Prognostic value of 1D, 2D, and volumetric tumor size increases in recurrent WHO grade 2 and 3 meningiomas: Radiological post-hoc analysis of the EORTC-BTG-1320 trial

Abstract

Background

Although the differential prognostic value of 1D, 2D, and volumetric meningioma size assessment has been reported, RANO meningioma criteria rely on bidimensional measurements.

Methods

In this post-hoc analysis of the EORTC-BTG 1320 trial, contrast-enhancing CNS WHO grade 2 and 3 meningiomas were assessed using 1D, 2D, and volumetric measurements. Different cutoff values for lesion size increase were compared 6 months after the start of antineoplastic treatment using Cox proportional hazards models to evaluate their association with overall survival (OS). Optimal cutoff values were identified using two criteria: maximal hazard ratio (HR) for death with statistical significance for median OS and the cutoff that maximized mean specificity and sensitivity for predicting 1-year OS.

Results

Among 57 evaluable patients, unidimensional 5 mm and 10 mm tumor size increase yielded the maximal HRs (HR = 3.41, 95% Confidence Interval (CI) 1.56–7.45, P < .01 and HR = 3.22, 95% CI 1.58–6.58, P < .01, respectively) for OS. A 6 mm tumor size increase maximized mean specificity and sensitivity (HR = 2.91, 95% CI 1.43–5.93, P < .01) for predicting 1-year OS. For tumor volume assessments, a 30% increase was associated with the maximal HR (HR = 3.69, 95% CI 1.64–8.31, P < .01) for OS whereas a 40% increase maximized the mean specificity and sensitivity (HR = 3.66, 95% CI 1.75–7.654, P < .01). Bidimensional measurements showed no significant OS association.

Conclusion

Unidimensional tumor measurements and tumor volume assessments show a stronger association with overall survival than bidimensional measurements in recurrent non-benign meningiomas. Integration of these methods into response assessment criteria for meningiomas should be considered.

Key Points.

• Unidimensional and volumetric tumor progression are strongly associated with overall survival in patients with CNS WHO grade 2 and 3 meningioma.

• Bidimensional tumor measurements, included in the current RANO response criteria for meningioma, are not significantly associated with overall survival.

• Unidimensional, bidimensional, and volumetric tumor measurements show strong correlations.

Importance of the Study.

The proposed response assessment criteria for clinical trials in patients with meningiomas are primarily aligned with those used for patients with glioma, relying on bidimensional measurements to identify tumor progression. However, the present study demonstrates that absolute unidimensional tumor measurements and tumor volume show a significantly stronger association with overall survival in CNS WHO grade 2 and 3 meningiomas compared to bidimensional measurements. Based on these findings, the measurement techniques used to assess tumor progression in clinical trials for patients with meningioma should be reconsidered.

In contrast to CNS WHO grade 1 meningioma, which can be treated successfully with surgery or radiation therapy alone, patients with CNS WHO grade 2 or 3 meningiomas more often experience recurrence and require additional treatment, mostly repeat surgery and radiotherapy, whereas the role of any systemic therapy remains controversial.¹ Previously, Macdonald criteria, developed for response assessment in high-grade glioma, or RECIST criteria, designed for solid tumors, have been used in clinical trials enrolling meningioma patients.^2,3

However, response assessment criteria developed for intraaxial brain tumors, such as gliomas, may not be sensitive enough to detect changes in meningioma size due to their extraaxial location, which is associated with an irregular shape, large initial sizes, linear components, varying growth rates, and often long progression-free survival (PFS) and overall survival (OS).⁴ The RANO working group meningioma subcommittee proposed specific criteria for meningioma response assessment.⁵ The threshold values for partial response (a 50% decrease in the diameter product) and progressive disease (a 25% increase in the diameter product) were adapted from intraaxial tumor criteria.^4,5 Moreover, a minor response category, defined as a diameter product reduction of more than 25% but less than 50%, was introduced to address the expected modest size change in meningiomas.⁵

