Accuracy of central neuro-imaging review of DIPG compared with histopathology in the International DIPG Registry

Abstract

Diffuse intrinsic pontine glioma (DIPG) is the most common pediatric brainstem tumor and carries a dismal prognosis. Diagnosis is based upon characteristic radiologic and clinical features, without routine histopathologic corroboration. On MRI, DIPG has been classically defined as an expansile, infiltrative mass involving >50% of the pons, which is hypointense or isointense on T1-weighted sequences, hyperintense on T2-weighted sequences, with minimal or absent contrast enhancement and frequent encasement of the basilar artery.^1–4 Recent reports, however, have demonstrated a more heterogeneous appearance of these tumors with varied enhancement, necrosis, and internal hemorrhage.⁴ Most patients present with a combination of cranial neuropathies, pyramidal tract dysfunction, and/or ataxia, often under six months duration.^5,6 Despite increasing evidence suggesting safety and feasibility of biopsy in DIPG,^7,8 its role at diagnosis remains controversial, with tissue acquisition usually reserved for patients with atypical imaging features, if required for clinical trial enrollment, or to identify molecular targets.^6,8

Studies of DIPG specimens from autopsy and biopsy (when available) have resulted in an improved understanding of the histopathologic spectrum of this disease, encompassing glioblastoma, anaplastic astrocytoma, and diffuse astrocytoma (WHO grade II-IV infiltrating gliomas).^9,10 Genome-wide sequencing analyses discovered recurrent mutations in the genes encoding histone H3.3 (H3F3A) or H3.1 (HIST1H3B) in most DIPG tumors,¹¹ with recent reclassification as diffuse midline glioma, H3K27M-mutant.¹² Histopathologic evaluation of pontine masses in a subset of pediatric patients, including some suspected to represent DIPG on the basis of clinico-radiologic features, have revealed discordant diagnoses, such as embryonal tumors, nondiffuse low-grade gliomas (LGGs), or rarely non-neoplastic tissue.^13–23 Pontine embryonal tumors exhibited radiologic properties not typically observed in DIPG, such as sharply-defined borders, diffusion restriction, leptomeningeal dissemination, and/or exophytic extension.^13,15,17 Brainstem pilocytic astrocytomas manifested as focally exophytic, well-circumscribed, densely enhancing, and/or partially cystic masses, with potential for excellent overall survival (OS).^3,22,23 Within a recently published cohort by Chiang et al. of pontine tumors lacking characteristic imaging features of DIPG that underwent biopsy, almost half had histopathologic and molecular profiles consistent with embryonal or low-grade glial neoplasms, with improved outcomes compared to tumors of infiltrative histology or historical controls with typical DIPG.¹⁹ In the absence of universal biopsy, reliable radiologic distinction between DIPG and other pontine tumors with potentially more favorable prognoses and treatment considerations is essential.

Furthermore, as most DIPG clinical trials do not require histologic confirmation, there is a critical need for precise, standardized radiologic eligibility criteria to ensure fair comparison of response and survival endpoints within and across study populations.^5,24 Although the median OS of children with DIPG remains less than 12 months, several reports have described survival beyond five years in approximately 2%-3% of patients.^21,25–28 However, earlier studies of very long-term survivors utilized inconsistent radiologic definitions of DIPG and often lacked corresponding tumor tissue, precluding histologic comparison.^25,27,28 An adequate understanding of the clinical heterogeneity and prognostic factors of this disease therefore similarly demands consistency and accuracy of neuroradiology assessment. Additionally, inter-observer variability in interpretation of diagnostic MRI as having typical or nontypical characteristics of DIPG has been demonstrated.^19,29 The importance of central neuro-imaging review for confirming eligibility to DIPG clinical trials has been increasingly recognized^4,5 and recommended by the Response Assessment in Pediatric Neuro-Oncology (RAPNO) working group,²⁴ yet its diagnostic accuracy has not been formally investigated. Evaluation of the precision of central neuro-imaging review is, therefore, critical to ensuring appropriate inclusion of a uniform population of patients with DIPG in prospective trials.

