Patterns of care and survival in patients with multifocal glioblastoma: A Danish cohort study

Abstract

Background

This Danish cohort study aims to (1) compare patterns of care (POC) and survival of patients with multifocal glioblastoma (mGBM) to those with unifocal glioblastoma (uGBM), and (2) explore the association of patient-related factors with treatment assignment and prognosis, respectively, in the subgroup of mGBM patients.

Methods

Data on all adults with newly diagnosed, pathology-confirmed GBM between 2015 and 2019 were extracted from the Danish Neuro-Oncology Registry. To compare POC and survival of mGBM to uGBM, we applied multivariable logistic and Cox regression analysis, respectively. To analyze the association of patient-related factors with treatment assignment and prognosis, we established multivariable logistic and Cox regression models, respectively.

Results

In this cohort of 1343 patients, 231 had mGBM. Of those, 42% underwent tumor resection and 41% were assigned to long-course chemoradiotherapy. Compared to uGBM, mGBM patients less often underwent a partial (odds ratio [OR] 0.4, 95% confidence interval [CI] 0.2–0.6), near-total (OR 0.1, 95% CI 0.07–0.2), and complete resection (OR 0.1, 95% CI 0.07–0.2) versus biopsy. mGBM patients were furthermore less often assigned to long-course chemoradiotherapy (OR 0.6, 95% CI 0.4–0.97). Median overall survival was 7.0 (95% CI 5.7–8.3) months for mGBM patients, and multifocality was an independent poor prognostic factor for survival (hazard ratio 1.3, 95% CI 1.1–1.5). In mGBM patients, initial performance, O[6]-methylguanine-DNA methyltransferase promotor methylation status, and extent of resection were significantly associated with survival.

Conclusions

Patients with mGBM were treated with an overall less intensive approach. Multifocality was a poor prognostic factor for survival with a moderate effect. Prognostic factors for patients with mGBM were identified.

Approximately 20% of patients with newly diagnosed histology-confirmed glioblastoma (GBM), the most aggressive primary brain tumor in adults, present with multifocal disease.^1–4 Multifocal GBM (mGBM) is typically defined as multiple synchronous contrast-enhancing lesions on T1-weighted MRI scan. A commonly used subcategory is multicentric GBM, which often refers to multiple widely separated contrast-enhancing lesions without a known anatomical tract for spread of disease between these.^2,5,6 Without clear clinical impact of subcategories^4–6 and with improved understanding of GBM evolution and migration^7–9 and technologic advances in imaging,¹⁰ multifocality, multicentricity, and all its variations are increasingly considered as manifestations of the same disease spectrum.^3,11

Although tumor focality and its spectrum are often presumed to be correlated with survival,^1,2,4,12,13 clinical research thus far is retrospective only, and has shown conflicting results of multifocality as adverse prognostic factor for survival.^2,6 In the largest series reporting on 7785 patients with mGBM from the United States National Cancer Data Base, multifocality was independently associated with a shorter survival.¹ The same result was observed in a registry-based Danish series reporting on 198 patients with mGBM treated between 2009 and 2014.¹³ In a single-center series on 231 mGBM patients treated at University Hospital Essen, Germany, specifically bihemispheric multifocality was independently associated with a shorter survival.⁴ Nonetheless, in other smaller series, multifocality was not independently associated with a shorter survival.^3,5 Present knowledge is thereby conflicting and mainly based on single-center series^3–5 or selected populations.^1,14,15 Additionally, O[6]-methylguanine-DNA methyltransferase (MGMT) promotor methylation status, a very important prognostic factor in the era of temozolomide,¹⁶ has not been systematically assessed in the present literature, limiting the interpretability of those results.^1,3–5,13 Moreover, possible variation in prognosis between mGBM patients has been explored very limitedly to date.^1,15

While patients with mGBM were eligible for the pivotal GBM trials (ie, multifocality was not an exclusion criterion), results for the mGBM subgroup have not been published.^17–19 Additionally, to the best of our knowledge, there are no randomized trials exclusive to mGBM patients. Standard of care for newly diagnosed mGBM, therefore, follows the general treatment recommendations for GBM, consisting of a maximal safe resection, followed by conventionally fractionated chemoradiotherapy (CRT) or alternatively hypofractionated short-course CRT in case of a poor prognosis, and adjuvant chemotherapy (ie, temozolomide).^18,20,21 Although patterns of care (POC) in mGBM have been sparsely described, it has been suggested that patients with mGBM receive an overall less intensive treatment approach compared to those with unifocal GBM (uGBM).^1,3,22 Contributing factors for this possible difference have not yet been studied, but could very well be related to additional technical challenges regarding both surgery and radiotherapy (RT).^11,12

