement

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

820

POLLOCK ET AL

Fig 4. Kaplan-Meier estimates of overall survival are shown for white and SS children over the three consecutive protocols.

in survival. The interpretation of the survival pattern for SS patients is less clear, although the outcome differences were smallest for the most recent protocol. The interpretation of the SS differences are complicated by the heterogeneity of the SS designation, which represents several geographically defined ethnic and cultural subsets for which we were unable to account, eg, Mexican, Puerto Rican, Cuban, South American, and European Spanish populations. Overall, the apparent reduction of race-specific survival differences no doubt reflects greater increases in survival for poor-risk patients. Other plausible reasons, either considered alone or in combination, may account for the differences in outcome. The survival differences might be related to unmeasured or unknown leukemia-cell biologic characteristics or other patient characteristics. However, our present-day classification methods allow us to stratify patients into risk groups having projected cure rates ranging from 20% to greater than 90%. Survival differences may be attributed to systematic differences in how therapy is administered or to differences in patient compliance with the treatment regimen. Prognosis has been closely tied to dose-intensity for drugs like oral mercaptopurine (6-MP)22 and might be affected by minor differences in compliance measured by missed or delayed administration of drugs. Adherence to a treatment regimen

could be influenced by a family’s sociodemographic and economic circumstances. In addition, compliance may be underreported because of the parents’ reluctance to report compliance problems to the treating physician for fear of the involvement of child protective service agencies. Leukemia therapy for the POG trials reported in this analysis was highly standardized, and most of the treatment was administered parenterally. There was no direct evidence suggesting that either institutional treatment bias (assessed by modeling institutional characteristics) or overall protocol noncompliance (assessed biannually by the protocol principal investigator) accounted for the observed survival differences. Nevertheless, undetected or underreported compliance problems may have contributed to the outcome differences we observed. Survival differences may be related to a disproportionate percentage of AA children coming from low socioeconomic status (SES) backgrounds. SES is strongly related to many factors that could affect outcome, including access to health care, supportive care, and nutritional status.23 For the children in the St Jude cohort,2 it had been suggested that improvements in nutritional status in the United States South explained the secular improvement in survival.24 However, Pui et al25 later reported that the proportion of children with subnormal nutritional status did not differ between the two time periods for their study; thus,

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

821

RACE DIFFERENCES IN SURVIVAL OF CHILDHOOD ALL Table 3. Variable

White v AA Univariate* AA v white Multivariate* AA v white WBC ⱖ 50,000 cells/mm3 Age ⬍ 3 years Age ⱖ 6 years Male v female Era‡ (I or II) v III Era‡ I v (II v III) Multivariate§ with ploidy AA v white WBC ⱖ 50,000 cells/mm3 Age ⬍ 3 years Age ⱖ 6 years Male v female Era‡ II v III DI ⬍ 1.16 White v SS Univariate* SS v white Multivariate* SS v white WBC ⱖ 50,000 cells/mm3 Age ⬍ 3 years Age ⱖ 6 years Male v female Era‡ (I or II) v III Era‡ I v (II v III) Multivariate§ SS v white WBC ⱖ 50,000 cells/mm3 Age ⬍ 3 years Age ⱖ 6 years Male v female Era‡ II v III DI ⬍ 1.16