Herein, we analyze a combination of histopathologic and centrally-reviewed diagnostic MRI data for patients referred to the International DIPG Registry (IDIPGR) based upon a suspected clinico-radiologic diagnosis of DIPG by respective local institutions. Our objectives were to (1) determine the prevalence of IDIPGR cases with histopathologic and/or centrally-reviewed radiologic evaluation inconsistent with DIPG and (2) assess the accuracy of central neuro-imaging review in confirming the diagnosis of DIPG using paired histopathology.

Methods

Study Population

Clinical, histopathologic, and radiologic data were abstracted retrospectively from the IDIPGR. Patients were referred to the registry as previously described,³⁰ after a diagnosis of DIPG at their local institution on the basis of clinical and imaging features. Patients were included in the histopathologic analysis if a specific diagnosis was available from biopsy and/or autopsy. Tumors diagnosed as “LGG”, “glioma,” or “astrocytoma” without further detail (n=24) were excluded due to inability to distinguish WHO grade I (inconsistent with DIPG) and grade II (potentially consistent with DIPG) histology (Supplementary Material S1). Patients were included in the radiologic analysis if diagnostic brain MRI was available with central review performed. Patients with both histopathologic and radiologic data were included in the combined analysis. Clinical data (age, gender, treatment, OS [time from diagnosis to death]) were abstracted using de-identified case-report forms. This study was approved by the IDIPGR scientific review board and Cincinnati Children’s Hospital IRB.

Histopathologic Evaluation

Histopathologic diagnoses from biopsy and/or autopsy were classified as consistent with DIPG if they included diffuse midline glioma, H3K27M-mutant, glioblastoma, anaplastic astrocytoma, diffuse astrocytoma, or high-grade glioma (HGG), not otherwise specified (NOS). WHO grade II-IV infiltrating gliomas diagnosed before 2016 with positive H3K27M-mutation testing (H3K27M-mutant immunohistochemistry and/or targeted sequencing) were reclassified as diffuse midline glioma, H3K27M-mutant.¹² Diagnoses other than the five above were classified as inconsistent with DIPG (i.e., nondiffuse LGGs, embryonal tumors, non-neoplastic tissue). One patient with inadequate tumor sampling on biopsy (reported as normal brain, but absent biopsy tract engagement with tumor was confirmed on postoperative imaging; OS: 5 months) was excluded. Central pathology review was performed by an experienced neuro-pathologist (C.F.) when tissue was submitted; otherwise, pathology reports from local institutions were used.

Radiologic Evaluation

Diagnostic brain MRIs were centrally reviewed by two board-certified, fellowship-trained pediatric neuro-radiologists with >25 years of postfellowship experience each, including extensive neuro-imaging evaluation in DIPG clinical trials (J.L.L, B.J.). T1-weighted, T2-weighted, T2-FLAIR, T1-postcontrast, diffusion, gradient-recalled echo, or susceptibility-weighted imaging were evaluated when available as described previously.⁴ Cases were categorized as: (i) characteristic of DIPG (arising from and involving ≥50% of the pons with diffuse pattern of involvement),^1–4 (ii) likely DIPG with some unusual features (often large areas of necrosis or hemorrhage), but otherwise characteristic of DIPG, or (iii) non-DIPG, alternative diagnosis suspected. Imaging features suggestive of alternative diagnoses included nonpontine origin, <50% pontine involvement, focally exophytic morphology of most of the lesion, sharply-defined margins, and/or marked diffusion restriction throughout the lesion. One reviewer would assess the case and fill out a standardized form. The second reviewer then assessed the case, referencing the primary reviewer’s form. Consensus opinion was utilized if there were discrepancies; all imaging features were ultimately assessed and agreed upon by both reviewers. Reviewers were blinded to clinical data, including histopathologic diagnosis, at initial review. All cases with imaging suggestive of alternative diagnoses and/or histopathology inconsistent with DIPG were re-reviewed for this study, without changes in impression.

Statistical Analyses

Clinical, histopathologic, and radiologic characteristics were summarized using descriptive statistics. Comparisons of OS, age, gender, and treatment between histopathology and imaging impression subgroups were performed using t-tests or one-way ANOVA, log-rank, and Chi-square analyses, conducted in SPSS (v26, IBM). Univariate analyses comparing imaging impression with histopathology were performed using the Freeman-Halton extension of the Fisher exact probability test and Chi-square analysis.

