Background: Screening for lung cancer with a low-dose computed tomography (CT) scan is estimated to prevent 3 deaths per 1000 individuals at high risk; however, false positive results and radiation exposure are relevant harms and deserve careful consideration. Screening candidates can only make an autonomous decision if doctors correctly inform them of the pros and cons of the method; therefore, this study aimed to evaluate whether doctors understand the test characteristics of lung cancer screening.
Methods: In a randomized trial 556 doctors (members of the Austrian Respiratory Society) were invited to answer questions regarding lung cancer screening based on online case vignettes. Half of the participants were randomized to the group 'solutions provided' and received the correct solutions in advance. The group 'solutions withheld' had to rely on prior knowledge or estimates. The primary endpoint was the between-group difference in the estimated number of deaths preventable by screening. Secondary endpoints were the between-group differences in the prevalence of lung cancer, prevalence of a positive screening results, sensitivity, specificity, positive predictive value, and false negative rate. Estimations were also compared with current data from the literature.
Results: The response rate was 29% in both groups. The reduction in the number of deaths due to screening was overestimated six-fold (95% confidence interval CI: 4-8) compared with the actual data, and there was no effect of group allocation. Providing the correct solutions to doctors had no systematic effect on their answers.
Conclusion: Doctors poorly understand the test characteristics of lung cancer screening. Providing the correct solutions in advance did not improve the answers. Continuing education regarding lung cancer screening and the interpretation of test characteristics may be a simple remedy.
Clinical trial registration: Clinical trial registered with www.clinicaltrials.gov (NCT02542332).
Keywords: Computed tomography; Lung cancer; Screening; Statistics.