New Predictive Model for Lung Cancer May Improve Accuracy in Prescreening Lung Nodules Prior to Resection
An investigative predictive model for lung cancer demonstrated more accuracy than the more commonly used Mayo Clinic model in prescreening lung nodules prior to resection, according to the results of a study reported by Deppen et al in the Journal of Thoracic Oncology. The investigators suggested that this new tool may fill the diagnostic void of validated models for this high-risk population.
Although it is well known that lung cancer continues to be the leading cause of cancer-related mortality in the United States, screening for the presence of lung modules with computed tomography has shown to be effective in reducing mortality. However, many patients prescreened for cancerous lung nodules subsequently undergo surgical resection for benign disease. Thus, a more effective predictive model for prescreening patients with lung nodules is needed.
With that in mind, Deppen and colleagues developed a predictive model and compared it with the more commonly used one from the Mayo Clinic. They hoped that such a comparative analysis would lead to a more accurate approach to the prescreening of lung nodules prior to resection.
Study Details
The predictive model (Thoracic Research Evaluation and Treatment model) was applied to two cohorts of patient records. The first, from the Vanderbilt University Medical Center, consisted of 492 patients with suspected or confirmed lung cancer. Patient records were evaluated from January 2005 to October 2010. Imaging data were abstracted from radiologist reports or original scans of the most recent preoperative computed tomography scans for lesion growth.
The second cohort was composed of 226 patient records from the Tennessee Valley Veterans Affairs who underwent a thoracic operation for known or suspected lung cancer between January 1, 2005, and December 31, 2013.
The ability of the predictive model developed by the investigators to discriminate between cancer and benign disease was evaluated by the area under the receiver operating characteristic curve. This factor was utilized to compare the Thoracic Research Evaluation and Treatment model and the Mayo Clinic model.
In total, the prevalence of lung cancer was 72% in the Vanderbilt University Medical Center cohort and 93% in the Tennessee Valley Veterans Affairs cohort. After resection, pathologic diagnosis of lung cancer was determined in 92% of patients and via surveillance in 8% of patients in the Vanderbilt University Medical Center cohort. All diagnoses in the Tennessee Valley Veterans Affairs cohort were determined pathologically.
Higher Accuracy and Better Calibration
After bootstrap adjustment, the Thoracic Research Evaluation and Treatment model resulted in an area under the curve of 0.87 (95% confidence interval [CI] = 0.83–0.92) and a Brier score of 0.12 (95% CI = 0.10–0.14) for patients in the Vanderbilt University Medical Center cohort. For patients in the Tennessee Valley Veterans Affairs cohort, the area under the curve was 0.89 (95% CI = 0.79–0.98), and the Brier score was 0.08 (95% CI = 0.06–0.10).
Using published coefficients to estimate lung cancer risk, the Mayo Clinic model resulted in an area under the curve of 0.80 (95% CI = 0.75–0.85) for patients in the Vanderbilt University Medical Center cohort. This number was significantly less (P < .001) than the area under the curve observed for patients in the investigative model. Thus, the Mayo Clinic model generally overestimated the risk of lung cancer compared with the Thoracic Research Evaluation and Treatment model.
Closing Thoughts
The results of this study suggested that the Thoracic Research Evaluation and Treatment model provides a high and consistent predictive tool for lung cancer based upon common clinical characteristics. In fact, it outperformed the Mayo Clinic model. As the prevalence of disease increased to 95%, as was found in the Tennessee Valley Veterans Affairs cohort, the accuracy of the Mayo Clinic model to discriminate malignancy decreased.
According to the investigators, “When examining the landscape of clinical prediction models for lung cancer, clinicians evaluating a patient immediately prior to surgery have no validated models for this high-risk population. The Thoracic Research Evaluation and Treatment model addresses that need.”
Eric L. Grogan, MD, MPH, of the Department of Thoracic Surgery, Vanderbilt University Medical Center, Nashville, Tennessee, is the corresponding author of the article in the Journal of Thoracic Oncology.
The study was supported by the Department of Veterans Affairs, the National Institutes of Health, and the Vanderbilt Institute for Clinical and Translational Research. The authors declared no potential conflicts of interest.
The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.