Results

Patient Cohort

A total of 704 patients were analyzed, with clinical and demographic characteristics summarized in Table 1. This cohort included patients with adequate histopathologic data (n = 286) and/or centrally-reviewed imaging (n = 569); 151 of these patients had both available and were also included in the combined analysis. Among the entire cohort, median age at diagnosis was 6.4 years, 53% were female, and median OS was 10 months. Eighty-nine percent of patients were treated with irradiation, 9% underwent re-irradiation, and 71% received systemic therapy.

Histopathology

Figure 1 summarizes the distribution of histopathologic diagnoses for 286 patients with tissue from biopsy (n = 147 [51%]), autopsy (n = 116 [41%]), or both biopsy and autopsy (n = 23 [8%]). Among 286 patients, 262 (92%) had histopathologic diagnoses consistent with DIPG, including diffuse midline glioma, H3K27M-mutant (n = 109), glioblastoma (n = 66), anaplastic astrocytoma (n = 51), diffuse astrocytoma (n = 21), and HGG NOS (n = 15). Forty diffuse gliomas had BRAF^V600E testing, negative in all; further molecular profiling was not available.

Fig. 1

Distribution of histopathologic diagnoses.

Open in new tabDownload slide

Notably, 23 (8%) patients had histopathologic diagnoses inconsistent with DIPG, most commonly nondiffuse low-grade glial neoplasms (pilocytic astrocytoma [n = 11], low-grade glioneuronal tumor [n = 1]) and embryonal tumors (primitive neuroectodermal tumor (PNET)/embryonal tumor, NOS [n = 7], medulloblastoma [n = 2]). Rare additional inconsistent diagnoses included germinoma (n = 1) and “gliocyte proliferation” (n = 1 [tissue was not submitted for central pathology review and imaging was not available to assess biopsy tract engagement with tumor; however, the patient died 10 months postdiagnosis, suggesting likely insufficiently sampled high-grade neoplasm]). Genomic testing was not available in most inconsistent cases.

One patient was diagnosed with a H3K27M-mutant, BRAF^V600E-wildtype ganglioglioma by locally-reviewed biopsy and autopsy (OS: 14 months). Given still uncertain biological significance of H3K27M mutations in the setting of noninfiltrative histology, yet likely propensity for more aggressive behavior,³¹ this case was not classified in either histopathologic category.

Pathology was centrally reviewed in 80 (28%) patients. In 49 patients, separate local pathology reports were also available, with zero discordance observed in classification as consistent versus inconsistent with DIPG, though specific diagnoses sometimes differed, and most inconsistent cases were only reviewed locally (Supplementary Material S2).

Twenty-one patients had paired tissue from biopsy and autopsy with sufficiently specific histopathologic diagnoses for comparison. All tumor pairs were histologically consistent with DIPG. In two cases, biopsy diagnosis was anaplastic astrocytoma and autopsy diagnosis was glioblastoma (H3K27M-mutation testing was not performed, and central pathology review was only possible on one postmortem sample). All other paired biopsy and autopsy specimens shared the same histopathologic diagnosis and grade.

When comparing patients with histopathologic diagnoses consistent or inconsistent with DIPG, there were no significant differences in age, gender, or median OS (Table 1). However, patients with histopathologic diagnoses inconsistent with DIPG had improved 5-year OS (10% versus 2%, p = 0.049) and a trend toward improved 2-year OS (19% versus 8%, p = 0.11). Among patients with histopathologic diagnoses discordant with DIPG, there was a trend toward higher median OS in nondiffuse LGGs compared to embryonal tumors (11 versus 7 months, p = 0.091).

Central Neuro-Imaging Review

Central neuro-imaging review was performed on diagnostic brain MRIs in 569 patients (including and expanding upon 400 IDIPGR cases previously described⁴). Among these 569 patients, 374 (66%) were classified as characteristic DIPG by imaging appearance, 155 (27%) as likely DIPG, but with some unusual imaging features, and 40 (7%) as non-DIPG, alternative diagnosis suspected.

