Virtual Library

Start Your Search

L. Billingham

Author of

  • +

    MS25 - Translating Research into Practice (Applied Statistics) (ID 42)

    • Event: WCLC 2013
    • Type: Mini Symposia
    • Track: Statistics
    • Presentations: 1
    • +

      MS25.2 - Statistical Requirements for Screening Trials (ID 580)

      14:00 - 15:30  |  Author(s): L. Billingham

      • Abstract
      • Presentation
      • Slides

      A screening programme is a set of procedures that can be applied to an asymptomatic population to enable early detection and treatment of disease. The success of the programme is dependent on this early intervention reducing morbidity and mortality associated with the disease. There has been much research to develop and assess potential screening programmes for lung cancer. In particular, there are a number of major clinical trials such as the National Lung Screening Trial in the United States, the United Kingdom Lung Cancer Screening Trial and the Dutch-Belgian NELSON trial, that assess the effectiveness of screening a high risk population with low dose computed tomography scans to reduce mortality from lung cancer. The key elements of any screening programme are: (i) identification of the target population for screening, (ii) the screening test that will be used to classify patients as likely or unlikely to have the disease, (iii) the frequency of applying the screening test, (iv) the diagnostic test that will be used to determine whether people are truly diseased or not, (v) the treatment that will be available for those diagnosed early. These choices will impact on the statistical design. The research question may be whether to introduce screening, to determine the frequency of screening, to identify the appropriate target population for screening or whether to add an additional screening tool to an existing screening modality. Randomised controlled trials are the gold standard for assessing a screening programme as they overcome major biases specific to screening namely lead-time bias, length bias, over-diagnosis bias and selection bias. The usual trial design parameters are important but special statistical issues arise in relation to screening trials. Statistical inferences should only be made back to the eligible population defined for the trial so the choice of eligibility criteria is important. Screening interventions can cause harm to individuals that are potentially disease-free so the target population is usually those who are at high risk of disease. Interventions such as CT scans which have both a relatively high risk and high level of harm should be targeted at a high risk population whilst those that are low risk to the individual such as sputum cytology could be targeted at a wider population. Harms also include the inconvenience and psychological effects of a false positive result. A high risk population is usually defined in terms of pack-years of smoking and time since quitting and simple patient characteristics such as age. Statistical models that predict risk of disease could enhance the identification of a high risk population but the accuracy of prediction should be considered in relation to its impact on the trial. The primary outcome measure to assess benefit of a lung cancer screening programme should be lung cancer specific mortality. Duration of follow-up is an important design parameter. It needs to be long enough to allow for the time lag before the impact of screening becomes apparent and not so long after the cessation of screening as to include a period of time when screening would have lost its impact. Appropriate statistical analysis of the primary outcome measure is essential to properly evaluate the benefits of screening. Typically the screening intervention will be compared to the control arm in terms of its ability to reduce cumulative mortality. If this is calculated over the entire period of screening and follow-up then the benefit may be underestimated. Comparing time-specific mortality rates is the recommended approach [1]. The analysis is also complicated by the problems of non-attendance and contamination but methods have been proposed to adjust for these [2]. Sample size calculations involve key design parameters including hypothesised mortality reductions, expected compliance and contamination, number of screening rounds and length of follow-up [3]. The accuracy of the screening test to predict disease in asymptomatic people is not only important for the feasibility and ethics but will also impact on sample size as inaccurate predictions will dilute the potential benefit of screening. Screening tests, such as those that involve biomarkers measured on a continuous measurement scale, should be rigorously developed and validated before assessing their clinical utility within a randomised controlled trial environment. References [1] Hanley JA; Measuring mortality reductions in cancer screening trials; Epidemiologic Reviews 2011; 33: 36-45. [2] Baker SG, Kramer BS, Prorok PC; Statistical issues in randomised trials of cancer screening; BMC Medical Research Methodology 2002; 2: 11. [3] Prorok PC, Marcus PM; Cancer screening trials: nuts and bolts; Seminars in Oncology 2010; 37(3): 216-223.

      Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.