Virtual Library

 x 
My Library Virtual Library FAQ

Start Your Search

G.M. Strauss



Author of

  • +

    P2.20 - Poster Session 2 - Early Detection and Screening (ID 173)

    • Event: WCLC 2013
    • Type: Poster Session
    • Track: Imaging, Staging & Screening
    • Presentations: 1
    • +

      P2.20-007 - Computed Tomography (CT) Screening for Lung Cancer: Does the Mortality Endpoint Provide Definitive Evidence for CT Superiority in the National Lung Screening Trial (NLST)? (ID 2014)

      09:30 - 09:30  |  Author(s): G.M. Strauss

      • Abstract

      Background
      NLST generated excitement reporting that CT-screening reduces lung cancer mortality in comparison to CXR-screening. In randomized population trials (RPTs) on cancer screening, disease-specific mortality is assumed to provide an unbiased measure of screening effectiveness. This is based upon the assumption that randomization produces comparison groups with equal probability of death from the target cancer, unless the intervention reduces risk. However, these assumptions have often been violated in the RPT setting, leading to uncertainty about the effectiveness of several screening interventions. Objectives of this analysis are to assess whether the mortality endpoint provides an unbiased measure of screening efficacy in NLST.

      Methods
      In NLST, 53,454 current/former smokers, age 55-74 years, were randomized to CT or CXR-screening. Participants were offered a prevalence screen followed by two annual incidence screens.

      Results
      There was a highly significant 20% lung cancer mortality reduction in the CT-group (TABLE). There was also a significant 12% increase in lung cancer incidence in CT-group. Because CT is more sensitive than CXR, higher lung cancer incidence is predictable based upon lead-time and length-biases, and possibly overdiagnosis. The significant three-fold excess of BAC in the CT-group may reflect some overdiagnosis. Moreover, the highly significant two-fold excess of stage I cancers may reflect conventional biases. However, incidence of the most virulent lung cancers, including small-cell lung cancer and NSCLC-not-otherwise-specified was 16% lower in experimental-group. There exists no plausible hypothesis to explain lower incidence of highly biologically aggressive lung cancers in the CT-group. These differences strongly suggest imbalances in randomization. This is not to suggest that CT is not superior to CXR. While survival has not been reported, there is a significant 29% reduction in case-fatality in CT-group. Moreover, despite higher incidence, there was a highly significant one-third reduction in the absolute number of stage IV cancers in CT-group.

      CT Group CXR Group p value Relative Risk (95% Confidence Interval)
      Population 26,722 26,732
      LUNG CANCER DEATHS
      mortality 356 (1.33%) 443 (1.66%) 0.0022 0.80 (0.70-0.92)
      fatality 356/1060 (33.6%) 443/941 (47.1%) <0.0001 0.71 (0.64-0.80)
      LUNG CANCER INCIDENCE AND STAGE
      incidence 1,060 (3.97%) 941 (3.52%) 0.0067 1.12 (1.03-1.23)
      stage I lung cancer 416 (1.56%) 196 (0.73%) <0.0001 2.12 (1.79-2.51)
      stage IV lung cancer 226 (0.85%) 335 (1.25%) <0.0001 0.67 (0.57-0.80)
      LUNG CANCER HISTOLOGIC SUBTYPES
      bronchioalveolar carcinoma (BAC) incidence 110 (0.41%) 35 (0.13%) <0.0001 3.14 (2.15-4.60)
      adenocarcinoma, squamous cell carcinoma, or large cell carcinoma incidence 664 (2.48%) 527 (1.97%) <0.0001 1.26 (1.12-1.41)
      small cell carcinoma or nonsmall cell carcinoma-not otherwise specified (NSCLC-NOS) incidence 268 (1.00%) 317 (1.18%) 0.046 0.84 (0.72-0.99)
      carcinoid incidence 6 (0.022%) 2 (0.007%) 0.18 3.00 (0.61-14.87)

      Conclusion
      In NLST, lower incidence of lethal lung cancer subtypes in CT-group predicts for lower lung cancer mortality, independent of CT-screening efficacy. Accordingly, lower lung cancer mortality does not provide an unbiased measure of CT-screening efficacy in NLST. While stage distribution and case-fatality advantages strongly support that CT is superior to CXR, further analysis, which must include survival, is required to judge the true effect of CT-screening. While a simple survival comparison of CT versus CXR would also be biased, an analysis comparing survival stratified by histologic subtype or other outcome predictors, would be useful in judging the effectiveness of CT-screening in NLST.

