Molecular marker testing and reporting completeness for adult-type diffuse gliomas in the United States

Abstract

Background

A newly developed brain molecular marker (BMM) data item was implemented by US cancer registries for individuals diagnosed with brain tumors in 2018—including IDH and 1p/19q-co-deletion statuses for adult-type diffuse gliomas. We thus investigated the testing/reporting completeness of BMM in the United States.

Methods

Cases of histopathologically confirmed glioblastoma, astrocytoma, and oligodendroglioma diagnosed in 2018 were identified in the National Cancer Database. Adjusted odds ratios (OR_adj) and 95% confidence intervals (CI) of BMM testing/reporting were evaluated for association with the selected patient, treatment, and facility-level characteristics using multivariable logistic regression. As a secondary analysis, predictors of MGMT promoter methylation testing/reporting among IDH-wildtype glioblastoma individuals were assessed. Key limitations of the BMM data item were that it did not include any details regarding testing technique or assay type and could not distinguish between a lack of testing and a lack of cancer registry reporting of testing results.

Results

Among 8306 histopathologically diagnosed adult-type diffuse gliomas nationally, overall BMM testing/reporting completeness was 81.1%. Compared to biopsy-only cases, odds of testing/reporting increased for subtotal (OR_adj= 1.38 [95% CI: 1.20–1.59], P < .001) and gross total resection (OR_adj=1.50 [95% CI: 1.31–1.72], P < .001). Furthermore, the odds were lowest at community centers (hospitals (67.3%; OR_adj=0.35 [95% CI: 0.26–0.46], P < .001) and highest at academic/NCI-designated comprehensive cancer centers (85.4%; referent). By geographical location, BMM testing/reporting completeness ranged from a high of 86.8% at New England (referent) to a low of 76.0 % in the West South Central region (OR_adj=0.57 [95% CI: 0.42–0.78]; P < .001). Extent of resection, Commission-on-Cancer facility type, and facility location were additionally significant predictors of MGMT testing/reporting among IDH-wildtype glioblastoma cases.

Conclusions

Initial BMM testing/reporting completeness for individuals with adult-type diffuse gliomas in the United States was promising, although patterns varied by hospital attributes and extent of resection.

There is growing evidence that molecular markers signify clinically relevant types of brain tumors. This has led to the integration of histopathological and molecular classification in the 2016 World Health Organization (WHO) classification of central nervous system (CNS) tumors (WHO-CNS4), followed by further refinement in the 2021 Fifth edition (WHO-CNS5).^1–3 Among these, the most widely researched and employed are Isocitrate Dehydrogenase 1/2 (IDH1/2) mutations 1p/19q co-deletion, and O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation statuses for adult-type diffuse gliomas.^3–6 These markers, among several others, are an integral part of modern diagnostic criteria and reflect our burgeoning understanding of the genetic origins of cancer.

Reflecting the importance of integrated diagnoses, multiple national registry stakeholders developed a site-specific data item to facilitate systematic reporting of Brain Molecular Markers (BMM; North American Association of Central Cancer Registries [NAACCR] Item #3816; Table 1).^7–9 We have previously detailed the development and validation of the novel BMM data item, showing high levels of completeness among tumor histopathologies relevant to each molecular marker.^10–12 The BMM data item allows for the unambiguous reporting of integrated histomolecular diagnoses that did not yet have unique International Classification of Diseases for Oncology, 3rd edition (ICD-O-3) morphology codes by facilitating the streamlining of collection standards. However, because of the complexities of modern cancer surveillance programs, the updated WHO classification and subsequent new data collection standards were only implemented for cases diagnoses beginning January 1, 2018.⁸ With the BMM data item now in practice in cancer registries, it is important to evaluate whether there were any barriers to BMM testing/reporting completeness.

