Adjuvant chemotherapy and overall survival in adult medulloblastoma

Abstract

Background

Although chemotherapy is used routinely in pediatric medulloblastoma (MB) patients, its benefit for adult MB is unclear. We evaluated the survival impact of adjuvant chemotherapy in adult MB.

Methods

Using the National Cancer Data Base, we identified patients aged 18 years and older who were diagnosed with MB in 2004–2012 and underwent surgical resection and adjuvant craniospinal irradiation (CSI). Patients were divided into those who received adjuvant CSI and chemotherapy (CRT) or CSI alone (RT). Predictors of CRT compared with RT were evaluated with univariable and multivariable logistic regression. Survival analysis was limited to patients receiving CSI doses between 23 and 36 Gy. Overall survival (OS) was evaluated using the Kaplan-Meier estimator, log-rank test, multivariable Cox proportional hazards modeling, and propensity score matching.

Results

Of the 751 patients included, 520 (69.2%) received CRT, and 231 (30.8%) received RT. With median follow-up of 5.0 years, estimated 5-year OS was superior in patients receiving CRT versus RT (86.1% vs 71.6%, P < .0001). On multivariable analysis, after controlling for risk factors, CRT was associated with superior OS compared with RT (HR: 0.53; 95%CI: 0.32–0.88, P = .01). On planned subgroup analyses, the 5 year OS of patients receiving CRT versus RT was improved for M0 patients (P < .0001), for patients receiving 36 Gy CSI (P = .0007), and for M0 patients receiving 36 Gy CSI (P = .0008).

Conclusions

This national database analysis demonstrates that combined postoperative chemotherapy and radiotherapy are associated with superior survival for adult MB compared with radiotherapy alone, even for M0 patients who receive high-dose CSI.

Medulloblastoma (MB) is the most common brain tumor in children. However, it is relatively rare in adults, with an estimated incidence of 0.6 per million.¹ The standard-of-care management for pediatric MB is postoperative radiotherapy (RT) with craniospinal irradiation (CSI) and posterior fossa or resection bed boost followed by adjuvant chemotherapy.^2,3 Adoption of adjuvant chemotherapy in the pediatric setting has been associated with improved disease control and has allowed successful CSI dose de-escalation in average-risk patients.^2–5

Even though chemotherapy is used routinely for pediatric patients, its benefit in adult MB is unclear. Data supporting adjuvant chemotherapy in the adult MB population are scarce. There are no randomized trials investigating the benefit of chemotherapy in adult MB, and evidence is generally limited to small retrospective series reported over several decades with conflicting results.^6–9 Furthermore, compared with children, adults may suffer less toxicity from CSI and greater morbidity from chemotherapy. This may lead to hesitancy in using postoperative chemotherapy, particularly in lower-risk patients, who are traditionally defined as those with M0 disease and near-total tumor resection^3,10 and those receiving high-dose CSI.⁸ In the absence of randomized trials for this rare disease, we used a large national database to evaluate the survival impact of adjuvant chemotherapy for adult MB, particularly in the setting of localized disease and high-dose CSI.

Importance of the Study

Evidence for appropriate postoperative management of adult medulloblastoma (MB) is limited. Use of adjuvant chemotherapy is routine in pediatric MB, although its benefit remains uncertain in adult MB. There are no randomized trials investigating adjuvant chemotherapy in adult MB, and data are generally limited to small retrospective series. We used a national database to analyze the survival benefit of adjuvant chemotherapy for MB in more than 400 adult patients. After adjustment for relevant demographic and clinical factors, this study found that the addition of adjuvant chemotherapy to craniospinal irradiation was associated with superior overall survival for adult MB. The survival benefit persisted in subgroup analyses of patients with M0 disease and those receiving high-dose craniospinal irradiation. In the absence of randomized data for this rare disease, our study supports the addition of adjuvant chemotherapy to craniospinal irradiation for the treatment of adult MB in patients who can tolerate multimodality treatment.

