Treatment-associated imaging changes in newly diagnosed MGMT promoter-methylated glioblastoma undergoing chemoradiation with or without cilengitide

Abstract Background Radiological progression may originate from progressive disease (PD) or pseudoprogression/treatment-associated changes. We assessed radiological progression in O6-methylguanine-DNA methyltransferase (MGMT) promoter-methylated glioblastoma treated with standard-of-care chemoradiotherapy with or without the integrin inhibitor cilengitide according to the modified response assessment in neuro-oncology (RANO) criteria of 2017. Methods Patients with ≥ 3 follow-up MRIs were included. Preliminary PD was defined as a ≥ 25% increase of the sum of products of perpendicular diameters (SPD) of a new or increasing lesion compared to baseline. PD required a second ≥25% increase of the SPD. Treatment-associated changes require stable or regressing disease after preliminary PD. Results Of the 424 evaluable patients, 221 patients (52%) were randomized into the cilengitide and 203 patients (48%) into the control arm. After chemoradiation with or without cilengitide, preliminary PD occurred in 274 patients (65%) during available follow-up, and 88 of these patients (32%) had treatment-associated changes, whereas 67 patients (25%) had PD. The remaining 119 patients (43%) had no further follow-up after preliminary PD. Treatment-associated changes were more common in the cilengitide arm than in the standard-of-care arm (24% vs. 17%; relative risk, 1.3; 95% CI, 1.004–1.795; P = .047). Treatment-associated changes occurred mainly during the first 6 months after RT (54% after 3 months vs. 13% after 6 months). Conclusions With the modified RANO criteria, the rate of treatment-associated changes was low compared to previous studies in MGMT promoter-methylated glioblastoma. This rate was higher after cilengitide compared to standard-of-care treatment. Confirmatory scans, as recommended in the modified RANO criteria, were not always available reflecting current clinical practice.


Neuro-Oncology
Cilengitide is an integrin inhibitor with antiangiogenic activity and was used in clinical trials combined with standard treatment, both for MGMT promoter-methylated and unmethylated glioblastomas. 5,6No survival benefit was found with the addition of cilengitide in either group.
Radiological and/or clinical worsening may be seen at any time posttreatment and may be the result of actual progressive disease (PD) or an effect of treatment-associated changes.[9][10][11] Angiogenesis inhibitors, such as cilengitide or bevacizumab, are supposed to block the formation of new vessels.Bevacizumab restores the permeability of leaky vessels through blocking effects on vascular endothelial growth factor (VEGF) signaling.In the large prospective AVAglio trial comparing standard treatment with bevacizumab to standard treatment with placebo in glioblastomas, considerably fewer patients in the intervention arm developed pseudoprogression compared to the control arm (2.2% vs. 9.3%). 12The effect of cilengitide on the rate of treatment-associated changes has not yet been investigated.
The response assessment in neuro-oncology (RANO) criteria of 2010 provided radiological recommendations to assess treatment response on anatomical MRI.In these criteria, the most common method to diagnose PD required an increase of ≥25% in the sum of products of the perpendicular diameters (SPD) of enhancing lesions. 13Because of the probability of early pseudoprogression, PD cannot be diagnosed in the first 3 months postradiation.
The proposal for modified RANO criteria for radiographic response assessment in glioblastoma of 2017 added an interval of 4 weeks or more between the first increase for preliminary PD and a second increase of ≥25% for confirmed PD. 14 In the case of no further increase, but rather stable or regressing disease, confirmed pseudoprogression is diagnosed.This proposed change from the 2010 criteria reflects the insight that treatmentassociated imaging changes can occur beyond three months postradiation.Another suggestion was to use the first postradiation MRI and not the early postoperative MRI as the baseline MRI, because of possible postsurgical artifacts and timing, as well as protocol variations between centers of the postsurgery MRI as patients are not yet included in clinical trials.
The main objective of this study was to subdivide new or increasing contrast-enhancing (CE) lesions into PD versus treatment-associated changes using the modified RANO criteria. 14Specifically, we aimed to determine the incidence of treatment-associated imaging changes as a function of: (a) timing after treatment and (b) the addition of cilengitide.