Hence, there is very limited knowledge on whether mGBM is a distinct type of GBM on biological and clinical levels. Likewise, questions regarding the optimal and possible need for differentiated treatment for mGBM are unanswered. We thus need to increase our understanding of the possible impact of tumor focality in GBM patients, required to enhance prognostication and clinical decision-making in the future. To do so, it is essential to study an unselected GBM population treated at multiple centers. The first aim of the present retrospective population- and registry-based Danish cohort study is, therefore, to compare POC and survival of mGBM to uGBM patients treated from 2015 to 2019 in conjunction with known prognostic factors. The second aim is to explore the association of patient-related factors with treatment assignment and prognosis, respectively, in patients with mGBM.

Methods

Nationwide Retrospective Cohort Study

This is a retrospective registry-based cohort study of histology-confirmed GBM patients operated in Denmark over a 5-year period. The cohort period started at 2015, which marks the commencement of mandatory MGMT-promotor methylation status assessment at diagnosis in Denmark, up to and including 2019, with completed follow-up on December 6, 2023. The RECORD statement was followed in the reporting of this study.²³

Neuro-oncology in Denmark

Neuro-oncology (surgery, RT, and chemotherapy) in Denmark has been traditionally centralized to 4 university hospitals (yearly volume of ca. 30–115 patients with GBM diagnosis per hospital). Multidisciplinary Tumor Boards have been anchored to these hospitals. The Danish Neuro-Oncology Group (DNOG) has been responsible for establishing and implementing national guidelines since 2008 (http://www.dnog.dk). These guidelines have followed international guidelines for general treatment recommendations. In addition, according to the DNOG guideline, tumor focality has been considered in neurosurgical planning. In clinical practice, this has meant that specifically the total tumor load and location(s), besides general factors, have been considered to plan or omit surgical intervention. Surgical intervention has often been restricted to biopsy and targeting 1 lesion. A gross total resection has been considered if the lesions were resectable and located in the same lobe. While multifocality has not been mentioned explicitly in the DNOG guideline regarding RT and chemotherapy indications, hypofractionated short-course radiotherapy (HFRT) with concomitant chemotherapy if possible, has been recommended in case of a poor prognosis in general. In clinical practice, this has meant that patients with a poor prognosis, assessed by several prognostic factors, have been offered short-course HFRT.²⁴ Additionally, for mGBM patients specifically, target volume and dose reductions have been considered to comply with the principles of partial brain irradiation. Postoperative RT has been administered with external beam RT using photon irradiation. Chemotherapy has been prescribed as temozolomide monotherapy, unless a patient had, for example, a contraindication or participated in a trial.

Danish Neuro-oncology Registry and Study Population

The Danish Neuro-Oncology Registry (DNOR) was used to identify all adults (≥18 years at the time of first surgery) with newly diagnosed, pathology-confirmed GBM in accordance with the WHO criteria at the time (SNOMED codes M94403, M94404, M94406, M94407, M94409, M94413, M94423), and furthermore used to extract all relevant data (http://www.dnog.dk/DNOR). All eligible patients according to the criteria above were included, irrespective of tumor focality. The only exclusion criterion applied was missing data on tumor focality. The date of first surgery defined the year of and patient’s age at diagnosis. The DNOR has been described by Hansen et al. showing very high completeness of patient registration and good validity of the registered patient data.²⁵

In the DNOR, mGBM has been defined as more than 1 contrast-enhancing lesion on T1-weighted MRI scan which are not connected by T2FLAIR abnormalities. Data on the size and location of the distinct lesions have been reported for up to 3 lesions and only for those measuring more than 10 mm. Isocitrate dehydrogenase (IDH) mutation status was not systematically obtained throughout the cohort period. Detailed information on the assessment method of IDH mutation and MGMT-promotor methylation status per patient has not been registered. In general, the assessment methods have changed during this cohort period. Only primary treatment modalities were assessed. Mortality data were automatically obtained from the Danish Civil Registration System.

