MD Anderson Research Highlights: ASTRO 2021 Special Edition
Key presentations focused on novel radiation therapy regimen, predictive biomarkers for immunotherapy and AI-based modeling
MD Anderson News Release October 25, 2021
The University of Texas MD Anderson Cancer Center’s Research Highlights provides a glimpse into recent studies in basic, translational and clinical cancer research from MD Anderson experts. This special edition features oral presentations by MD Anderson researchers at the 2021 American Society for Radiation Oncology (ASTRO) Annual Meeting (Oct. 24-27) on novel therapeutic and diagnostic approaches, including partial breast irradiation, evaluating PD-L1 levels as biomarkers to better predict response to immunotherapy, and deep learning and biomechanical models.
Partial breast irradiation is effective with better cosmetic outcomes and less toxicity compared to whole breast irradiation (Abstract 1026)
For patients with breast cancer, twice-daily accelerated partial breast irradiation (APBI) may lead to poorer long-term cosmetic outcomes — such as volume loss, indentation at the site of the lumpectomy, asymmetries and deformation — compared with daily whole breast irradiation (WBI). However, Benjamin Smith, M.D., Jay Reddy, M.D., Ph.D., and a team of researchers hypothesized a once-daily regimen of APBI would yield improved cosmetic results and long-term quality of life by reducing side effects without compromising tumor control.
In a multi-institutional, phase II trial, researchers enrolled 149 women with ductal carcinoma in situ (DCIS) or early invasive, estrogen receptor-positive breast cancer and treated them with the Optimizing Preventative Adjuvant Linac-based (OPAL) radiation regimen, a once-daily hypofractionated radiation therapy that delivers higher doses of radiation therapy over a shorter period of time, to evaluate cosmesis, local control, breast pain and toxicity. The researchers compared these patients to 176 matched patients from their previous trial that evaluated patients who received WBI and additional boost dose.
Patients on the OPAL regimen experienced less toxicity and treatment-related adverse events compared to the WBI cohort. Within six months of starting the regimen, 14.1% of patients experienced grade 2 or greater toxicity compared to 71% in the WBI cohort. At two years after radiation, cosmetic outcomes also were significantly better in patients receiving the OPAL regimen — 93% of patients in the OPAL cohort reported excellent or good cosmetic outcome compared to 77% of WBI patients. Overall, the results of the study showed the OPAL regimen was associated with improved patient-reported cosmetic outcomes, functional status, breast pain and physician-reported cosmetic outcomes.
Tracking PD-L1 expression on circulating stromal cells throughout chemoradiation predicts survival outcomes for unresectable stage 3 NSCLC (Abstract 20)
Chemoradiation followed by immunotherapy is the current standard of care to treat patients with locally advanced, unresectable non-small cell lung cancer (NSCLC). While this strategy has demonstrated a significant impact on patient survival and cure rates, only a small percentage of patients benefit. To understand which patients would benefit from combination therapy, Steven Lin, M.D., Ph.D., and a team of researchers investigated whether tracking the expression of immune checkpoint protein PD-L1 on cancer associated macrophage-like cells (CAMLs), a type of circulating stromal cell found specifically in the bloodstream of numerous solid-tumor malignancies, throughout chemoradiation could serve as a predictive biomarker for patient responses.
To evaluate correlations between PD-L1 expression and progression free survival (PFS) or overall survival (OS), researchers compared three groups of patients with stage 3 unresectable NSCLC: those treated with chemoradiation alone as the control, chemoradiation and immunotherapy drug atezolizumab, and chemoradiation and immunotherapy drug durvalumab.
PD-L1 levels in pre-treatment blood and tumor samples were not predictive of outcomes in any patient cohort. Similarly, PD-L1 levels from blood samples taken after chemoradiation were not predictive for PFS or OS in the group of patients who received chemoradiation alone.
However, in patients who received atezolizumab or durvalumab, high PD-L1 expression on CAMLs was predictive of better PFS and OS compared to patients with low PD-L1. Chemoradiation plus atezolizumab achieved a PFS hazard ratio, which estimates the survival of a treated population compared to the control group, of 2.8 and an OS hazard ratio of 4.3. Chemoradiation plus durvalumab achieved a PFS hazard ratio of 5.2 and an OS hazard ratio of 5.9. Overall, their data showed that high PD-L1 levels on CAMLs after chemoradiation were statistically significant in predicting superior clinical outcomes for patients receiving anti-PD-L1 immunotherapies.
Deep learning-based segmentation and biomechanical model-based dose accumulation improve accuracy for radiation therapy (Abstract 86)
Deformations of the stomach and duodenum can occur over the course of radiation therapy to treat liver cancer. These deformations can lead to discrepancies between the planned radiation dose and the actual delivered dose. To track the accumulation of dose within the body and to estimate the impact of discrepancies, radiation oncologists use deformable image registration (DIR) — which creates a composite of two or more diagnostic images to identify anatomical and morphological changes in organ shape, size and position. However, this approach can be challenging for traditional intensity-based registration algorithms due to low contrasts in the images. A research team led by Kristy Brock, Ph.D., used deep learning segmentation and biomechanical models of the liver, stomach and duodenum to improve the accuracy of dose accumulation and the understanding of the dose-toxicity relationship.
Retrospective dose accumulation was conducted on 75 patients with primary and metastatic liver cancer treated with external beam radiotherapy guided by daily CT-on-rails (CTOR). The study compared the physician-drawn contours on CTOR to the AI-based contours. The results — presented by Guillaume Cazoulat, Ph.D. — showed that larger differences were estimated between planned and delivered doses using the contours-based DIR compared to the traditional intensity-based method. With the traditional approach, the normal tissue complication probability (NTCP) calculated for the duodenum found that 25% of patients had a greater than 5% difference between planned and accumulated dosage. The biomechanical model-based DIR found this degree of difference in 38% of patients, suggesting it is a more sensitive approach to estimate discrepancies.
Researchers plan to apply the fully automatic workflow method to a larger cohort of patients to help them better understand the relationship between delivered dose and toxicity outcomes in an effort to develop new adaptive treatment strategies and minimize radiation-induced toxicities in the gastrointestinal organs.