Methods
The EORTC-CENTRIC study (NCT00689221) was a multicenter, open-label, phase-3 randomized clinical trial to investigate the effect of cilengitide on overall survival (OS) in patients with MGMT promoter-methylated glioblastoma; in this trial, the addition of cilengitide to standard-of-care chemoradiation did not result in an OS benefit.Details on the trial design and its main results have been published. 5Informed consent was available from all participants for the initial study.As stated in the original paper, the CENTRIC study was approved by the institutional review boards or independent ethics committees of the participating institutions and competent authorities according to country-specific regulations. 5The current study is a secondary analysis of the imaging data and the original consent also covers additional studies of the imaging data.
For this retrospective study of a prospectively collected cohort, 2 clinical researchers (CMF and MF) reviewed 60% and 40%, respectively, of the MRIs of the CENTRIC database to determine the rates of PD and treatmentassociated changes according to the modified RANO criteria. 14The first 10 patients were assessed by both readers and then discussed in the group (CMF, MF, TJS, NG, PL) to test the agreement between the 2 raters and verify the radiological criteria.All cases wherein the reader was uncertain about the correct evaluation were discussed in this same group.

Importance of the Study
Response evaluation in irradiated gliomas remains difficult due to treatment-associated changes.A modification of the 2010 response assessment in neuro-oncology (RANO) criteria was proposed in 2017 to improve radiological response assessment.We reviewed a large, multicenter cohort of patients with O 6 -methylguanine-DNA methyltransferase (MGMT) promoter-methylated glioblastoma treated with standard chemoradiation with or without cilengitide.With modified RANO, we found a low rate of treatment-associated changes (21%) compared to previous studies, which in part may be related to the stringent criteria.Eighty-eight of 274 patients with progression (32%) had treatment-associated changes.The interpretation was difficult in 119 patients (43%) because no confirmatory scan was made after preliminary PD.
The rate of treatment-associated changes was highest during the first 3 months after RT and higher after cilengitide than after standard-of-care treatment.
Our findings may facilitate further research to better understand the pathophysiology of treatmentassociated changes.This study may offer practical insights for the design of future clinical trials.

Inclusion Criteria
All patients with the availability of: (a) 3 or more follow-up MRIs in total and at least 2 after the end of RT, and (b) all relevant clinical information, were included.Relevant clinical data consisted of data on treatment allocation, RT, progression-free survival (PFS), and second-line therapy.
We hypothesized that the excluded patients with missing data showed early progression and examined if their time from diagnosis to the PFS date was less than three months.Furthermore, we noted their treatment allocation group.

Definition of Baseline, Response, and Outcome
As proposed in the modified RANO criteria, the baseline MRI was defined as the first MRI after the end of RT. 14 In case of a (partial or complete) response during follow-up, the best response MRI before preliminary PD was defined as the new nadir and the new baseline MRI.
To detect early pseudoprogression, we also compared the first MRI after RT (baseline MRI as defined before) to the pre-RT MRI.In the case of preliminary PD on the first MRI after RT compared to the pre-RT MRI and no postsurgical artifacts that hindered the evaluation, the pre-RT MRI was defined as the baseline MRI.
Preliminary PD was defined as a first ≥25% increase of the SPD of the contrast-enhancing lesion compared to baseline on the axial T1-weighted MRI after gadolinium-based contrast agent administration.The outcome PD was defined as a second ≥25% increase of the SPD on at least 1 follow-up MRI four weeks later.A new measurable lesion outside of the radiation field was immediately considered to be PD according to the 2010 RANO criteria. 13A new measurable lesion within the radiation field was added to the SPD until a maximum of five lesions.The outcome of "treatment-associated changes" was defined as either stable disease (SD, <25% increase or <50% decrease on at least 2 follow-up MRIs, each 1 at least 4 weeks after the previous); or partial or complete response (>50% decrease on at least 1 follow-up MRI 4 weeks later) after preliminary PD. Figure 1 depicts an example of PD and treatmentassociated changes.
The response included partial or complete response (PR/CR), preliminary partial or complete response (pPR, pCR) or pseudoresponse (PsR) as defined by the modified RANO criteria. 14We classified the outcomes of patients with no RANO progression within the available follow-up time as (preliminary) partial response (pPR, PR), (preliminary) complete response (pCR, CR), stable disease (SD), or pseudoresponse (PsR).The best response of patients with no measurable disease at baseline was SD.For patients who did not have further follow-up MRIs after preliminary PD, we classified the outcome as unconfirmed PD.
Certain cases could not be classified according to the above-mentioned criteria since there was only 1 follow-up scan after preliminary PD without a clear increase or