The primary aim of this study was to evaluate the practice patterns of BMM testing/reporting completeness in the United States. As a secondary aim, we also investigated the practice patterns in testing/reporting completeness of the new MGMT promoter methylation data item—a key predictive biomarker—among IDH-wildtype glioblastoma individuals.^11,13–15

Methods

Data Sources and Variables

The National Cancer Database (NCDB) reports data for more than 85% of newly diagnosed primary malignant brain tumor individuals in the United States, drawn from more than 1500 Commission-on-Cancer-accredited cancer programs in facilities across the United States.¹⁶ Individuals were recorded with histopathologically confirmed adult-type diffuse gliomas—including glioblastomas (International Classification for Disease, Oncology 3rd edition [ICD-O-3] morphology/behavior code 9440/3), astrocytomas (ICD-O-3 morphology/behavior codes 9400/3 or 9401/3), and oligodendrogliomas (ICD-O-3 morphology/behavior codes 9450/3 or 9451/3)—were identified in the NCDB for diagnoses made between January 1–December 31, 2018. At the time of analysis, 2018 was the latest year for which data were reported by the NCDB, and the reporting of the BMM data item only began for individuals diagnosed starting on January 1, 2018. The BMM variable and a few new ICD-O-3 codes (e.g. H3K27-mutant diffuse midline glioma) that allowed for the identification of integrated diagnoses were implemented in United States cancer registries as of January 1, 2018. Because the NCDB data were reported from patients diagnosed in 2018, the WHO CNS grade would likely have been reported using the histology-based revised WHO-CNS4 criteria published in 2016. Only cases with a histological WHO grade of 2, 3, or 4 were included in the analyses. Additionally, for brain tumor cases diagnosed starting January 1, 2018, data were reported for a new MGMT promoter methylation site-specific data item (NAACCR #3889). The BMM variable required IDH status for the coding of diffuse astrocytoma and glioblastoma histologies, and both IDH status and 1p/19q co-deletion status for the coding of oligodendrogliomas.

Statistical Analysis

The primary outcome of interest was the status of BMM testing/reporting completeness. The BMM variable did not distinguish between a lack of testing and a lack of registry reporting of testing results, therefore testing and completeness were combined into a single outcome for analysis. As a secondary analysis, we evaluated the status of MGMT promoter methylation testing/reporting completeness among individuals diagnosed with an IDH-wildtype glioblastoma (as identified by a BMM code of “05”). BMM and MGMT testing/reporting were both defined as complete if tests were ordered or reported, regardless of the result, and incomplete otherwise. The independent variables of interest included demographic (age at diagnosis, sex, race/ethnicity, and prior cancer diagnosis) and socioeconomic (insurance status and ZIP code-level median household income) factors, in addition to extent of resection (EOR; categorized as biopsy-only, subtotal resection [STR], and gross total resection [GTR]), histological WHO grade, tumor histopathological type, primary site (supratentorial, infratentorial, overlapping lesion, vs brain, and not otherwise specified [NOS]), and by the reporting facility characteristics (type, location, and urbanicity). Descriptive statistics were calculated for each independent variable, categorized by testing/reporting completeness status. The association between testing/reporting completeness and each independent variable was assessed using univariable logistic regression. Additionally, all independent variables were evaluated together in an unbiased fashion using multivariable logistic regression to account for potential confounding. Variables were selected based on clinical relevance and included in multivariable models regardless of statistical significance at the univariable level. All observations with missing data for any independent variable were excluded. Because NCDB suppresses the facility type and location data items for cases below the age of 40 years, such cases were analyzed separately from the primary analysis. To investigate MGMT promoter methylation testing/reporting completeness among IDH-wildtype glioblastoma, the univariable and multivariable models were repeated in a secondary analysis. All statistical analyses were conducted using R 4.1.0 and a conservative two-sided α = 0.001 to help reduce false positives. This study received approval from the Duke University School of Medicine Institutional Review Board and was conducted in accordance with the Declaration of Helsinki. The NCDB Participant User Files contain de-identified national data for which consenting was not applicable.

