Imaging timing after surgery for glioblastoma: an evaluation of practice in Great Britain and Ireland (INTERVAL-GB)- a multi-centre, cohort study

This was a multicenter, retrospective cohort study of patients with newly-diagnosed glioblastoma undergoing surgery and oncological therapy. The study was conducted in 26 neuro-oncology centers in the UK and Ireland. The study protocol has previously been published and provides a detailed overview of the study design [16‐18]. The analysis plan was changed once after protocol publication; changing the definition of ‘compliance’ to imaging guidelines from a binary variable (Compliant and noncompliant) to a continuous variable (time spent compliant), to acknowledge that many patients spent periods of time as compliant, and noncompliant, during the follow-up period. Local audit and Caldicott guardian approval was obtained at each unit before data collection could commence.

Data was captured consecutively on adult patients (aged ≥ 18 years) with a new histopathological diagnosis of glioblastoma (according to the diagnostic criteria at the time of diagnosis – the 2016 World Health Organisation [WHO] Classification) [19], who underwent surgery and any active oncological treatment between August 31, 2018, and February 1, 2019. Data collection took place between November 11, 2021, and May 22, 2022. Collaborators identified eligible patients by searching historic Multidisciplinary Team Meetings (MDTMs), histopathological and/or surgical records locally against the inclusion and exclusion criteria (Supplementary Table 1).

We used the Castor (Castor, NY, USA) online database to securely collect and store data. Data was collected by local investigators from a combination of the patient’s clinical, radiological and histopathological records. The variable domains consisted of baseline clinical and radiological variables, surgical and histopathological variables, adjuvant and supportive treatment details, follow-up MRI details and outcome measures. Follow-up MRI outcomes were determined by the overall scan report reported by a neuroradiologist at each participating center- and classified into three groups: stable, recurrence/further growth, and pseudoprogression. To be entered, reports had to be confirmed by MDTM agreement. The minimum data requirement for a case to be included was the baseline characteristics, surgery date, and post-operative follow-up details.

The primary outcome was compliance with follow-up MRI surveillance schedules defined in the latest NICE and EANO guidelines (see Supplementary Table 2 for definitions of compliance) [5, 6]. In summary, the NICE guidelines recommend imaging every three to six months post-surgery for two years, and EANO recommend a scheduled scan every three months for two years. If imaging was not completed prior to these dates, a patient was defined as being non-compliant, until the next scan was completed (correlating with time spent compliant to imaging recommendations). The secondary outcomes were OS and PFS. OS was defined as the date of surgery to date of death from any cause. PFS was defined as the time from date of surgery until MRI evidence of tumor recurrence, validated by MDTM agreement. Additional secondary outcomes included initiation of second-line chemotherapy, and re-intervention (repeat surgery or repeat radiotherapy). Follow-up was defined as until date of last clinic appointment, scan (MRI/CT), or date of death. Patients were censored from PFS if they progressed, or death occurred.

The required sample size was derived following a three-center pilot study of 123 patients. The null hypothesis of the study was that more than half of patients (> 50%) will not be scanned in accordance with NICE Guidelines, and that groups scanned according to these guidelines will demonstrate improved OS (superiority assumption), and PFS. This identified a 34% proportion of patients compliant with NICE guidelines, and a median survival of 9.6 months. A hazard ratio of 1.35 was used to set the minimal clinically important difference for the study, which roughly equates to 3 months of OS benefit (similar survival benefit to Temozolomide in Stupp trial). Assuming a mortality rate of 85% of patients during the 24-month follow-up period, to achieve 80% power, with a 5% type 1 error rate, the sample size using a 2 sample, means superiority calculation was 456 patients (approximately 22 Neuro-Oncology units, assuming an average patient list of 20 patients per center).

Data analysis was carried out using R version 4.0.2 using the gpplot2, survminer, survival, and forcats, packages [20‐22]. Compliance to NICE and EANO guidelines was summarised using descriptive statistics. This consisted of 3 groups- a group that had every follow-up scan within the recommended time period (Fully compliant); a group that had mixed periods of compliance with imaging, and mixed periods of non-compliance during follow-up (Inbetween); and a group that had no imaging carried out within the recommended timeframe (Never compliant). Fully compliant was further defined as: every scheduled follow-up scan occurring in equal or more frequently than the imaging recommendations (i.e. a patient having three scans, all of which occurred two months apart would be fully compliant). Inbetween was defined as at least one period of time where imaging occurred less frequently than recommended, alongside at least one period of recommended imaging (i.e. a patient who had three follow-up scans, the first two being three-six months apart, followed by a third scan nine months after last scan. Never compliant was defined as having no scheduled imaging within the recommended period (i.e. a patient who had two follow-up scans, both 9 months apart). In the event that a patient had no post-operative imaging, but had survived long enough to have imaging, they were defined as never compliant.