Association of Race and Survival: Cox Proportional Hazards Regression Results P

PHR

95% CI

⬍.001

1.80†

1.53-2.11

⬍.001 ⬍.001 ⬍.001 ⬍.001 .008 ⬍.001 ⬍.001

1.68† 2.37 0.64 2.07 1.18 0.71 0.66

1.43-1.97 2.07-2.71 0.55-0.74 1.79-2.40 1.04-1.33 0.60-0.84 0.57-0.75

.009 ⬍.001 ⬍.001 ⬍.001 .009 .02 ⬍.001

1.42† 2.16 0.59 2.16 1.23 0.77 2.72

1.12-1.80 1.78-2.61 0.47-0.73 1.76-2.65 1.03-1.46 0.64-0.92 1.99-3.72

⬍.001

1.23†

1.13-1.34

⬍.001 ⬍.001 ⬍.001 ⬍.001 .005 .013 .06

1.26† 2.57 0.60 2.14 1.19 0.70 0.72

1.15-1.37 2.24-2.95 0.51-0.69 1.85-2.48 1.05-1.35 0.60-0.83 0.62-0.82

⬍.001 ⬍.001 ⬍.001 ⬍.001 .042 .001 ⬍.001

1.33† 2.32 0.54 2.21 1.20 0.74 2.35

1.19-1.49 1.92-2.79 0.44-0.68 1.81-2.70 1.01-1.42 0.62-0.89 1.76-3.13

No. of Patients in Regression Model

No. of Deaths

4,699

1,067

4,688

1,063

2,922

517

4,681

1,038

4,670

1,034

2,935

525

*For POG 8036, POG 8602, and POG 9005/6. †PHR ⫺ 1 ⫽ the adjusted excess mortality. ‡Era: I ⫽ POG 8036, II ⫽ POG 8602, III ⫽ POG 9005/6. §For POG 8602 and POG 9005/6.

change in nutritional status over time was not a likely explanation. In another report, Mexican-American children with ALL did not have significantly poorer survival rates than did white children.8 Secular changes in access to public assistance programs such as Medicaid, Head Start, food stamps, and better employment opportunities through equal opportunity/affirmative action may have played a role. For adult malignancies, inferior outcome in minority populations has been directly attributed to correlates of SES.26 Adults of lower SES often present with a more

advanced stage disease at diagnosis,27 may have more restricted access to effective therapies,28 and may be less compliant with their therapy. In contrast, these factors are less applicable to children with leukemia. In our cohort, there was almost no difference in lag-time, a measure of delay of diagnosis by race. This variable was collected for all newly diagnosed patients treated on POG frontline protocols. There was also no difference in lag-time over the three treatment eras. Overall, 97% of children achieved a complete remission with induction therapy. Reduced access to effective therapies and low compli-

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

822

POLLOCK ET AL

Table 4.

Prognostic Importance of Adding Individual Factors to a Regression Model That Includes Race, Age, Leukocyte Count, DI, and Era of Treatment Race

Prognostic Factor

AA v white cALLA⫺ Platelet count ⱖ 100 ⫻ 109/L CNS blast count ⱖ 5/mm3 Protocol deviation t(9;22) t(4;11) t(1;19) Trisomy 21 Trisomy 4/10 Trisomy 8 Annual volume, ⬎ 25 patients/yr SS v white cALLA⫺ Platelet count ⱖ 100 ⫻ 109/L CNS blast count ⱖ 5/mm3 Protocol deviation t(9;22) t(4;11) t(1;19) Trisomy 21 Trisomy 4/10 Trisomy 8 Annual volume, ⬎ 25 patients/yr

Prognostic Factor

PHR

95% CI

PHR

95% CI

1.79 2.13 2.07 1.69 1.34 1.51 1.51 1.43 1.47 1.50 1.80

1.51-2.13 1.67-2.71 1.63-2.60 1.26-2.28 1.01-1.80 1.13-2.03 1.12-2.04 1.07-1.92 1.10-1.97 1.12-2.01 1.51-2.13

1.02 0.83 1.36 1.39* 5.41* 2.22* 0.94 0.70* 0.24* 1.20 1.10

0.85-1.22 0.66-1.05 0.93-1.97 1.11-1.72 3.85-7.59 1.15-4.29 0.62-1.42 0.54-0.91 0.13-0.42 0.82-1.75 0.95-1.28

1.28 1.18 1.17 1.56 1.22 1.26 1.25 1.26 1.26 1.24 1.28

1.17-1.41 1.01-1.37 1.01-1.37 1.34-1.81 1.04-1.43 1.07-1.47 1.06-1.46 1.08-1.48 1.07-1.47 1.06-1.46 1.17-1.41

0.98 0.82 1.34 1.26* 6.32* 2.64* 0.77* 0.70* 0.23* 1.06 1.06

0.82-1.18 0.64-1.06 0.89-2.01 1.01-1.59 4.49-8.89 1.44-4.84 0.93-0.58 0.54-0.90 0.13-0.40 0.72-1.57 0.92-1.23

*P ⬍ .05.