Results

For the primary analysis of BMM testing/reporting completeness patterns, a total of 8306 histopathologically diagnosed individuals with an adult-type diffuse glioma and who had complete data were evaluated. Among these individuals, the BMM data item was tested/reported for 81.1% (n = 6733; Table 1). Individuals who were younger than 40 years old when diagnosed (n = 1573 complete cases) were excluded from primary analyses because the NCDB suppresses several facility variables for such individuals; however, 86.7% of individuals <40 years old had BMM testing/reporting completeness (n = 1048; Supplementary Table 3).

National Practice Patterns for Testing/Reporting Completeness of BMMS Among Individuals With Adult-Type Diffuse Gliomas

With respect to patient demographics and socioeconomics—including age, sex, race/ethnicity, prior history of cancer, ZIP code-level household income, and insurance status—BMM testing/reporting completeness did not vary significantly across characteristics in multivariable analysis (Table 1). For example, 77.0% of Hispanic individuals had BMM testing/reporting completeness as compared to 81.6% of non-Hispanic White individuals, but this finding was not significantly different when accounting for individuals’ other attributes (OR_adj 0.73 [95% CI: 0.59–0.92] P = .007). Similarly, 82.5% of individuals from the richest quartile of households had BMM testing/reporting completeness compared to 78.3% of individuals from the quartile in lower-income counties, but this was not significant in multivariable analysis (OR_adj 1.13 [95% CI: 0.96–1.33] P = .146; Table 2). BMM testing/reporting completeness was also associated with insurance status: As compared to privately insured individuals, uninsured individuals showed slightly lower levels (76.7% vs 81.9%; OR_adj 0.93 [95% CI: 0.67–1.31] P = .661) and individuals with non-Medicare/Medicaid governmental insurance was slightly higher levels (86.8% vs 81.9%; OR_adj 1.70 [95% CI: 1.10–2.72] P = .022) in univariable—but not multivariable—analyses.

The relationships between individuals’ tumor and treatment features and BMM testing/reporting completeness were also evaluated. Although, in unadjusted analyses, tumors reported to be of an oligodendroglioma histopathology appeared to be more likely to have BMM testing/reporting completeness (93.1 vs 80.2%, OR_adj 3.36 [95% CI: 2.31–5.10] vs glioblastoma histopathology, P < .001; Table 2), that was not the case following multivariable adjustment. BMM testing/reporting completeness was independent of tumors’ primary site, tumor sequence, treatment location, or histological WHO grade (all P > .001; Table 2). Notably, individuals who individuals underwent either STR (82.6% vs. 76.4%; OR_adj 1.38 [95% CI: 1.20–1.59] P < .001) or GTR (83.7% vs 76.4%; OR_adj 1.50 [95% CI: 1.31–1.72] P < .001) were significantly more likely to have BMM testing/reporting completeness than individuals whom only received surgical biopsy—including after accounting for individuals’ other characteristics (Figure 1).

Notably, BMM testing/reporting completeness levels nationally varied significantly by the diagnosing facility’s features. For instance, in terms of geographical location, BMM testing/reporting completeness ranged from a high of 86.8% for individuals in New England (including Connecticut, Massachusetts, Maine, New Hampshire, Rhode Island, and Vermont; referent) to a low of 76.0% for individuals in the West South Central region (Arkansas, Louisiana, Oklahoma, and Texas; OR_adj 0.57 [95% CI: 0.42–0.78]; P < .001; Table 2). Additionally, the type of hospital was also associated with BMM testing/reporting completeness, with 85.4% of individuals at academic/research facilities having BMM tested and reported (referent), whereas only 67.3% at community (OR_adj 0.35 [95% CI: 0.26–0.46] P < .001) and 75.1% at comprehensive community (OR_adj 0.52 [95% CI: 0.45–0.59] P < .001) cancer programs had testing/reporting completeness (Table 2 and Figure 1). Odds of BMM testing/reporting completeness were not associated with facility urbanicity nor by whether the reporting facility was the initial diagnosis hospital or not.

