Abstract
This quality improvement study examines disparities in proton radiotherapy use in patients enrolled in Children’s Oncology Group Trials
Proton radiotherapy is a high-cost, limited resource, for which the most robust evidence-based indications are in pediatric cancers owing to its ability to reduce long-term adverse effects compared with photon therapy.1,2,3 Most proton radiotherapy facilities are located in metropolitan areas or at major academic centers, and therefore require physician referral to the center for treatment. In addition, radiotherapy involves daily treatment for up to 8 weeks, requiring patients to potentially relocate to obtain treatment. Owing to the potential barriers to proton radiotherapy access, we aimed to assess the association of race and socioeconomic factors with proton use in patients with solid malignant diseases enrolled on Children’s Oncology Group (COG) prospective trials.
Methods
Data were provided by the COG. Patients were included if they received external beam radiotherapy on COG solid tumor prospective trials that allowed physicians to choose the radiation modality (eg, proton or photon radiotherapy) and were closed to accrual as of May 2018. Patients treated after 2010 were included. Patients in trials that only allowed emergency palliative radiotherapy or total body irradiation, not from the United States, or with missing radiotherapy modality were excluded. The race/ethnicity data were generally provided by the clinician or clinical research coordinator. Race was categorized as White non-Hispanic, White Hispanic, Black, Other, and unknown. The Dana-Farber/Harvard Cancer Center institutional review board approved this study.
Univariate and multivariable logistic regression models identified factors associated with radiation modality. P < .05 was considered statistically significant. Analyses were conducted using SAS statistical software (version 9.4, SAS Institute). The analysis was performed between July and December of 2019.
Results
The final cohort included 1240 patients. From 2010 to 2018, 1054 of 1240 (85%) and 186 of 1240 (15%) patients received photon and proton radiotherapy, respectively. Table 1 shows baseline characteristics by radiotherapy modality.4 In univariate analysis, Black pediatric patients were less likely to receive proton radiotherapy than non-Hispanic White patients (OR, 0.4; 95% CI, 0.22-0.75; P = .004) (Table 2).4 Proton radiotherapy use varied by enrolling institution location and was less frequent among patients enrolled on non-CNS vs CNS trials, patients with shorter distance between enrolling and radiotherapy institution, and patients with metastatic disease (Table 2). Insurance type, neighborhood-poverty level, and gender were not associated with radiation modality.
Table 1. Patient Characteristics by Radiation Modality.
| Variable | No. (%) | ||
|---|---|---|---|
| Overall cohort (n = 1240) | Photon therapy (n = 1054) | Proton therapy (n = 186) | |
| Protocol (disease) | |||
| ACNS1123 (germ cell tumor) | 61 (5) | 43 (4) | 18 (10) |
| ACNS0221 (low-grade glioma) | 21 (2) | 21 (2) | 0 (0) |
| ACNS0331 (medulloblastoma) | 112 (9) | 99 (9) | 13 (7) |
| ACNS0332 (medulloblastoma/PNET) | 199 (16) | 158 (15) | 41 (22) |
| ACNS0333 (AT/RT) | 35 (3) | 24 (2) | 11 (6) |
| ANBL12P1 (neuroblastoma) | 82 (7) | 72 (7) | 10 (5) |
| ANBL09P1 (neuroblastoma) | 30 (2) | 23 (2) | 7 (4) |
| ARST08P1 (rhabdomyosarcoma) | 115 (9) | 113 (11) | 2 (1) |
| ARST0921 (rhabdomyosarcoma) | 6 (0.5) | 5 (0) | 1 (1) |
| ARST0332 (soft-tissue sarcoma) | 121 (10) | 116 (11) | 5 (3) |
| ARST0531 (rhabdomyosarcoma) | 186 (15) | 160 (15) | 26 (14) |
| ARST0331 (rhabdomyosarcoma) | 22 (2) | 20 (2) | 2 (1) |
| ARET0321 (retinoblastoma) | 11 (1) | 10 (1) | 1 (1) |
| AEWS1031 (Ewing sarcoma) | 239 (19) | 190 (18) | 49 (26) |
| CNS trial | |||
