DNA methylation subclasses predict the benefit from gross total tumor resection in IDH-wildtype glioblastoma patients

Abstract

Background

DNA methylation-based tumor classification allows an enhanced distinction into subgroups of glioblastoma. However, the clinical benefit of DNA methylation-based stratification of glioblastomas remains inconclusive.

Methods

Multicentric cohort study including 430 patients with newly diagnosed glioblastoma subjected to global DNA methylation profiling. Outcome measures included overall survival (OS), progression-free survival (PFS), prognostic relevance of EOR and MGMT promoter methylation status as well as a surgical benefit for recurrent glioblastoma.

Results

345 patients (80.2%) fulfilled the inclusion criteria and 305 patients received combined adjuvant therapy. DNA methylation subclasses RTK I, RTK II, and mesenchymal (MES) revealed no significant survival differences (RTK I: Ref.; RTK II: HR 0.9 [95% CI, 0.64–1.28]; p = 0.56; MES: 0.69 [0.47–1.02]; p = 0.06). Patients with RTK I (GTR/near GTR: Ref.; PR: HR 2.87 [95% CI, 1.36–6.08]; p < 0.01) or RTK II (GTR/near GTR: Ref.; PR: HR 5.09 [95% CI, 2.80–9.26]; p < 0.01) tumors who underwent gross-total resection (GTR) or near GTR had a longer OS and PFS than partially resected patients. The MES subclass showed no survival benefit for a maximized EOR (GTR/near GTR: Ref.; PR: HR 1.45 [95% CI, 0.68–3.09]; p = 0.33). Therapy response predictive value of MGMT promoter methylation was evident for RTK I (HR 0.37 [95% CI, 0.19–0.71]; p < 0.01) and RTK II (HR 0.56 [95% CI, 0.34–0.91]; p = 0.02) but not the MES subclass (HR 0.52 [95% CI, 0.27–1.02]; p = 0.06). For local recurrence (n = 112), re-resection conveyed a progression-to-overall survival (POS) benefit (p < 0.01), which was evident in RTK I (p = 0.03) and RTK II (p < 0.01) tumors, but not in MES tumors (p = 0.33).

Conclusion

We demonstrate a survival benefit from maximized EOR for newly diagnosed and recurrent glioblastomas of the RTK I and RTK II but not the MES subclass. Hence, it needs to be debated whether the MES subclass should be treated with maximal surgical resection, especially when located in eloquent areas and at time of recurrence.

Maximal and safe resection is the cornerstone of glioblastoma therapy. Although tumors have usually already infiltrated far into the surrounding tissue, resulting inevitably in local and discontinuous recurrences, extent of resection (EOR) remains one of the main factors determining a more favorable prognosis.^1–3 Recent studies provided evidence that maximized EOR benefits the survival outcome regardless of the O6-methylguanine DNA-methyltransferase (MGMT) promoter methylation status.³ Hence, it is assumed that every glioblastoma patient will benefit from maximal resection if feasible. Surgery is routinely followed by combined radiotherapy and chemotherapy with the DNA-alkylating agent temozolomide,⁴ for which the MGMT promoter methylation status is a known predictive marker.^5–7

In the past years, global DNA methylation profiling—the use of arrays to determine DNA methylation patterns across the genome—has been developed into a tool that increases the accuracy of exact molecular classification of central nervous system tumors with the potential to further stratify patients, beyond the assessment of routine clinicopathological characteristics.^8,9 Using the DNA methylation-based classifier, glioblastomas can be assigned to different subclasses, including receptor tyrosine kinase (RTK) I, RTK II, RTK III, H3.3 G34-mutant, midline, MYCN and mesenchymal (MES).⁸ In adult patients, the most common methylation subclasses comprise RTK I, RTK II, and MES. For patient cohorts, which were mainly selected for clinical trials, the survival outcome was comparable between these three subclasses^10–12 and it was indicated that the prognostic benefit of a methylated MGMT promoter may be limited to the RTK II subclass in patients treated with monotherapy.¹⁰ However, further prognostic significances of these epigenetic subgroups for the treatment of glioblastoma patients are largely unknown.