Neuro-Oncology
decrease.To provide a meaningful classification for these patients and to prevent the formation of a biased set of "unclassifiable" cases, we handled these cases as follows: in the case of only 1 follow-up scan at least eight weeks after the preliminary PD scan showing SD (<15% increase) and the date of PFS was not the date of the preliminary PD scan, then we classified the course as treatmentassociated changes.In the case of only 1 follow-up scan fewer than 8 weeks later and/or the lesion increased >15% and/or the PFS date corresponded to the preliminary PD date, then the outcome was defined as unconfirmed PD.In these cases, true progression was suspected but the modified RANO criteria for PD were not met.
A new treatment introduced before the end of available follow-up could influence the outcome.Therefore, we excluded patients with an outcome defined as SD or treatment-associated changes who started a second treatment before the last available scan (n = 6).Salvage surgery on a lesion that developed on the PFS date was regarded as a second treatment.Otherwise, a second surgery was ignored.All lesions that were labeled as "(probably) outside the radiation field" were reviewed by an experienced neuro-oncologist (TJS).Of note, planning information on radiation fields was not available for comparison; we estimated the extent of the radiation field from the T2-weighted FLAIR hyperintense and contrast-enhancing regions on postoperative MRI.Difficult cases, and problems with the use of the response criteria, were discussed and solved in consensus meetings between the first 2 authors (MF, CMF) and the 2 senior authors (NG, TJS).
We evaluated the MRIs according to the modified RANO criteria.The most used criteria in previous clinical trials are however the RANO criteria of 2010, which served also in the CENTRIC study. 13To compare the results based on the modified RANO criteria of 2017 with the RANO criteria of 2010, we also give an overview of the results according to the 2010 RANO criteria.Table 1 summarizes the main differences between the 2010 and 2017 RANO criteria.

Analysis
Time-to-progression (TTP) was defined as the time from the end of RT to preliminary PD and categorized into 3 subgroups: before 3, from 3 to 6, and after 6 months after RT.We calculated a relative risk with a 95% CI and P-value for the development of treatment-associated changes in the cilengitide group compared to the control group.We used a chi-square test with 95% CI and P-value to compare the patients with missing or no missing values between the allocation groups.To compare the survival curves of patients with PD and treatment-associated changes, the log-rank test was used.

Results
In total, we reviewed 4010 MRIs from 545 patients randomized into the CENTRIC study.The patient characteristics are given in the original article. 5In brief, men were the predominant group (53%), the median age was 58 years, and almost half had undergone a gross total resection (49%).Of all patients, 49.9% were randomized into the cilengitide arm, and 50.1% in the control arm.Due to missing MRIs or clinical data, 121 patients were excluded from the analysis.Of the remaining 424 patients, 221 patients (52%) had been randomized into the cilengitide arm and 203 patients (48%) into the control arm.The median SPD at baseline or nadir was 0 mm 2 ; interquartile range (IQR), 565.3 (0-565.3).RANO measurable disease at baseline or nadir was present in 283 patients (67%).RANO response led to a new baseline MRI in 102 patients (24%).In total, 141 patients (33%) had no measurable disease at baseline.

Confirmatory scan required
No Yes SPD = sum of products of perpendicular diameters; SD = stable disease, PR = partial response, CR = complete response.
as unconfirmed PD, mostly because a clinical decision to diagnose PD was made.The overall rate of treatmentassociated changes was 21% (88 of 424 patients).Seven of the 15 patients with only 1 follow-up MRI after preliminary PD were classified as treatment-associated changes and 8 as unconfirmed PD.
We observed a response to treatment in 101 patients, mostly partial response (n = 61).The outcome analysis showed no progression within the available follow-up time in 150 patients (35% of 424 patients).Figure 2 depicts the patient selection process and the detailed outcome groups.
Figure 3 shows the outcome after preliminary PD for different time periods.Of all patients who developed preliminary PD 3 months after RT, 54% were diagnosed with treatment-associated changes and 46% with PD or unconfirmed PD.However, of all the patients with preliminary PD after 6 months after RT, 13% had treatment-associated changes, and 87% had PD or unconfirmed PD.
In the cilengitide group of 221 patients, 54 patients (24%) developed treatment-associated changes, and 30 patients (14%) developed PD, whereas in the control group of 203 patients, 34 patients (17%) developed treatmentassociated changes, and 37 patients (18%) PD.The relative risk for the development of treatment-associated changes in the cilengitide group compared to the control group was 1.3 (95% CI 1.004-1.795;P = .047).
Five patients were diagnosed with PD based on new measurable lesions outside of the radiation field.

Patients with Missing Data
The progression date (PFS) of 100 of the 121 excluded patients occurred within 3 months after diagnosis and after 3 months in another 21 patients.Numerically, more excluded patients had been randomized into the control group than into the intervention group (70 vs. 51 patients, chi-square test 3.8, P = .05).