Among 40 patients with imaging features suggestive of non-DIPG diagnoses, the most common reason for radiologic exclusion was nonpontine tumor origin (n = 25), with the following distribution/details of primary tumor locations: medullary (n = 11), cerebral peduncle (n = 2), thalamic or bi-thalamic (n = 5), cerebellar (n = 3), tectal (n = 2), and gliomatosis with brainstem involvement (n = 2). Additional imaging features suggestive of alternative diagnoses included sharply-defined disease margins (n = 8), primarily exophytic tumor growth (n = 9), marked diffusion restriction throughout (n = 4), and/or <50% pontine involvement (n = 3). Although not the sole reason for exclusion, a subset of tumors also demonstrated unusual (circumscribed/homogenous or extensive) enhancement patterns (n = 3) or dissemination (n = 4 [spinal, posterior fossa, and/or leptomeningeal metastases]).

There were no significant differences in patient age or gender between imaging impression groups (Table 1). Imaging impression correlated with prognosis, with significant differences in median OS as well as two- and five-year survival rates. Patients with imaging features suggestive of alternative diagnoses had prolonged OS compared to patients with likely DIPG or imaging characteristic of DIPG (median OS: 15 versus 10 versus 11 months respectively, p < 0.001; 2-year OS: 34% versus 11% versus 6%, p < 0.001; 5-year OS: 21% versus 5% versus 1%, p < 0.001) (Table 1, Figure 2B).

Fig. 2

A. Combined radiologic and histopathologic analysis. B–C. OS based on centrally-reviewed imaging impression (B) and combined imaging impression and histopathology (C).

Open in new tabDownload slide

Combined Histopathologic and Radiologic Analysis

There was a significant association between the three central neuro-imaging review impressions and histopathology (consistent or inconsistent with DIPG) among 151 patients included in the combined analysis (p < 0.001, Figure 2A). Additionally, when central neuro-imaging review impression was dichotomized to characteristic of or likely DIPG versus suggestive of an alternative diagnosis, there was a significant correlation with histopathology (consistent or inconsistent with DIPG) (p < 0.001).

Among 77 patients with imaging classified as characteristic of DIPG, 76 (99%) had histopathologic diagnoses consistent with DIPG (diffuse midline glioma, H3K27M-mutant, glioblastoma, anaplastic astrocytoma, diffuse astrocytoma, and HGG NOS; Supplementary Material S3). Among 57 patients classified as likely DIPG with some unusual imaging features, 55 (96%) had histopathologic diagnoses consistent with DIPG (Supplementary Material S4). Imaging features of cases classified as characteristic or likely DIPG with consistent histopathology are summarized in Supplementary Material S5, most with >2/3 pontine involvement (92%), T1-hypointensity (95%), and T2-hyperintensity (98%), yet variable diffusion restriction, heterogeneity, hemorrhage, enhancement, and necrosis. There was no significant difference in distribution of histologic grades II-IV between patients classified radiologically as characteristic versus likely DIPG and none of the visually-assessed imaging features correlated with histologic grade (Supplementary Materials S5 and S6). Three patients within these two imaging impression groups with histopathologic diagnoses inconsistent with DIPG all had embryonal tumors (PNET) by local pathology report (tissue not submitted for central review; H3K27M-mutation testing was negative in one; no additional molecular testing was available). MRIs of these embryonal tumors (Figure 3) demonstrated infiltrative pontine masses, with absent diffusion restriction and enhancement in one, large amount of hemorrhage limiting evaluation of diffusion restriction in the second, and a localized zone of diffusion restriction in the third. Diagnostic spine imaging was negative in one case and not available in the other two. OS for these three patients ranged from zero to seven months.

Fig. 3

Embryonal tumors not excluded by imaging criteria. (T2, T2-weighted; FLAIR, fluid-attenuated inversion recovery; +C, contrast-enhanced; T1, T1-weighted; DWI, diffusion-weighted image; GRE, gradient-recalled echo).

