Association of clinical trial enrollment and survival using contemporary therapy for pediatric acute lymphoblastic leukemia

Abstract

While early studies reported superior survival for cancer patients enrolled on clinical trials, recent findings are inconclusive. We investigated the association between enrollment on contemporary trials and event-free survival (EFS) in pediatric B-cell acute lymphoblastic leukemia (B-ALL). In a retrospective cohort of 274 children (1-21 years) treated for B-ALL from 2008 to 2015, 55.5% enrolled with no disparity in enrollment by age, sex, or ethnicity. Three-year EFS was similar for enrolled and not enrolled patients (90.1% [95% CI, 82.5-94.5] versus 86.5% [95% CI, 77.7-92.0]). Clinical trial enrollment did not affect pediatric B-ALL survival, albeit in a limited-size cohort treated at a single academic institution.

1 |. INTRODUCTION

A series of population-based studies from 1970s to 1990s reported that children with leukemia enrolled on a clinical trial experienced greater survival than children not enrolled on a trial.^1–3 Advances in treatment for childhood leukemia have substantially improved survival. Control arms on recently completed pediatric B-cell acute lymphoblastic leukemia (B-ALL) trials from the Children’s Oncology Group (COG) and other consortia report 5-year event-free survival (EFS) of 75–94% compared to 19–60% in earlier decades.^4–7 Given these improvements, the continued association of trial enrollment with survival, especially in the era of minimal residual disease (MRD) based risk-stratified therapy, is unclear. Even with these treatment advances, survival for certain demographic subgroups lag behind, including adolescents and young adults (AYAs, aged ≥15–39 years⁸) and Hispanic patients.^9,10 These subgroups are underrepresented on trials,¹¹ a proposed contributing factor to their poorer survival.⁹ However, no reports have investigated whether enrollment influences survival for these subgroups. We hypothesized that in the contemporary treatment era of MRD risk-stratified B-ALL therapy, trial enrollment is not associated with EFS or overall survival (OS) and that poorer survival among AYAs and Hispanic patients is not due to lower enrollment.

2 |. RESULTS

We analyzed survival of a retrospective cohort of patients 1–21 years old with Philadelphia chromosome negative B-ALL, diagnosed between 2008 and 2015 at a single academic center, with available diagnostic flow cytometry. All flow cytometry data were independently reviewed (M.O.) to ensure a uniform cohort of confirmed B-ALL. During the assessment period, trials included AALL0331, AALL0932, AALL0232, and AALL1131. Patients not enrolled on study were treated on institutional protocols (not at the discretion of physicians) with modified regimens based on CCG1991, CCG1961, or AALL0232 (all studies available at https://clinicaltrials.gov, Supplementary Table S1 for treatment descriptions). Demographic, disease, and treatment data were extracted (Table 1). Endpoints included trial enrollment (yes/no), EFS (relapse, second malignancy, or death), and OS. Patients receiving hematopoietic cell transplant were censored at the time of transplant. Following standard analyses of distribution between groups, a multivariable logistic regression model was constructed using candidate predictors (Table 1) with trial enrollment as the endpoint. For survival analyses, Kaplan–Meier curves with associated log-rank tests were constructed. Following confirmation of the proportional hazards assumption via Schoenfeld residuals, multivariable Cox proportional hazard models were constructed incorporating enrollment status, MRD, and other known predictors of EFS/OS. The study was approved by the Institutional Review Board. All statistical tests were two-sided, with significance set at P ≤ 0.05 and performed with STATA software 14.2 (Stata statistical software, release 14; College Station, TX).

TABLE 1.

Description of cohort by trial enrollment status

	Overall		Not enrolled		Enrolled
	n	(%)	n	(%)	n	(%)	Pa
Total cohort	274	100	122	45.5	152	55.5	-
Age (mean, median [range])	7.4, 5.2 [1.2–20.0]		7.9, 5.6 [1.6–20.0]		6.9, 4.8 [1.2–19.0]		0.11
Age <15	238	87.0	103	84.4	135	88.8	0.26
Age ≥15	36	13.1	19	15.6	17	11.2
Sex
Female	135	49.3	61	50.0	74	48.7	0.83
Male	139	50.7	61	50.0	78	51.3
Ethnicity
Not-Hispanic	78	28.5	28	23.0	50	32.9	0.07
Hispanic	196	71.5	94	77.0	102	67.1
Down syndrome
Not Down syndrome	267	97.4	119	97.5	148	97.4	1.00
Down syndrome	7	2.6	3	2.5	4	2.6
NCI/Rome risk
SR	167	60.9	74	60.7	93	61.2	0.93
HR	107	39.1	48	39.3	59	38.8
Presenting WBC (max)
WBC (mean, median [range])	39.1, 9.0 [0.3–771.8]		36.3, 8.3 [0.3–771.8]		41.2, 11.5 [0.8–562]		0.43
WBC <50 K/μl	222	81.0	100	82.0	122	80.3	0.72
WBC ≥50 K/μl	52	19.0	22	18.0	30	19.7
CNS status
CNS1	244	89.1	108	88.5	136	89.5	0.38
CNS2	28	10.2	14	11.5	14	9.2
CNS3	2	0.7	0	0	2	1.3
Cytogenetic Categoryb
Average	157	57.3	79	64.8	78	51.3	0.02
Favorable	113	41.2	43	35.2	70	46.1
Unfavorable	4	1.5	0	0	4	2.6
EOI MRD
<0.01%	165	60.2	66	54.1	99	65.1	0.15
≥0.01%	40	14.6	22	18.0	18	11.8
Unknown	69	25.2	34	27.9	35	22.7
For MRD+ (median, [range])	0.26 [0.012–36]		0.22 [0.026–36]		0.30 [0.012–22.4]		0.91

NCI, National Cancer Institute; WBC, white blood cell; CNS, central nervous system; EOI MRD, end of induction minimal residual disease (bone marrow).

