5MO - Prospective validation of a radiomics signature for chemoradiotherapy lung cancer patients

Background

Radiomics refers to the comprehensive quantification of tumour phenotypes and is a promising field of scientific study with a large amount of activity in recent times. However, to date no clinical level of evidence 1 has been provided for any of the many radiomics signatures published in the literature. The purpose of this study is to provide that evidence by prospectively validating a prognostic radiomics signature for chemoradiotherapy lung cancer patients [1].

Methods

A total of 228 chemoradiotherapy lung cancer patients were collected from the observational SDC lung clinical trial (NCT01855191). Tumour volumes within the CT scans of these patients were segmented automatically using a deep learning model (validated DICE agreement with oncologists of 0.88) and used as input to the radiomics signature. The primary outcome was overall survival. The signature was used to classify patients as responders or non-responders (high/low score based on of the threshold proposed by Aerts et al.). Predefined statistical tests were performed to prospectively validate the performance of the model.

Results

Discrimination of the model was assessed by Harrell’s concordance index (c-index = 0.64; 95% CI: 0.59-0.70). Kaplan-Meier survival curves differed significantly between responders/non-responders (log-rank test; p = 0.002). The calibration slope (β) on the linear predictor of the signature in a Cox proportional hazards model was 1.27 (H0: β = 1, p = 0.37), indicating a valid relative risk model. A joint test of the coefficients of the individual variables of the signature (p = 0.028) indicated that the performance in the prospective validation cohort could be improved by adjusting the model coefficients.

Conclusions

To the best of our knowledge this study demonstrates the first clinical evidence level 1 for any radiomics signature. This has implications for the wider field as it demonstrates that other signatures could also be prospectively validated. This signature could be practically used as a clinical decision support tool to evaluate the likelihood of response to chemoradiotherapy, or as a stratification tool in future clinical trials. [1] Aerts et al. Nature Communications. 2014;5:4006.

Clinical trial identification

NCT01855191.

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

R. Leijenaar: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment: Oncoradiomics. F. Zerka: Full/Part-time employment: Oncoradiomics. A. Vaidyanathan: Full/Part-time employment: Oncoradiomics. B. Miraglio: Full/Part-time employment: Oncoradiomics. N. Tsoutzidis: Full/Part-time employment: Oncoradiomics. W. Vos: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment, Officer/Board of Directors: Oncoradiomics. S. Walsh: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment, Officer/Board of Directors: Oncoradiomics. P. Lambin: Advisory/Consultancy, Shareholder/Stockholder/Stock options, Officer/Board of Directors: Oncoradiomics.