Huang et al. retrospectively analyzed longitudinal MRI data from patients with recurrent meningioma treated with various systemic therapies in routine clinical practice or as part of clinical trials.⁶ They compared different response assessment criteria using 1D and 2D measurements, as well as volumetric tumor analysis, to identify which imaging metrics had the strongest correlation with OS. Their findings indicated that volumetric criteria and a 10 mm change in maximal diameter were most strongly associated with OS.⁶ Other studies have similarly suggested that volumetric analysis of MRI data may be a more sensitive method for assessing slow-growing brain tumors.⁷

The purpose of this study was to compare the most commonly used response assessment techniques, including 1D, 2D, and volumetric tumor measurements, using a prospectively enrolled multicenter study cohort, with the goal of validating the cutpoint selection methodology and key findings previously published by Huang et al.⁶

Methods

Patient Cohort and Study Design

The current study was a post-hoc analysis of the prospectively conducted multicenter EORTC-BTG-1320 study (NCT02234050).⁸ Different imaging response assessment criteria, including 1D, 2D, and volumetric tumor measurements, were compared to identify the technique with the strongest association with OS in patients with recurrent CNS WHO grade 2 and 3 meningiomas. The study received approval from the local ethics committees, and all patients provided written informed consent. The design and outcomes of the EORTC-BTG-1320 trial have previously been published in detail.⁸ In summary, the trial was conducted as an open-label, multicenter, randomized, phase II study involving patients with recurrent CNS WHO grade 2 and 3 meningioma with no more option for local therapy (resection or radiotherapy) after maximal feasible surgery and radiotherapy (n = 90). Patients were randomly assigned in a 2:1 ratio to receive either intravenous trabectedin or local standard of care. Although no significant differences in PFS and OS were observed between the treatment groups, the trial provided a large homogenized data set of these rare tumor entities, allowing for the evaluation of response assessment criteria based on standardized magnetic resonance imaging (MRI) follow-ups every 9 weeks.⁸

This post-hoc analysis includes data from 90 patients. Of these, 57 patients survived until the 6-month landmark, had at least two scans, a baseline MRI scan, and another one within the first 6 months and were therefore eligible for further analysis. Alongside the MR images, clinical data such as CNS WHO tumor grade, treatment regime, age, sex, and OS were collected. OS was defined as time in days from randomization to death or the date of censoring in days. For patients who were still alive at the end of the study, the date of the last tumor assessment was used as the censoring date.

Response Assessment Criteria

The open-source software ITK-Snap (version 3.8.0) was used to obtain tumor size measurements and perform volumetric segmentations on contrast-enhanced T1-weighted MRI sequences.

In accordance with the RANO criteria, target lesions were defined as contrast-enhancing tumor lesions.^4,5 In cases with multiple tumor lesions, up to five target lesions were identified by selecting the largest tumorous lesions and, among them, those with the most clearly defined margins, as these enable the most reproducible measurements.^4,5 The sum of the selected target lesions was used for further analysis. All measurements were centrally performed. For 1D measurements, the longest tumor diameter was recorded. For 2D measurements, a second dimension perpendicular to the longest tumor diameter was added, and the product of these two measurements was calculated. Tumor volume was determined using a semi-automated segmentation tool and was manually adjusted by a senior neuroradiologist.

The cutoff values for response assessment were aligned with the manuscript of Huang et al.⁶, as shown in Table 1.

Table 1.

Defined cutoff values for response assessment

Progression	1D	2D	Volume
	Sum of tumor diameter increase	Sum of product of tumor diameter increase	Tumor volume increase of:
	5 mm, 10 mm 5%, 10%	15%, 25%	20%, 30%, 40%, 50% and 60%

Statistical Analysis

Continuous variables were summarized using descriptive statistics, including the number of valid observations (n), median, interquartile range (IQR), and number of missing values (unknown). Categorical variables were presented as absolute counts and percentages, with missing data treated as a separate category. A Spearman rank correlation coefficient was calculated to assess the pairwise relationship between 1D, 2D, and volumetric measures at baseline as well as the increase within the first 6 months.