Materials and Methods

Data Source and Study Population

We conducted an analysis using the National Cancer Data Base (NCDB) registry.¹¹ The NCDB currently captures approximately 70% of all newly diagnosed malignancies in the United States annually from facilities accredited by the Commission on Cancer. The data used in this study are derived from a de-identified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are neither responsible for the analytical or statistical methodology used nor the conclusions drawn from these data by the investigators. This study was granted an exemption from the Yale Human Investigations Committee.

Inclusion criteria were patients aged 18 years and older who were diagnosed between 2004 and 2012 with histologically confirmed MB without prior cancer diagnosis. We limited our sample to patients who underwent surgical resection, received adjuvant CSI, and did not have extra-neuraxial metastases (M4 disease) in order to avoid ascertainment bias from noncoded adjuvant treatment data and to isolate patients who were being treated with curative intent. We excluded patients who died or were lost to follow-up within 2 months of diagnosis to account for immortal time bias.¹² We divided the remaining sample into cohorts receiving adjuvant radiotherapy and chemotherapy (CRT) or adjuvant radiotherapy (RT) alone. The flowchart summarizing patient inclusion and exclusion can be found in Fig. 1.

Fig. 1.

Consolidated standards of reporting trials diagram. CSI = craniospinal irradiation.

Statistical Analysis

Patient characteristics and facility covariates included age (continuous), year of diagnosis (continuous), sex (male vs female), Charlson Deyo Score (CDS) (0 vs 1 or 2), histologic subtype (desmoplastic vs large cell vs classic/NOS), Chang modified M stage (M0 vs M1-3), tumor size (<4 cm vs ≥4 cm), extent of resection (no macroscopic residual vs macroscopic residual vs residual unknown), race (white vs other), facility volume (“high-volume” vs “low-volume”), insurance (private vs nonprivate), distance from facility (<20 miles vs ≥20 miles), time to radiotherapy initiation from diagnosis (TTR) (days, continuous), and CSI dose (“low dose” [23 to <30 Gy] vs “high dose” [30–36 Gy]). A high-volume facility was defined as a facility that treated at least 8 cases of MB during the study time period. Facilities that treated fewer than 8 cases were defined as low volume. Eight cases were chosen a priori as a threshold to approximate an average of approximately 1 case per year. Chemotherapy treatment details, including chemotherapy sequence as well as use of multiagent vs single-agent therapy, were examined in the CRT group. Chemotherapy was designated as concurrent (CCRT) if start time occurred within 14 days of RT start or sequential (SCRT) otherwise.

Demographic, clinicopathologic, temporal, and treatment details were compared between patients in the RT and CRT cohorts using chi-square tests and Wilcoxon rank-sum tests. Multivariable logistic regression models using fixed effects and mixed effects adjusting for facility were constructed to determine predictors of CRT compared with RT with associated odds ratios (ORs). An all-inclusive logistic regression model was constructed to include all variables captured. Models were evaluated for goodness of fit with the Hosmer-Lemeshow test.

Overall survival (OS) was evaluated using the time from diagnosis until death after excluding patients who did not have documented CSI doses ranging from 23 to 36 Gy (Fig. 1). The Kaplan-Meier estimator with the log-rank test was used to compare OS between the RT and CRT cohorts. Multivariable Cox proportional hazards models were used to calculate adjusted hazard ratios (HRs) and their 95% confidence intervals (CIs). Three model versions were constructed. The first was all-inclusive of variables analyzed in the study. The second and third models were constructed using forward and backward stepwise selection, respectively, with entry threshold probability P < .05 and removal threshold probability P < .1. Schoenfeld residuals were calculated for each model to test if the proportional hazards assumption had been violated. Among the CRT cohort, we compared survival by chemotherapy characteristics and CSI dose.