Extent of Resection

Extent of resection (EOR) was deducted from codes for the surgical procedure and result of the direct postoperative MRI (≤72 h) following the principles described by Karschnia et al.²⁶ Scoring of a complete and near-total resection was MRI-scan-based; complete resection if the MRI showed no residual contrast-enhancing tumor; near-total resection if the MRI showed nonmeasurable residual contrast-enhancing tumor (diameter <10 mm). Scoring of a partial resection and biopsy was based on a surgical code and MRI-scan result, if available; partial resection, if a code for resection was registered and no direct postoperative MRI scan was made or in combination with a direct postoperative MRI scan showing measurable contrast-enhancing residual tumor (both diameters ≥10 mm); biopsy, if a code for biopsy was registered and no direct postoperative MRI scan was made or in combination with a direct postoperative MRI scan showing measurable contrast-enhancing residual tumor.

Radiotherapy and Chemotherapy

RT and chemotherapy were considered primary treatment if registered as such and if started within 12 weeks, that is, 84 calendar days, of the previous treatment modality without progressive disease reported in the meantime. RT was classified into 3 types (planned prescribed dose); conventionally fractionated long-course RT based on a fraction size of 1.8–2 Gy and/or >15 fractions (eg, 54–60 Gy in 1.8–2.0 Gy per fraction)¹⁷; HFRT based on a fraction size of >2 Gy to a total dose of >30 Gy (eg, 40 Gy in 15 fractions or 34 Gy in 10 fractions)^27,28; and other. Planned chemotherapy was considered as concomitant treatment if registered as such and/or if it overlapped with RT. It was considered as adjuvant treatment if registered as such, and/or if it not overlapped with RT.

Statistical Analysis

All statistical analyses were performed using SPSS version 28. Significance was set at P < .05. Patient-related factors were compared between patients with mGBM and uGBM using the Pearson Chi-square or Mann–Whitney U test, where appropriate. To compare the EOR between mGBM and uGBM patients, we applied multivariable multinomial logistic regression analysis to correct for potential confounders (ie, year of diagnosis, sex, age at diagnosis, initial WHO performance status [PS] before surgery, tumor location, tumor with or without midline cross, and center). To compare the assignment to long-course CRT (ie, planned treatment) between mGBM and uGBM patients, we applied multivariable logistic regression analysis to correct for potential confounders (ie, year of diagnosis, sex, age at diagnosis, initial WHO-PS, tumor location, tumor with or without midline cross, MGMT-promotor methylation status, EOR, and center). Overall survival (OS) was measured from first surgery to death or last follow-up and was estimated using the Kaplan–Meier method. An event was defined as death from any cause, with patients censored at the last follow-up. The log-rank test was used to compare survival trends between subgroups. Multivariable Cox regression analysis was used to compare survival between mGBM and uGBM patients, while correcting for potential confounders (ie, sex, age at diagnosis, initial WHO-PS, tumor location, tumor with or without midline cross, MGMT-promotor methylation status, EOR, and center). In secondary analysis, the variable “IDH mutation status” was added. The proportional hazards assumption of the Cox models was examined by using the “log-minus-log” approach and stratified for variables not meeting the assumption. Lastly, to analyze the association of patient-related factors with treatment assignment and prognosis, we established multivariable logistic and Cox regression models in the subgroup of mGBM patients, respectively.

Ethics Statement

The study was exempted from review by the Central Denmark Region Committees on Health Research Ethics; no informed consent from the individual patients was needed. The study was approved by the Danish Patient Safety Authority (Reg.no. 31-1521-174) and the DNOR (Reg.no. DNOR-2020-03-02). The study was registered on the internal registry of research projects of the Central Denmark Region (Reg.no. 1-16-02-74-20).

Results

Study Population

The cohort consisted of 1346 newly diagnosed GBM patients. Tumor focality was not reported in 3 patients, resulting in a study population of 1343 patients, of whom 231 (17.2%) had multifocal disease at diagnosis. For 197 out of 231 mGBM patients, data on at least 2 lesions were collected; these data showed that the topography was unilobar in 48 patients and multilobar in 149 patients. For the remaining 34 mGBM patients, only data on 1 lesion were collected and topography was therefore unknown.

Patient and Tumor Characteristics

The patients with multifocal and unifocal disease had a median (range) age of 68.0 (30.8–93.5) and 66.6 (18.1-89.7) years at diagnosis, respectively. Patients with mGBM had a significantly lower initial WHO-PS compared to those with uGBM (P-value 0.02; Table 1). Compared to uGBM patients, frontal and parietal lobe tumor location (P-value <.001) and midline cross (P-value <.001) was more frequent in mGBM patients. A methylated MGMT-promotor was equally frequent in mGBM and uGBM patients (49% and 44%), respectively.