Radiological Criteria
Here we give an overview of the results according to the 2010 RANO criteria: Progressive disease within the first 12 weeks following the end of RT can only be diagnosed in the case of a new lesion outside of the radiation field or by neuropathological confirmation.Therefore, we would have had to wait for further follow-up in 106 of 107 patients who progressed in our study within the first three months after RT before diagnosing progression or treatment-associated changes.In the last patient, a new lesion outside of the radiation field was found and progression could have been called.All 167 patients with preliminary PD after 12 weeks after RT (those with treatment-associated changes, PD and unconfirmed PD) would have been diagnosed with PD, without the need for a confirmatory scan.This difference in response assessment may also explain why we found no progression in 35% of patients in our study, although a PFS date was noted for these patients in the original CENTRIC data.

Survival
The mean overall survival (OS) of the patients according to modified RANO with treatment-associated changes was 997.1 days/32.8months (95% CI 905.0-1089.2) and for those with (unconfirmed) progressive disease 822.2 days/27.0months (95% CI 760.6-883.7).The P-value of the logrank test was 0.003.The Kaplan-Meier curve is included in   disease (with a confirmatory scan) was 877.8 days/28.8months (95% CI 775.2-980.4).The comparison with those with treatment-associated changes produced a P-value of .115(log-rank test).We supposed that the patients with unconfirmed PD showed clear clinical progression and therefore did not undergo a confirmatory scan.We compared the OS of the patients with unconfirmed and confirmed PD, which showed a numerical lower OS for the patients with unconfirmed PD (mean OS 766.8 days/25.2months) (95% CI 698.8-834.7),P-value of log-rank test .151).

Discussion
In this retrospective analysis of a multicenter, openlabel, phase-3 randomized clinical trial in patients with MGMT promoter-methylated glioblastoma treated with cilengitide, we reviewed MRI data from 545 patients according to the modified RANO criteria.After the exclusion of patients with missing data, we found an increased risk for treatment-associated changes in the cilengitide group (24% of 221 patients) compared to the control group (17% of 203 patients).These numbers produced a relative risk of 1.3 (95% CI = 1.004-1.795).
6][17][18] This discrepancy between our rate and previous rates could be the consequence of the use of the more stringent, modified RANO criteria in our study with the necessity to await a confirmatory scan before diagnosing treatment-associated changes.The previously mentioned author groups assessed tumor progression according to the Macdonald's, the 2010 RANO, or their own criteria. 13,22urthermore, the incidence of treatment-associated changes may be lower in the treatment arm due to a similar antiangiogenic effect of cilengitide to that of bevacizumab in previous studies. 12We found the contrary: a higher incidence in the treatment arm.Of note, cilengitide has a different mechanism of action compared to bevacizumab with a less prominent, if any, effect on the blood-brain barrier. 6,23Local hypoxia in tumors leads to the binding of hypoxia-inducible factor to the vascular endothelial growth factor (VEGF) gene and the release of circulating VEGF.VEGFs bind to VEGF-receptors (VEGFR) on endothelial cells, which stimulates their aggregation and proliferation, and leads to the expression of integrins.Integrins expressed in various cell types enable the cells to link to extracellular matrix proteins and to migrate and form for example new vessels.Bevacizumab binds to VEGF and hinders the linkage to the VEGFR-1 and 2, whereas the integrin inhibitor cilengitide blocks the linkage of the integrins αvβ3 and αvβ5 to ligands or matrix proteins. 6,23,246][27][28][29][30] Integrins αvβ3 and αvβ5 interact with VEGFR-2 on endothelial (tumor) cells, [31][32][33][34] whereas VEGFR-1 interacts with β1 integrins. 35A higher incidence of treatment-associated changes after cilengitide compared to standard treatment may be explained by a transient normalization of the blood-brain barrier by cilengitide improving perfusion and an in vivo observed increase in VEGFR-2-dependent angiogenesis. 5,30,34,36Considering, therefore, the VEGF-pathway, αvβ3 and αvβ5 blockage could lead to more imaging changes related to treatmentassociated changes compared to a blockage of VEGFR-1 and 2, because the development of treatment-associated changes may rely more on VEGFR-1 and the development of pseudoresponse may depend on VEGFR-1 blocking mechanisms.
Treatment-associated changes developed mostly within three months after RT (66% of all treatment-associated changes; 54% of all patients).In earlier studies, 12-64% of all patients with treatment-associated changes had these changes within 3 months after RT. 15,16,[18][19][20][21] The limitations of our study are the retrospective nature, the single assessment of a part of the database by 2 nonradiologists and no information on the radiation fields.However, the exact radiation fields were only necessary in the small number of patients (n = 5) with a probable out-of-field lesion.This limitation is therefore considered negligible.To ensure the quality of the assessment, the researchers were trained by 2 experienced neurooncologists.In addition, repeated consensus meetings for the review of cases with these researchers and neurooncologists were held.Finally, additional discussions with a European expert panel on oncological neuroradiology, the EORTC Brain tumor imaging committee, took place.The retrospective design led to 43% unconfirmed PD outcomes with no further follow-up, because the original study used the 2010 RANO criteria for outcome evaluation and, consequently, did not require a confirmatory scan for increasing MRI lesions. 5However, in the most recent RANO recommendations, a confirmatory scan is advised in the first 3 months after radiotherapy, but not mandatory; at later time points, it is not advised. 37In our data, we saw that the patients with "unconfirmed PD" (i.e.without a confirmatory scan) had survival that was numerically worse than for "confirmed PD, " supporting the notion that most cases of "unconfirmed PD"-cases were true PD.
We did not add volumetric data as the dataset was too heterogeneous-and thus insufficient-for efficient and robust methods of automated volume segmentation.Adding volumetric data in future studies could extend and expand our findings.
The prespecified assessment criteria and the large, multicenter, multinational sample constitute the strengths of this study.The consensus meetings and the close supervision by 2 experienced neuro-oncologists (TJS, NG) increased the interrater reliability of the rating.
This study illustrates the constraints of the 2010 RANO criteria. 13The lack of a clear definition of treatment-associated changes forces clinicians to rely on clinical status and experience to make a diagnosis, and therefore many different definitions of treatment-associated changes are described in the literature.In this study, 119 patients were diagnosed with unconfirmed PD in the absence of a confirmatory scan.