ance are less of an issue for children than for adults because, in the case of children, it would be grounds for medical neglect resulting in separation of a child from the parents. Because all children in this cohort were treated on a POG therapeutic protocol, reduced access cannot explain our study results. Variation of outcomes may be related to differences in the response to therapy. Genetic heterogeneity, with race as a proxy, may explain the variation in the response to chemotherapy. Racial differences in drug response may be attributed to functional polymorphisms in the key enzymes that metabolize chemotherapeutic agents used by POG to treat ALL, namely methotrexate (MTX) and 6-MP. Genetic differences in dihydrofolate reductase (DHFR) and thiopurine S-methyltransferase, key enzymes responsible for metabolizing MTX and 6-MP, may covary by race. In patients with T-cell ALL and B-precursor ALL, Matherly et al29 reported that higher DHFR levels were independently associated with higher risk of treatment failure and that expression of DHFR was higher in AA patients compared with white patients. Schmiegelow et al30 reported that low levels of the cytotoxic metabolites of

6-MP and MTX directly correlate with a higher rate of relapse. A small subset of patients from our cohort was enrolled onto POG companion pharmacology studies.31,32 We found that intracellular levels of methotrexate in RBCs were lower for AA and SS patients than they were for white patients over the course of treatment (data are not shown). Even after adjustment for known prognostic factors and measures of compliance, racial differences in survival of childhood B-precursor ALL remain. The magnitude of these differences has decreased with each successive clinical trial, whereas the prevalence of poor prognostic features at the time of diagnosis has remained the same. Differences in outcome may be directly tied to how children of different races metabolize chemotherapeutic agents, specifically MTX and 6-MP. Future studies are needed to determine whether survival is dependent on pharmacogenetics and whether pharmacogenetic differences are related to race and ethnicity. If so, then further improvements in survival for children with ALL, specifically for AA and SS children, may require individualized chemotherapy dosing based on specific patient pharmacogenetic profiles.

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17

823

RACE DIFFERENCES IN SURVIVAL OF CHILDHOOD ALL

REFERENCES 1. Camitta BM, Pullen J, Murphy S: Biology and treatment of acute lymphocytic leukemia in children. Semin Oncol 24:83-91, 1997 2. Pui CH, Boyett JM, Hancock ML, et al: Outcome of treatment for childhood cancer in black as compared with white children: The St Jude Children’s Research Hospital experience, 1962 through 1992. JAMA 273:633-637, 1995 3. Novakovic B: U.S. childhood cancer survival, 1973-1987. Med Pediatr Oncol 23:480-486, 1994 4. Sather H, Honour R, Sposto R, et al: Differences in the presentation of acute lymphoblastic leukemia and non-Hodgkin’s lymphoma in black children and white children: A report from the Epidemiology Committee of the Children’s Cancer Study Group, in Magrath IT, Conor GT, Ramot B (eds): Pathogenesis of Leukemias and Lymphomas: Environmental Influences. New York, NY, Raven Press, 1984, pp 179-189 5. Falletta JM, Boyett J, Pullen DJ, et al: Clinical and phenotypic features of childhood acute lymphocytic leukemia in whites, blacks, and Hispanics: A Pediatric Oncology Group study, in Magrath IT, Conor GT, Ramot B (eds): Pathogenesis of Leukemias and Lymphomas: Environmental Influences. New York, NY, Raven Press, 1984, pp 191-195 6. Kalwinsky DK, Rivera G, Dahl GV, et al: Variation by race in presenting clinical and biologic features of childhood acute lymphoblastic leukaemia: Implications for treatment outcome. Leuk Res 9:817-823, 1985 7. Walters TR, Bushore M, Simone J: Poor prognosis in Negro children with acute lymphocytic leukemia. Cancer 29:210-214, 1972 8. Hord MH, Smith TL, Culbert SJ, et al: Ethnicity and cure rates of Texas children with acute lymphoid leukemia. Cancer 77:563-569, 1996 9. Borowitz MJ, Carroll AJ, Shuster JJ, et al: Use of clinical and laboratory features to define prognostic subgroups in B-precursor acute lymphoblastic leukemia: Experience of the Pediatric Oncology Group. Recent Results Cancer Res 131:257-267, 1993 10. Smith M, Bleyer A, Crist W, et al: Uniform criteria for childhood acute lymphoblastic leukemia risk classification. J Clin Oncol 14:680-681, 1996 (letter) 11. Look A, Roberson P, Williams D, et al: Prognostic importance of blast cell DNA content in childhood acute lymphoblastic leukemia. Blood 65:1079-1086, 1985 12. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958 13. Peto R, Pike MC, Armitage P, et al: Design and analysis of randomized clinical trials requiring prolonged observation of each patient: Analysis and examples. Br J Cancer 35:1-39, 1977 14. Cox DR: Regression models and life tables. J R Stat Soc 34:187-220, 1972 15. Pullen J, Boyett J, Shuster J, et al: Extended triple intrathecal chemotherapy trial for prevention of CNS relapse in good-risk and poor-risk patients with B-progenitor acute lymphoblastic leukemia: A Pediatric Oncology Group study. J Clin Oncol 11:839-849, 1993 16. Land VJ, Shuster JJ, Crist WM, et al: Comparison of two schedules of intermediate-dose methotrexate and cytarabine consolidation therapy for childhood B-precursor cell acute lymphoblastic leukemia: A Pediatric Oncology Group study. J Clin Oncol 12:1939-1945, 1994

