Glioblastoma and Radiotherapy: a multi-center AI study for Survival Predictions from MRI (GRASP study)

BACKGROUND

The aim was to predict survival of glioblastoma at eight months after radiotherapy (a period allowing for completing a typical course of adjuvant temozolomide), by applying deep learning to the first brain MRI after radiotherapy completion.

METHODS

Retrospective and prospective data were collected from 206 consecutive glioblastoma, IDH-wildtype patients diagnosed between March 2014-February 2022 across 11 UK centers. Models were trained on 158 retrospective patients from three centers. Holdout test sets were retrospective (n=19; internal validation), and prospective (n=29; external validation from eight distinct centers).Neural network branches for T2-weighted and contrast-enhanced T1-weighted inputs were concatenated to predict survival. A non-imaging branch (demographics/MGMT/treatment data) was also combined with the imaging model. We investigated the influence of individual MR sequences; non-imaging features; and weighted dense blocks pretrained for abnormality detection.

RESULTS

The imaging model outperformed the non-imaging model in all test sets (area under the receiver-operating characteristic curve, AUC p=0.038) and performed similarly to a combined imaging/non-imaging model (p>0.05). Imaging, non-imaging, and combined models applied to amalgamated test sets gave AUCs of 0.93, 0.79, and 0.91. Initializing the imaging model with pretrained weights from 10,000s of brain MRIs improved performance considerably (amalgamated test sets without pretraining 0.64; p=0.003).

CONCLUSIONS

A deep learning model using MRI images after radiotherapy, reliably and accurately determined survival of glioblastoma. The model serves as a prognostic biomarker identifying patients who will not survive beyond a typical course of adjuvant temozolomide, thereby stratifying patients into those who might require early second-line or clinical trial treatment.

Factors associated with survival in small cell lung cancer: an analysis of real-world national audit, chemotherapy and radiotherapy data

BACKGROUND

The mainstay of treatment for small cell lung cancer (SCLC) involves platinum doublet chemotherapy but the optimal duration, 4 vs. 6 cycles, is not known. Concurrent thoracic radiotherapy followed by prophylactic cranial irradiation (PCI) is recommended for fit individuals with limited stage. However, outside of clinical trials, the efficacy of sequential thoracic radiotherapy and PCI for extensive stage is uncertain.

METHODS

This retrospective, observational, cohort study used English national lung cancer data to determine the factors associated with survival for all people diagnosed with SCLC. More precisely, for individuals who received chemotherapy, we examined survival by the chemotherapy duration, thoracic radiotherapy dose and the use of PCI.

RESULTS

In total 6,438 people were diagnosed with SCLC. We identified that male sex (OR 0.7; 95% CI: 0.62-0.80), increasing age (P=0.01) greater comorbidity (P≤0.01), extensive stage (OR 0.21; 95% CI: 0.19-0.25) and worse performance status (PS2 vs. PS0 adjusted OR 0.38 95% CI: 0.31-0.48) were associated with reduced 1-year survival. Receipt of chemotherapy augmented survival. We analysed data for 1,761 people who had received chemotherapy. Thoracic radiotherapy (≥30 Gy for extensive stage and ≥40 Gy for limited stage) and PCI were independently associated with better survival (P≤0.01 for each), but 6 cycles of chemotherapy instead of 4 was not (limited stage adjusted OR 0.97; 95% CI: 0.48-1.97) extensive stage adjusted OR 1.34; 95% CI: 0.81-2.21).

CONCLUSIONS

Extending chemotherapy beyond 4 cycles to 6 does not augment survival. Appropriately prescribed thoracic radiotherapy and PCI can prolong survival in both limited and extensive stage SCLC.

Continuous Hyperfractionated Accelerated Radiotherapy (CHART) for Non-small Cell Lung Cancer (NSCLC): 7 Years' Experience From Nine UK Centres

AIM

Continuous hyperfractionated accelerated radiotherapy (CHART) remains an option to treat non-small cell lung cancer (NSCLC; NICE, 2011). We have previously published treatment outcomes from 1998-2003 across five UK centres. Here we update the UK CHART experience, reporting outcomes and toxicities for patients treated between 2003 and 2009.

MATERIALS AND METHODS

UK CHART centres were invited to participate in a retrospective data analysis of NSCLC patients treated with CHART from 2003 to 2009. Nine (of 14) centres were able to submit their data into a standard database. The Kaplan-Meier method estimated survival and the Log-rank test analysed the significance.

RESULTS

In total, 849 patients had CHART treatment, with a median age of 71 years (range 31-91), 534 (63%) were men, 55% had undergone positron emission tomography-computed tomography (PET-CT) and 26% had prior chemotherapy; 839 (99%) patients received all the prescribed treatment. The median overall survival was 22 months with 2 and 3 year survival of 47% and 32%, respectively. Statistically significant differences in survival were noted for stage IA versus IB (33.2 months versus 25 months; P = 0.032) and IIIA versus IIIB (20 months versus 16 months; P = 0.018). Response at 3 months and outcomes were significantly linked; complete response showing survival of 34 months against 19 months, 15 months and 8 months for partial response, stable and progressive disease, respectively (P < 0.001). Age, gender, performance status, prior chemotherapy and PET-CT did not affect the survival outcomes. Treatment was well tolerated with <5% reporting ≥grade 3 toxicity.

CONCLUSION

In routine practice, CHART results for NSCLC remain encouraging and we have been able to show an improvement in survival compared with the original trial cohort. We have confirmed that CHART remains deliverable with low toxicity rates and we are taking a dose-escalated CHART regimen forward in a randomised phase II study of sequential chemoradiotherapy against other accelerated dose-escalated schedules.

The late radiotherapy normal tissue injury phenotypes of telangiectasia, fibrosis and atrophy in breast cancer patients have distinct genotype-dependent causes

The relationship between late normal tissue radiation injury phenotypes in 167 breast cancer patients treated with radiotherapy and: (i) radiotherapy dose (boost); (ii) an early acute radiation reaction and (iii) genetic background was examined. Patients were genotyped at single nucleotide polymorphisms (SNPs) in eight candidate genes. An early acute reaction to radiation and/or the inheritance of the transforming growth factor-β1 (TGFβ1 −509T) SNP contributed to the risk of fibrosis. In contrast, an additional 15 Gy electron boost and/or the inheritance of X-ray repair cross-complementing 1 (XRCC1) (R399Q) SNP contributed to the risk of telangiectasia. Although fibrosis, telangiectasia and atrophy, all contribute to late radiation injury, the data suggest that they have distinct underlying genetic and radiobiological causes. Fibrosis risk is associated with an inflammatory response (an acute reaction and/or TGFβ1), whereas telangiectasia is associated with vascular endothelial cell damage (boost and/or XRCC1). Atrophy is associated with an acute response, but the genetic predisposing factors that determine the risk of an acute response or atrophy have yet to be identified. A combined analysis of two UK breast cancer patient studies shows that 8% of patients are homozygous (TT) for the TGFβ1 (C-509T) variant allele and have a 15-fold increased risk of fibrosis following radiotherapy (95% confidence interval: 3.76–60.3; P=0.000003) compared with (CC) homozygotes.