BackgroundEstimation of the risk of malignancy in pulmonary nodules detected by CT is central in clinical management. The use of artificial intelligence (AI) offers an opportunity to improve risk prediction. Here we compare the performance of an AI algorithm, the lung cancer prediction convolutional neural network (LCP-CNN), with that of the Brock University model, recommended in UK guidelines.MethodsA dataset of incidentally detected pulmonary nodules measuring 5–15 mm was collected retrospectively from three UK hospitals for use in a validation study. Ground truth diagnosis for each nodule was based on histology (required for any cancer), resolution, stability or (for pulmonary lymph nodes only) expert opinion. There were 1397 nodules in 1187 patients, of which 234 nodules in 229 (19.3%) patients were cancer. Model discrimination and performance statistics at predefined score thresholds were compared between the Brock model and the LCP-CNN.ResultsThe area under the curve for LCP-CNN was 89.6% (95% CI 87.6 to 91.5), compared with 86.8% (95% CI 84.3 to 89.1) for the Brock model (p≤0.005). Using the LCP-CNN, we found that 24.5% of nodules scored below the lowest cancer nodule score, compared with 10.9% using the Brock score. Using the predefined thresholds, we found that the LCP-CNN gave one false negative (0.4% of cancers), whereas the Brock model gave six (2.5%), while specificity statistics were similar between the two models.ConclusionThe LCP-CNN score has better discrimination and allows a larger proportion of benign nodules to be identified without missing cancers than the Brock model. This has the potential to substantially reduce the proportion of surveillance CT scans required and thus save significant resources.

No abstract

BackgroundThe UK has poor lung cancer survival rates and high early mortality, compared to other countries. We aimed to identify factors associated with early death, and features of primary care that might contribute to late diagnosis.MethodsAll cases of lung cancer diagnosed between 2000 and 2013 were extracted from The Health Improvement Network database. Patients who died within 90 days of diagnosis were compared with those who survived longer. Standardised chest X-ray (CXR) and lung cancer rates were calculated for each practice.ResultsOf 20 142 people with lung cancer, those who died early consulted with primary care more frequently prediagnosis. Individual factors associated with early death were male sex (OR 1.17; 95% CI 1.10 to 1.24), current smoking (OR 1.43; 95% CI 1.28 to 1.61), increasing age (OR 1.80; 95% CI 1.62 to 1.99 for age ≥80 years compared to 65–69 years), social deprivation (OR 1.16; 95% CI 1.04 to 1.30 for Townsend quintile 5 vs 1) and rural versus urban residence (OR 1.22; 95% CI 1.06 to 1.41). CXR rates varied widely, and the odds of early death were highest in the practices which requested more CXRs. Lung cancer incidence at practice level did not affect early deaths.ConclusionsPatients who die early from lung cancer are interacting with primary care prediagnosis, suggesting potentially missed opportunities to identify them earlier. A general increase in CXR requests may not improve survival; rather, a more timely and appropriate targeting of this investigation using risk assessment tools needs further assessment.

IntroductionLung cancer screening (LCS) by low-dose computed tomography (LDCT) offers an opportunity to impact both lung cancer and coronary heart disease mortality through detection of coronary artery calcification (CAC). Here, we explore the value of CAC and cardiovascular disease (CVD) risk assessment in LCS participants in the Lung Screen Uptake Trial (LSUT).MethodsIn this cross-sectional study, current and ex-smokers aged 60–75 were invited to a ‘lung health check’. Data collection included a CVD risk assessment enabling estimation of 10 year CVD risk using the QRISK2 score. Participants meeting the required lung cancer risk underwent an ungated, non-contrast LDCT. Descriptive data, bivariate associations and a multivariate analysis of predictors of statin use are presented.ResultsOf 1005 individuals enrolled, 680 were included in the final analysis. 421 (61.9%) had CAC present and in 49 (7.2%), this was heavy. 668 (98%) of participants had a QRISK2≥10% and QRISK2 was positively associated with increasing CAC grade (OR 4.29 (CI 0.93 to 19.88) for QRISK2=10%–20% and 12.29 (CI 2.68 to 56.1) for QRISK2≥20% respectively). Of those who qualified for statin primary prevention (QRISK2≥10%), 56.8% did not report a history of statin use. In the multivariate analysis statin use was associated with age, body mass index and history of hypertension and diabetes.ConclusionsLCS offers an important opportunity for instituting CVD risk assessment in all LCS participants irrespective of the presence of LDCT-detected CAC. Further studies are needed to determine whether CAC could enhance uptake and adherence to primary preventative strategies.