Lung cancer screening model misses 41% fewer cancers than NLST model

An update to a National Cancer Institute (NCI) lung cancer risk prediction model was found to be more sensitive than the National Lung Screening Trial (NLST) criteria for lung-cancer detection, according to a study published Feb. 21 in the New England Journal of Medicine.

Christine D. Berg, MD, of NCI in Rockville, Md., and colleagues modified the 2011 model from the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial in order to make it applicable to NLST data, and also to compare the model’s ability to estimate risk with NLST criteria.

“Use of an accurate model that incorporates additional risk factors to select persons for screening may identify more persons who have lung cancer or in whom lung cancer will develop,” wrote Berg et al.

Since the NLST demonstrated a 20 percent reduction in mortality from lung cancer screening, some organizations have begun applying NLST criteria to identify persons to be screened, explained the authors. The previous PLCO model also attempted to predict risk in an effort to maximize the efficiency of a screening program by looking at a number of predictors, such as age, body-mass index, family history of lung cancer, chronic obstructive pulmonary disease, chest radiography from the previous three years, smoking history and more.

In the updated model, which the authors called PLCO_M2012, previous radiography was excluded, while the variables “race or ethnic group” and “status with respect to a personal history of cancer” were added.

PLCO_M2012 was developed and validated using data from the more than 80,000 people in the PLCO control and intervention groups who had ever smoked. Area under the receiver operating characteristic curve (AUC) was evaluated in both a development data set and a validation data set. A total of 14,144 people in the validation data set met NLST criteria, and when comparing the models, follow-up was truncated at six years since the follow-up in the PLCO data exceeded the follow-up in the NLST.

Results showed the AUC for the development and validation data sets were 0.803 and 0.797, respectively. “A predictive model with an AUC in this range may be of value in providing individual-level information and in population-level screening programs,” wrote the authors.

PLCO_M2012 had a sensitivity of 83 percent, compared with 71.1 percent for NLST criteria, reported Berg and colleagues. Positive predictive value was also improved in the PLCO_M2012, while specificity was similar with NLST criteria.

Using PLCO_M2012, 41.3 percent fewer cancers were missed. Among 37,332 smokers in the PLCO intervention group, the updated PLCO model selected 81 more people for screening who received a diagnosis of lung cancer in follow-up than did the NLST criteria, according to the authors. After factoring in overdiagnosis and mortality rates, they determined that 12 additional lung cancer deaths would have been prevented in this group had PLCO_M2012 criteria been used rather than NLST criteria.

“Because the mortality reduction from CT screening effectiveness did not vary according to lung-cancer risk, it appears that use of the PLCO_M2012 to select persons for lung-screening programs could potentially be an effective method leading to improved cost-effectiveness of screening with additional deaths from lung cancer prevented,” concluded the authors.