Associations between OS from randomization and clinically relevant factors (eg, sex, age, CNS WHO grade, treatment scheme) were evaluated by Cox proportional-hazards regression. The analysis on the tumor change was conducted following Huang et al.⁶ by using a 6-month landmark time point, where a Cox proportional hazards model was employed to analyze the relationship between tumor growth, defined as relative change from the baseline measurement to the last MRI scan prior to the 6 months mark, and OS. OS beyond the 6-month landmark was defined as the time from this landmark to death or last follow-up. This means that only subjects who survived at least 6 months from randomization could be included in this analysis, that is, day 0 in these conditional survival analyses would correspond to 6 months from randomization in the EORTC-BTG-1320 trial.⁸ Both univariable and multivariable Cox regression models were performed. Univariable denotes models with just one independent variable. To assess the validity of the proportional hazards assumption, Schoenfeld residuals and Kaplan–Meier curves were visually inspected.

In this analysis, each imaging criterium refers to the change in tumor size within the 6 months after randomization. The optimal cutoff value for the imaging criteria was determined by dichotomizing the change from baseline to 6 months and choosing the threshold yielding the maximum hazard ratio among the subset of all pre-fixed thresholds with statistical significance (P < .01). The same thresholds were used as by Huang et al.⁶ As an alternative approach for identifying the optimal threshold, we looked at 1-year survival (after the 6-month landmark) and chose the cutoff where the mean of specificity and sensitivity were maximized. The results were visualized using a receiver operating curve (ROC) curve. The area under the curve (AUC) for each ROC was reported for each model. Estimation was conducted using the Nearest Neighbor estimation from survivalROC package.

For both analyses, we assumed a negative correlation between tumor growth and survival, for example, higher increase in tumor growth between baseline and 6 months corresponds to lower conditional overall survival after 6 months. To validate the resulting cutoffs, we conducted a multivariable Cox regression, including clinically relevant covariates such as age, sex, WHO CNS grade as well as baseline tumor size. Furthermore, associations between OS and other variables such as age, sex, WHO CNS grade, and treatment were assessed using log-rank tests and Cox regression analyses.

To account for multiple comparisons across the three different methods (1D, 2D, and volume), a more stringent P-value of less than .01 was considered significant. All statistical analyses were performed using R version 4.3.1, along with the packages “survival,” “compare,” and “survivalROC.”

Results

Patient Characteristics

Supplementary Table 1 provides an overview of the clinical and demographic characteristics of the patient cohort used for the 6-month landmark survival analysis. Patients underwent a median number of three MRI (interquartile range, IQR, 2–4), with a median time between examinations of 65 days (IQR 62–75.5). Median time between baseline and first follow-up MRI was 70 days (IQR 63.75–77), while the last MRI examination was performed, on medium, 118 days (IQR 62–185) after baseline MRI. Sex (P = .54), age (P = .6), CNS WHO grade (P = .24), and treatment scheme (P = .79) showed no association with OS from randomization.

Comparison of Cutoff Values for Different Response Assessment Techniques Associated with OS from 6-month Landmark

A 6-month landmark was used, including all patients who were still alive and had at least two MRI examinations within this time frame. Please note that now all subsequent time-to-event analysis include only patients who were alive at 6 months. Furthermore, OS and PFS are now defined from the 6-landmark onwards, that is, in the subsequent analysis, time 0 corresponds to the 6-month after randomization. A total of 57 patients were included in this analysis, with a median time between baseline and the last MRI examination before the 6-month landmark of 130 days (IQR 79–148 days). Sex (P = .54), age (P = .78), CNS WHO grade (P = .41), and treatment scheme (P = 0.75) showed no association with OS from the 6-month landmark.

Median changes in the sum of tumor diameters (1D measurements), the sum of the product of the tumor diameters (2D measurements), and the sum of tumor volumes are given in Table 2.

Table 2.

Median change of tumor measurements within the first 6 months after randomization. Only patients with OS more than 6 months were included.