Propensity score matching was performed with one-to-one nearest-neighbor matching without replacement, with bootstrapping to identify matched cohorts representing the 2 treatment modalities. Matching was performed by using both variables found to be independent predictors of OS on multivariable analysis and those representing traditional risk factors. Age was dichotomized into categories for matching purposes using the median age of the population. Covariate balance was evaluated by using standardized differences of means.¹³ Cox proportional hazards regression models were used to determine significant contributors to OS in the matched cohort, adjusting for propensity quintile.¹⁴

We conducted a sensitivity analysis for survival, expanding the analysis to patients with documented receipt of CSI regardless of dose. A second sensitivity analysis was performed using propensity score-matched pairs. We conducted additional sensitivity analyses testing varying thresholds for high-volume facility from 6 to 12 total patients treated. Finally, we conducted subgroup survival analyses of M0 patients, patients who received 36 Gy CSI and patients who were both M0 and received 36 Gy CSI.

All tests were 2-sided, and a P value <.05 was considered statistically significant. All analyses were performed using Stata SE version 13.

Results

Patient Characteristics and Treatment Patterns

We identified 751 patients who met our inclusion criteria (Fig. 1). Patients were treated at 342 unique facilities, among which 14 (4.1%) were high-volume. The maximum number of cases treated at one facility over the evaluated time period was 12, with high-volume facilities treating 123 patients (16.4%). The median age of the entire cohort was 29 years. Clinical and demographic characteristics are found in Table 1.

Table 1.

Demographic, clinical, and pathologic characteristics for patients receiving adjuvant craniospinal irradiation (RT) and those receiving CSI and chemotherapy (CRT)

	Total (n = 751)	RT (n = 231)	%	CRT (n = 520)	%	P Value
Age (y), Median	29	33		27		<.001
Range	18–85	18–85		18–78
Year of Diagnosis						.002
2004–2006	231	90	39.0%	141	61.0%
2007–2009	270	81	30.0%	189	70.0%
2010–2012	250	60	24.0%	190	76.0%
Sex						.44
Male	446	142	31.8%	304	68.2%
Female	305	89	29.2%	216	70.8%
Charlson-Deyo Score
0	669	210	31.4%	459	68.6%	.28
1–2	82	21	25.6%	61	74.4%
Histology						.95
Classic / NOS	615	188	30.6%	427	69.4%
Desmoplastic	112	35	31.3%	77	68.7%
Large cell	24	8	33.3%	16	66.7%
M stage						.02
M0	660	211	32.0%	449	68.0%
M1-3	51	7	13.7%	44	86.3%
Unknown M stage	40	13	30.8%	27	69.2%
Tumor size						.53
<4 cm	277	89	32.1%	188	67.9%
≥4 cm	474	142	30.0%	332	70.0%
Extent of resection						.006
No macroscopic residual	122	26	21.3%	96	78.7%
Macroscopic residual	50	10	20.0%	40	80.0%
Unknown residual	579	195	33.7%	384	66.3%
Race						.13
White	623	186	29.9%	437	70.1%
Other	122	45	36.9%	77	63.1%
Unknown Race	6	0	0%	6	100.0%
HVF						.03
High volume (≥8 total cases)	123	48	39.0%	75	61.0%
Low volume (<8 total cases)	628	183	29.1%	445	70.9%
Insurance
Private	465	138	29.7%	327	70.3%	.41
Nonprivate	286	93	32.5%	193	67.5%
Distance from facility (miles)						.02
<20	453	125	27.6%	328	72.4%
≥20	298	106	35.6%	192	64.4%
CSI dose						<.001
23 to <30 Gy	121	15	12.4%	106	87.6%
30–36 Gy	356	129	36.2%	227	63.8%
Other/unknown dose	274	87	31.8%	187	68.3%
Time-to-RTa (wk), Median	5.9	5.9		6.0		.38
Chemotherapy sequence
Concurrent	–	–	–	249	47.9%
Sequential	–	–	–	271	52.1%
Chemotherapy agents
Single-agent	–	–	–	136	26.2%
Multiagent	–	–	–	341	65.6%
Unknown no. of agents	–	–	–	43	8.3%

Abbreviations: CSI, craniospinal irradiation; no., number; NOS, not otherwise specified; RT, radiotherapy; wk, week(s); y, year(s).