Open in new tabDownload slide

Among 17 patients with imaging features suggestive of alternative diagnoses, eight (47%) had histopathologic diagnoses inconsistent with DIPG, predominantly embryonal tumors, and nondiffuse low-grade gliomas. Reasons for radiologic exclusion are summarized in Figure 2A and Supplementary Material S7. Two patients had primary pontine embryonal tumors (PNET/embryonal tumor NOS), which were excluded due to well-circumscribed, noninfiltrative appearance (Figure 4A and B); both also exhibited uncharacteristic extensive spinal metastases. The remaining six patients had nonpontine brainstem tumor origin in addition to other uncharacteristic imaging features. Three pilocytic astrocytomas and one low-grade glioneuronal neoplasm were medullary-origin tumors with well-defined borders and/or dorsal focally exophytic morphology (Figure 5A). The medulloblastoma demonstrated cerebellar origin with marked diffusion restriction (Figure 4C), and the germinoma had tectal/pineal origin with exophytic appearance.

Fig. 4

Three embryonal tumors excluded as non-DIPG by imaging criteria. (For cases A. and B., diffusion imaging was not available.).

Open in new tabDownload slide

Fig. 5

Example cases excluded by nonpontine origin.

Open in new tabDownload slide

Eight (47%) patients had histopathologic diagnoses consistent with infiltrating grade II-IV gliomas; however, all were excluded radiologically due to nonpontine origin (Figure 5B); these tumors exhibited primary bi-thalamic, thalamic, cerebellar, or cerebral peduncle origin, therefore representing nonpontine diffuse midline gliomas.

The aforementioned unique H3K27M-mutant ganglioglioma had imaging features suggestive of an alternative diagnosis (nonpontine origin, sharply-defined margins, exophytic morphology; Figure 5C).

A significant difference in likelihood of patients undergoing diagnostic biopsy was observed between imaging impression subgroups (non-DIPG = 83%, likely DIPG = 56%, characteristic of DIPG = 36%; p < 0.001).

Additional survival results.

—When comparing the above eight patients with imaging features suggestive of alternative diagnoses and histopathology inconsistent with DIPG with 131 patients with imaging characteristic of or likely DIPG and histopathology consistent with DIPG, there were significant differences in median OS (27 versus 11 months, p = 0.004) and 2-year OS (57% versus 11%, p < 0.001) (Figure 2C). The two longest survivors in the former group had brainstem pilocytic astrocytomas (OS:12 years and >13 years [still alive]). The two patients with pontine embryonal tumors with imaging suggestive of non-DIPG diagnoses experienced poorer outcomes, with OS of five and 27 months.

Of 129 patients in the latter group (both centrally-reviewed imaging and histopathology consistent with DIPG) with sufficient survival data, six (4.7%) survived beyond five years. Histopathologic diagnoses included diffuse midline glioma, H3K27M-mutant (n = 2), anaplastic astrocytoma (n = 1 [H3K27M-mutation testing not performed]), and diffuse astrocytoma (n = 3 [negative H3K27M-mutation testing in two, not performed in one]). Median age at diagnosis was 18 years (range: 6-27) and treatments included upfront radiotherapy (n = 6), systemic therapy (n = 5), and re-irradiation (n = 1).

H3K27M-mutational status and central neuro-imaging review

.—Tumor H3K27M-mutational status was available for 75 (50%) patients from the combined analysis. Among 40 patients with characteristic DIPG by imaging with H3K27M-mutation testing available, 33 (83%) were H3K27M-mutant. Similarly, among 31 patients classified as likely DIPG with some unusual imaging features, 24 (77%) were H3K27M-mutant. Only four patients with imaging features suggestive of alternative diagnoses had H3K27M-mutation testing, positive in two. There was no significant difference in H3K27M mutation prevalence between the three imaging impression groups (p = 0.28), between patients in the former two imaging groups (characteristic versus likely DIPG, p = 0.59), or between patients with imaging characteristic or likely DIPG versus suggestive of alternative diagnoses (p = 0.15).

Disseminated disease.

—Diagnostic spine MRI results were available in a subset of patients from local reports and/or central review when imaging was submitted. Among patients in the combined analysis with brain imaging characteristic of or likely DIPG, 64 had spine MRIs, with metastases in six (9%), all grade III-IV infiltrating gliomas. Of patients with brain imaging suggestive of alternative diagnoses, nine had at least partial spinal imaging, with dissemination reported in five, including four with inconsistent histopathology (aforementioned two embryonal tumors, pilocytic astrocytoma, germinoma). Intracranial leptomeningeal disease at diagnosis was extremely rare, identified in 1.5% (8/529) centrally-reviewed brain MRIs classified as characteristic or likely DIPG (five had corresponding histopathology, all grade III-IV gliomas). Leptomeningeal dissemination was observed in one case with imaging suggestive of an alternative diagnosis (tissue not available), as noted above.