Here, in a multicenter cohort of 430 IDH-wildtype glioblastomas, we show that patients with MES tumors gain no significant advantage in overall survival (OS) or progression-free survival (PFS) from a gross total or near gross total resection. In contrast, OS and PFS are significantly prolonged by gross-total resection or near-gross total resection in patients with RTK I and RTK II tumors. We further show that a re-resection is only beneficial in RTK I and RTK II subclasses. Thus, by DNA methylation classification, we provide the basis for molecular stratification of glioblastoma patients who will or will not benefit from maximal resection at diagnosis or tumor recurrence. Our findings raise the question whether maximal resection of MES glioblastomas should be aimed for, especially when there is a risk of postoperative deficit.

Materials and Methods

Study Population

Glioma tissue from 430 patients who were newly diagnosed with IDH-wildtype glioblastoma, and who underwent surgery at University Medical Center Hamburg-Eppendorf, University Hospital Frankfurt, or Charité University Hospital Berlin (all Germany) was analyzed. Patients underwent surgery between January 2012 and December 2021. Informed written consent was obtained from all patients. Diagnosis was based on the WHO classification.¹³ The EOR was stratified into gross total resection (GTR), near GTR, and partial resection (PR) or stereotactic biopsy. A GTR was defined as a complete removal of contrast-enhancing parts, a near GTR as a removal of more than 90% of the contrast-enhancing parts, whereas a resection of lower than 90% was defined as PR/biopsy. The EOR of contrast-enhancing parts was evaluated by MRI performed up to 48 h after surgery. OS was calculated from diagnosis until death or last follow-up, and PFS from diagnosis until progression according to Response Assessment in Neuro-Oncology (RANO) criteria based on local assessment.¹⁴ This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.¹⁵

Inclusion Criteria

Inclusion criteria were defined based on DNA methylation profiling results and clinical data as presented in Figure 1. Methylation profiling results were submitted to the molecular neuropathology (MNP) methylation classifier v11b4 hosted by the German Cancer Research Center (DKFZ).⁸ Patients were included if the calibrated score for methylation class family glioblastoma, IDH-wildtype was >0.84.¹⁶ In addition, patients with a score below 0.84 but above 0.7 with a combined gain of chromosome 7 and loss of chromosome 10 or amplification of epidermal growth factor receptor (EGFR) were included in accordance with cIMPACT criteria.¹⁷ Furthermore, to clearly separate the subclasses RTK I, RTK II, and MES, a class member score of ≥0.5 for one of these three subclasses was required. Patients had to further meet the following clinical criteria: supratentorial tumor localization, adjuvant treatment after surgery, age above 18 years, and availability of data for OS and PFS.

DNA Methylation Profiling

DNA was extracted from tumors and analyzed for genome-wide DNA methylation patterns using the Illumina EPIC (850k) array. Processing of DNA methylation data was performed with custom approaches as previously described.¹⁶ Classification was performed using the MNP brain tumor classifier v11b4 of the DKFZ.^8,16 Evaluation of the MGMT promoter methylation status was made from the classifier output v11b4 using the STP27-method.

3D Volumetric Segmentation

We analyzed T1-weighted as well as T2-weighted FLAIR (Fluid attenuated inversion recovery) magnetic resonance imaging (MRI) axial images before surgery. The program BRAINLAB was used. The region of interest was delineated in every slice, enabling a multiplanar 3D reconstruction. The volume of contrast enhancement and FLAIR hyperintensity was assessed in cm³.

Statistical Analysis

Differences in continuous variables were analyzed with the Mann–Whitney U test and differences in proportions were analyzed with the chi-square-test or Fisher exact test. Variables with possible prognostic effect were assessed by log-rank test and all survival curves were visualized as results from the Kaplan Meier analysis. Overall and progression-free survival, hazard ratios (HRs), and 95% confidence interval (CI) were computed for each group Cox proportional hazards regression model. The potential prognostic variables were age (continuous), gender (male, female), tumor location (frontal, parietal, temporal, occipital), hemisphere (left, right, both), number of lobe involvement (one, > one), extent of resection (GTR, near GTR, partial resection/biopsy), MGMT promoter status (methylated, nonmethylated), Karnofsky Performance Score (continuous), volume of contrast-enhancing tumor (continuous) and FLAIR lesion (continuous), and DNA methylation subclass (RTK I, RTK II, MES). All variables associated with OS or PFS with a p-value less than 0.05 in univariate analysis were included in the multivariable model. Multicollinearity was established between risk factors using correlation coefficient and Variance Inflation Factors (VIF). To avoid multicollinearity, a correlation coefficient >0.8 between any two variables was set as the threshold for multicollinearity. In addition, the threshold for VIF measured on the set of variables was set ≥ 4 to indicate the presence of multicollinearity. In the analyses of this study, calculated parameters did not reach these thresholds. In general, a two-sided p-value less than 0.05 was considered statistically significant. All analyses were performed using SPSS Inc. (Chicago, IL, USA). Data illustrations were performed using GraphPad Prism 9.