^aRepresents time in weeks from surgery to initiation of radiotherapy.

Overall, 231 patients (30.8%) had RT alone, and 520 patients (69.2%) had CRT. Of the 477 patients with documented CSI doses between 23 and 36 Gy, 356 (74.6%) received CSI ≥30 Gy. Six-hundred sixty patients (87.9%) had M0 disease. On univariable analysis, variables associated with CRT compared with RT use were younger median age (27 vs 33 y, P < .001), later year of diagnosis (OR: 1.10/y; 95% CI: 1.03–1.17), M1-3 disease versus M0 disease (86.3% vs 68.0%, P = .02), unknown extent of resection versus macroscopic residual disease versus no macroscopic residual disease (66.3% vs 80.0% vs 78.7%, P = .006), low-volume versus high-volume facility (70.9% vs 61.0%, P = .02), distance <20 miles versus ≥20 miles from facility (72.4% vs 64.4%, P = .02), and low-dose versus high-dose CSI (87.6% vs 63.8%, P = <.001) (Table 1).

Multivariable Analysis of CRT Predictors

On multivariable, multilevel mixed-effects logistic regression, independent predictors of lower CRT use were age (OR: 0.94/ y; 95% CI: 0.92–0.97), distance ≥20 miles from treatment facility (OR: 0.51; 95% CI: 0.29–0.90), and high-dose CSI (OR: 0.25; 95% CI: 0.12–0.51) (Table 2). Independent predictors of higher CRT use were M1–3 disease (OR: 3.40; 95% CI: 1.15–10.06) and increasing TTR (OR: 1.10/ wk; 95% CI: 1.02–1.18) (Table 2). A likelihood-ratio test comparing a fixed-effects logistic regression model (not shown) with the multilevel model met statistical significance for these data (P = .03), indicating that the multilevel model provided a better fit for the data.

Table 2.

Multivariable logistic regression analysis of predictors of adjuvant chemotherapy use

Variablea	OR	95% CI	P Value
Age, y (continuous)	.94	.92–.97	<.001
Year of diagnosis (continuous)	1.00	.88–1.13	.97
Sex (Male)
Female	.89	.52–1.52	.68
Charlson Deyo Score (0)
1–2	2.26	.95–5.36	.06
M stage (M0)
M1-3	3.40	1.15–10.06	.03
Histology (Classic/NOS)
Desmoplastic	.71	.36–1.40	.32
Large cell	1.79	.43–7.46	.43
Tumor size (<4 cm)
≥4 cm	1.52	.90–2.59	.12
Extent of resection (no macroscopic residual tumor)
Macroscopic residual tumor	.96	.28–3.27	.95
Unknown residual tumor	.41	.17–.96	.04
Race (white)
Other	.64	.31–1.32	.23
Facility volumeb (Low-volume)
High-volume	.96	.49–1.85	.90
Insurance (private)
Nonprivate	.65	.38–1.13	.13
Distance from facility (<20 miles)
≥20 miles	.51	.29–.90	.02
CSI dose (23 to <30 Gy)
30–36 Gy	.25	.12–.51	<.001
Time-to-RT, wk (continuous)c	1.10	1.02–1.18	.01

Abbreviations: CI, confidence interval; CSI, craniospinal irradiation; NOS, not otherwise specified; OR, odds ratio; RT, radiotherapy; wk, week(s); y, year(s).

^aReference variables are in parentheses.

^bHigh-volume designated for facilities that treated at least 8 patients.

^cTime in weeks from surgery to initiation of radiotherapy.

Chemotherapy Characteristics

Of patients receiving adjuvant chemotherapy (n = 520), 249 (47.9%) received CCRT, and 271 (52.1%) received SCRT. Three hundred forty-one patients (65.6%) received multiagent chemotherapy, 136 patients (26.2%) received single-agent chemotherapy, and 43 patients (8.3%) received an unknown number of agents.