1D measurements (sum of diameter)	Median (IQR)
Relative increase in %	10 (5 - 19)
Absolute increase in mm	6 (2 - 12)
2D measurements (sum of product of diameter)
Relative increase in %	25 (14 - 51)
Tumor volume
Relative increase in %	36 (18 - 87)

To compare cutoff values for different response assessment techniques (1D, 2D, tumor volume) showing the strongest association with OS, we applied two different approaches:

• (A) Conditional overall survival from 6-month landmark

First, a Cox regression model from the 6-month landmark was fitted for a predefined set of cutoff values. Table 3 provides an overview on the number of patients with an increase above each cutoff as well as the model summary.

Table 3.

Summary of uniform Cox regressions from 6-month landmark time point with various cutoff values for 1D, 2D, and volumetric tumor measurements. A P-value < .01 was considered as significant. Only patients with OS over 6 months were included in the analysis.

Increase in the sum of tumor diameters	Number of patients (n; %)	Hazard Ratio	Lower CI	Upper CI	P-value
1D measurements
20% increase	14 (25%)	1.845	0.846	4.023	.124
10% increase	27 (47%)	1.735	0.858	3.525	.128
10 mm increase	20 (35%)	3.228	1.584	6.58	.001*
5 mm increase	31 (54%)	3.414	1.563	7.455	.002*
2D measurements
25% increase	31 (54%)	1.802	0.862	3.768	.118
15% increase	41 (72%)	2.022	0.828	4.937	.122
Volumetric measurement
60% increase	20 (35%)	3.187	1.55	6.552	.0016*
50% increase	24 (42%)	3.351	1.623	6.919	.0011*
40% increase	25 (44%)	3.664	1.754	7.654	.0005*
30% increase	31 (54%)	3.697	1.644	8.314	.0016*
20% increase	41 (72%)	3.789	1.322	10.86	.0132

An increase in tumor diameter of 5 mm and 10 mm using 1D measurements demonstrated a significant association with OS, with hazard ratios of 3.41 (95%CI: 1.56–7.46, P = .002) and 3.23 (95%CI: 1.58–6.58, P = .001), respectively.

Additionally, tumor volume increases of 30%, 40%, 50%, and 60% were also significantly associated with OS, with hazard ratios of 3.7 (95%CI: 1.64–3.31, P = .001), 3.66 (95%CI: 1.75–7.65, P = .001), 3.35 (95%CI: 1.62–5.93, P = .002), and 3.19 (95%CI: 1.55–6.55, P = .0016), respectively. The cutoffs 5 mm increase in 1D-diameter and 30% increase in volume, showed the highest hazard ratios for their measures, while statistically significant.

For 2D measurements, the predefined cutoff values of 15% and 25% increase in the sum of the tumor diameter products did not show a significant association with OS from 6-month landmark.

• (B) One-year conditional overall survival from 6-month landmark

The optimal cutoff for 1-year survival and corresponding ROC was calculated. We found that the relative increase in volumetric measure had the highest AUC at 0.72. Other AUC values are shown in Supplementary Table 2andSupplementary Figure 1.

For the 1D measure, the best cutoffs were an absolute increase of 6 mm or a relative increase of 10% compared to the baseline tumor with a mean sensitivity and specificity of 0.691 and 0.597, respectively. The optimal cutoff for 2D was a relative increase of 20% (mean sensitivity and specificity = 0.579) and for the volumetric it was 40% (mean sensitivity and specificity = 0.711). Supplementary Table 2 shows the complete summary of all cutoff values, their performance measures, and AUC for each measure.

When implementing these cutoffs in a univariable Cox regression, only two values were statistically significant. In the 1D measurement category, an absolute increase of 6 mm in the sum of tumor diameters resulted in a hazard ratio of 2.91 (95% CI: 1.43–5.93) with a P-value of P = .003 (c-index 0.650), as illustrated in the Kaplan–Meier curve in Figure 1.

Figure 1.

Kaplan Meier plot for overall survival from 6-month landmark by absolute 1D tumor increase.