Results

Study Population

A total of 430 patients who were newly diagnosed with a glioblastoma were enrolled in this study. After applying the aforementioned inclusion criteria, 345 patients were available for further analysis. The mean age of the study population was 61.4 years. 122 patients (40.0%) were female. GTR was achieved in 38.4% and near GTR in 26.6% of patients, whereas 35.0% underwent partial resection or biopsy (Supplemental Table 1). In 153 patients (50.2%), a methylated MGMT promoter was diagnosed. The majority (88.4%) received radiochemotherapy as adjuvant treatment after surgery.

Clinical Data of DNA Methylation Subclasses

After DNA methylation profiling, patients were stratified according to their methylation subclass RTK I (31.1%), RTK II (41.0%) and MES (27.9%) (Figure 1A, Supplemental Table 1). Basic clinical characteristics, such as location, contrast enhancement volume, FLAIR volume and EOR did not differ between the subclasses (Supplemental Table 1). Patients with RTK I tumors were significantly older than patients with RTK II or MES tumors (Figure 1B, Supplemental Table 1), without a significant difference in sex distribution (Figure 1C). Karnofsky performance status after diagnosis (p = 0.64) and MGMT promoter methylation status (p = 0.83) were comparable in all three groups (Supplemental Table 1).

Impact of EOR in DNA Methylation Subclasses

305 patients (88.4%) received radiochemotherapy after surgery and were included for further survival analysis (Figures 1–3). At a mean (SD) follow-up time of 14.5 (15.3) months, 179 deaths (58.6%) were observed. During the study period, 170 patients (55.7%) showed tumor progression. The median length of survival for all patients from the time of index surgery was 15.8 months (range 1–97 months). After stratifying the study population which were treated with combined therapy according to their methylation subclass, survival analyses revealed no significant differences for OS (p = 0.06, Figure 1D) and PFS (p = 0.52, Supplemental Figure 1A) between these groups. Patients who received a GTR (Figure 2A and 2B) had a significantly longer OS (HR, 2.18; 95% CI, 1.46–3.24; p < 0.01) (Figure 2C) and PFS (HR, 1.78; 95% CI, 1.19–2·65; p < 0.01) (Supplemental Figure 1B) using Cox proportional hazards regression model (Supplemental Tables 2 and 3). When examining the impact of maximal EOR in each methylation subclass a significant prolongation of OS in RTK I (HR, 2.87; 95% CI, 1.36–6.08; p < 0.01) (Figure 2D) and RTK II tumors was found (HR, 5.09; 95% CI, 2.80–9.26; p < 0.01) (Figure 2E). However, a prolonged OS for a GTR or near GTR was not observed in the MES subclass (HR, 1.45; 95% CI, 0.69–3.09; p = 0.33) (Figure 2F). A significantly longer PFS of a GTR or near GTR was seen for RTK II tumors, but not the RTK I and MES subclasses (Supplemental Figures 1C-E). Detailed information for univariate and multivariate analysis for each subclass are available for OS in Supplemental Tables 4–6 and for PFS in Supplemental Tables 7–9.

Impact of MGMT Promoter Methylation Status in DNA Methylation Subclasses

Of 305 combined-treated patients, 153 (50.2%) had a methylated MGMT promoter status, and DNA methylation subgroups were equally distributed among methylated and nonmethylated tumors (Figure 3A). As expected, a methylated MGMT promoter status was predictive of favorable OS (HR, 0.51; 95% CI, 0.37–0.72; p < 0.01) (Supplemental Table 2) and PFS (HR, 0.44; 95% CI, 0.31–0.62; p < 0.01) (Supplemental Table 3) in the combined cohort. These results were also observed for the individual DNA methylation subclasses by univariate analysis (Supplemental Tables 4–9). However, after adjusting for covariates using Cox regression model, the MGMT promoter methylation status remained an independent factor for a favorable outcome in RTK I (HR, 0.37; 95% CI, 0.19–0.71; p < 0.01) and RTK II (HR, 0.56; 95% CI, 0.34–0.91; p = 0.02) tumors, but not the MES subclass (HR, 0.52; 95% CI, 0.27–1.02; p = 0.06) (Figure 3B). Similar results were observed for PFS (Supplemental Figure 2, Supplemental Tables 7–9).

