In patients with stage I lung cancer, adding noncancerous features from CT chest imaging predicts overall survival better than clinical characteristics alone, according to a paperin the American Journal of Roentgenology.
Modeling that incorporates noncancerous imaging features captured on chest computed tomography (CT) along with clinical features, when calculated before stereotactic body(SBRT) is administered, improves survival prediction, compared with modeling that relies only on clinical features, the authors report.
“The focus of the study was to look at the environment in which the cancer lives,” said senior author Florian J. Fintelmann, MD, radiologist at Massachusetts General Hospital and associate professor of radiology at Harvard Medical School, both in Boston. “This is looking at parameters like the aortic diameter, body composition – that is, the quantification and characterization of adipose tissue and muscle – coronary artery calcifications, andquantification.”
CT images are used by radiation oncologists to determine where the radiation should be delivered. “There is more information from these images that we can utilize,” he said.
Survival estimates in patients with state I lung cancer now rely on biological age, ECOG (Eastern Cooperative Oncology Group) score, and the presence of comorbidities, Dr. Fintelmann said.
This retrospective investigation involved 282 patients with a median age of 75 years. There were 168 women and 114 men. All patients had stage I lung cancer and were treated with SBRT between January 2009 and June 2017.
Investigators analyzed pre-treatment chest images with CT. They assessed(CAC) score (see above image), pulmonary artery (PA)-to-aorta ratio, emphysema, and several measures of body composition (skeletal muscle and adipose tissue). They developed a statistical model to link clinical and imaging features with overall survival.
An elevated CAC score (11-399: HR, 1.83 [95% confidence interval, 1.15-2.91]; ≥ 400: HR, 1.63 [95% CI, 1.01-2.63]), increased PA-to-aorta ratio (HR, 1.33 [95% CI, 1.16-1.52], per 0.1-unit increase) and decreased thoracic skeletal muscle (HR, 0.88 [95% CI, 0.79-0.98], per 10 cm2/m2 increase) were independently associated with shorter overall survival, investigators observed.
In addition, 5-year overall survival was superior for the model that included clinical and imaging features and inferior for the model restricted to only clinical features. Of all features, the one that emerged the most predictive of overall survival was PA-to-aorta ratio.
In this single-center study of stage I lung cancer patients who were undergoing SBRT, increased CAC score, increased PA-to-aorta ratio, and decreased thoracic skeletal muscle index were independently predictive of poorer overall survival.
“Our modeling shows that these imaging features add so much more [to predicting overall survival],” Dr. Fintelmann said. “The strength of this study is that we show the utility [of the model] and how it exceeds the clinical risk prediction that is currently standard of care. We think this will benefit patients in terms of being able to counsel them and better advise them on their medical decisions.”
This proof-of-concept investigation requires external validation, Dr. Fintelmann stressed. “External data for validation is the next step,” he said, noting he and co-investigators welcome data input from other investigators.
Elsie Nguyen, MD, FRCPC, FNASCI, associate professor of radiology, University of Toronto, responded by email that the study shows that imaging features supplement clinical data in predicting overall survival.
“This study demonstrates the value of extracting non–cancer related computed tomography imaging features to build a model that can better predict overall survival as compared to clinical parameters alone (such as age, performance status and co-morbidities) for stage I lung cancer patients treated with SBRT,” Dr. Nguyen wrote.
“Coronary artery calcium score, pulmonary artery-to-aorta ratio, and sarcopenia independently predicted overall survival,” she wrote. “These results are not surprising, as the prognostic value of each of these imaging features has already been established in the literature.”
Dr. Nguyen pointed out the power in the sum of these imaging features to predict overall survival.
“However, the results of this study demonstrate promising results supportive of the notion that combining clinical and imaging data points can help build a more accurate prediction model for overall survival,” she wrote. “This is analogous to the Brock University (in St. Catharines, Ontario) calculator for solitary pulmonary nodules that calculates malignancy risk based on both clinical and imaging data points. However, external validation of these study results at other centers is first required.”
Dr. Fintelmann and Dr. Nguyen have disclosed no relevant financial relationships.
A version of this article first appeared on.