Survival Analysis

For the survival analysis, 468 patients were included (Fig. 1). One-hundred forty-two patients (30.3%) had RT, and 326 patients (69.7%) had CRT. Median follow-up was 5.0 years. At the time of data extraction, 86 deaths (18.4% of patients) had occurred. Of the patients in the RT cohort, 99 (69.7%) were alive. Of the patients in the CRT cohort, 283 (86.8%) were alive. Median survival times were not met in either cohort. Five-year estimated OS for the entire population analyzed was 81.6%. Five-year estimated OS was improved in the CRT cohort compared with the RT cohort (86.1% vs 71.6%; P < .0001; Fig. 2). On univariable analysis, other variables associated with poorer OS were older age (HR: 1.02/y; 95% CI: 1.004–1.04; P = .02), desmoplastic histology compared with classic/NOS histology (HR: 1.86; 95% CI: 1.13–3.06; P = .01), and nonprivate insurance (HR: 1.76; 95% CI: 1.14–2.70; P = .01) (Table 3). On multivariable survival analysis, CRT was independently associated with improved OS in all models. For the Cox regression model that included all captured variables, CRT was associated with improved OS (HR: 0.53; 95% CI: 0.32–0.88; P = .01), and variables associated with worse OS were desmoplastic histology compared with classic/NOS histology (HR: 1.94; 95% CI: 1.14–3.32; P = .02) and nonprivate insurance (HR: 1.90; 95% CI: 1.16–3.10; P = .01) (Table 3). Both forward and backward stepwise selection yielded identical models that demonstrated independent associations for improved OS with CRT, and poorer OS for desmoplastic histology compared with classic/NOS histology and nonprivate insurance (Supplementary material, Table SA1). Sensitivity analysis including patients who received CSI, regardless of dose, demonstrated similar survival outcomes on univariable analysis (Supplementary material, Fig. SA4), and multivariable Cox regression (data not shown). Sensitivity analysis using varying thresholds for high-volume facility demonstrated similar survival outcomes (data not shown).

Fig. 2.

Plot of overall survival for all patients included in survival analysis by adjuvant therapy, n = 468.

Table 3.

Univariable and multivariable predictors of mortality

All-inclusive Model	Univariable Analysis			Multivariable Analysis
Variablea	HR	95% CI	P Value	HR	95% CI	P Value
Adjuvant treatment (RT)
CRT	.42	.27–.64	<.001	.53	.32–.88	.01
Age, y (continuous)	1.02	1.004–1.04	.02	1.02	.997–1.04	.09
Sex (Male)
Female	.89	.58–1.38	.61	1.03	.64–1.67	.89
Charlson Deyo Score (0)
1 or 2	1.40	.76–2.58	.28	1.33	.69–2.55	.39
M Stage (M0)
M1-3	.93	.40–2.14	.86	1.12	.47–2.66	.80
Histology (classic / NOS)
Desmoplastic	1.86	1.13–3.06	.01	1.94	1.14–3.32	.02
Large cell	1.65	.52–5.28	.40	1.50	.45–4.98	.51
Tumor size (<4 cm)
≥4 cm	1.24	.79–1.94	.35	1.29	.79–2.11	.31
Extent of resection (no macroscopic residual tumor)
Macroscopic residual tumor	.41	.05–3.31	.40	.34	.04–2.80	.32
Unknown residual tumor	1.34	.61–2.96	.47	1.22	.54–2.76	.63
Race (white)
Other	1.09	.61–1.93	.78	.99	.50–1.94	.97
Facility volume (low-volume)b
High-volume	1.08	.63–1.86	.77	.82	.47–1.43	.48
Insurance type (private)
Nonprivate	1.76	1.14–2.70	.01	1.90	1.16–3.10	.01
Distance from facility (<20 miles)
≥20 miles	1.20	.78–1.85	.40	1.17	.72–1.90	.54
CSI dose (23 to <30 Gy)
30–36 Gy	1.20	.72–1.99	.49	.91	.51–1.64	.76
Time-to-RT, wk (continuous)c	.97	.92–1.02	.29	.97	.91–1.02	.23

RT cohort includes patients receiving adjuvant CSI without chemotherapy, and CRT cohort includes patients receiving CSI and chemotherapy. Analysis includes only patients receiving CSI between 23 and 36 Gy.

