Hormuzd Katki

PhD

NCI

Senior Investigator

NLST (Learn more about this study)

NLST-188

Jan 19, 2016

Heterogeneity of screening outcomes across participating centers in the National Lung Screening Trial

The National Lung Screening Trial (NLST) demonstrated a 20% reduction in lung cancer mortality with three annual screens of high-risk individuals using low-dose computed tomography (LDCT). However, there may be substantial variation in screening outcomes among the participating study centers. Screening outcomes can be efficacy, CT-positivity, CT-false-positivity, number needed to screen to prevent 1 death (NNS), efficiency (number of false-positive CTs per prevented death), and overdiagnosis (number of extra lung cancers diagnosed per prevented death). Variation in outcomes might reflect differences in local clinical practices between centers related to the patient characteristics, implementation of the screening protocol, the determination of a positive finding, and the subsequent investigational follow-up of potentially positive findings.

We plan to contrast outcomes from different centers, in particular, CT-positivity versus NNS. As CT-positivity increases, we expect NNS to decrease. This may allow evidence-based decisions on the CT-positivity that is required to achieve a certain NNS (and also screening efficiency, etc.), which might help develop more cost-effective CT lung screening guidelines. We have also developed methodology to project the impact of CT lung screening for the US (currently under review). We plan to project the impact of different decisions on CT-positivity (and other outcomes) to projected outcomes from CT lung screening in the US.

Many medical centers will not use NLST criteria to evaluate CT images, but instead use Lung-RADS. Lung-RADS criteria were retrospectively applied to the NLST, and diagnostic characteristics have been estimated (Pinsky et al, Ann Int Med, 2015). We wish to also define CT-positivity using the Lung-RADS data and repeating all analyses. We plan to project the impact of Lung-RADS on the NLST (if Lung-RADS had been used), and for CT lung screening in the US.

Aim 1: Evaluate variation in screening outcomes (such as: efficacy, CT-positivity, CT-false-positivity, number needed to screen to prevent 1 death (NNS), efficiency (number of false-positive CTs per prevented death), and overdiagnosis (number of extra lung cancers diagnosed per prevented death)) across the NLST centers. We will calculate these outcomes for each center and examine their distribution univariately. Then we will meta-regress the outcomes on each other. If necessary, we will consider a random-effects meta-regression to account for overdispersion beyond a fixed-effects model. We may adjust for patient-level lung-cancer risk factors as confounders. We will use our methodology to project the impact of different decisions on CT-positivity (and other outcomes) to projected outcomes from CT lung screening in the US.

Aim 2: Re-define CT-positivity based on retrospectively-applied Lung-RADS criteria, and re-do the analyses of Aim 1. We plan to project the impact of Lung-RADS on the NLST (if Lung-RADS had been used), and for CT lung screening in the US.