Discussion

As DIPG largely remains a clinico-radiologic diagnosis without routine tissue acquisition, reliable imaging distinction between DIPG and other pontine tumors with potentially divergent outcomes and treatment paradigms is critical. To our knowledge, this is the first study to formally evaluate the diagnostic accuracy of central neuro-radiology review, utilizing a combination of histopathologic and MRI data from a multi-institutional cohort of cases submitted to the IDIPGR based upon a suspected diagnosis of DIPG. Findings also provide valuable insight into the prevalence and corresponding features of IDIPGR-enrolled tumors with histopathologic and/or centrally-reviewed radiologic assessments inconsistent with DIPG.

Central neuro-imaging review, performed by experienced neuro-radiologists and based upon clearly-defined criteria, demonstrated high accuracy in radiologically confirming the diagnosis of DIPG and distinguishing DIPG from other, histologically discordant pontine tumors. Among patients classified as having imaging characteristic of or likely DIPG, 98% (131/134) had tumor histopathology from biopsy and/or autopsy consistent with DIPG (grade II-IV infiltrating gliomas [glioblastoma, anaplastic astrocytoma, diffuse astrocytoma, HGG NOS], most now reclassified as H3K27M-mutant diffuse midline glioma). Despite increased recognition of the need for central imaging review to verify eligibility on DIPG clinical trials in the absence of histologic confirmation,^4,5,24 the accuracy of this process has not previously been extensively investigated. Furthermore, concerns about potential misdiagnosis and inclusion of patients with uncharacteristic imaging features on earlier studies have precluded comparison of treatment efficacy and limited understanding of prognostic factors in this disease.^5,25,27,28 Our results suggest that centralized review with strict, standardized criteria can be relied upon for radiologic confirmation of DIPG, thereby ensuring uniform patient populations within and across future trials.

Appropriate exclusion of non-DIPG brainstem tumors by central neuro-imaging review was similarly demonstrated. Among 17 patients in the combined analysis for whom centralized imaging interpretation revealed features indicative of alternative diagnoses, approximately half had histopathology inconsistent with DIPG–most commonly embryonal tumors or pilocytic astrocytomas (all remaining patients with histopathology consistent with infiltrating grade II-IV gliomas were excluded due to nonpontine tumor origin, instead likely representing nonpontine diffuse midline gliomas). Our findings corroborate prior reports describing embryonal tumors and pilocytic astrocytomas presenting in the pons and throughout the brainstem, but often exhibiting radiologic properties distinguishable from DIPG, especially well-circumscribed disease borders, focally exophytic morphology, and/or dense enhancement (in the latter).^{3,13,15,17,22,23} While all nondiffuse LGGs and most embryonal tumors were correctly excluded by centralized review utilizing the above radiologic criteria, embryonal tumors (PNET/embryonal tumor, NOS) comprised the rare (2% [3/134]) cases of discordance between central neuro-imaging classification (characteristic or likely DIPG) and inconsistent histopathology. Specimens were not available for central pathology review and molecular testing was limited, impeding further interpretation, but the potential for misidentification of pontine embryonal tumors as DIPG is illustrated and deserves exploration in future research. These results also support inclusion of diffusion-weighted imaging in DIPG trials to aid in radiologically excluding pontine embryonal tumors, given marked diffusion restriction classically associated with this histology, including of primary pontine location^15,19; however, this sequence was not uniformly obtained and one misdiagnosed embryonal tumor demonstrated uncharacteristically absent diffusion restriction. Additionally, previous studies of brainstem embryonal tumors have consistently described very young ages at presentation (under three years)^{13–15,20,21}; in our cohort, patients with histologically-confirmed embryonal tumors displayed similar trends toward lower median age at diagnosis compared to patients with histopathology consistent with DIPG (4 versus 7 years), but this did not reach statistical significance, albeit possibly limited by the former’s small sample size. Presenting age may, therefore, be considered in conjunction with aforementioned essential MRI criteria to guide differentiation of these entities, especially in less certain cases.

