Screening for lung cancer has a long and controversial history. Successful screening is predicated on two fundamental principles.1 First, the screening test should be able to detect disease in an early preclinical phase before symptoms develop. Second, treatment should be available and more effective when provided during the preclinical phase. On the surface it would appear that screening for lung cancer passes both of these tests, given our experience with treating ‘early’ versus ‘late’ stage lung cancer that is clinically detected. However, if we acknowledge that at least some cases of clinically detected stage I and II lung cancer might represent disease that is relatively indolent biologically as opposed to ‘early’, then the possibility exists that early detection will not alter the natural history of lung cancer and result in more frequent cure. Fortunately, the hypothesis that lung cancer screening with CT scanning reduces mortality is currently being evaluated in several large randomised controlled trials in both the USA and Europe.2–6

Another set of prerequisites for any screening programme is that both the screening test and any downstream testing should be convenient for patients, relatively inexpensive and without a high risk of complications. Recent attention has focused on previously discounted risks from the radiation exposure that are associated with undergoing CT screening and follow-up.7 Smith-Bindman et al estimated that the median effective radiation dose associated with a single non-contrast CT examination of the chest (8 mSV) was approximately equal to that of 117 chest x-rays and that, in 40-year-old women, one radiation-induced cancer would ultimately develop for every 720 who underwent CT.8 In a population-based analysis, Berrington de Gonzalez et al calculated that the approximately 72 million CT scans performed in the USA in 2007 would ultimately be responsible for 29 000 future radiation-induced cancers including 4100 related to CT scans of the chest.9

Other downstream risks associated with CT screening for lung cancer include complications of surgery and non-surgical biopsy. CT-guided needle biopsy is generally regarded as safe, but pneumothorax requiring chest tube placement occurs in about 5% of cases.10 Likewise, while fatal complications of video-assisted thoracoscopic wedge resection are infrequent (<1%), persistent air leaks and postoperative pneumonia can complicate 5–10% of cases.11

In this context, anything that can be done to minimise complications associated with screening or downstream testing will help tilt the balance in favour of benefits over harms. For over 10 years, beginning with the Early Lung Cancer Action Project,2 studies of lung cancer screening have implemented follow-up protocols that attempt to minimise downstream testing and its potential complications. However, practices for follow-up of incidentally-detected or screening-detected nodules have not been standardised for use in clinical settings or prospectively validated. The objective of follow-up is to identify malignant nodules promptly to permit timely surgical resection while minimising the frequency of unnecessary surgery (or biopsy) in the overwhelming majority of patients who have benign nodules or the smaller group of patients with relatively indolent malignant nodules that may represent cases of overdiagnosis.

In the absence of evidence-based recommendations for follow-up, the current standard of care is to use the expert consensus-based recommendations of the Fleischner Society12 which specify the frequency and duration of CT follow-up depending on the presence of risk factors for lung cancer and the size of the nodule. In general, the recommended follow-up is more aggressive for patients with risk factors and larger nodules because the prevalence of malignancy is higher in these groups and because it is more difficult to detect growth in smaller lesions over shorter time intervals. This difficulty is related to both technical and human factors. Because one doubling in volume corresponds to a 26% increase in diameter, it is nearly impossible in most clinical practice settings reliably to identify growth corresponding to one doubling in volume in a nodule that initially measures 4 mm in diameter, which would measure 5 mm in diameter after one doubling time.

The concept of volume doubling time (VDT) was introduced first by Collins et al13 in a study of pulmonary metastases and later studied empirically in patients with lung nodules by Nathan et al,14 Weiss et al,15 Steele and Buell16 and Geddes.17 Importantly, the concept of VDT assumes that the tumour grows exponentially, which is to say that the tumour doubles in volume at a constant rate. While there is some empirical evidence to support the exponential growth hypothesis for nodules when they measure between 10 mm and 30 mm in diameter, it is likely that growth is even faster for tumours earlier in their natural history, while growth is probably slower than exponential when tumours become very large and outgrow their blood supply. Computer-assisted methods of volumetric analysis were initially described and applied to the problem of characterising small pulmonary nodules over 10 years ago,18 19 but have not been widely adopted in clinical practice.

More recently, functional or molecular imaging with positron emission tomography (PET) has gained favour in thoracic oncology for characterisation of (primarily) incidentally detected lung nodules, initial staging of patients with non-small cell carcinoma and prognostication. Interestingly, greater uptake of fluorodeoxyglucose (FDG) in a malignant nodule has been shown be associated with faster growth rates20 and worse survival.21 The association between FDG uptake and growth suggests that PET might be used to distinguish between malignant and benign nodules. An important limitation of this approach is that PET is thought to be less sensitive for identifying malignancy in nodules measuring <8–10 mm in diameter.