Regarding tumor volume, an increase of 40% in the sum of tumor volumes resulted in a hazard ratio of 3.66 (95% CI: 1.72–6.97) and a P-value of P < .001 (c-index 0.657). The Kaplan–Meier curve in Figure 2 illustrates the association of a 40% increase in the sum of tumor volume with survival.

Figure 2.

Kaplan Meier Plot for Overall Survival from 6-month landmark by relative volumetric tumor increase.

In the 2D measurement category, no cutoff value was significantly associated with predicted OS at the 6-month landmark. The optimal cutoff of 20% increase had an estimated hazard ratio of 1.85 (95% CI: 0.85–4.02) and a P-value of P = .12.

After adjusting for age, sex, and histological grade, the cutoff for the tumor volume remained significantly associated with survival (HR = 3.81, 95% CI: 1.79–8.12, P < .001). The optimal cutoff for the 1D absolute difference was also statistically significant with a hazard ratio of 3.55 (95% CI: 1.55–8.12, P < .001) after adjusting. None of the other determined cutoffs showed a statistically significant association with survival after adjusting for these variables (see Supplementary Tables 3and4).

Moreover, Spearman rank correlation revealed a strong correlation in tumor size at baseline across the 1D, 2D, and volumetric measurements (1D and 2D measurements: r = 0.8, P < .001; 1D and volumetric measurements: r = 0.8, P < .001; 2D and volumetric measurements: r = 0.89, P < .001). The relative change in tumor size within the first 6 months after randomization was also highly correlated among the three measures (1D and 2D measurements: r = 0.87, P < .001; 1D and volumetric measurements: r = 0.68, P < .001; 2D and volumetric measurements: r = 0.66, P < .001).

Discussion

The use of clinically relevant cutoff values for response assessment is of utmost importance when conducting clinical trials or making decisions in routine clinical practice. The RANO working group meningioma subcommittee recently proposed response assessment criteria for patients with meningioma, with cutoff values largely adopted from established response assessment criteria for low- and high-grade gliomas.^4,5 Additionally, the subcommittees recommended incorporating volumetric tumor measurements into the response assessment of meningiomas. This recommendation was based on a recently published retrospective study that demonstrated a stronger association between tumor volumes and OS compared to bidimensional tumor measurements in recurrent meningiomas of all grades undergoing systemic therapy.⁶

To support these recommendations and validate previous findings, the current study was designed as a post-hoc analysis of a multicenter, prospective clinical trial enrolling patients with recurrent CNS WHO grade 2 and 3 meningiomas,⁸ with the aim of comparing different response assessment criteria based on 1D, 2D and volumetric tumor measurements on MR images, to serve as an independent external validation of the methodological framework in Huang et al.⁶

Initially, based on each patient’s individual OS, a set of predefined cutoff values for the response criteria was simultaneously compared (approach A). The threshold yielding the maximum hazard ratio with statistical significance was determined, similar to the study by Huang et al.⁶ Alternatively, we determined the optimal threshold for predicting 1-year survival (approach B), maximizing the mean of specificity and sensitivity.

Despite the strong correlation between all response assessment techniques both approaches revealed that, consistent with previously published findings,⁶ tumor volume and absolute unidimensional tumor measurements showed the strongest association with OS. This was true regardless of whether the individual OS or predicted 1-year OS was used.

An increase of 5 mm or 10 mm in tumor diameter revealed comparable associations with OS (HR 3.41, 95%CI: 1.56–7.46, P = .002 and HR 3.22, 95%CI: 1.58–6.58, P = .001, respectively) using the individual OS (approach A). Maximizing the mean specificity and sensitivity for predicting 1-year survival (approach B) revealed a cutoff value of 6 mm (HR 3.12, 95%CI: 1.46–6.65, P = .003). Due to known high interreader variability in detecting small changes in lesion size, an absolute 10 mm cutoff appears to be the most reliable when using unidimensional measurements, consistent with the findings of Huang et al.⁶