National Practice Patterns for Testing/Reporting Completeness of MGMT Promoter Methylation Status Among IDH-Wildtype Glioblastoma Individuals

In the secondary analysis, a total of 5418 IDH-wildtype glioblastoma cases were identified, with 3933 (72.6%) showing complete MGMT promoter methylation testing and reporting. In terms of patient features, MGMT testing/reporting was associated with age: The odds of testing/reporting completeness generally increased as patient age decreased, with the highest testing/reporting levels among individuals ages 45–49 years (79.3% vs 69.7% in 70+ years olds; OR_adj 2.19 [95% CI: 1.54–3.15], P < .001) and the lowest among individuals ages 70+ years (referent; Supplementary Tables 1 and 2). Approximately 58.6% of uninsured individuals and 67.8% individuals with Medicaid exhibited testing/reporting completeness for MGMT compared to 75.3% of privately insured individuals, although this was not significant in the multivariable model (Supplementary Tables 1 and 2). Additionally, Individuals managed at facilities in every location other than the Mountain (Arizona, Colorado, Idaho, Montana, New Mexico, Nevada, Utah, and Wyoming) and Pacific (Alaska, California, Hawaii, Oregon, and Washington) regions showed significantly lower levels of testing/reporting completeness compared to individuals managed in New England (87.4%; referent), with the lowest being West South Central (51.6%; OR_adj 0.16 [95% CI: 0.11–0.23)] P < .001). The highest level of testing/reporting completeness was at academic cancer centers (referent) with 77.2% of IDH-wildtype glioblastomas showing testing/reporting completeness for MGMT promoter methylation, while at community cancer programs only 52.6% exhibit testing/reporting completeness (OR_adj 0.40 [95% CI: 0.28-0.57] P < .001; Table 1 and Supplementary Table 1). EOR was again strongly associated with MGMT promoter methylation testing and completeness, with 77.1% of STR (OR_adj 1.66 [95% CI: 1.41–1.95] P < .001) and 73.6% of GTR (OR_adj 1.41 [95% CI: 1.21–1.64] P < .001) individuals having complete testing/reporting compared to biopsy only (66.0%; referent).

Discussion

We have previously shown that up to 20%–25% of histopathologically coded oligodendrogliomas, 6%–9% of astrocytomas, and 9% of glioblastomas that were reported in cancer registry data were potentially misclassified (when compared to their molecular markers such as 1p/19q status), which illustrated the value of a dedicated BMM variable that enables cancer registries to report integrated histomolecular diagnoses.¹⁷ Here, we found that the testing/reporting completeness levels for BMM in its first year of implementation by US cancer registries were high overall: 81% of adult individuals with adult-type diffuse gliomas (and 87% of individuals <40 years old; Supplementary Table 3). These levels compared favorably to the earlier 1p and 19q site-specific factors, which, although registry collection was mandatory, was only reported for 43%–50% of oligodendrogliomas at the time of their introduction (2010–2012), and showed a decrease in chromosomes 1p (9.3% to 9.0%) and 19q (9.5% to 9.0 %) loss of heterozygosity completeness and an increase in MGMT promoter methylation testing/reporting completeness from 7.7% to 10.3%.¹⁸ In 2019, the NCI’s SEER program evaluated the reporting features of SSDIs by registrars for cancers diagnosed in 2018. For 3 IDH-wildtype glioblastoma test cases, among 164–171 cancer registrars, 8%–9% erroneously reported a BMM code of “99” (eg, not tested or not reported), and 3%–4% reported a code of “88” (eg, not available). Likewise, for a test case of BMM “test ordered, results not in chart”, 20% of 179 registrars reported a code of “99” and 3% “88”. These findings suggest that a substantive proportion of BMM not tested/reported represented a lack of appropriate registry reporting, as opposed to a lack of testing.