Abbreviations: CI, confidence interval; CRT, chemotherapy and radiotherapy; CSI, craniospatial irradiation; HR, hazard ratio; RT, radiotherapy; wk, week(s).

^aReference variables are in parentheses.

^bHigh-volume designated for facilities that treated at least 8 patients.

^cTime in weeks from surgery to initiation of radiotherapy.

Propensity score matching produced a cohort of 127 patients, which was well-matched on all factors found to be significantly associated with OS on multivariable Cox proportional hazards regression and on traditional risk factors including age, CDS, M-stage, and residual tumor status (Supplementary material, Table SA2). Propensity score matching confirmed a survival benefit of CRT over RT (HR: 0.49; 95% CI: 0.29–0.86; P= .01). Five-year OS was 84.1% (95% CI: 76%–90%) for the matched CRT cohort and 73.6% (95% CI: 63%–81%) for the RT cohort (Supplementary material, Fig. SA3).

Subgroup Survival Analyses

Survival outcomes by adjuvant treatment for patient M-stage and CSI dose subgroups are found in Fig. 3. Among M0 patients analyzed for survival (n = 411), 129 (31.4%) received RT, and 282 (68.6%) received CRT. Among patients receiving 36 Gy CSI (n = 301), 116 (38.5%) received RT, and 185 (61.5%) received CRT. Among patients with M0 disease who were receiving 36 Gy CSI (n = 254), 104 (40.9%) received RT, and 150 (59.1%) received CRT. Five-year estimated OS was improved in the CRT versus RT cohort for patients with M0 disease (85.7% vs 72.0%; P < .0001), for patients receiving 36 Gy (86.8% vs 71.7%; P = .0007), and for patients with M0 disease who received 36 Gy (86.2% vs 70.8%; P = .0008). Sensitivity survival analysis for patients receiving between 30 and 36 Gy CSI (n = 350) demonstrated similar outcomes (data not shown). Among patients in the CRT cohort (n = 326), there was a nonsignificant trend for improved OS and receipt of multiagent chemotherapy compared with single-agent chemotherapy (HR: 0.56; 95% CI: .29–1.09; P = .09). There was no statistically significant association with OS and receipt of CCRT compared with SCRT (HR: 1.18; 95% CI: 0.62–2.23; P = .61) or receipt of high-dose CSI compared with low-dose CSI (5-year OS: 87.3% vs 83.6%; P = .47).

Fig. 3.

Plot of overall survival by adjuvant therapy for patient subgroups. (A) Patients with M0 disease (n = 411). (B) Patients receiving 36 Gy craniospinal irradiation (CSI) (n = 301). (C) Patients with M0 disease who received 36 Gy CSI (n= 254). Radiation therapy (RT) cohort included patients receiving adjuvant CSI without chemotherapy; chemotherapy and radiotherapy (CRT) cohort included patients receiving CSI and chemotherapy. Analysis includes only patients receiving CSI between 23 and 36 Gy.

Discussion

This analysis is the first of its kind to examine adjuvant treatment survival outcomes for adult MB on a national scale. In this analysis of more than 400 adult patients with MB, we found substantially superior OS for patients with adult MB who received combined postoperative chemotherapy and radiotherapy compared with RT alone (14.5% absolute 5 y OS benefit). We found that the OS benefit of CRT persists after adjusting for multiple potential confounders and in a risk factor-matched analysis. The OS benefit of CRT also persisted in the subsets of patients who received high-dose CSI to 36 Gy and/or had M0 disease.