  • +

    P3.20 - Poster Session 3 - Early Detection and Screening (ID 174)

    • Event: WCLC 2013
    • Type: Poster Session
    • Track: Imaging, Staging & Screening
    • Presentations: 1
    • +

      P3.20-007 - Chest X-ray (CXR) Screening for Lung Cancer: Overdiagnosis, Endpoints, and Randomized Population Trials (RPTs) (ID 2009)

      09:30 - 09:30  |  Author(s): G.M. Strauss

      • Abstract

      Background
      CXR-screening has long been considered ineffective because RPTs have failed to demonstrate significant mortality reductions in populations randomized to CXR-screening. While these studies demonstrate significant survival advantages associated with CXR-screening, these advantages are widely believed to be spurious, due to the interpretation that CXR-screening leads to substantial overdiagnosis. Indeed, the overdiagnosis hypothesis is the only alternative to the conclusion that CXR-screening saved lives in these trials. The objective of this analysis is to assess the magnitude of overdiagnosis in the context of CXR-screening, and to determine whether survival or mortality provides the best measure of efficacy in existing RPTs.

      Methods
      The Memorial-Sloan-Kettering and Johns-Hopkins Lung Projects were designed to assess the effectiveness of sputum cytology, since participants were randomized to annual CXR-screening and sputum cytology every four-months or to annual CXR alone. In Mayo-Lung-Project (MLP), following a normal prevalence CXR, 9,192 smokers were randomized to an experimental group (EG) undergoing CXR and cytology every 4-months for 6-years followed by 3-years of observation, or to a control group (CG) observed for 9-years. Controls were advised to obtain annual CXR.

      Results
      In MSKLP and JHLP, there were no differences between EG and CG regarding mortality, survival, or stage distribution. However, 5-year survival in both trials was about three-fold superior to contemporary national statistics. This suggests a beneficial effect of CXR, assuming that overdiagnosis did not confound survival in EG and CG in both studies. In MLP, lung cancer mortality was 6% higher in EG (p=0.62), although 5-year survival was significantly superior in EG (34% vs. 13%; p=0.0021). The fact that lung cancer incidence was 30% higher in EG (p=0.013) led to the hypothesis that overdiagnosis was responsible for the mortality/survival discrepancy. However, MLP data are inconsistent with overdiagnosis. Long-term survival was achieved only among those undergoing resection. Half of resected cancers were apparently cured, while no unresected patient was cured (7-year survival: 50% vs. 0%: p<0.0001). Indeed, follow-up extended to >20 years in MLP continues to show higher incidence and mortality in EG. While interpreted as supporting overdiagnosis, long-term follow-up results are primarily informative as to randomization adequacy. Mathematical modeling studies indicate that in MLP-EG higher lung cancer incidence and mortality was due to population heterogeneity, due to problems with randomization, not to overdiagnosis.

      Conclusion
      Abundant evidence supports that the magnitude of overdiagnosis is minimal in the context of CXR-screening. Moreover, evidence from RPTs indicates that CXR-screening is superior to no screening. The survival endpoint is not meaningless in RPTs if one can exclude or adjust for confounding due to overdiagnosis and with sufficiently long follow-up. Because overdiagnosis was not a confounder of survival in MLP, survival most accurately reflected CXR effectiveness. Indeed, long-term survival provided an unbiased surrogate for cure. While the National Lung Screening Trial demonstrates that CT-screening is superior to CXR-screening, CXR is more widely accessible and relatively inexpensive. CXR-screening has the potential to substantially reduce the global burden of lung cancer mortality.