^aPearson chi-square, Fisher exact test, or K-S test; significance set at P < 0.05 (bolded).

^bFavorable includes “double trisomy” 4+10 (n = 45) and ETV6-RUNX1 (n = 69). Unfavorable includes hypodiploid (<44 chromosomes, n = 1), rMLL (n = 3), iAMP21 (n = 0).

Of the 274 patients in the cohort, 152 (55.5%) enrolled on a trial. We found no significant differences between enrolled and not enrolled patients, except the cytogenetic classification (Table 1). In multivariable analysis, no candidate predictors, including AYA and Hispanic ethnicity, were significantly associated with enrollment (AYA odds ratio [OR] = 0.67, [95% confidence interval [95% CI] 0.29–1.54], P = 0.35; Hispanic OR = 0.61, [95% CI 0.35-1.06], P = 0.08). At a median follow-up of 3.1 years, trial enrollment was not associated with EFS (Fig. 1A, 3-year EFS enrolled 90.1% [95% CI 82.5-94.5] vs. not enrolled 86.5% [95% CI 77.7-92.0%]) or OS (Fig. 1B, 3-year OS enrolled 95.8% [95% CI 89.9-98.3] vs. not enrolled OS 96.2% [95% CI 90.1-98.6%]). In multivariable analysis of EFS, enrollment was not significantly associated with EFS (hazard ratio [HR] = 0.61, [95% CI 0.29-1.30], P = 0.20) (Supplementary Table S1). In models inclusive of AYA, ethnicity, and enrollment, only AYA was associated with poorer EFS (Supplementary Table S2).

FIGURE 1.

Kaplan–Meier survival curves of event-free survival (A) and overall survival (B) for not enrolled (blue solid line) and enrolled (red dashed line) patients. LRT, log-rank test P-value

3 |. DISCUSSION

While prior studies from the 1970s to 1990s showed superior survival for pediatric leukemia patients enrolled on clinical trials,^1–3 recent studies have contradicted earlier results.^12–14 The current study supports the absence of an association between trial enrollment and survival for pediatric ALL in the modern MRD risk-stratified treatment era and in the context of the most recent generation of trials. In this cohort, 55.5% of patients enrolled on trial, similar to 55.8% of pediatric ALL patients nationwide enrolled onto COG studies.⁴ In contrast to previous studies,¹¹ we report similar trial enrollment among at-risk AYA and Hispanic subgroups. Our findings are the first to suggest that lower enrollment rates on trials do not contribute to poorer survival for AYA and Hispanic patients.

Several reasons could account for the differences between our and previously reported results. First, high survival rates with contemporary regimens limit the interventional arms’ potential effect sizes, making it less likely to observe a survival benefit. Second, the relationship between enrollment and survival in prior studies may be confounded by disproportionately greater enrollment at large cancer centers, potentially affecting delivery of care. Third, earlier studies may have found a survival benefit because of selection biases, with healthier patients meeting eligibility criteria for trial enrollment.¹⁴

The study’s principal strength is the investigation of a cohort of patients with a uniform diagnosis treated at a single institution, an approach that minimizes institutional and disease-specific confounders. The duration of follow-up encompasses a time frame within which most events in pediatric B-ALL have been shown to occur.¹⁵ Our findings are consistent with the lack of association between survival and enrollment reported in a previous study¹³ that largely included patients from an era predating the use of MRD for risk stratification. However, in contrast to previous reports, our study is unique in its incorporation of MRD in predictive models, its inclusion of the most recent clinical trials using MRD-based risk stratification, and its ethnic diversity allowing for a sizable Hispanic representation.

As a retrospective cohort, the study is limited by incomplete MRD results (no differences were detected in unknown MRD between enrolled and not enrolled groups, Table 1), a lower than expected representation of AYA patients and those with adverse cytogenetics, and no available documentation of reasons for patient nonenrollment. In addition, the sample size and follow-up time were not sufficient to detect small differences in rates of events. Nevertheless, the association between enrollment and survival for children with B-ALL with a focus on at-risk AYA and Hispanic subgroups is presented here for the first time. Given that this is a single academic institution study, caution is warranted in generalizing these results, especially in situations where MRD is not available as standard of care for patients not enrolled on a study.

In summary, we found that trial enrollment is not itself associated with improved survival using recent MRD risk-stratified approaches to pediatric B-ALL therapy. However, while enrollment itself might not confer a survival benefit, trial participation is critically important for continued improvement in therapy for childhood ALL. Our findings are thus highly relevant to inform provider discussions of trials with patients and families and to aid in the interpretation of the applicability of trial data to the general population. Additional studies are necessary to better understand barriers to enrollment and to prospectively validate these findings in diverse, multicenter cohorts of sufficient size to detect subtle differences in survival.

Abbreviations:

AYA

adolescent and young adult

B-ALL

B-cell acute lymphoblastic leukemia

COG

Children’s Oncology Group

EFS

event-free survival

MRD

minimal residual disease

NCI

National Cancer Institute

OS

overall survival