Socioeconomic Disparities Associated With MGMT Promoter Methylation Testing for Patients With Glioblastoma

Abstract

This cohort study evaluates the national practice patterns of MGMT gene testing and identified potential factors associated with access to testing.

Silencing of the DNA-repair MGMT gene via promoter methylation is presently the only clinically relevant predictive biomarker for patients with glioblastoma and, increasingly, a critical eligibility criterion for clinical trial participation.^1,2 Consequently, National Comprehensive Cancer Network guidelines recommend testing all newly diagnosed glioblastomas.³ Promoter methylation status of MGMT is particularly important for clinical decision-making for patients who are elderly or frail, who are less able to tolerate multimodal therapy and for whom temozolomide can be withheld due to its limited benefit in MGMT-unmethylated cases.² Herein we evaluate the national practice patterns of MGMT testing and identify potential factors associated with access to testing.

Methods

The National Cancer Database contains data on more than 70% of patients with newly diagnosed cancer in the US. With Partners HealthCare institutional review board approval, patients (≥40 years old) diagnosed with histologically confirmed glioblastoma, World Health Organization grade IV, between 2010 and 2016 were identified using International Classification of Diseases for Oncology, Third Edition coding and collaborative staging brain-specific variables.⁴ Only patients with complete data encoded for MGMT testing (38.2% of identified cases) from their diagnosing institution were included. Patient informed consent was waived by the institutional review board owing to the deidentified nature of the data analyzed in this study.

The primary outcome was receipt of MGMT promoter methylation testing. Lack of testing was defined as “test not done (test not ordered and not performed).” Univariable associations were assessed by Fisher exact or χ² tests. Multivariable logistic regression was used to evaluate the associations between patients’ demographic and socioeconomic characteristics and testing, and between testing and receiving chemotherapy. Multivariable Cox regression was used to assess the association of testing with overall survival (OS). Stata, version 15.1 (StataCorp), was used for analyses (2-sided α = .05).

Results

Out of a total of 12 830 patients evaluated, 56.9% with newly diagnosed glioblastoma received MGMT testing, increasing to 73.6% by 2016. Patients who were elderly, uninsured, insured through Medicaid, from poorer quartiles of households or low-population urban/rural areas, or diagnosed at community programs independently received less testing (Table 1). Among patients with Karnofsky Performance Scale (KPS) data (n = 1937), testing was more likely for favorable status (KPS ≥70; 1013/1474 [68.7%]) than poor status (KPS <70; 284/463 [61.3%]; P = .003). Patients who were Hispanic (52.5%) or Black non-Hispanic (52.5%) were less likely to be tested than those who were White non-Hispanic (57.3%) or Asian/Pacific Islander (56.5%; P = .01), but not independently so following multivariable adjustment.

Table 1. Factors Associated With MGMT Promoter Methylation Testing in Patients With Glioblastoma.

Characteristic	Total No.	% With MGMT testinga	Testing % in 2016	Multivariable logistic regression of having MGMT testingb
				aOR (95% CI)	P value
Age at diagnosis, y
40-49	1258	61.5	81.2	1.27 (1.08-1.48)	.004
50-59	3437	58.3	74.3	0.98 (0.87-1.10)	.72
60-69	4251	57.3	74.2	1 [Reference]
70-79	2875	55.2	70.9	0.95 (0.84-1.08)	.45
≥80	1009	50.0	69.0	0.83 (0.70-0.99)	.04
Primary payer
Uninsured	386	42.8	60.9	1 [Reference]
Private insurance	5841	60.8	76.1	1.78 (1.39-2.28)	<.001
Medicaid	802	53.9	74.9	1.30 (0.98-1.74)	.07
Medicare	5476	54.7	71.9	1.49 (1.14-1.93)	.003
Median household income by zip code, $
<38 000	1693	49.7	63.5	1 [Reference]
38 000-47 999	2775	53.2	69.9	1.11 (0.96-1.29)	.15
48 000-62 999	3523	57.2	74.5	1.23 (1.06-1.42)	.006
≥63 000	4819	61.4	78.5	1.31 (1.13-1.52)	<.001
Cancer program type
Community	466	44.9	56.2	1 [Reference]
Comprehensive community	3794	44.3	66.8	0.99 (0.80-1.23)	.94
Academic/NCI-designated	6532	65.0	79.8	2.21 (1.78-2.73)	<.001
Integrated network	2038	57.5	72.2	1.65 (1.31-2.08)	<.001

Abbreviations: aOR, adjusted odds ratio; NCI, National Cancer Institute.