Taken together, our study provides histopathologic validation of specific imaging criteria for the radiologic diagnosis of DIPG (infiltrative, mostly T1-hypointense/T2-hyperintense tumors with >50% [usually >67%] pontine involvement) or exclusion as non-DIPG (well-circumscribed and/or focally exophytic morphology), which should be utilized both in prospective trials (performed by central radiology review) and routine clinical practice. In addition to illustrating the diagnostic accuracy of these collective criteria, our findings underscore their importance for appropriately identifying non-DIPG pontine tumors with potentially more favorable prognoses and treatment considerations. Among 286 patients enrolled on the IDIPGR with biopsy and/or autopsy data available, 23 (8%) had histopathology inconsistent with DIPG, often nondiffuse LGGs and embryonal tumors. Correspondingly, of 569 patients with centrally-reviewed diagnostic MRI, 40 (7%) had imaging features suggestive of an alternative diagnosis. Improved outcomes were observed among patients with histopathology and/or radiology inconsistent with DIPG. Centralized interpretation of diagnostic MRI was prognostic of OS, with significantly prolonged survival demonstrated in patients classified radiologically as having features suggestive of non-DIPG diagnoses compared to patients with imaging features characteristic of or likely DIPG, expanding upon similar results described by Fonseca et al.²¹ Furthermore, within the combined analysis, patients with centrally-reviewed imaging features suggestive of alternative diagnoses and histopathology inconsistent with DIPG exhibited significantly improved OS compared to patients with imaging features characteristic of or likely DIPG and histopathology consistent with DIPG. Given the higher proportion of histologic data coming from diagnostic biopsy as opposed to autopsy in patients classified by central neuro-radiology review as non-DIPG, local institutions’ suspicion of DIPG and pursuit of upfront tissue acquisition were likely appropriately guided by application of the above criteria.

Among cases with paired histopathologic and radiologic data, patients with brainstem pilocytic astrocytomas had the most favorable outcomes (two of three alive >10 years). Conversely, patients with embryonal tumors experienced uniformly poor prognoses and worse outcomes than nondiffuse LGGs; none survived beyond 30 months and all but one died within one year despite receiving irradiation and/or chemotherapy. Our findings are consistent with prior studies describing potential for prolonged OS in patients with pontine pilocytic astrocytomas which were histologically confirmed and/or suspected based on noninfiltrative imaging appearance,^3,19,21–23 as well as poor outcomes among children with biopsy- or autopsy-proven brainstem embryonal tumors.^13,14 The favorable prognostic impact of centralized radiologic interpretation and/or histopathology incongruous with DIPG is thus likely driven by the improved survival observed in nondiffuse LGGs on average; however, survival was variable within this histology (assessed across the entire cohort) and genomic profiling was not available to further characterize tumors with worse outcomes. Appropriate recognition of pontine embryonal tumors is also important, despite uniformly poorer prognoses, given different treatment considerations, such as intensive chemotherapy, autologous stem cell transplant, and craniospinal irradiation.^15,20 Therefore, whereas the role of surgical biopsy remains in question in DIPG,^6–8 tissue acquisition should be strongly encouraged in patients with pontine tumors for which DIPG is considered unlikely radiologically to determine diagnosis, prognosis, and therapy.

Incorporating centralized neuro-imaging review for eligibility verification on DIPG prospective trials and retrospective research endeavors will enhance our understanding of the clinical heterogeneity and prognostic factors underlying this disease. Survival beyond five years has been demonstrated in approximately 2-3% of patients across previous studies^21,25–28; however, interpretation of earlier reports has been limited by unreliable clinico-radiologic definitions of DIPG and scarcity of corresponding tumor tissue for histologic comparison.^19,25 With application of stricter imaging criteria for inclusion as DIPG, as performed by Hoffman et al.,²⁵ a subset of previously described very long-term survivors are suspected to have alternative diagnoses due to the presence of uncharacteristic radiologic features (well-marginated, exophytic, enhancing, and/or nonpontine tumor appearance).^27,28 In our combined analysis, six (4.7%) patients with both centrally-reviewed imaging features and histopathology consistent with DIPG survived more than five years. Notably, median age at diagnosis was 18 years and five underwent adjuvant systemic therapy, in accordance with known associations between older age and receipt of systemic therapy with prolonged OS in DIPG.²⁵ Utilizing paired centralized radiologic and histopathologic data, our findings validate this rare subset of very long-term survivors of DIPG (including two with biopsy-proven H3K27M-mutant diffuse midline glioma) who deserve continued investigation.