Up to this point, there has been a scarcity of data to guide adjuvant treatment decision-making for MB in adults given its relative infrequency compared with the pediatric population. Only one prospective study of adult MB has been reported to date in which chemotherapy use was determined by presence of high-risk features such as M1-4 disease or residual tumor (n = 36).⁸ The largest reported retrospective study of adult MB to date, Padovani et al., demonstrated no survival benefit to the addition of adjuvant chemotherapy (n = 253).¹⁵ Several small institutional series reporting on adult MB have suggested a survival benefit to adjuvant chemotherapy,^6,9,16 while others have not.^7,17,18 Call et al. demonstrated a nonsignificant trend for improved survival with chemotherapy, but only in adults with high-risk MB (n = 66).⁶ The ability to evaluate adjuvant chemotherapy based on the available data is limited by small sample sizes and heterogeneous populations and chemotherapy regimens. Additionally, many patients in these studies did not receive regimens containing cisplatin, an agent that is currently used commonly and is associated with improved disease control in pediatric MB.^2,3,19

Due to the lack of high-quality evidence and potential differences in risks and benefits of adjuvant therapies in children and adults, there may be disagreement on the use of adjuvant chemotherapy for adult MB. While CSI is associated with side effects in all patients, children are particularly prone to long-term and deleterious morbidity from CSI.^20,21 Conversely, chemotherapy can be more toxic in adults than children, which may encourage less incorporation of chemotherapy for adult MB. Indeed, while the vast majority of pediatric MB patients receive adjuvant chemotherapy,³ our analysis demonstrates that adjuvant chemotherapy is excluded in approximately one-third of adult MB patients receiving postoperative RT. Our analysis also reveals that adults more commonly received high-dose CSI (74% of patients with documented doses) and that receipt of high-dose CSI was associated with the exclusion of chemotherapy compared with low-dose CSI (64% CRT vs 88% CRT). Additionally, patients with M0 disease were less likely to receive chemotherapy compared with those having M1-3 disease (68% CRT vs 86% RT). These trends highlight ongoing uncertainty in national clinical practice regarding the necessity of adjuvant chemotherapy in adult MB, particularly in the setting of higher CSI dose and lower-risk adults. Our analysis, however, suggests that the benefit of adjuvant chemotherapy persists in these patient subsets. Of the subsets analyzed, patients who had both M0 disease and who received high-dose CSI were the least likely to receive CRT, yet CRT continued to be associated with improved survival in this subgroup.

Among patients who received CRT, 5-year OS was favorable regardless of receipt of high-dose versus low-dose CSI. It is important to highlight that this study does not attempt to evaluate the safety of CSI dose de-escalation in adult MB as there is likely a selection bias limiting the ability to draw conclusions regarding this finding. We encourage further study in this regard.

On multivariable Cox regression analysis, other factors associated with poorer survival were tumor histology and insurance type. In this population, desmoplastic histology predicted poorer survival, a finding that is at odds with much of the pediatric literature in which desmoplastic histology has been associated with better prognosis.²² Desmoplastic histology appears to be more common in adult MB than in children,¹⁵ and its prognostic significance in adults is not as clearly established. One series suggested improved outcomes in adults with desmoplastic histology compared with classic histology,⁷ but several others did not.^15,23,24 The genetic underpinnings and behavior of adult MB tumor histologies may be different from those of pediatric tumors, which might explain differences in prognostic status.^15,25 Interpretation of our results in this context, however, may be diminished by the combined coding of classic and NOS histology as one variable in the NCDB. Large-cell histology has been associated with poorer survival in MB.²² The impact of this histology was difficult to assess, however, given the small number of patients with large-cell histology in our study (n = 24). The impact of histology should be interpreted cautiously, however, because of lack of centralized pathology review in this national analysis and likely institutional variability in experience with histopathologic subtyping of MB. Further study is needed to determine the prognostic significance of tumor histology in adult MB.

Lack of private insurance was also associated with poorer survival. This finding has been observed previously in population analyses across multiple cancer disease sites and may reflect poorer access and lower quality of health care received in general.²⁶ Survival in these patients may be improved through future health care reforms aimed at promoting access to high-quality health services.

