Skip to Main Content
Principal Investigator
Samir Soneji
Dartmouth College
Position Title
Assistant Professor
About this CDAS Project
NLST (Learn more about this study)
Project ID
Initial CDAS Request Approval
Feb 25, 2014
Assessing the Translation of CT Screening from Clinical Trial to Everyday Practice
The recently completed multi-center randomized National Lung Screening Trial (NLST) concluded that computed tomography (CT) screening reduced lung cancer death by 20% and all-cause death by 7%, compared to chest X-ray screening. While encouraging, the NLST findings leave a critical question of population health unanswered: Will the translation of CT screening from NLST to everyday practice reduce the burden of lung cancer? Often, the benefits observed in randomized control trials (RCT) far exceed the benefits observed in everyday practice because the trial population differs substantively from that of the general population and medical care is performed in the community setting rather than in the high-volume academic medical centers of the RCT. For example, perioperative mortality following carotid endarterectomy was substantially higher than observed in 2 national trials, even in the institutions participating in the RCTs. In the case of lung cancer screening, only 0.5% of surgical patients died within 2 months of surgery in NLST, compared to 4% nationally.

This project seeks to improve clinical practice by assessing the generalizability of NLST to everyday practice in three ways. First, although the most likely explanation for the effect of CT screening on lung cancer mortality in NLST was early detection and better early stage survival, the 7% reduction in all-cause mortality suggests additional mechanisms. This project will help uncover those mechanisms and quantify how CT screened proved efficacious in NLST. By assessing how CT screening reduced death in NLST, this project will help to determine how CT screening may also reduce death in everyday practice. For example, if medical and surgical complications prove to be an additional important factor that led to the efficacy of CT screening in NLST, hospitals in everyday practice may improve their screening programs by quality improvement efforts to minimize these complications.

Second, for lung cancer screening in everyday practice, major complications resulting from invasive diagnostic procedures and surgical treatment may represent the key barriers in achieving the mortality reductions observed in the NLST. This project will characterize national major complication rates and motivate additional studies regarding the effectiveness of specific health care systems in diagnosis and treatment. Additionally, this project will help to identify quality improvement targets for lung cancer diagnosis- and treatment- related complications that would enable hospitals in everyday practice to achieve similar results as those observed in NLST. This project is timely because lung cancer screening has already begun, although we do not yet know how high contemporary complication rates are in everyday practice.

Third, I will assess the generalizability of NLST findings to the target population of NLST-eligible US adult smokers. The NLST trial sample differs in substantively important ways compared to the national population of screen-eligible smokers and these differences may substantially mute the effect of CT screening on mortality. This project will provide nationally representative estimates of the effect of CT screening.

Aim 1. To determine the mechanisms by which CT screening reduced mortality from lung cancer and all other causes of death in NLST. Hypotheses: In NLST [a] CT screening reduced mortality from lung cancer through earlier stage at diagnosis and mortality from other causes through earlier detection of other cancers and respiratory disease and [b] the benefits of CT screening were partially offset by higher complication-related mortality among patients screened by CT, compared to chest X-ray. Applying causal mediation analysis on full individual-level NLST data, I will statistically decompose the effect of CT screening into the effects of earlier detection of lung cancer, medical/surgical complications, & detection of clinically significant incidental findings.

Aim 2. To assess major complication rates from [1] lung cancer-related invasive diagnostic procedures among Medicare (2002-2009) patients in NLST and non-NLST hospitals and [2] lung cancer surgery among SEER- Medicare (2002-2009) patients receiving care in NLST and non-NLST hospitals located within SEER registry areas, accounting for stage at lung cancer diagnosis. Hypotheses: [a] Major complications from invasive diagnostic procedures will be higher in non-NLST hospitals than NLST hospitals and [b] Major surgical complications will be greater for non-NLST hospitals than NLST hospitals located in SEER registry areas. To assess major complications, I will create consecutive annual cohorts of all patients receiving invasive diagnostic procedures and surgery (regardless of whether screen-detected) and follow patients forward in time to ascertain complication-related hospitalization or death.

Aim 3. To assess if the findings of NLST are generalizable to the US population of NLST-eligible smokers. Hypotheses: [a] The efficacy of CT screening in NLST will diminish if complication rates approach levels observed in everyday practice and [b] Age and smoking status are important treatment effect moderators that differ between NLST and everyday practice; standardizing NLST results to the national population of NLST- eligible smokers will diminish the effect of CT screening. Applying a statistical simulation model, I will subject NLST participants to the range of complications rates observed in everyday practice and reassess the efficacy of CT screening. Applying statistical propensity score-related analysis, I will standardize NLST results to a national target population of US NLST-eligible smokers and reassess the effectiveness of CT screening.


William Black, MD (Dartmouth College)

Related Publications