| No | 812 (65) | 709 (67) | 103 (55) |
| Yes | 428 (35) | 345 (33) | 83 (45) |
| Sex | |||
| Male | 736 (59) | 623 (59) | 113 (61) |
| Female | 504 (41) | 431 (41) | 73 (39) |
| Race | |||
| White | |||
| Non-Hispanic | 769 (62) | 641 (68) | 128 (74) |
| Hispanic | 140 (11) | 115 (12) | 25 (15) |
| Black | 161 (13) | 149 (16) | 12 (7) |
| Other | 44 (4) | 37 (4) | 7 (4) |
| Unknown | 126 (10) | 112 | 14 |
| Insurance type | |||
| Limited-income insurancea | 391 (32) | 341 (35) | 50 (30) |
| Other insuranceb | 749 (60) | 633 (65) | 116 (70) |
| Unknown | 100 | 80 | 20 |
| Population-level poverty ratec | |||
| ≥20% Below poverty level | 217 (18) | 182 (17) | 35 (19) |
| <20% Below poverty level | 1022 (82) | 871 (83) | 151 (81) |
| Unknown | 1 | 1 | 0 |
| Distance between enrolling institution and radiotherapy facility, miles | |||
| <25 | 1091 (88) | 1000 (95) | 91 (49) |
| ≥25 | 149 (12) | 54 (5) | 95 (51) |
| Metastatic disease | |||
| Yes | 250 (20) | 234 (22) | 16 (9) |
| No | 990 (80) | 820 (78) | 170 (91) |
Abbreviations: AT/RT, atypical teratoid/rhabdoid tumors; CNS, central nervous system; PNET, primitive neuroectodermal tumor.
Limited-income insurance includes Medicaid, California Children's Services, and self-pay/uninsured.
Other insurance includes Medicare, Medicaid plus private insurance, Medicaid plus Medicare, private insurance, and military-sponsored insurance.
Calculated from the first 3 digits of each patient’s 5-digit zip code using the weighted average of 5-digit zip code poverty rate from 2015 American Community Survey.4
Table 2. Univariate and Multivariable Logistic Regression Models for Receipt of Proton Radiotherapy.
| Adjustment variables | No. for univariate analysis | Univariate models | Multivariable model (n = 1114)a | ||
|---|---|---|---|---|---|
| OR (95% CI)b | P value | OR (95% CI)b | P value | ||
| Race | 1114 | ||||
| White | |||||
| Non-Hispanic | 1 [Reference] | NA | 1 [Reference] | NA | |
| Hispanic | 1.09 (0.68-1.75) | .72 | 1.56 (0.89-2.73) | .12 | |
| Black | 0.40 (0.22-0.75) | .004 | 0.35 (0.17-0.72) | .004 | |
| Other | 0.95 (0.41-2.17) | .90 | 0.65 (0.24-1.81) | .41 | |
| CNS trial | 1240 | ||||
| No | 1 [Reference] | NA | NA | NA | |
| Yes | 1.66 (1.21-2.27) | .002 | NS | NS | |
| Insurance type | 1140 | ||||
| Other insurancec | 1 [Reference] | NA | NA | NA | |
| Limited-income insuranced | 0.80 (0.56-1.14) | .22 | NS | NS | |
| Population-level poverty ratee | 1239 | ||||
| <20% Below poverty level | 1 [Reference] | NA | NA | NA | |
| ≥20% Below poverty level | 1.11 (0.74-1.66) | .61 | NS | NS | |
| Metastatic disease | 1240 | ||||
| No | 1 [Reference] | NA | 1 [Reference] | NA | |
| Yes | 0.33 (0.19-0.56) | <.001 | 0.44 (0.24-0.80) | .007 | |
| Enrolling institution location | 1240 | ||||
| Middle Atlantic | 1 [Reference] | NA | 1 [Reference] | NA | |
| East North Central | 0.82 (0.48-1.39) | .45 | 0.45 (0.23-0.91) | .03 | |
| East South Central | 0.26 (0.11-0.60) | .002 | 0.32 (0.12-0.83) | .02 | |
| Mountain | 0.16 (0.06-0.44) | <.001 | 0.16 (0.05-0.49) | .002 | |
| New England | 0.79 (0.39-1.62) | .52 | 1.11 (0.48-2.60) | .80 | |
| Pacific | 0.31 (0.16-0.59) | <.001 | 0.28 (0.12-0.61) | .002 | |
| South Atlantic | 0.55 (0.32-0.95) | .03 | 0.52 (0.26-1.04) | .07 | |
| West North Central | 0.53 (0.28-1.03) | .06 | 0.47 (0.21-1.06) | .07 | |
| West South Central | 0.78 (0.41-1.48) | .45 | 0.93 (0.43-2.03) | .86 | |
| Sex | 1240 | ||||
| Male | 1 [Reference] | NA | NA | NA | |
| Female | 0.93 (0.68-1.28) | .67 | NS | NS | |
| Distance between enrolling institution and radiotherapy facility, miles | 1240 | ||||
| ≥25 | 1 [Reference] | NA | 1 [Reference] | NA | |
| <25 | 0.05 (0.04-0.08) | <.001 | 0.05 (0.03-0.08) | <.001 | |
Abbreviations: CNS, central nervous system; NA, not applicable; NS, not significant; OR, odds ratio.