Information regarding specific chemotherapeutic agents, dosing, and sequence was limited in the dataset, and certain combinations may confer more benefit than others. Our data suggested a trend toward improved survival with multiagent chemotherapy compared with single-agent chemotherapy, but no survival improvement with CCRT compared with SCRT. Data regarding the comparative effectiveness of various chemotherapy agents in the adult MB population are scarce. The aforementioned prospective trial initially used a mechlorethamine, vincristine, procarbazine, and prednisone (MOPP)-like regimen and then transitioned to a cisplatin-based regimen in later years of enrollment.⁸ Other series have reported regimens commonly used in pediatric MB including vincristine, cisplatin, lomustine, and/or cyclophosphamide.^15,16,24,27 As a concurrent agent, most series report the use of weekly vincristine, which has been extrapolated from pediatric data.²⁸

A limitation of this study is the inherent selection bias characteristic of retrospective cohort studies. Although our methods incorporated techniques such as multivariable regression and propensity score matching to adjust for multiple confounding factors (eg, age, comorbidity index, and M-stage), there are limits to the granularity of variables captured in the NCDB. Important staging and risk-stratifying details such as postoperative imaging quality and timing, presence of centralized pathologic review, cerebrospinal fluid sampling, and molecular analysis are not defined in the NCDB. There was also a significant amount of missing data assessing the extent of surgery, a known risk factor for pediatric MB.¹⁰ This could have an unmeasured confounding effect on survival and limits our ability to analyze outcomes by traditional risk groupings. The prognostic value of residual tumor in adult MB is less clear.^8,15,29 In our study population, we found a balanced distribution of known values of macroscopic residual tumor across our treatment cohorts and no prognostic impact of this variable on univariable or multivariable analysis. We thus believe it unlikely that missing data could be responsible for artificially inflating the observed association of chemotherapy with survival. Furthermore, assessment of residual tumor in this national analysis should be interpreted with caution, given the lack of centralized imaging and pathology review and the likely institutional variability in the experience of evaluating residual tumor. Another limitation of our study is that we were only able to analyze overall—not cause-specific—survival, and it is possible that some patient deaths were attributable to non-MB–related causes. Additionally, data on relapse and disease progression are not available in the NCDB, and it will be important for future studies to investigate both progression-free and overall survival.

While adjuvant chemotherapy was associated with a robust survival benefit in this analysis, we are unable to capture patient susceptibility to toxicity, performance status, and quality of life information, which must be weighed in treatment decision-making. Clinical judgment regarding the suitability of a particular patient for chemotherapy should continue to be made on an individualized basis.

Conclusions

This analysis demonstrates that combined adjuvant chemotherapy and radiotherapy are associated with improved survival in adult MB compared with RT alone on a national scale. The survival benefit of adjuvant chemotherapy persisted in patients with M0 disease and patients receiving high-dose craniospinal irradiation. This study emphasizes the need for a prospective randomized clinical trial of standardized postoperative RT with or without standardized chemotherapy that includes centralized review of imaging and pathology. In the absence of randomized clinical data, this study supports the use of adjuvant chemotherapy in addition to CSI in adult MB patients who can tolerate multimodality treatment.

Supplementary Material

Supplementary material is available at Neuro-Oncology Journal online (http://neuro-oncology.oxfordjournals.org/).

Funding

No funding has been received to support this research

Conflict of interest statement. Dr. Park received honoraria and travel reimbursement from Varian Medical Systems, unrelated to the current work. Dr. Lester-Coll has received funding from Elekta, unrelated to the current work. Dr. Yeboa has received travel reimbursement from Eli Lilly, unrelated to the current work. Dr. Yu has received a research gift from 21st Century Oncology, unrelated to the current work. Dr. Roberts has received travel funding for a IBA-sponsored workshop on proton RT for pediatric cancers, unrelated to the current work.