2017 HSR&D/QUERI National Conference

1026 — Comparison of observed harms and expected mortality benefit for Veterans in VHA's low-Dose CT lung cancer screening demonstration project

Lead/Presenter: Tanner Caverly, COIN - Ann Arbor
All Authors: Caverly TJ (Center for Clinical Management Research; Ann Arbor VA) Hayward RA (Center for Clinical Management Research; Ann Arbor VA) Wiener RS (Center for Healthcare Organization & Implementation Research; Bedford VAMC) Buzenberg K (Ann Arbor VA) Slatore CG (Center to Improve Veteran Involvement in Care; VA Portland Health Care System) Tanner NT (Ralph H. Johnson VAMC Charleston, SC) Yun S (Ann Arbor VA) Fagerlin A (Salt Lake City VA Center for Informatics Decision Enhancement and Surveillance)

Objectives:
VHA's lung cancer screening demonstration project identified a high rate of false-positives following an initial low-dose computed tomography (LDCT) screen. We sought to put these findings into context by studying, for Veterans enrolled in the project, how observed harms compared to the expected mortality benefit.

Methods:
Between 7/2013 and 6/30/2015, the project enrolled 2,106 Veterans to receive LDCT screening across 8 academic VA's. To put the observed harms following initial LDCT screening into context, we individualized the expected mortality benefit of a single screen for each Veteran (applying the trial-based relative risk reduction in lung cancer mortality with screening to each Veteran's predicted baseline lung cancer mortality, calculated by inputting data from VA's Corporate Data Warehouse into the validated Bach risk model). Finally, we separated Veterans into quintiles of risk to assess, for each quintile: number of lung cancer cases observed; screening effectiveness (number needed to undergo a single initial screen to prevent 1 lung cancer death or NNS); and screening efficiency (ratio of false positives and unnecessary procedures per lung cancer death prevented).

Results:
The 1.5% prevalence of lung cancer in the project correlated closely with predicted prevalence and 29 of 31 lung cancer cases occurred in the 3 highest quintiles of risk. The initial LDCT screen was least effective for Veterans in quintile 1 (NNS of 3,660) and most effective for Veterans in quintile 5 (NNS of 365). Similarly, the initial screen was least efficient for Veterans in quintile 1 (2,131 false-positives and 52 unnecessary diagnostic procedures per lung cancer death prevented) and much more efficient for those in higher risk quintiles (e.g., only 205 false-positives and 13 unnecessary diagnostic procedures per death prevented in quintile 5).

Implications:
A single LDCT for eligible Veterans at lower lung cancer risk was much less effective and efficient, while effectiveness and efficiency for Veterans at higher risk exceeded that of established cancer screening programs (e.g., annual mammography for 10 years beginning at age 50 has an estimated NNS of 1,000 with 490-670 false-positives and 70-100 unnecessary biopsies per cancer death prevented).

Impacts:
This study supports the importance of using a risk-based approach to select Veterans for LDCT screening.