Table 1.

Comparison of patient and tumor characteristics of multifocal GBM to unifocal GBM patients in the nationwide Danish cohort 2015–2019 (n = 1343)

Characteristic	Unifocal, n = 1112 n (%)	Multifocal, n = 231 n (%)	P-value
Sex  Male  Female	573 (51.5) 539 (48.5)	114 (49.4) 117 (50.6)	.5
Age  ≤40 years  41–50 years  51–60 years  61–70 years  71–80 years  ≥81 years	42 (3.8) 84 (7.6) 237 (21.3) 365 (32.8) 319 (28.7) 65 (5.8)	8 (3.5) 16 (6.9) 45 (19.5) 79 (34.2) 72 (31.2) 11 (4.8)	.6
Initial WHO-PS before surgery  0  1  2  3  4  Unknowna	323 (29.0) 447 (40.2) 220 (19.8) 91 (8.2) 28 (2.5) 3 (0.3)	56 (24.2) 87 (37.7) 47 (20.3) 32 (13.9) 8 (3.5) 1 (0.4)	.02
Tumor locationb  Temporal  Frontal  Parietal  Occipital  Deep-seated  Unknowna	385 (34.6) 351 (31.6) 175 (15.7) 102 (9.2) 97 (8.7) 2 (0.2)	48 (20.8) 98 (42.4) 50 (21.6) 19 (8.2) 16 (6.9) 0 (0)	<.001
Tumor crosses midline  Yes  No  Unknowna	146 (13.1) 928 (83.5) 38 (3.4)	79 (34.2) 146 (63.2) 6 (2.6)	<.001
MGMT methylation-status  Methylated  Unmethylated  Not donea	489 (44.0) 578 (52.0) 45 (4.0)	112 (48.5) 102 (44.2) 17 (7.4)	.08
IDH-mutation status  Mutated  Wild type  Not done	31 (2.8) 818 (73.6) 263 (23.7)	5 (2.2) 164 (71.0) 62 (26.8)	.5

IDH = Isocitrate dehydrogenase; MGMT = O[6]-methylguanine-DNA methyltransferase; PS = performance status.

^aPatients with a variable unknown/not done are excluded from the respective comparative analyses.

^bPer patient report of tumor location for largest lesion.

Treatment Assignment

Twenty-five percent of the mGBM patients and 14% of the uGBM patients did not proceed to adjuvant treatment and were thus managed with surgery only, respectively (Figure 1A and Supplementary Table 1). Fifty percent of the mGBM patients and 65% of the patients with uGBM were treated with trimodality treatment (surgery, radiation, and chemotherapy, concomitant and/or adjuvant), respectively.

Figure 1.

Percentages of primary oncologic treatment (A), extent of resection (B), and postoperative radiotherapy type (C) in newly diagnosed GBM patients in the nationwide Danish cohort 2015–2019 (n = 1343), specified for tumor focality. CRT = chemoradiotherapy; CT = chemotherapy; RT = radiotherapy; S = surgery.

Extent of Resection in mGBM Versus uGBM

Of the mGBM patients, 58% underwent a biopsy, 23% a partial, 9% a near-total resection, and 10% a complete resection (Figure 1B). In comparison, of the uGBM patients, 24% underwent a biopsy, 24% a partial, 23% a near-total resection, and 30% a complete resection. In multivariable analysis, patients with a multifocal tumor, as opposed to those with a unifocal tumor, were less likely to undergo a partial resection (odds ratio [OR] 0.4, 95% confidence interval [CI] 0.2–0.6, P-value <.001), a near-total resection (OR 0.1, 95% CI 0.07–0.2, P-value <.001), and a complete resection (OR 0.1, 95% CI 0.07–0.2, P-value <.001), all compared with a biopsy (Supplementary Table 2).

Predictors of Tumor Resection in mGBM

In the subgroup of mGBM patients, all 16 patients with a deep-seated tumor underwent a biopsy and were excluded from this multivariable analysis (Table 2). Another 7 patients were excluded due to missing data. Of the remaining 208 mGBM patients included in this analysis, 94 underwent a tumor resection (partial, near-total, or complete resection). In multivariable analysis, patients with a poorer initial WHO-PS and those with a tumor with midline cross were less likely to undergo a tumor resection. Treatment in center C was associated with a higher likelihood of tumor resection.