Outcome after Local Recurrence in DNA Methylation Subclass

We identified 68 patients who were operated for local recurrent glioblastoma. Clinical characteristics are detailed in Supplemental Table 10. To investigate the benefit of a recurrence resection as a therapeutic option prior to second line therapy, we compared these 68 patients who underwent a re-resection and received adjuvant therapy with 44 patients without a re-resection before starting second line therapy (Figure 4A). Both groups included patients with local recurrence and previously completed combined radiochemotherapy. Survival analysis revealed a favorable progression-to-overall survival (POS) time for the re-resection cohort (HR, 2.56; 95% CI, 1.34–5.01; p < 0.01) (Figure 4B). Here, it must be taken into consideration that patients who underwent re-resection were more likely to receive combined therapy as second line adjuvant therapy (p < 0.01, Supplemental Table 10). After stratifying according to the respective methylation subclasses obtained from the primary tumor, the survival benefit was valid for patients with RTK I (HR, 2.82; 95% CI, 1.12–7.08; p = 0.03) (Figure 4C) and RTK II tumors (HR, 3.47; 95% CI, 1.55–7.78; p < 0.01) (Figure 4D), but not MES subclass (HR, 1.62; 95% CI, 0.61–4.29; p = 0.33) (Figure 4E).

Heterogeneity of DNA Methylation Subclasses

Since former studies reported on intratumoral DNA methylation heterogeneity within glioblastomas,^18,19 we investigated whether the methylation subclasses were stable during the course of disease progression. We therefore performed global DNA methylation analysis on 22 matched samples obtained from surgery at initial diagnosis and surgery at first recurrence. In 21 patients with a valid classifier output, we observed a switch of the DNA methylation subclass in 5 (23.8%) cases (Figure 4F). A change of the MGMT promoter methylation status occurred in 1 (4.8%) case.

Outcome of Patients ≥ 70 Years of Age in DNA Methylation Subclasses

From the main cohort treated with radiochemotherapy after surgery, we identified 70 patients who were 70 years of age and older, and who had a Karnofsky performance score of at least 60% (Figure 5A). Survival analyses revealed a significantly longer OS for the MES subclass (HR, 0.35; 95% CI, 0.13–0.90; p = 0.03) (Figure 5B). In addition, the EOR (HR, 2.39; 95% CI, 1.07–5.34; p = 0.03) and Karnofsky performance score (HR, 0.96; 95% CI, 0.94–0.99; p < 0.01) were independent prognostic factors in this distinctive patient cohort (Supplemental Table 11).

Outcome after Monotherapy in DNA Methylation Subclasses

Of all 345 patients, 40 patients (10.3%) were treated with monotherapy after surgery, of whom 27 (67.5%) received radiation and 13 (32.5%) TMZ (Figure 5C). After stratifying these patients according to their methylation subclasses, basic clinical characteristics were balanced between these groups (Supplemental Table 12). Within the monotherapeutically treated patients, the methylation subclass was not associated with OS, neither for treatment with TMZ (HR, 0.45; 95% CI, 0.09–2.34; p = 0.34) (Figure 5D top) nor radiation (HR, 1.24; 95% CI, 0.29–5.13; p = 0.77) (Figure 5D bottom) (Supplemental Tables 13–15).

Discussion

In recent years, DNA methylation-based tumor profiling facilitates a more precise classification and distinction of brain tumor subgroups. For glioblastoma, RTK I, II, and MES are the most common subclasses. RTK I tumors are enriched for platelet-derived growth factor receptor A (PDGFRA) gene amplification, whereas RTK II tumors frequently harbor amplification of epidermal growth factor receptor (EGFR) gene, while MES tumors have no typical recurrent mutations.^12,20,21 Although an overlap was seen between the methylation-based subclassification and transcriptome analysis,²² DNA methylation does not completely match with RNA-based subclasses or the proteogenomic and metabolomic landscape of glioblastomas.²¹ Nevertheless, thus far no real translation to the clinical practice from these extensive subclassification has been achieved. Our study investigated the utility of global methylation profiling (850K) to provide a methylation-based therapy guidance for glioblastoma patients.