Various cutoff values for tumor volume (30%–60% tumor volume increase) showed a significant association with OS, with 30% increase in tumor volume being the cutoff with the maximal HR with individual OS in approach A (HR 3.69, 95%CI: 1.46–8.31, P < .001), while a 40% increase in tumor volume was the optimal cutoff for predicted 1-year OS (HR 3.45, 95%CI: 1.75–7.65, P < .001) in approach B. These findings align with the recommendation of the proposed response assessment criteria for meningioma by the RANO working group and the results of Huang et al.^5,6

Interestingly, despite the established response assessment criteria for meningioma, which recommend bidimensional measurements, neither approach used in this study revealed a significant correlation between bidimensional measurements of tumor progression and patient OS. The fact that unidimensional and volumetric measurements show a more significant association with overall survival may reflect the non-spherical growth pattern of this extra-axial, often irregularly shaped tumor type, which may be more clearly captured by assessing 1D and volumetric tumor growth. Also, previous findings have shown that unidimensional tumor measurements and tumor volume assessments have a better association with meningioma patient OS than bidimensional.⁶ Moreover, the classification of tumor 3D volume growth rates was associated with patient outcomes and clinical status of meningioma patients in this study cohort. This suggests that tumor volume assessment could help identify early signs of drug activity in meningioma patients by reflecting volume stabilization.⁹ Based on these results, the measurement method in the response assessment criteria for meningioma should be reconsidered or supplemented by alternative methods, such as the simple and effective absolute unidimensional tumor measurement or the implementation of routine tumor volume segmentations, potentially supported by AI tools integrated into everyday clinical practice.

The main limitations of our study include the retrospective study design, the lack of a prospective validation cohort for cutoff definition, and the relatively small sample size. Specifically, identifying and testing cutoff values within the same dataset introduces a potential risk of overfitting, which could limit the generalizability of our results. However, our analysis was conducted using an independently accrued, prospectively collected multicenter trial cohort (EORTC-BTG-1320),⁸ applying the same methodology and predefined thresholds as used in the previous study by Huang et al.⁶ In doing so, we performed an external validation not of new numerical cutoffs, but of the methodology for determining them. The reproducibility of key associations—namely, the stronger correlation of unidimensional and volumetric tumor measurements with overall survival, compared to bidimensional measurements—across two methodologically aligned but distinct cohorts, supports the robustness of these results. Nonetheless, our findings should be regarded as hypothesis-generating. Further prospective studies with larger cohorts and formal external validation arms will be necessary to confirm the clinical utility of the proposed cutoff values.

However, CNS WHO grade 2 and 3 meningiomas are rare tumor subtypes, making this cohort uniquely valuable for testing response assessment in this specific tumor group. Unfortunately, only a minor number of patients (11%) in this study exhibited tumor size reduction at some point of the study, so imaging cutoffs for partial or minor response could not be tested. Another limitation of our study is the use of landmark analysis, which inherently conditions on survival to a fixed time point and may introduce bias by excluding patients with early events. This design restricts generalizability and limits inference from the baseline for the entire cohort. However, we took deliberate steps to mitigate these concerns, including clearly defining the landmark and applying two complementary analytic approaches and Cox regression models, and verifying the proportional-hazards assumption by examining Schoenfeld residuals. The 6-month landmark approach as well as the cutoff values used in this study to assess overall survival (OS) were adopted directly from a previous publication.⁶ We refrained from selecting a 3-month landmark due to the short median time period between randomization and follow-up MRI (64 days), with only minimal observed changes in tumor size. Huang et al. further defined a 12-month landmark in addition to the 6-month landmark. Patients in our cohort had a lower median survival of 312 days because we exclusively included patients with recurrent CNS WHO grade 2 and 3 meningiomas, which are known to have significantly shorter survival compared to those with CNS WHO grade 1 meningiomas. As a result, using a 12-month landmark would have reduced our study population from the initial 90 patients to 38. Due to the limited statistical power with this smaller sample size, we only used the 6-month landmark in the current study. Thus, additional different time points should be assessed in future trials. Another limitation of this study is that OS was assessed as a predefined secondary endpoint in the EORTC 1320 study without distinguishing tumor-related deaths from other causes, resulting in the unavailability of specific data on tumor-related mortality.