No significant difference in H3K27M-mutational status was identified between the centralized radiologic impression groups. Approximately 80% of tumors classified as having imaging features either characteristic of or likely DIPG tested positive for the H3K27M mutation, consistent with prior reports describing difficulty distinguishing between H3K27M-mutant and wildtype DIPG tumors on the basis of radiologic visual assessment alone.^4,19,21 DIPG clinical trials evaluating treatments that specifically target H3K27M-mutant biology should therefore consider requiring tissue for molecular confirmation.

Our study was limited by the small fraction of tumors submitted for central pathology review, requiring reliance on local reports and impeding confirmation of histologic diagnoses, especially in most cases inconsistent with DIPG. Additionally, H3K27M-mutation testing was not available in all cases, including most nondiffuse low-grade gliomas, and it was not possible to assess H3K27 trimethylation loss, precluding further genomic classification. Incorporation of comprehensive molecular profiling, including universal H3K27M-mutation and trimethylation assays, as well as detailed methylation testing in reported pontine embryonal tumors, will be critical in future research. While central radiology reviewers were blinded to clinicopathologic information, the second reviewer had access to the first reviewer’s impression, which may confer bias; however, concordance between these two reviewers in independently distinguishing characteristic or likely DIPG versus non-DIPG was high in a subset of patients (0.967).⁴ Moreover, the precision of consensus opinion adjudicating rare discrepant cases suggests a similar collaborative approach can be applied when interpreting challenging MRIs. Finally, although not possible in the present analysis, it will be important to explore clinical symptomatology correlates of imaging impressions.

Despite these limitations, our study demonstrates the high diagnostic accuracy of centralized imaging review for radiologic confirmation of DIPG in clinical trials, and provides histopathologic validation of specific criteria for use in routine clinical practice as well. Precise radiologic distinction between DIPG and other pontine tumors is essential, as almost 10% of patients enrolled on the IDIPGR had histopathology and/or centrally-reviewed diagnostic MRI features inconsistent with DIPG, often representing nondiffuse LGGs and embryonal tumors. In the absence of routine tissue acquisition, central neuro-imaging review with strict criteria can be relied upon for appropriate inclusion of a uniform population of patients with DIPG in prospective trials as well as appropriate exclusion of alternative pontine tumors with potentially more favorable prognoses, treatment considerations, and thresholds for biopsy.

Acknowledgements

We thank the patients and families for their invaluable contribution to this research. We thank the IDIPGR operations team for data support and the organizations above for financial support of the IDIPGR.

Funding

Supporting foundation organizations include: The Cure Starts Now Foundation, The Cure Starts Now Australia, Brooke Healey Foundation, Wayland Villars Foundation, Aidan’s Avengers, Aubreigh’s Army, Austin Strong, Cure Brain Cancer, Jeffrey Thomas Hayden Foundation, Laurie’s Love Foundation, Love Chloe Foundation, Musella Foundation, Pray Hope Believe, Reflections Of Grace, Storm the Heavens Fund, Whitley’s Wishes, Gabriella’s Smile Foundation, The Gold Hope Project, The Isabella and Marcus Foundation, Lauren’s Fight for Cure, Robert Connor Dawes Foundation, Ryan’s Hope, Benny’s World, Lily Larue Foundation, Marlee’s Mission, RUN DIPG, American Childhood Cancer Organization, The DIPG Collaborative, Snapgrant.com, the Kyler Strong Foundation, Keris Kares.

Conflict of Interest statement. None.

Authorship statement. Conception and design: J.L.L., M.A.L, M.F., C.F., M.D. Data acquisition: All. Data analysis/interpretation: J.L.L., M.A.L, M.F., C.F., M.D., A.L., B.C. Drafting/revising critically for important intellectual content: All. Final approval of submission: All.

References