Unscheduled scans were not considered in the compliance definition- if an unscheduled scan occurred, the patient would remain compliant, and the time window for the next scan re-set, as if they had a scheduled scan (i.e. would have 3–6 months to have a further scheduled scan before being considered non-compliant). Continuous variables were analysed using mean (standard deviation [SD]), or median (interquartile range [IQR]) [23]. Differences in characteristics of compliant and non-compliant groups were presented with descriptive statistics. PFS and OS was estimated using Kaplan–Meier survival method. To identify if imaging frequency was associated with increased survival, and to account for any differences in characteristics associated with survival benefit, we conducted multivariable, Cox regression analysis, incorporating compliance as a variable alongside variables known to affect glioblastoma outcomes (age, performance status, extent of resection [EOR], and Stupp protocol treatments) [24‐29]. As compliance is not a binary variable, and can change over time (i.e. a patient who is initially compliant with imaging guidelines for the first 12 months of follow-up before becoming non-compliant), compliance was classified as a time varying covariate, with periods of compliance a separate component [30]. This accounts for each individual period spent compliant and noncompliant to imaging, to account for unequal or differential periods of follow-up, minimising bias (i.e. a patient who has spent 20 months compliant and two months non-compliant would be reflected proportionally in analysis). Two separate cox regression models were completed, with compliance according to guideline (NICE and EANO) included separately as a variable in each model (to avoid collinearity). We used Hazard Ratios (HR) and a 95% confidence interval to measure effect size. For each analysis conducted, the number of patients with sufficient data entry was used, giving varying numbers for each analysis point. As biopsy groups have a different scanning recommendation, we carried out additional sensitivity analysis by removing the group, then repeating the analysis.

There were 818 patients identified from 26 out of 32 neuro-oncology centers in the UK and Ireland, of which 754 met the minimum data requirement (Table 1). The median number of patients per center was 33 (IQR 17–42) (Supplementary Table 3). Surgical and adjuvant treatments are summarised in Table 1, and detailed in Supplementary Table 4. The median follow-up time after surgery was 10.5 months (IQR 5.3–19.4 months). The total follow-up time was 10,100 (median, 12.5 months/person, IQR 5.2–19.4) person-months of follow-up.

The post-surgical imaging schedules are summarised in Supplementary Table 5 and 6. In total, 522 (69.2%) patients had a post-operative MRI scan within 7 days of surgery. Of these, 462 (88.5%) were a post-operative scan to assess extent of resection, and 60 (11.5%) were for radiotherapy planning without an extent of resection scan. An MRI scan within the 72-h recommended NICE guidelines after resection (excluding biopsy) was seen in 78.0% (N = 407/522), and within the 48-h EANO guidelines in 64.2% (N = 335/522). The follow-up MRI scan timings after radiotherapy to assess for progression are presented in Fig. 1, and Supplementary Tables 5 and 6. At every follow-up scan, 40–50% showed disease progression. Pseudoprogression was seen in 8.3% of scans (median, IQR 7.8%-9.8%).

In total, 398 (52.8%) of patients had imaging intervals that were fully compliant with NICE guidelines, 70 (9.3%) had intervals that moved between recommended and noncompliant scanning, and 286 (37.9%) were never compliant. In total, for the EANO guidelines, 188 (24.9%) of patients were fully compliant; 217 (28.8%) moved between recommended and noncompliant scanning, and 349 (46.3%) were never compliant. Compliance differed depending on EOR, and Stupp protocol completion (Fig. 2, Supplementary Table 7).

In total, 684 (90.7%) progressed during the study period. The median PFS was 6.7 months (95% CI 6.2–7.3 months) (Supplementary Fig. 1). 641 patients (85.0%) died during the study period. The median OS using the KM estimate method was 11.4 months (95% CI 10.4–12.4 months), with expected differences depending on EOR, and STUPP protocol adherence (Supplementary Fig. 1). In total, of those that progressed radiologically, 305 (79.0%) had first progression detected using scheduled MRI, with 86 (21.0%) having progression first detected on unscheduled MRI. Of those who died, 305 (47.6%) had progression on MRI before death, with 336 (52.4%) without MRI confirmation of progression. Treatments for progression are shown in Supplementary Table 4. Patients with full and in between compliance were more likely to receive second-line chemotherapy, and re-operation (Supplementary Table 8). Survival differed between the three compliance groups (Fig. 3).

The multivariable cox regression analysis results are shown in Table 2. On multivariable analysis including age, performance status, EOR, and completion of STUPP protocol, compliance with NICE recommendations (time varying covariate) was associated with increased OS (HR 0.56, 95% CI 0.46–0.66, P < 0.001). Compliance with EANO recommendations was also associated with increased OS (HR 0.54, 95% CI 0.45–0.63, P < 0.001). Compliance with NICE recommendations was not associated with increased PFS (HR 0.93, 95% CI 0.79–1.10, P = 0.349). Compliance with EANO recommendation was not associated with increased PFS (HR 0.99, 95% CI 0.84–1.16, P = 0.874). After repeated subgroup multivariable analysis excluding the biopsy group, provided similar results (Supplementary Table 9).