Table 2.

Multivariable logistic regression analysis to investigate the association between patient-related factors and assignment to tumor resection in patients with multifocal GBM

Characteristic	Groups	Events (n)a	OR	95% CI	P-value
Age at diagnosis	≤50 year	14	Ref
	51–60	19	0.6	0.2–2.1	.4
	61–70	35	0.9	0.3–2.8	.8
	71–95	26	0.4	0.1–1.3	.1
Initial WHO-PS	0	35	Ref
	1	31	0.2	0.1–0.6	.001
	2	19	0.4	0.1–0.98	.046
	3-4	9	0.1	0.03–0.4	<.001
Tumor locationb	Frontal	43	Ref
	Temporal	23	1.2	0.5–3.0	.7
	Parietal	19	0.8	0.3–1.8	.5
	Occipital	9	1.1	0.3–3.5	.9
Topography	Unilobar	22	Ref
	Multilobar	52	0.8	0.4–2.0	.7
	Unknownc	20	1.7	0.5–5.9	.4
Tumor crosses midline	No	73	Ref
	Yes	21	0.3	0.2–0.7	.005
Center	A		Ref
	B		0.3	0.1–1.1	.08
	C		2.8	1.1–7.3	.04
	D		0.6	0.3–1.4	.2

Included in analysis: n = 208. Patients with a deep-seated tumor were excluded from the model due to low number of events, that is, all 16 patients with a deep-seated mGBM underwent biopsy (in other words zero tumor resections). Model adjusted for year of diagnosis. CI = confidence interval; OR = odds ratio; PS = performance status.

^aEvent defined as mGBM patient assigned to tumor resection; a total of 94 events were reported.

^bPer patient report of tumor location for largest lesion.

^cData on tumor location only provided for 1 lesion.

Conventionally Fractionated Chemoradiotherapy in mGBM Versus uGBM

The long-course CRT regimen was utilized in 41% and 58% of the mGBM and uGBM patients, respectively (Figure 1C). In multivariable analysis, patients with mGBM, as opposed to those with uGBM, were less likely to be assigned to long-course CRT (OR 0.6, 95% CI 0.4–0.97, P-value .03, Supplementary Table 3).

Predictors of Conventionally Fractionated Chemoradiotherapy in mGBM

A total of 207 mGBM patients were included in this multivariable analysis (24 excluded due to missing data, Table 3). Of those, 85 were assigned to long-course CRT. None of the patients with an initial WHO-PS of 4 (ie, before surgery) were assigned to long-course CRT. In multivariable analysis, patients with an age over 70 years, an initial WHO-PS of 2–4, a tumor with midline cross, and a biopsy-only were less likely to be assigned to long-course CRT. Treatment in center B was associated with a higher likelihood of assignment to long-course CRT.

Table 3.

Multivariable logistic regression analysis to investigate the association between patient-related factors and assignment to conventionally fractionated “long-course” chemoradiotherapy in patients with multifocal GBM

Characteristic	Groups	Events (n)a	OR	95% CI	P-value
Age at diagnosis	≤50 year	13	Ref
	51–60	25	0.4	0.09–1.7	.2
	61–70	38	0.5	0.1–2.1	.4
	71–95	9	0.04	0.008–0.2	<.001
Initial WHO-PS	0	33	Ref
	1	37	0.5	0.2–1.2	.1
	2–4b	15	0.1	0.05–0.4	<.001
Topography	Unilobar	23	Ref
	Multilobar	49	0.7	0.3–1.9	.5
	Unknownc	13	1.1	0.3–4.5	.9
Tumor crosses midline	No	66	Ref
	Yes	19	0.3	0.1–-0.7	.005
MGMT methylation status	Methylated	41	Ref
	Unmethylated	44	1.1	0.5–2.4	.8
Extent of resection	Complete	18	Ref
	Near total	8	0.3	0.06–1.8	.2
	Partial	27	0.8	0.2–3.1	.7
	Biopsy	32	0.3	0.08–0.9	.04
Center	A		Ref
	B		5.3	1.2–24.0	.03
	C		1.3	0.4–4.1	.7
	D		2.0	0.8–5.2	.1

Included in analysis: n = 207. Model adjusted for year of diagnosis. CI = confidence interval; MGMT = O[6]-methylguanine-DNA methyltransferase; OR = odds ratio; PS = performance status.