17. Camitta B, Mahoney D, Leventhal B, et al: Intensive intravenous methotrexate and mercaptopurine treatment of higher-risk non-T, non-B acute lymphocytic leukemia: A Pediatric Oncology Group study. J Clin Oncol 12:1383-1389, 1994 18. Mahoney DH Jr, Camitta BM, Leventhal BG, et al: Repetitive low dose oral methotrexate and intravenous mercaptopurine treatment for patients with lower risk B-lineage acute lymphoblastic leukemia: A Pediatric Oncology Group pilot study. Cancer 75:2623-2631, 1995 19. Pollock BH: Quality assurance for interventions in clinical trials: Multicenter data monitoring, data management, and analysis. Cancer 74:2647-2652, 1994 20. Vietti J, Shuster J, Amylon M, et al: Progress against childhood cancer: The Pediatric Oncology Group experience. Pediatrics 89:597600, 1992 21. Bhatia S, Sather H, Zhang J, et al: Ethnicity and survival following childhood acute lymphoblastic leukemia (ALL): Follow-up of the Children’s Cancer Group (CCG) cohort. Proc Am Soc Clin Oncol 18:568a, 1999 (abstr 2190) 22. Relling MV, Hancock ML, Boyett JM, et al: Prognostic importance of 6-mercaptopurine dose intensity in acute lymphoblastic leukemia. Blood 93:2817-2823, 1999 23. Viana MB, Murao M, Ramos G, et al: Malnutrition as a prognostic factor in lymphoblastic leukaemia: A multivariate analysis. Arch Dis Child 71:304-310, 1994 (letter) 24. Pinkle D: Acute lymphoid leukemia outcomes in black and white children. JAMA 274:379-380, 1995 25. Pui CH, Hancock ML, Smith KR, et al: Acute lymphoid leukemia outcomes in black and white children. JAMA 274:380, 1995 (letter) 26. Dayal HH, Chiu C: Factors associated with racial differences in survival for prostatic carcinoma. J Chronic Dis 35:553-560, 1982 27. Eley JW, Hill HA, Chen VW, et al: Racial differences in survival from breast cancer: Results of the National Cancer Institute Black/White Cancer Survival Study. JAMA 272:947-954, 1994 28. Bach PB, Cramer LD, Warren JL, et al: Racial differences in the treatment of early-stage lung cancer. N Engl J Med 341:1198-1205, 1999 29. Matherly LH, Taub JW, Wong SC, et al: Increased frequency of expression of elevated dihydrofolate reductase in T-cell versus Bprecursor acute lymphoblastic leukemia in children. Blood 90:578-589, 1997 30. Schmiegelow K, Schroder H, Gustafsson G, et al: Risk of relapse in childhood acute lymphoblastic leukemia is related to RBC methotrexate and mercaptopurine metabolites during maintenance chemotherapy: Nordic Society for Pediatric Hematology and Oncology. J Clin Oncol 13:345-351, 1995 31. Graham ML, Shuster JJ, Kamen BA, et al: Red blood cell methotrexate and folate levels in children with acute lymphoblastic leukemia undergoing therapy: A Pediatric Oncology Group pilot study. Cancer Chemother Pharmacol 31:217-222, 1992 32. Graham ML, Shuster JJ, Kamen BA, et al: Changes in red blood cell methotrexate pharmacology and their impact on outcome when cytarabine is infused with methotrexate in the treatment of acute lymphocytic leukemia in children: A Pediatric Oncology Group study. Clin Cancer Res 2:331-337, 1996

Information downloaded from jco.ascopubs.org and provided by at VANDERBILT UNIVERSITY on July 21, 2014 from Copyright © 2000 American Society of Clinical Oncology. All rights reserved. 129.59.115.17