• Clinical Epidemiology
• Lung Cancer

We thank Rogers et al1 for their comments on our recent article.2 Although it is possible that there may be some under-reporting of chest X-ray (CXR) requests in The Health Improvement Network primary care data set, it is highly unlikely that the associations that we identified in our paper2 are influenced by this because the practice of under-reporting of CXRs is unlikely to be related to deaths from early cancer. For this reason, if anything, our findings may even underestimate the true associations. We know that almost 40% of patients with lung cancer are diagnosed via an emergency route,3 so we were expecting that the patients who died early would be less likely to have undergone primary care CXR. Our CXR rates are corrected for background lung cancer incidence in our multivariable analysis and there was no association between high background lung cancer rates and higher CXR rates. The focus of this paper was on early death from lung cancer, hence a comment on long-term survival and the factors that influence this are outside the remit of the paper.

Although we agree with Rogers et al that reliance on primary care to identify lung cancer with no additional tools to identify those at greatest risk is unlikely to be successful, we do not feel that a more liberal CXR requesting policy is likely to be the best answer. Iyen-Omofoman et al4 derived and validated a risk assessment tool, which had greater discriminatory ability compared with current National Institute of Health and Care Excellence (NICE) guidelines. The use of a predictive score meant that for each case of lung cancer detected 119 CXRs would need to be performed, when compared with 421 CXRs per case using current NICE guidelines. Others have produced similar tools.5 ,6 As suggested in our paper, we believe that the most promising approach is to provide primary care colleagues with validated tools that allow better targeting of investigations.