Selection processes for lung cancer screening

Abstract

A risk-prediction model is a more eff ective method of selecting patients for lung cancer screening than are the criteria used by the US National Lung Screening trial (NLST), according to the results of a new study. Organisations such as the American Cancer Society have endorsed screening of high-risk individuals for lung cancer (assuming that highquality radiology and treatment are available). Consequently, institutions and universities in the USA have started to establish screening programmes. The NLST showed a 20% reduction in cancer mortality with screening and low-dose CT. US screening programmes frequently use the NLST criteria—specifi cally, individuals aged 55–74 years, with a history of smoking of at least 30 packyears, and, for ex-smokers, having quit no more than 15 years previously (or some combination of these factors). Researchers led by Martin Tammemagi (Brock University, St Catharines, ON, Canada) modifi ed the lung cancer risk-prediction model previously developed for the Prostate, Lung, Colorectal and Ovarian Cancer (PLCO) screening trial. The model takes into account variables such as age, body-mass index, chronic obstructive pulmonary disease status, and smoking-related factors. The researchers applied the NLST screening criteria and the risk-prediction model to the PLCO population, and compared the results. The risk-prediction model proved more sensitive (83·0% vs 71·1%) and had a higher positive predictive value (4·0% vs 3·4%) than the NLST criteria. “Most importantly”, notes Tammemagi, “the risk-prediction model missed 41·3% fewer cancers than the NLST criteria”. Research recently published by Jiemin Ma and colleagues suggested that if all eligible Americans were screened for lung cancer, 12 000 deaths would be averted each year. If the risk-prediction model rather than the NLST criteria were used in screening programmes, a further 2750 lives would be saved (although in a real-world setting, such high numbers are unlikely to be realised). “It is a signifi cant but small step forward”, affi rms Michael Seckl (Imperial College London, London, UK). But lung cancer screening with low-dose CT throws up a huge number of false-positives—around 95%. “What we really need is some kind of additional test to determine risk, a blood test, for example, or a urine test”, Seckl added. Forthcoming costeff ectiveness studies will help to shape the future of lung cancer screening (the NLST data suggested that 320 screenings would avert one death, which compares favourably with similar studies for mammograms), as will the recommendation of the US Preventive Services Taskforce, expected later in 2013.