In this issue of Thorax (see page 854), Thirumaran and colleagues describe their experience with a consecutive group of patients evaluated for haemoptysis who had a normal chest radiograph.1 Over a 4-year period these investigators identified 270 patients who were evaluated for this problem. They made two important observations. First, they found that 26 of the 270 patients (9.6%) had respiratory tract malignancies as the cause of the haemoptysis. Of these 26 respiratory tract malignancies, 20 were non-small cell lung cancer (NSCLC) and 2 were small cell lung cancer. Secondly, they performed both fibreoptic bronchoscopy (FOB) and chest CT on the majority of the 270 patients evaluated. Overall, at least 269 FOBs and 257 CTs were performed to diagnose these 22 lung cancers. These observations lead to a series of fundamentally important questions for doctors to consider.

Why should haemoptysis with a normal chest radiograph be a concern? Haemoptysis can be massive and life-threatening, by causing either airway compromise or haemodynamic insufficiency. Fortunately, haemoptysis is rarely massive. I work in a busy Intensive Care Unit (ICU; 19 beds and >1500 admissions per year) in a large urban hospital (907 beds and about 47 000 admissions per year) and we see probably only 1–2 cases of massive, life-threatening haemoptysis per year. In the study by Thirumaran et al, of the 275 episodes of haemoptysis, none was massive; the majority were described as simply streaks of bleeding. Although doctors should be aware of the risk of massive haemoptysis, most cases of haemoptysis represent minor bleeding which stops spontaneously. The real concern about haemoptysis with a normal chest radiograph is understanding the cause. As Thirumaran and colleagues point out, haemoptysis should be considered a possible signal of an underlying serious illness, particularly lung cancer. Doctors should not be lulled into believing that lung cancer has been firmly excluded by the finding of a normal chest radiograph. Quekel and colleagues found that in 49 of 259 patients (19%) with NSCLC presenting as a nodular lesion on chest radiograph the cancer had been “missed”—that is present but not identified by the radiologist—on previous chest radiographs.2 Shah et al also documented cases of NSCLC “missed” during an initial chest radiograph.3 “Missed” cases of NSCLC on chest radiograph were usually small tumours in the upper lobes with often superimposed structures, although endobronchial cancers can also be “missed” on a chest radiograph. The 22 lung cancers documented in this study by Thirumaran et al were “missed” by chest radiography. The consequence of an NSCLC “missed” on chest radiograph is a delay in diagnosis with possibly failure to provide a surgical cure.2 3

How should general practitioners (GPs) consider haemoptysis as a presenting symptom? Haemoptysis, even with a normal chest radiograph, should be considered a signal of lung cancer. This is an important message because it has been estimated that a GP in the UK might see a new lung cancer case only once in every 8 months.4 Respiratory symptoms are common presenting complaints to GPs, but respiratory symptoms are usually due to benign causes. Consequently, GPs probably do not consider lung cancer often in their daily practice. However, of all the respiratory symptoms seen in general practice, haemoptysis seems to be most clearly linked to lung cancer.4 One weakness in the study by Thirumaran et al is that it is unclear how many patients in their catchment area actually presented to their GP with haemoptysis. I suspect that the number of patients presenting with haemoptysis to GPs (with either an unremarkable chest radiograph or with no chest radiograph obtained) in the West Yorkshire region during the years of this study was probably (substantially) higher than the number actually referred for evaluation at the Dewsbury and District Hospital. Education about the concerns related to haemoptysis at the local level is clearly needed.

