New chest x-ray predictive model identifies positive lung cancer diagnosis
A study published in the June issue of the Journal of Thoracic Oncology has identified several nonradiographic and radiographic predictors of lung cancer, and presents an accurate model for estimating the probability of lung cancer in individuals with suspicious radiographs.

Chest radiographs are routinely employed in clinical practice, and radiographic findings that are abnormal suspicious (AS) for lung cancer occur commonly, according to the authors. However, the majority of abnormal suspicious (AS) radiographic abnormalities are not cancer.

Martin Carl Tammemagi, MD, of Brock University in St. Catherines, Ontario, and his team of U.S. researchers examined the chest radiographs obtained through the National Cancer Institute's (NCI) Prostate Lung Colorectal Ovarian Cancer (PLCO) screening trial. The PLCO intervention arm had 77,465 individuals, of whom 12,314 were AS and of these 232 (1.9 perrcent) had lung cancer (were true positive).

Overall, the investigators found that older age, lower education levels and a longer smoking history were all associated with a true positive diagnosis for lung cancer in those individuals with an abnormal screening chest radiograph. Other factors that contributed to a true positive diagnosis include a family history of lung cancer and a suspicious mass in the upper/middle chest region.

The model had a receiver operator characteristic (ROC) area under the curve (AUC) of 86.4 percent. The authors said that the model excluding the smoking variables had an ROC AUC of 77.1 percent and excluding all nonradiographic variables had an ROC AUC of 73.3 percent. Smoking and nonsmoking nonradiographic variables significantly added to prediction.

"The factors will be particularly valuable to those healthcare providers and clinicians identifying patients with abnormal chest x-rays that might indicate possible lung cancer," Tammemagi said. "An earlier diagnosis is expected to lead to a more favorable outcome for the patient, so it is our hope that predictors will assist clinicians in calling for the most necessary and timely tests."

Tammemagi and colleagues said that these findings may be of value for screening, research and patient and clinician decision-making.

This study was conducted in collaboration with the NCI, Georgetown University in Washington, D.C., and the University of Minnesota in Minneapolis.