Similarly, we observed a high level of testing/reporting completeness (73%) for the new MGMT promoter methylation variable among IDH-wildtype glioblastomas—a tumor type for which MGMT promoter methylation status plays a key prognostic and predictive role. This testing/reporting level represents a stark improvement from the earlier MGMT site-specific factor, for which cancer registries only reported data for 12% of glioblastomas cases at the time of its introduction (2010–2012). Among glioblastoma cases from 2010 to 2016, MGMT promoter methylation testing rates improved from 57% to 74% of newly diagnosed individuals.¹⁹ Among individuals histopathologically diagnosed with adult-type diffuse gliomas (ie, glioblastoma, astrocytoma, or oligodendroglioma) in 2018, the testing/reporting completeness for BMM was generally high at 81%, while the testing/reporting completeness of MGMT promoter methylation status among individuals with IDH-wildtype glioblastoma was modestly lower at 72%. These newest estimates reflect a dramatic increase in the testing/reporting completeness of site-specific factors from 2012 to 2018.

Individuals’ sociodemographic characteristics—including their sex, race/ethnicity, insurance status, and ZIP-code level household income—did not seem to influence the levels of either BMM or MGMT promoter methylation testing/reporting completeness. Although some of these factors were associated with testing/reporting completeness in the individual unadjusted models, these effects were no longer significant in the adjusted models after controlling for the individuals’ other covariates. Importantly, EOR was significantly associated with the testing/reporting completeness of both BMM and MGMT promoter methylation. Individuals who received STR or GTR demonstrated an ~50% higher odds of having BMM and MGMT promoter methylation status testing/reporting completeness as compared to those who underwent biopsy only. Because EOR is unlikely to have affected BMM reporting practices by cancer registrars, it likely reflects differences in the underlying testing practices for molecular markers. One possible explanation is that biopsies may be more likely to result in insufficient tissue for molecular testing. Additionally, age was found to be a significant predictor of MGMT promoter methylation testing/reporting completeness, with individuals ages 45–49 years showing more than double the odds of testing/reporting completeness as compared to patients in the oldest group—despite the clinical value of MGMT promoter methylation testing being most clear in older patients. In our previous analysis of barriers to MGMT promoter methylation testing in glioblastoma patients, we also observed a significant decrease in testing levels associated with increased age.¹⁹ One important, unmeasured confounder may be patient’s performance status, whereby older patients with poor performance status may be less likely to undergo resection, and therefore may have less tissue available for MGMT promoter testing. Unfortunately, a more rigorous dissection of this observation is prevented by the limitations of the NCDB data; however, a closer examination using other data sources is warranted.

The characteristics of the reporting hospital, including geographic location and Commission-on-Cancer facility type, were significantly associated with both BMM and MGMT methylation testing/reporting completeness. With respect to facility location, BMM testing/reporting completeness ranged from a low of 76% of individuals in the West South Central region to 87% of individuals managed in the New England region. Because the lack of testing and lack of cancer registry reporting of testing results are coded together, it was not possible to discern whether this geographical variability was driven by differences in testing patterns or cancer registry reporting patterns. Using an earlier version of the MGMT site-specific factor that did distinguish between lack of testing and a lack of reporting, we previously found that MGMT promoter methylation testing levels themselves varied geographically, with a similar pattern to that observed in the new MGMT variable, suggesting that some of the regional variability seen here may also reflect differences in testing practices and regional differences in facility type distribution.¹⁹ Although the training and accreditation of cancer registrars are standardized nationally in the United States, it may also be possible that reporting strategies were not completely harmonious nationally.²⁰ Additionally, BMM testing/reporting completeness levels varied significantly by hospital type: With the highest levels (85% of individuals) at academic and NCI-designated comprehensive cancer centers (ie, facilities with postgraduate medical education programs and >500 newly diagnosed cancer cases per year) and the lowest (66% of individuals) observed at community cancer programs (ie, facilities with between 100 and 500 newly diagnosed cancer cases each year). As with facility geographical location, it was not possible to directly parse the contribution of lack of testing and lack of reporting. In both cases, increased education around the importance of both conducting and reporting guideline-recommended molecular testing for diffuse gliomas may be merited to help reduce the observed differences in testing/reporting levels. Finally, among those individuals who sought care at multiple hospitals, there was no difference in testing/reporting levels between the initial hospital or subsequent hospitals. Neither did we observe differences in testing/reporting levels by the extent of urbanicity or rurality of the managing hospital.

