The limited detection of lung cancer on chest X-rays

In a study of everyday clinical practice, the diagnosis 'lung cancer' was not made on the chest X-ray initially in one-fifth of the cases, even though in retrospect the lesions had been visible on the chest X-ray. In nearly half of these cases, the diagnosis had been missed at least twice on succeeding X-rays of the patient in question. Most often, superimposed structures are responsible for having missed lesions on the chest X-ray; this was the case in 71% of the patients in whom a lesion was missed. The intra- and inter-observer variation in the detection of small tumours on the chest X-ray are quite large (kappa: 0.38 and 0.48, respectively). Providing clinical information or previous chest X-rays for purposes of comparison does not improve the observer performance. Separate evaluation by two assessors ('double reading') also does not improve the results significantly. Simultaneous evaluation ('dual reading') improves the sensitivity slightly (from 28 to 37%), but has hardly any effect on the specificity (from 93 to 92%), thus improving the results somewhat. The results with computed tomography are much better (high sensitivity but low specificity). A CT-scan of the chest seems indicated in case of a suspicion of lung cancer and/or when there is a (slight) suspicion of lung cancer on the basis of the chest X-ray.