Research Article: Computed tomography based radiomic signature as predictive of survival and local control after stereotactic body radiation therapy in pancreatic carcinoma

Date Published: January 18, 2019

Publisher: Public Library of Science

Author(s): Luca Cozzi, Tiziana Comito, Antonella Fogliata, Ciro Franzese, Davide Franceschini, Cristiana Bonifacio, Angelo Tozzi, Lucia Di Brina, Elena Clerici, Stefano Tomatis, Giacomo Reggiori, Francesca Lobefalo, Antonella Stravato, Pietro Mancosu, Alessandro Zerbi, Martina Sollini, Margarita Kirienko, Arturo Chiti, Marta Scorsetti, Qinghui Zhang.

http://doi.org/10.1371/journal.pone.0210758

Abstract

To appraise the ability of a radiomics signature to predict clinical outcome after stereotactic body radiation therapy (SBRT) for pancreas carcinoma.

A cohort of 100 patients was included in this retrospective, single institution analysis. Radiomics texture features were extracted from computed tomography (CT) images obtained for the clinical target volume. The cohort of patients was randomly divided into two separate groups for the training (60 patients) and validation (40 patients). Cox regression models were built to predict overall survival and local control. The significant predictors at univariate analysis were included in a multivariate model. The quality of the models was appraised by means of area under the curve and concordance index.

A clinical-radiomic signature associated with Overall Survival (OS) was found significant in both training and validation sets (p = 0.01 and 0.05 and concordance index 0.73 and 0.75 respectively). Similarly, a signature was found for Local Control (LC) with p = 0.007 and 0.004 and concordance index 0.69 and 0.75. In the low risk group, the median OS and LC in the validation group were 14.4 and 28.6 months while in the high-risk group were 9.0 and 17.5 months respectively.

A CT based radiomic signature was identified which correlate with OS and LC after SBRT and allowed to identify low and high-risk groups of patients.

Partial Text

Patients affected by pancreatic adenocarcinoma have a typically un-favourable prognosis, with a 5-year overall survival (OS) rate as low as 6% [1]. Surgery, as a treatment of choice, leads to 5-year OS rates of about 20 to 25%. Nevertheless, a large fraction of the total patients, are unfit to surgery already at diagnosis (due to the stage of the disease or other concomitant exclusion criteria) [2]. For this reason, chemo-radiotherapy is a frequently adopted solution for many patients but with the drawback of some significant rate of severe toxicity (grade 3 to 4) and a still low rate of survival. The median OS is in the range of 5 to 15 months and the 2-year OS is about 30% [3]. The role of stereotactic body radiotherapy (SBRT) was investigated [4], but only few trials were published with reports about significant late toxicity rates [5–11]. Our institutional experience [12] demonstrated that SBRT with a fractionation of 45Gy in 6 sessions is an effective and safe therapeutic option for non-operable as well as for isolated local recurrences with a median OS of 13 months for the non-operable cases. 1 and 2 year OS were 59±7% and 18±9% respectively. In the same study freedom from local progression (FFLP) was 87±6% at 2 years [12]. The comparative assessment of the clinical use of SBRT in the management of pancreatic cancer demonstrated (Table 5 in that original publication), with a quite variable range of fractionation regimens, very consistent findings in terms of FFLP and median OS times ranging from 8 to 20 months, again consistent with our findings [12].

The main characteristics of the cohort of 100 patients are summarized in Table 1. The mean volume of the target was 24.59±17.6 cm3 (range: 2.8–109.5 cm3). Concerning the SBRT treatment, all patients respected the planning aims for target coverage. Table 2 summarizes the data for the training and validation sets.

This study aimed to appraise the correlation between some radiomic signatures and the clinical outcome in a retrospective analysis of 100 patients with pancreas cancer treated according the institutional protocols. As noticed, textural analysis has been rarely applied to pancreatic cancer and almost no efforts, so far, were put in the radiomic assessment of outcome data [17–20] and none in association to SBRT. In this study, planning CT scans were used as a basis of the analysis to derive the textural features which demonstrated the possibility to be modelled vs OS and LC. The methodology adopted in the present study is derived from similar other investigations [27–28] and aims to simplicity. The panel of features available for testing did not included higher order texture (e.g. wavelets). Our hypothesis was that a radiomics signature, if existing, should be found within the set of most robust and easy to compute classes of features. As discussed in [20], the use of conventional planning CT and the clinical target volumes (with minor pre-processing edits) is another factor of simplicity which can guarantee a straightforward implementation of the radiomic methodology in the clinical practice. Although final scope of a “predictive” tool is the classification of patients in risk groups (or any similar stratification) prior to therapy, in practice, the models has to be trained and tuned retrospectively on cohorts of patients where the outcome of treatments is known. In this perspective, our investigation covers the first elements of the entire chain, i.e. the determination of a (potentially) useful signature, in a retrospective investigation on a cohort of patients completely treated (with radiotherapy and/or chemotherapy). Further prospective studies shall be designed to validate the predictive models on un-treated cohort of patients to measure their reliability and performance. Another limiting factor in the study was the exclusion, from the predictive model, of any clinical factor which might have influenced the outcome. In practice, all patients received the same SBRT treatment but some different chemotherapy regimen. The inclusion of clinical factors might strengthen the predictive value of the predictive models but, in the present study, the aim was to identify, if existing, a purely radiomic feature. Further studies will be devoted to the refinement of the models and the elimination (or explicit inclusion) of all the valuable factors. Concerning chemotherapy, 55% of the patients received it prior to SBRT. These pre-treated patients were evaluated after the end of chemotherapy with Torax-abdominal CT scan with the evidence of partial response (56%), stable disease (31%) or local progression limited to pancreatic cancer (13%). No patient showed distant progression of disease, therefore all patients were eligible for SBRT.

A radiomics signature made of simple clinical and textural features allowed to generate a predictive model for OS and LC in patients affected by advanced pancreatic cancer treated with SBRT. A fair discrimination power was found applying the model to training and validatin samples. Further validation studies would be advisable to confirm these findings.

 

Source:

http://doi.org/10.1371/journal.pone.0210758

 

Leave a Reply

Your email address will not be published.