Limitations

Our study was constrained by multiple limitations attributed to the usage of NCDB data. In particular, the new molecular SSDIs do not distinguish between the lack of molecular testing and the lack of cancer registry reporting of testing results—consequently, the observed differences could arise from either or both. Additionally, the NCDB does not report time intervals more granular than a year and, at the time of analysis, molecular SSDI data were only available for individuals diagnosed in 2018, precluding evaluation of how testing and reporting rates have changed over time. he NCDB does not capture details regarding molecular marker testing technique, assay type, or timing, including whether IDH status was assessed by immune-histochemistry or sequencing, and how 1p/19q co-deletion status was evaluated. Furthermore, the NCDB does not presently report data for other commonly used molecular markers (e.g., CDKN2A/B, ATRX, TP53, and H3K27 trimethylation). The NCDB suppresses facility type and location information for individuals diagnosed at age <40 years, so as to help ensure de-identification of patients. After conducting a sensitivity analysis in this age cohort (sans the suppressed variables), no significant associations with BMM testing or reporting was observed in this age cohort (Supplementary Tables 3 and 4).

Conclusions

Novel BMM and MGMT promoter methylation site-specific data items were implemented by US cancer registries for brain tumor diagnoses made starting January 1, 2018. Among individuals diagnosed with adult-type diffuse gliomas in the NCDB, we found the initial testing/reporting completeness of BMM of was to be promisingly high, with 81% was of individuals 40 years and older and 87% of individuals younger than 40 years. Although the initial testing/reporting of completeness of the new MGMT promoter methylation data item among IDH-wildtype glioblastoma was relatively lower (73%), these results this compared favorably to the levels reported by the MGMT variable analyses from previous years. While individuals’ sociodemographic factors were not strongly associated with molecular testing, BMM and MGMT testing/reporting completeness were significantly associated with EOR as well as the geographic location and cancer program type of the reporting hospital. Key limitations of the BMM data item were that it did not include any details regarding testing technique or assay type and that it could not distinguish between a lack of testing and a lack of cancer registry reporting of testing results. Nevertheless, the BMM integrated diagnosis data item overall exhibited promising levels of testing/reporting completeness in the United States in its first year of implementation.

Funding

CBTRUS has been funded by the Centers for Disease Control and Prevention under Contract No.75D30119C06056 Amendment/Modification No:00002, the American Brain Tumor Association, Novocure, the Musella Foundation, The Sontag Foundation, National Brain Tumor Society, the Pediatric Brain Tumor Foundation, the Uncle Kory Foundation, and the Zelda Dorin Tetenbaum Memorial Fund, and private and in-kind donations. The research services of JSB-S, KW, and GG were provided by the Division of Cancer Epidemiology and Genetics of the NCI. Contents are solely the responsibility of the authors and do not necessarily represent the official views of either the CDC or of the NCI. JBI gratefully acknowledges funding support from the National Cancer Institute (K12CA090354) and Conquer Cancer Foundation/Sontag Foundation.

Acknowledgements

JSB-S is a full-time paid employee of the NIH/NCI. GC and KW are full-time contractors of the NIH/NCI. Contents are solely the responsibility of the authors and do not necessarily represent the official views of either the CDC or of the NCI. The data used in the study are derived from a de-identified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigator.

Conflicts of Interest

There are no conflicts of interest relevant to this manuscript to report.

Authorship

conceptualization & supervision: CN, JBI, QO; methodology: CN, JBI, QO; formal analysis and investigation: CN, JBI; critical interpretation of results & writing: CN, JBI, QO; read and approved final version: All authors.

References

Author notes