All factors were tested in the multivariable model using backward selection, and factors significant at P < .05 were retained.
ORs were calculated based on the odds of having received proton therapy.
Other insurance includes Medicare, Medicaid plus private insurance, Medicaid plus Medicare, private insurance, and military-sponsored insurance.
Limited-income insurance includes Medicaid, California Children's Services, and self-pay/uninsured.
Calculated from the first 3 digits of each patient’s 5-digit zip code using the weighted average of 5-digit zip code poverty rate from 2015 American Community Survey.4
All factors significant on univariate analysis were significant on multivariable analysis, except CNS trial enrollment (Table 2). After adjusting for enrolling institution location, distance to radiotherapy facility, and presence of metastatic disease, Black patients remained less likely to receive proton radiotherapy compared with non-Hispanic White patients (OR, 0.35; 95% CI, 0.17-0.72; P = .004) (Table 2). Of note, 2-way interaction terms between covariates were tested in the multivariable analysis, and none were statistically significant.
Discussion
Black pediatric patients were less likely to receive proton radiotherapy, despite being enrolled on national prospective trials in which treatment is highly standardized. Measures of socioeconomic status, including household-level (insurance) and area-level poverty, were not associated with proton radiotherapy. In addition, location of enrolling institution was associated with proton radiotherapy and patients treated with proton radiotherapy traveled further for treatment, possibly reflecting sparse, unequal distribution of proton radiotherapy centers. Notably, the number of proton radiotherapy centers rose from 9 to 28 during the study period. We controlled for primary site (CNS vs non-CNS), but not for malignant disease (including cancers with differences in incidence by race, eg, Ewing sarcoma), due to concern for overfitting.
Limitations of this study include indirect measurements of socioeconomic status, lack of granular geographic information, and limited power owing to the relatively fewer patients who received proton radiotherapy. Although prior epidemiologic studies have shown income and insurance status, but not race, to be associated with proton radiotherapy use, these included patients from an earlier era when few proton centers were open.5,6 The reported racial disparities are concerning, particularly as this cohort represents a selected population of patients treated in prospective national clinical trials. Our results suggest that a racial disparity in proton radiotherapy use may exist; additional confirmatory investigation is needed.
References
- 1.Yock TI, Bhat S, Szymonifka J, et al. Quality of life outcomes in proton and photon treated pediatric brain tumor survivors. Radiother Oncol. 2014;113(1):89-94. doi: 10.1016/j.radonc.2014.08.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Ladra MM, Szymonifka JD, Mahajan A, et al. Preliminary results of a phase II trial of proton radiotherapy for pediatric rhabdomyosarcoma. J Clin Oncol. 2014;32(33):3762-3770. doi: 10.1200/JCO.2014.56.1548 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kahalley LS, Peterson R, Ris MD, et al. Superior intellectual outcomes after proton radiotherapy compared with photon radiotherapy for pediatric medulloblastoma. J Clin Oncol. 2019:Jco1901706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.American Community Survey (ACS) 2015; https://www.census.gov/programs-surveys/acs. Accessed December 27, 2019.
- 5.Odei B, Frandsen JE, Boothe D, Ermoian RP, Poppe MM. Patterns of care in proton radiation therapy for pediatric central nervous system malignancies. Int J Radiat Oncol Biol Phys. 2017;97(1):60-63. doi: 10.1016/j.ijrobp.2016.09.011 [DOI] [PubMed] [Google Scholar]
- 6.Shen CJ, Hu C, Ladra MM, Narang AK, Pollack CE, Terezakis SA. Socioeconomic factors affect the selection of proton radiation therapy for children. Cancer. 2017;123(20):4048-4056. doi: 10.1002/cncr.30849 [DOI] [PubMed] [Google Scholar]