How accurately does haemoptysis with a normal chest radiograph predict an occult lung cancer? In this series, there were 22 cases of lung cancer in 270 patients, reflecting a prevalence of lung cancer of 8.2% for this clinical scenario.1 Others have reported that the prevalence of lung cancer for this clinical scenario might be 5–6%.5 It is interesting to consider the reported prevalence of lung cancer in patients presenting with haemoptysis and a normal chest radiograph in the context of the ongoing controversy over screening for lung cancer. In studies published since 2004 using low dose CT (LDCT) for lung cancer screening, performed in asymptomatic populations older than age 40 usually with some smoking history, the baseline prevalence of lung cancer varied from 1.3% to 2.8%.6 7 8 9 10 11 12 13 Clinicians accept the value of searching for a lung cancer in a patient presenting with haemoptysis and a normal chest radiograph, but not in an asymptomatic patient older than 40 with a smoking history.14 This observation suggests that perhaps the benefit of screening for lung cancer could be demonstrated in a subset of asymptomatic patients with a somewhat higher prevalence of lung cancer. The implications of this insight are important from a public health perspective. In the USA, the number of lung cancers continues to increase, most lung cancers present in an advanced stage, and the projected 5-year survival for lung cancer remains disappointingly low.14 Identifying lung cancer early enough with a safe screening test to allow curative surgical resection would seem to be an attractive strategy for addressing the lung cancer problem. The approach over the last decade to earlier identification of lung cancer has relied on using more sophisticated chest imaging techniques, for example LDCT versus chest radiography. Improved imaging alone, though, might not be the solution. Ongoing studies using LDCT to screen for lung cancer have also attempted to identify higher risk patients using smoking history.15 Other approaches to risk assessment besides age and smoking history are also being explored, such as analyses of exhaled breath volatiles.16 Improved imaging applied to patients with a known higher prevalence of disease is a promising approach to solving the lung cancer screening dilemma.

How should a doctor evaluate a patient presenting with haemoptysis and a normal chest radiograph? Thirumaran and colleagues performed both FOB and chest CT in >90% of the patients who presented with haemoptysis and a normal chest radiograph. This approach seems clinically reasonable, because chest radiography “misses” small lung cancers and endobronchial lung cancers. However, the combined use of FOB and CT resulted in hundreds of procedures being performed, suggesting that a more efficient process for identifying lung cancers is needed. There are a variety of ways, such as cost-effectiveness analyses, to look at the efficiency of diagnostic strategies. One intriguing approach to compare the efficiency of diagnostic strategies is to calculate the number of tests needed to establish a diagnosis (NTND). This approach appeals to clinicians because it reflects how they solve problems on a daily basis. In a hypothetical analysis of a cohort of patients presenting with haemoptysis and a normal chest radiograph the NTND was calculated for various diagnostic approaches.5 Strategies employing either FOB or CT first, followed by the other test if non-diagnostic results were obtained, resulted in a substantial NTND, just as Thirumaran and colleagues reported.1 However, a novel approach incorporating the use of sputum cytology as an initial diagnostic test resulted in a substantially lower NTND than either the FOB or CT first approach, while still maintaining a high degree of diagnostic accuracy. Sputum cytology can identify endobronchial lung cancer with a reasonably low false-positive rate. An abnormal sputum cytology would be followed by FOB. An unremarkable sputum cytology could be followed by CT, or by serial chest radiographs.

Haemoptysis with a normal chest radiograph should raise concerns at several levels. For the individual patient and their doctor, it should raise the suspicion of lung cancer. GPs see many patients with respiratory symptoms, most of which are due to relatively benign causes. Haemoptysis, though, seems to be a more accurate predictor of underlying lung cancer than other respiratory symptoms, such as cough. For the overall healthcare system, a clearer understanding of which patients would benefit from screening for lung cancer is obviously needed. It is curious that the baseline prevalence of lung cancer in LDCT screening trials ranges from 1.3% to 2.8%, yet the value of lung cancer screening in the asymptomatic patient with a smoking history has not been established. Hopefully, the results of the National Lung Screening Trial, a 50 000 patient study evaluating lung cancer mortality in patients undergoing screening with LDCT currently underway in the USA, will clarify this situation. Alternatively, perhaps concentrating on integrating other methods into the screening process, which would identify patients with a higher prevalence of lung cancer, might increase the value of LDCT screening. At present, clinical practice accepts that there are subsets of patients, such as those with haemoptysis and a normal chest radiograph, with an expected higher prevalence of lung cancer who warrant careful evaluation. Clinicians should be aware, though, that there are strategies for searching for lung cancer in these patients which seem to be more efficient than performing CT and FOB on every patient.