However, an advantage of the current trial is that the analysis is based on the prospectively conducted EORTC-BTG-1320 trial, reflecting a highly homogenous study population consisting of patients with recurrent CNS WHO grade 2 and 3 meningiomas. All patients were required to demonstrate radiologically documented progression of any tumorous lesion at trial enrollment, with a > 25% increase in the past year or the appearance of a new lesion, thus representing a cohort of fast-growing meningiomas. Additionally, the use of a standardized MRI protocol across all participating centers ensured uniform imaging data, ideal for volumetric tumor assessment. The use of isometric T1-weighted post-contrast MR sequences with a slice thickness ≤ 1.5 mm, as recommended in the standardized brain tumor imaging protocol (BTIP), allowed for highly accurate measurements, capturing even very small changes in tumor size that might be missed with MR examinations using greater slice thickness or a higher distant factor.¹⁰

Conclusion

Our study, along with previous findings, demonstrates that unidimensional tumor measurements and tumor volume assessments provide a stronger correlation with overall survival in recurrent CNS WHO grade 2 and 3 meningioma patients than bidimensional measurements. Consideration should be given to incorporating alternative approaches in the current response assessment criteria, such as the straightforward and effective unidimensional tumor measurement or routine tumor volume segmentations, especially in light of potentially integrating AI tools into routine diagnostics in the future.

Funding

This study received no specific funding.

Conflict of interest statement. JF has received honoraria for lectures and consultation from the following for-profit companies: Novarits, Seagen, Sanova, Servier. LB has no conflict of interest. ELR has received research grants from Bristol Meyers Squibb (BMS), and honoraria for lectures or advisory board participation or consulting from Astra Zeneca Daiichi, Bayer, Biodexa/Sitoxi, Janssen, Leo Pharma, Pfizer, Pierre Fabre, Roche, Seattle Genetics, and Servier. AS has no conflict of interest. RR has no conflict of interest. GL reports a relationship with consulting or advisory role funding from ABBVIE, Bayer, Novartis, Orbus Therapeutics, BrainFarm, Celgene, CureTeq, GlaxoSmithKline, Health4U, Braun, Janssen, BioRegio Stern, Servier, Novocure, and travel funding from Roche and Bayer, Servier. JMSS has no conflict of interest. PB has received honoraria for lectures, consultation or advisory board participation from Servier, “Ekspertpanelet” and “Moloklinikken.” MB has no conflict of interest. VG has no conflict of interest. TG has no conflict of interest. FS is co-founder and shareholder of Heidelberg Epignostix GmbH. WW has no conflict of interest. GM has received honoraria for lectures, consultation, or advisory board participation from the following for-profit companies: Brainlab, Accuray, Servier, Astra-Zeneca, Novocure, Pfizer. MW has received research grants from Novartis, Quercis, and Versameb, and honoraria for lectures or advisory board participation or consulting from Bayer, Curevac, Medac, Novartis, Novocure, Orbus, Philogen, Roche, Sandoz, and Servier. FK has no conflict of interest. MP has received honoraria for lectures, consultation, or advisory board participation from the following for-profit companies: Bayer, Bristol-Myers Squibb, Novartis, Gerson Lehrman Group (GLG), CMC Contrast, GlaxoSmithKline, Mundipharma, Roche, BMJ Journals, MedMedia, Astra Zeneca, AbbVie, Lilly, Medahead, Daiichi Sankyo, Sanofi, Merck Sharp & Dome, Tocagen, Adastra, Gan & Lee Pharmaceuticals, Janssen, Servier, Miltenyi, Böhringer-Ingelheim, Telix, Medscape, OncLive.

Authorship

Contribution to study design and its implementation: JF, LB, ELR, AS, RR, GL, JMSS, PB, MB, VG, TG, FS, WW, GM, MW, FK, MP. Data analysis and interpretation: JF, MP. Writing of first manuscript draft: JF, MP. All authors read, edited, and approved the final version of the manuscript.

Data availability

Underlying data can be provided upon reasonable request to the corresponding author and approval from relevant regulatory authorities.