Virtual Library
Start Your Search
Pranjal Vaidya
Author of
-
+
P2.04 - Immunooncology (Not CME Accredited Session) (ID 953)
- Event: WCLC 2018
- Type: Poster Viewing in the Exhibit Hall
- Track:
- Presentations: 1
- Moderators:
- Coordinates: 9/25/2018, 16:45 - 18:00, Exhibit Hall
-
+
P2.04-17 - Pre-Therapy Radiomic Features Can Distinguish Hyperprogression from Other Response Patterns to PD1/PD-L1 Inhibitors in NSCLC (ID 13642)
16:45 - 18:00 | Presenting Author(s): Pranjal Vaidya
- Abstract
Background
Immune checkpoint inhibitors (ICIs) can lead to durable responses in a fraction of patients with advanced non-small cell lung cancer (NSCLC), however a majority of patients do not respond to these agents. Of the non-responders, a subset of patients who have a dramatic increase in their tumor growth rates after ICI therapy have been previously described in the literature. There are currently no clinically validated biomarkers to identify these hyperprogressors (HPs). We sought to evaluate whether radiomic features of the tumor on baseline CT scans from patients with advanced NSCLC treated with ICIs could distinguish hyperprogressors from non-responders (NRs) and responders (Rs).
a9ded1e5ce5d75814730bb4caaf49419 Method
We retrospectively reviewed the charts of 336 patients with advanced NSCLC who received monotherapy with a PD1/PD-L1 inhibitor. For patients who developed progressive disease within 3 cycles of ICI therapy, pre-baseline, baseline and post treatment scans were used to calculate tumor growth kinetics (TGK). TGK pre and post ICI was calculated by dividing the sum of the largest diameters of target lesions per RECIST criteria by the interval time between scans. The ratio of post- treatment TGK and pre-treatment TGK was used to identify hyperprogressors (ratio ≥2, N=28). Intratumoral and peritumoral radiomic features using annular 2 mm rings up to 10 mm from the center of the tumor on the baseline CT were extracted for hyperprogressors (N=28), responders (N=28) and non-responders (N=29). In the training cohorts, a total of 925 features that were differentially expressed in hyperprogressors vs responders (N=28) and hyperprogressors vs non-responders (N=28) were investigated.
4c3880bb027f159e801041b1021e88e8 Result
Top 5 predictive radiomic features (from the Haralick, Laws and Gabor texture families) were identified using a minimum redundancy maximum relevance (mRMR) feature selection algorithm and their performance was validated using an independent test set of hyperprogressors vs responders (N=28) and hyperprogressors vs non-responders (N=29). Linear Discriminant Analysis (LDA) classifier was able to distinguish hyperprogressors vs non-responders with an area under the receiver operating curve (AUC) of 0.85 and a sensitivity of 0.92. The same classifier separated hyperprogressors and responders with an AUC and sensitivity of 0.58 and 0.85 respectively.
8eea62084ca7e541d918e823422bd82e Conclusion
Intratumoral and peritumoral radiomic textural features on baseline CT scans were able to distinguish hyperprogressors from those with other patterns of response to ICIs, particularly other non-responders. These radiomic features may serve as a predictive marker for patients who develop accelerated disease progression with ICIs. Further validation of these results in multi institutional cohorts is warranted.
6f8b794f3246b0c1e1780bb4d4d5dc53
-
+
P3.16 - Treatment of Early Stage/Localized Disease (Not CME Accredited Session) (ID 982)
- Event: WCLC 2018
- Type: Poster Viewing in the Exhibit Hall
- Track:
- Presentations: 1
- Moderators:
- Coordinates: 9/26/2018, 12:00 - 13:30, Exhibit Hall
-
+
P3.16-10 - Radiomic Features on CT are Prognostic of Recurrence as well as Predictive of Added Benefit of Adjuvant Chemotherapy in ES-NSCLC (ID 14270)
12:00 - 13:30 | Presenting Author(s): Pranjal Vaidya
- Abstract
Background
Early-Stage non-small cell lung cancer (ES-NSCLC) accounts for approximately 40% of NSCLC cases, with 5-year survival rates varying between 31-49%. The decision to offer adjuvant chemotherapy for these patients is primarily dependent on several clinical and visual radiographic factors as there is a lack of biomarkers which can accurately stratify and predict disease risk
a9ded1e5ce5d75814730bb4caaf49419 Method
Retrospective chart review between 2005-14 yielded 315 ES-NSCLC patients who underwent surgery with the primary tumor having relapsed in 75 cases. From the entire cohort, 74 underwent adjuvant chemotherapy. This cohort was randomly divided into a training(N=60) and validation(N=255). A total of 248 intratumoral(IT) and peritumoral(PT) radiomic textural features were extracted for every patient. The most stable, significant and uncorrelated features were selected from training cohort using LASSO Cox-regression model. Performance of imaging features was evaluated using hazard ratio(HR) and concordance index(CI). Linear Discriminant Classifier(LDA) was trained using top imaging features and performance of predicted labels was assessed using Kaplan-Meier survival curves and log-rank test.
4c3880bb027f159e801041b1021e88e8 Result
Top nine radiomic textural features (from the Haralick, Collage, Laws, Gabor texture families) included a combination of four IT and five PT from 0-12mm distance outside the nodule. The features were prognostic of recurrence (N=255, CI=0.66, HR =1.8, p<0.05). To evaluate the predictive model, subset analysis was performed on the test set. The imaging feature based classifier was able to identify low and high risk groups in the surgery alone setting (N=181, CI=0.73, HR=4.4, p<0.005), potentially identifying patients who might have benefitted from adjuvant chemotherapy. Meanwhile, in the group of patients who received adjuvant chemotherapy following surgery, the classifier did not identify any difference between high and low risk groups (N=74, CI=0.69, HR=1, p>0.05).
8eea62084ca7e541d918e823422bd82e Conclusion
We identified radiomic features from within and outside lung nodule that were prognostic of recurrence and also predictive of added benefit of adjuvant chemotherapy in ES-NSCLC.
6f8b794f3246b0c1e1780bb4d4d5dc53