^aEvent defined as patient assigned to conventionally fractionated chemoradiotherapy; a total of 85 events were reported.

^bThere were no mGBM patients assigned to long-course CRT with an initial WHO-PS of 4.

^cData on tumor location only provided for 1 lesion.

Overall Survival in mGBM Versus uGBM

At a median follow-up of 11 (range 0–105) months, 1290 out of 1343 patients had died. Median OS was 11.5 (95% CI 10.2–11.8) months, with 6-month and 1-year survival rates of 70% and 48%, respectively. For patients with mGBM, the median OS was 7.0 (95% CI 5.7–8.3) months and for those with uGBM, the median OS was 12.6 (95% CI 11.1–12.9) months. Figure 2 depicts the survival probability over time for patients with mGBM and uGBM, stratified for initial WHO-PS. In multivariable analysis, multifocality was significantly associated with a shorter survival; multifocal versus unifocal tumor hazard ratio [HR] 1.3, 95% CI 1.1–1.5, P-value .001 (Supplementary Table 4). Secondary analysis including the variable of IDH mutation status gave the same results (HR 1.3, 95% CI 1.1–1.6, P-value <.001).

Figure 2.

Survival probabilities over time for newly diagnosed patients with multifocal and unifocal GBM, stratified for initial WHO performance status before surgery. mGBM = multifocal glioblastoma; PS = performance status; uGBM = unifocal glioblastoma.

Prognostic Factors in mGBM

In the subgroup of patients with mGBM, the median OS was 7.4 (95% CI 4.6–9.4) and 6.5 (95% CI 4.0–8.0) months in case of MGMT-promotor methylated and unmethylated tumor, respectively (P-value .004) (Supplementary Figure 1). The median OS was 8.4 (95% CI 6.2–9.8) months in case of a tumor without midline cross and 4.9 (95% CI 2.4–5.6) months with midline cross (P-value .02) (Supplementary Figure 2). A total of 207 patients were included in this multivariable analysis (24 excluded due to missing data), of whom 206 had died at the time of follow-up data collection (Table 4). In multivariable analysis, a poorer initial WHO-PS and an MGMT-promotor unmethylated tumor were independently associated with a shorter survival. There was a nonsignificant trend toward shorter survival in case of a tumor with midline cross at diagnosis, with increasing number of affected lobes, and with a lesser EOR.

Table 4.

Prognostic factors for survival in patients with multifocal GBM as assessed in multivariable Cox regression analysis

Characteristic	Groups	Events (n)a	HR	95% CI	P-value
Age at diagnosis	≤40 years	7	Ref
	41–50	11	0.4	0.2–1.2	.1
	51–60	43	0.7	0.3–1.8	.5
	61–70	70	0.9	0.4–2.0	.7
	71–80	66	1.5	0.6–3.6	.4
	81–95	9	1.3	0.4–3.8	.6
Sex	Male	103	Ref
	Female	103	0.9	0.6–1.2	.3
Initial WHO-PS	0	51	Ref
	1	77	1.0	0.7–1.6	.8
	2	44	1.8	1.1–2.9	.01
	3	28	2.9	1.6–5.2	<.001
	4	6	16.5	5.9–46.3	<.001
Tumor locationb	Frontal	40	Ref
	Temporal	89	1.1	0.7–1.7	.7
	Parietal	47	1.1	0.7–1.7	.5
	Occipital	16	1.0	0.6–1.8	1.0
	Deep seated	14	1.0	0.5–1.9	.9
Topography	Unilobar	43	Ref
	Multilobar (2 distinct locations)	110	1.3	0.8–1.9	.2
	Multilobar (3 distinct locations)	25	1.3	0.7–2.4	.3
	Unknownc	28	0.8	0.5–1.5	.5
Tumor crosses midline	No	136	Ref
	Yes	70	1.2	0.9–1.7	.3
MGMT methylation statusd	Methylated	107	Ref
	Unmethylated	99	1.8	1.3–2.5	<.001
Extent of resection	Complete	24	Ref
	Near total	17	0.9	0.4–1.7	.9
	Partial	49	1.2	0.7–2.1	.4
	Biopsy	116	1.6	1.0–2.7	.051

Included in analysis: n = 207. Model stratified for center. CI = confidence interval; HR = hazards ratio; MGMT = O[6]-methylguanine-DNA methyltransferase; PS = performance status.