We present the following major findings: (1) The DNA methylation subclasses RTK I, RTK II and MES have a comparable OS and PFS after surgery and standard adjuvant therapy. (2) A survival benefit of a maximized EOR exists in the RTK I and RTK II subclasses, but not the MES subclass. (3) A methylated MGMT promoter is an independent prognostic factor for the OS in RTK I and RTK II tumors but did not reach statistical significance in the MES subclass even though a survival prolongation was observed. (4) For recurrent glioblastoma, a re-resection is associated with a prolonged POS in RTK I and RTK II tumors. (5) Elderly patients of 70 years of age and older who receive radiochemotherapy have a favorable OS and PFS in the MES subclass.

The Neurooncology Working Group (NOA)-8 and EORTC-26101 trials showed a comparable survival outcome between the RTK I, RTK II, and MES methylation subclasses,^10,11 which we also demonstrated in our study. Only MES tumors tended to have a longer OS, which may relate to a higher number of tumor-infiltrating lymphocytes,^12,23 resembling a more immunogenic tumor. However, the relationship of infiltrating lymphocytes and prognosis in glioblastoma remains controversial.^12,24–26

Surgical resection remains the cornerstone in glioblastoma therapy. It is well established that the EOR represents an independent factor for OS and PFS, regardless of the IDH and MGMT promoter methylation status.^1,3,27 Therefore, the surgical treatment goal remains to achieve a GTR of the contrast-enhancing tumor parts without a postoperative neurological deterioration. However, DNA methylation profiles were never considered as a prognostic factor for the survival benefit of maximized resected tumors. Conducting 430 global methylation profiles we found that the survival benefit of gross-totally resected glioblastomas was limited to RTK I and RTK II methylation subclasses. In patients with tumors classified as MES subclass and treated with adjuvant radiochemotherapy, we found no significant difference in OS and PFS between biopsied, partially resected, or gross totally resected tumors. These results raise the question of whether patients with a MES glioblastoma should invariably undergo maximal resection when postoperative neurologic deterioration is likely to occur, thus preventing adjuvant therapy. Conversely, this also implies that achieving GTR in RTK I and RTK II tumors is of particular importance. However, it should be noted that DNA methylation profiling to determine the subclass postsurgery takes approximately 1 week. Therefore, rapid testing for DNA methylation subclasses is desired, preferably preoperative, or intraoperative in cases with a glioblastoma-suspected MRI scan. To this end, a promising approach was pursued by Djirackor and colleagues who achieved a high accuracy of intraoperative DNA methylation-based classification of CNS tumors using Nanopore ultra-low coverage whole genome sequencing and obtained results within 120 min in most cases.^28,29 In this rapidly evolving field, additional approaches are currently being presented that raise the hope of intraoperatively determination of DNA methylation subclasses.^30,31

Besides the well-defined prognostic impact of EOR, the importance of the MGMT promoter methylation status is undeniable, since favorable survival has been historically demonstrated in patients with a methylated MGMT promoter.^32–34 In our cohort who underwent combined treatment after surgery, a prolonged survival of MGMT-methylated tumors was seen in all patients, especially in the RTK I and RTK II subclass, while there was no difference in the proportion of a methylated MGMT promoter between the subclasses. For MES tumors, it needs to be addressed that the survival outcome of MGMT-methylated tumors was more favorable than tumors with a nonmethylated MGMT promoter, even though a statistical significance was not reached. Therefore, choosing adjuvant therapy dependent on the MGMT promoter methylation status seems to be justified for MES tumors. Since it remains difficult to find the optimal therapeutic regimen for patients of older age, we defined a subgroup of septuagenarians (70 years and older and Karnofsky score of at least 60%) treated with radiochemotherapy. In this subgroup, the MES subclass was associated with a favorable survival. These results underscore the previously mentioned findings that treatment response, whether monotherapy or radiochemotherapy, and the benefit of a methylated MGMT promoter depend on DNA methylation subclasses as well as patient characteristics.