Neuro-Oncology
If different criteria were used, such as the 2010 RANO criteria, they might have been diagnosed as progressive.Some of these 119 may have had treatment-associated changes, as we have seen that 88 patients in this study were diagnosed as such after initial preliminary PD.
In conclusion, the rate of treatment-associated changes in the CENTRIC database was low compared to previous studies in newly diagnosed MGMT promoter-methylated glioblastoma.This could be the consequence of the stringent diagnostic criteria used.We observed more treatment-associated changes after cilengitide compared to standard treatment alone.The added value of new criteria sets such as the proposed modified RANO criteria of 2017 needs to be validated, ideally in a prospective setting.

Figure 1 .
Figure 1.Illustration of treatment-associated changes and progressive disease.Top row 1: Patient example for treatment-associated changes.On the first MRI (1.baseline), the baseline scan was performed 3 weeks after the end of radiotherapy.The first follow-up MRI 7 weeks later (1.baseline + 7 weeks) shows preliminary progressive disease.The 2 follow-up MRIs show a decrease after 14 weeks (1.baseline + 14) compared to "1.Baseline+7 weeks" and a stable lesion after 22 weeks (1.baseline + 22).This patient was treated with cilengitide in addition to standard therapy.Bottom row 2: Patient example for progressive disease.The baseline MRI (2. baseline) was performed 2 weeks before the start of radiotherapy.This patient developed preliminary progressive disease on the first MRI after radiotherapy 11 weeks later (2.baseline + 11 weeks).The 2 follow-up MRIs show further progression of the lesion at weeks 14 and 27 (2.baseline + 14 and 2. baseline + 27).This patient was part of the control group without cilengitide.

Figure 4 .Figure 2 .
Figure 2. Flowchart of the inclusion process and outcome categories.

FliesFigure 3 .
Figure 3. Outcome after preliminary progressive disease, depicted per time period and per treatment group (cilengitide or control arm).Time was measured from the end of radiotherapy (RT) to preliminary progressive disease.The group "(Unconfirmed) Progressive Disease" includes the outcome groups with unconfirmed progressive disease (who were mostly considered clinically to have progressive disease), and those with progressive disease according to modified RANO.

Figure 4 .
Figure 4. Kaplan-Meier curve to compare the survival of patients with treatment-associated changes and those with (unconfirmed) progressive disease.The P-value of the log-rank test was .003.

Table 1 .
Summary of the 2010 RANO and 2017 Modified RANO Criteria >50% decrease in SPD followed by SD or further decrease (PR) Sustained complete disappearance of all enhancing lesions (CR)