^aEvent defined as death due to any cause; a total of 206 events were reported.

^bPer patient report of tumor location for largest lesion.

^cData on tumor location only provided for one lesion.

^dSlightly increasing hazard ratio with increasing follow-up time.

Discussion

This retrospective registry-based Danish cohort study covering 2015 until 2019, was set out to contribute to the sparse evidence on mGBM, and in that way increase informed decision-making regarding the treatment of future patients with mGBM. There is a general lack of uniform definitions in studies reporting on multifocality and multicentricity.^2,4–6 One of the contributing factors is the translation of the proposed definitions into a clear radiologic assessment.^2,29 In more detail, to distinguish between unifocality and multifocality, some reports added the requirement of a certain distance between contrast-enhancing lesions,^15,29,30 while other reports added this requirement to distinguish between multifocality and multicentricity.^2,11,31 The latter distinction was, furthermore, made in different reports by the presence or absence of connecting T2FLAIR abnormalities between contrast-enhancing lesions^5,29,32,33 and/or lesions being located in distinct lobes or in both hemispheres.^4,32,34,35 The definition for multifocal disease in the present report is, therefore, in line with what has been used for mGBM, with a tendency toward what has occasionally been used for multicentric GBM. This is reflected by the mGBM incidence (17%) and survival (median OS 7.0 months) in this cohort that seem to be slightly lower than, but close to, that previously reported for multifocality.^2,6 Our results should be interpreted with this in mind.

This study has several strengths and limitations inherent to the registry-based design. First, our findings only regard those patients with a pathology-confirmed GBM diagnosis according to the WHO criteria at the time. Notably, as a surgical procedure, including biopsy, may have been more often omitted in those patients with multiple contrast-enhancing lesions instead of a single, there may be a potential selection bias in our study population. Second, due to the treatment assignment bias inherent to our study design, we cannot conclude on the optimal treatment for mGBM patients, nor untangle treatment effects in survival outcomes. Third, although there were found high agreements for MRI characteristics, including tumor focality, in a previous validation study of the used DNOR,²⁵ there remains an uncertainty in the misclassification rate. Finally, we would like to list relevant missing data in this study, for example, progression-free survival rates, comorbidity, use of corticosteroids, surgical morbidity and mortality, number of contrast-enhancing lesions and total tumor volume, actual treatment decisions, and patient preferences. We could, therefore, not correct for all factors that possibly influenced our outcome measures, nor investigate all relevant end points. Despite these shortcomings, we are able to increase knowledge on POC for mGBM patients in a real-world setting, and to compare mGBM to uGBM patients while acknowledging the most important confounding factors. Moreover, the completeness of our nationwide study population and data, especially with regard to the systematic assessment of MGMT-promotor methylation status at diagnosis (assessed in 95% of patients), strengthens our findings and their relevance.

A poorer initial performance in patients with mGBM has been reported previously^1,3 and may reflect a larger disease burden. This finding is also in line with the view of mGBM as more advanced disease, that is, a late diagnosis. During the disease course of GBM patients, it is not uncommon that additional contrast-enhancing lesions become visible with time.³⁶ It should be stressed though that it is currently unclear whether mGBM is a reflection of a late diagnosis or indicates a tumor that is particularly fast in migration, or is both. Very limited preclinical research possibly indicated a monoclonal origin,^9,37–39 a prevalent mesenchymal subtype,^9,39 and different molecular subtypes between synchronous tumor foci in mGBM.^9,39 Furthermore, genetic differences were found when comparing tumor foci from mGBM patients to those from uGBM.^40–42 It will thus be a task for further research to elucidate on the biological processes and their chronological order in mGBM.

Our findings support the hypothesis that patients with mGBM receive less intensive treatment than those with uGBM. In the present cohort, only 50% of the patients with mGBM were treated with a combination of surgery, RT, and chemotherapy versus 65% of the patients with uGBM, in line with the literature.^{1,3,4,15,43–45} mGBM was, furthermore, independently associated with an overall less intensive treatment approach, thus far only examined univariately in 1 other study.¹ To date in literature, the overall less intensive treatment approach in mGBM patients has been suggested to result from a poorer PS,¹ additional technical challenges,^11,12 and a presumed poorer prognosis^1,2,4,12 in comparison to uGBM patients. An in-depth study, including actual treatment decisions, could aid in bridging this knowledge gap and optimizing the possible decision to proceed with or omit treatment.