Ashraf et al have attempted to capitalise on technological advances in computer-assisted measurement and molecular imaging to improve the characterisation of screening-detected lung nodules,22 presumably in an effort to reduce the rate of false positive screening evaluations and thereby reduce the risk of unnecessary biopsy or surgery (see page 315). In this retrospective analysis of data from the Danish Lung Cancer Screening Trial, they used semiautomated computer software (accompanied by manual measurements in a small number of nodules) that enabled them to estimate growth rates based on measurements performed on the initial scan in comparison with a follow-up scan 3 months later. In addition, they reviewed the results of PET imaging performed within 3 months of nodule detection. Subjects included 53 participants with 54 indeterminate pulmonary nodules measuring up to 20 mm in diameter, including 35 solid nodules, 9 semi-solid nodules and 10 ground glass opacities. In this sample the prevalence of malignancy was 37%. Almost 60% of the nodules were identified during the baseline (prevalence) round of screening.

The authors found that both greater FDG uptake and shorter VDT were significantly associated with malignancy. Using thresholds of equal to or greater than the mediastinal blood pool for FDG uptake and less than 365 days for VDT, they found that either technique, when evaluated in isolation, identified malignancy with a sensitivity of 70% (95% CI 48% to 85%) and a specificity of 91% (95% CI 77% to 97%). In a multivariable analysis, both FDG uptake and VDT were independently associated with malignancy. Furthermore, all 10 nodules with high FDG uptake and short VDT were malignant, while only 2 of 30 nodules (7%) with low FDG uptake and long VDT were malignant.

Although larger confirmatory studies are needed, these data suggest that patients with rapidly growing hypermetabolic nodules should be referred immediately for surgical resection, provided that three additional conditions are met: (1) the nodule is located in the periphery and accessible via video-assisted thoracoscopy (VATS); (2) there is no medical contraindication to VATS wedge resection; and (3) the suspicion of endemic mycosis or tuberculosis is not high. In such cases, many surgeons are correct to argue that needle biopsy adds little to the evaluation.

For patients with nodules that are not hypermetabolic and have a long VDT, I agree with Afshar et al that a follow-up scan in 1 year is probably sufficient provided that the patient understands and accepts the uncertain risks associated with delayed diagnosis and treatment of malignancy that will occur in as many as 21% of cases (the upper limit of the 95% CI around the false negative rate of 7%). Recent results from the Dutch-Belgian randomised trial of lung cancer screening, although not directly comparable, support this approach because <0.3% of participants with a negative evaluation that included volumetric measurement were eventually found to have lung cancer in this study.23 In patients with larger nodules or equivocal findings, a follow-up scan in 3–6 months should still be considered. In order to minimise radiation exposure and the attendant risks highlighted by Smith-Bindman and Berrington de Gonzalez, a low-dose thin-section unenhanced protocol with limited longitudinal coverage should be employed, as suggested by the Fleischner Society.12 Why this important and sensible recommendation has not been widely implemented in clinical practice is a worthy, indeed urgent, topic for quality improvement committees.

Lastly, for patients with discordant findings on FDG-PET and volumetric analysis, the likelihood of cancer is intermediate to high (57%, 95% CI 33% to 79%). Ashraf et al recommend repeating the CT scan in 3 months, but this seems redundant and unnecessary if rapid growth has already been identified. Likewise, increased FDG uptake on PET suggests that one is likely to be dealing with an active infectious or inflammatory process that requires further investigation, even if the finding is technically a false positive one for malignancy. I would therefore recommend tissue sampling by needle biopsy or bronchoscopy in this group, although it would not be wrong to perform VATS wedge resection in patients with larger nodules or borderline FDG uptake or VDT.

A remaining question is to what extent these findings apply to patients in current practice with incidentally detected lung nodules. It is not unreasonable to consider these patients as being similar to those with nodules that are detected during the prevalence round of CT screening. Not surprisingly, malignancy was significantly more likely in prevalent nodules than in incident nodules in the study by Ashraf et al, although the difference was not significant after adjustment for FDG uptake and VDT. Nevertheless, the lower prevalence of malignancy among nodules detected during baseline screening suggests that the combined criteria of FDG uptake and VDT will have a better negative predictive value and worse positive predictive value when applied to patients with incidentally detected nodules.

Going forward, practices for characterising small pulmonary nodules should be evaluated in randomised controlled trials in which the intervention is compared with the current standard of care (Fleischner Society guidelines), so trade-offs between benefits and harms can be quantified for the benefit of patients and the clinicians who counsel and care for them. Such information is of immediate importance for managing patients with incidentally detected pulmonary nodules, but it will take on additional urgency if CT screening for lung cancer is ultimately found to be effective in reducing mortality.