Despite numerous efforts, there is still no standard of care for recurrent glioblastoma, and the decision to re-operate depends on the general condition of the patient and the local or multifocal recurrence.³⁵ In our study, we compared a group of patients with a local recurrence of the same age, Karnofsky index and MGMT promoter methylation status who underwent re-resection and adjuvant therapy or conservative treatment alone. In the RTK I and RTK II subclasses, but not in the MES subclass, longer survival was observed between the aforementioned groups. Thus, the decision to re-operate at the time of recurrence might have a more favorable survival advantage in RTK I and RTK II than in MES glioblastomas. Because determining the methylation subclass at the time of initial surgery is currently impractical, methylation-based decision making at the time of disease recurrence may be useful.

Although the methylation subclass was changed in 24.8% of tumors this epigenetic imprint appears more stable than transcriptional subtyping, in which 46.0% of tumors had an altered subtype.³⁶ Till date, it is largely unknown how switching DNA methylation subclasses can potentially affect the prognosis of glioblastoma patients, as the study of temporal heterogeneity has been limited to transcriptional subtypes. In our study, transition had no effect on patient outcome, and the MES subclass remains the most stable tumor subclass.

In summary, we show that patients with MES tumors gain no significant advantage in OS or PFS from GTR or near-GTR compared to patients with RTK I and RTK II tumors, which have a prolonged OS and PFS if a maximized resection is performed. We further demonstrate that a re-resection is only beneficial in RTK I and RTK II tumors for local recurrences. Thus, by methylation classification, we provide the basis for molecular stratification of glioblastoma patients who will or will not benefit from maximal resection at diagnosis or at tumor recurrence.

Limitations

This study has several limitations. The retrospective cohort includes patients from three major neuro-oncology centers rather than from a prospective randomized clinical trial. However, this would also be difficult to ethically justify to evaluate the benefit of maximal resection. Our cohort consisted of patients only amendable for biopsy and in which complete resection was anticipated but uncertain due to the conduct of surgery and the feedback obtained by intraoperative neuromonitoring, hence explaining the relatively low rate of gross-total resected tumors. An independent validation cohort is lacking, but a comparison across centers showed consistent results. No data for toxicity and adverse events during adjuvant therapy was available. Additionally, treatment decisions for surgical resections at time of local recurrence could be influenced by surgeon and patient perspectives, which is not documented in this study and the numbers within subgroups of patients that received monotherapy or were treated at relapse were small and statistical analysis may be of limited reliability. In the OS analysis between DNA methylation subgroups and the prognostic impact of MGMT promoter methylation status, we found a trend for the MES subclass in both categories, but this did not reach statistical significance. It is worth considering that this effect could also be achieved in a larger cohort or with a smaller number of censored data.

Conclusion

This study evaluated the prognostic relevance of DNA methylation profiling in a large, multi-institutional study population diagnosed with glioblastoma who received standard of care treatment and was followed until tumor recurrence and death. Although overall survival did not differ between DNA methylation subclasses, our results showed that the known survival benefit of maximized EOR applies to RTK I and RTK II tumors but not to the MES subclass. Therefore, it must be discussed whether the MES subclass should invariably be treated with maximal surgical resection, especially at tumor progression.

Acknowledgment

U.S. was supported by the Fördergemeinschaft Kinderkrebszentrum Hamburg. F.L.R. received a Research Grant from Illumina Inc. A.E. is thankful for the support within the interdisciplinary graduate school “Innovative Technologies in Cancer Diagnostics and Therapy” funded by the City of Hamburg. We thank all the patients who gave informed consent and without whom this research would not have been possible.

Conflict of interest statement. None declared.

Authorship statement. Dr Drexler and Dr Ricklefs had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: Drexler, Ricklefs, Schüller, Dührsen, Westphal. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: Drexler, Ricklefs, Dührsen, Lamszus, Schüller, Westphal. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Drexler, Ricklefs, Eckhardt, Sauvigny. Obtained funding: Ricklefs, Westphal, Schüller. Administrative, technical, or material support: Eckhardt, Filipski, Siewert, Sauvigny. Supervision: Ricklefs, Dührsen, Schüller, Westphal.

Data availability

Individual patient-level data are available for the retrospective cohorts upon request.

References

Author notes