^aMGMT testing percentages across each of the displayed variables were significantly different (for all χ² tests, P < .001), including across MGMT testing percentages in 2016 (n = 2822; for all χ² tests, P < .02).

^bMultivariable model was additionally adjusted (data not shown) for patients’ sex, race/ethnicity, Charlson-Deyo comorbidity index, histology, tumor site, and size, which were not associated with testing, as well as patient’s county population/rurality (urban/rural, <20 000, aOR, 0.81 vs metro, >1 000 000; 95% CI, 0.69-0.95; P = .009), year of diagnosis, extent of resection, and cancer program location, which were significantly associated with testing. Primary payer included other government insurance (n = 184) and not available (n = 141) (data not shown).

Whereas 5618 patients out of 7303 (76.9%) with MGMT-tested glioblastoma received chemotherapy (2272/2936 [77.4%] of MGMT-methylated and 3346/4367 [76.6%] of MGMT-unmethylated cases), only 3739 (67.6%) untested patients received chemotherapy (Table 2). Lack of testing was associated with similar unadjusted OS as MGMT-unmethylated tumors and with an intermediate adjusted OS between MGMT-methylated and MGMT-unmethylated tumors.

Table 2. Association of MGMT Promoter Methylation Testing With the Management of Patients With Glioblastoma.

MGMT testing	Multivariable logistic regression of receiving chemotherapya		Overall survivalb		Multivariable Cox regression for overall survival
	aOR (95% CI)	P value	Median (95% CI), mo	2 y, % (95% CI)	aHR (95% CI)	P value
None	1 [Reference]		13.8 (13.3-14.4)	23.9 (22.1-25.8)	1 [Reference]
Yes
Methylated	1.43 (1.26-1.63)	<.001	18.9 (17.7-19.8)	38.2 (35.4-41.0)	0.73 (0.67-0.80)	<.001
Unmethylated	1.31 (1.17-1.47)	<.001	14.2 (13.7-14.6)	20.5 (18.7-22.4)	1.11 (1.04-1.20)	.004

Abbreviations: aHR, adjusted hazard ratio; aOR, adjusted odds ratio.

^aMultivariable model was additionally adjusted for patients’ age at diagnosis, sex, race/ethnicity, Charlson-Deyo comorbidity index, year of diagnosis, median household income and county population/rurality of patient’s residence, primary payer, histology, tumor site and size, extent of resection, and cancer program type and location (data not shown).

^bOverall survival was estimated by Kaplan-Meier methods and evaluated for patients treated with resection and radiotherapy with ≥1 month of follow-up to account for survival bias (n = 4419; 83.0% died). The Cox model was also adjusted for age, sex, tumor size, chemotherapy, and extent of resection (data not shown).

Discussion

Using US national data, we found that patients with newly diagnosed glioblastoma who were uninsured, insured through Medicaid, from poorer households or low-population urban or rural areas, or diagnosed at community hospitals were disproportionately less likely to receive MGMT testing. Although testing rates improved by 2016 across all settings—likely reflecting the growing awareness of the clinical importance of MGMT status (eg, NOA-08, Nordic-ISRCTN81470623, RTOG-0525) and its integration into national guidelines—testing still lagged for the aforementioned patient populations.^1,2

Additionally, testing was underused for patients who were elderly or who had poor performance status, populations that would most benefit from MGMT testing-guided deescalation of combined therapy. These testing disparities appeared to influence patients’ management: patients who were untested received less chemotherapy and had intermediate adjusted survival between that of patients who were MGMT methylated and MGMT unmethylated, suggesting that untested glioblastomas comprise a mix of methylation statuses whose management could be improved by appropriate testing. Notable limitations included the database’s limited details about test types and timing.

The results of this study indicate that substantial socioeconomic and care setting disparities exist in the testing of this important biomarker for patients with glioblastoma. With MGMT status increasingly becoming a critical trial eligibility criterion, failure to test in a timely manner will compound existing barriers to patients’ access to clinical trials.^5,6 Future studies are needed to define how these disparities in MGMT promoter methylation testing arise and to help develop strategies to ensure equitable access to quality oncologic care for patients with glioblastoma.