We, furthermore, explored which factors contributed to the variation in treatment between patients with multifocal disease, by applying multivariable analyses in the subgroup of mGBM patients. The variation in surgical procedures between mGBM patients was reflected by 58% undergoing a biopsy, 23% a partial, and 19% a near-total/complete resection. Similarly, the variation in postoperative RT type was reflected by 41% of mGBM patients assigned to long-course CRT and 25% to short-course HFRT. Of note, all mGBM patients with a deep-seated tumor underwent biopsy, and all mGBM patients with a WHO-PS of 4 were not assigned to long-course CRT, showing clear consensus in treatment management in these subgroups. For the remaining mGBM patients, our findings suggest that hospital of treatment, besides other expected factors, is one of the explanatory factors for the variation in assignment to tumor resection and long-course CRT. Large variation in surgical procedures between centers has been described before for GBM in general^46,47 and for mGBM specifically; for example, 0%–45% near-total/complete resection in mGBM patients.^3,4,15,45 Variation between centers in postoperative RT regimens is less clear with only a few publications reporting fairly similar percentages of mGBM patients assigned to long-course CRT and lower percentages for short-course HFRT^1,15,45 compared to our findings. Taken together, it is likely that differences in the management of mGBM patients exist between centers and that those differences probably start at the selection of patients for surgery. This could relate to the evidence gap and a lack of consensus on the best treatment approach in mGBM.^1,11 A thorough understanding of possible differences between a larger number of centers, including at least functional and patient-reported outcomes, could contribute to the optimization of mGBM management.

Our results support the hypothesis that patients with mGBM at diagnosis have a shorter survival than those with uGBM in line with the previous Danish cohort study covering 2009–2014 and the largest published series.^1,13 In more detail, patients with newly diagnosed mGBM in our cohort had a median OS of 7.0 (95% CI 5.7–8.3) months comparable to literature,^{1,3–5,14,15,32,33} and mGBM was an independent poor prognostic factor for survival with a small effect size (HR of 1.3, 95% CI 1.1–1.5) also in line with literature.^1,6 For example, having a multifocal tumor, instead of a unifocal tumor, had approximately a similar negative effect on survival as a 1-point drop in WHO-PS as illustrated in Figure 1. To conclude, our data show that multifocality had a moderate adverse effect on prognosis.

This is one of the first contributions addressing prognostic factors among mGBM patients.^1,14 Our results show the striking variation in survival between patients with mGBM as is also seen in GBM patients in general.¹⁶ To illustrate, when stratifying for initial WHO-PS, the median OS for mGBM patients in our cohort ranged from 1.4 to 13.6 months. Besides confirmation of initial WHO-PS, MGMT-promotor methylation status and EOR as independent prognostic factors for mGBM patients specifically, we showed a possible role for tumor topography (unilobar, multilobar) and tumor crossing midline as prognostic factor. In more detail on the latter, patients with a mGBM and tumor crossing midline had an even shorter survival in univariable analysis (median OS 4.9 months with and 8.4 months without midline cross) though not in multivariable analysis. An mGBM with midline cross could reflect bihemispheric disease, which has been shown to be associated with an even shorter survival⁴ although sparsely studied. Our findings thus stress the importance of taking into account multiple factors when estimating the prognosis of patients with mGBM.

To conclude, our results probably reflect a tendency to treat mGBM patients with an overall less aggressive approach. Not surprisingly, there is variation in how we (can) treat patients with mGBM though, and intensive trimodality treatment was feasible in selected mGBM patients. Considering the limitations inherent to our study design, we furthermore conclude that multifocality is probably an independent poor prognostic factor with a moderate effect. In addition, survival differences between subgroups of mGBM patients are larger than the survival difference between mGBM and uGBM patients. We, therefore, argue that multifocality cannot stand on its own as reason to divert from standard of care, and standard of care is recommended whenever technically feasible. The new information generated in this study can support estimating an mGBM patient’s prognosis. This will, in turn, be helpful to guide future treatment decision-making and clinical trial design.

Conflict of interest statement

CAH reports personal fees from Bristol Myers Squibb and GlaxoSmithKline, outside the submitted work.

Funding

This work was supported by Danish Comprehensive Cancer Center—Brain Tumor Center/Danish Cancer Society (R295-A16770) and Danish Comprehensive Cancer Center (2021-08).