Staging of non-small cell lung cancer (NSCLC) has undergone a significant evolution, from plain chest radiographs to anatomical imaging, invasive techniques and, most recently, metabolic imaging using positron emission tomography (PET) scans. Even the literature regarding PET imaging has undergone significant evolution. Initial reports were characterised by compelling yet anecdotal images. This was followed by approximately 10 years of studies showing that mediastinal staging by PET was superior to computed tomography (CT) which, of course, was not surprising because CT had already been shown to be notoriously misleading in many situations. Eventually authors began addressing the clinically more relevant question of whether PET can replace invasive mediastinal staging. The article by Tournoy and colleagues¹ in this issue of Thorax illustrates how far we have come (see page 696). Not only does this study use the most sophisticated technology—an integrated PET/CT scanner—but, more importantly, the authors have elevated the science a notch by thoughtfully evaluating nuances of scan interpretation in order to maximise what can be gained from this staging modality.

The overall scientific quality of the study by Tournoy and colleagues is good. An appropriate gold standard was used by requiring a surgical staging procedure after a negative needle staging test (transoesophageal ultrasound with needle aspiration, transbronchial needle aspiration, etc, which carry a 20–30% false negative rate).² The authors should also be commended for looking at enlarged and normal size nodes separately, since PET uptake in smaller nodules is more difficult to detect. In addition, the careful evaluation of different objective criteria to try to improve the reliability of the PET interpretation is a valuable addition. On the other hand, reporting results on a per node basis statistically biases the results in favour of PET. Furthermore, this makes the data less applicable clinically because we must decide how to manage patients, not individual nodes. Additionally, lumping together mediastinal and hilar nodes biases the study in favour of PET because it avoids a distinction that can be difficult to make on PET. This also makes the data less clinically applicable because involved N1 nodes are generally treated differently from involved N2 nodes. An additional criticism is that the final assessment of the nodes is vague (which would also tend to bias the results in favour of PET). It is unclear whether patients with a negative invasive staging went on to resection, and whether further nodal sampling at the time of resection was done. Finally, it appears that node sampling was limited (average of 2.01 nodes per patient, including both mediastinal and hilar node stations).

In the end, the study by Tournoy et al¹ shows that evolution does not necessarily mean we are making progress. The study shows that PET is not adequate to avoid invasive staging in most instances despite the sophisticated technology of CT/PET, the evaluation of objective criteria and the inherent bias of a per node analysis. The key test parameters for a clinician in deciding whether a test is reliable enough to guide the management of an individual patient is the false positive or false negative rate. This study shows that a positive PET scan must be confirmed (false positive rate of 21% with enlarged nodes and 50% with normal size nodes).¹ Furthermore, the study shows that a negative PET result carries a significant false negative rate (9% with normal size nodes and 10% with enlarged nodes).¹ These results are generally consistent with that of other authors, which have shown a false negative rate of PET in the mediastinum of 20–30% in patients with discrete enlarged mediastinal nodes^3,4 and of 25% in patients with a normal mediastinum and enlarged N1 nodes.^4–6 However, the false negative rate of PET in the mediastinum is <5% for patients with a peripheral clinical stage I tumour.^5,6,7,8,9,10

It is interesting to revisit the evolution of CT imaging. The early CT studies, involving limited numbers of patients and first generation scanners, reported sensitivity and specificity rates of >90%. Later studies, involving many hundreds of patients and fourth generation scanners, found sensitivity and specificity rates of approximately 60–70%.^11,12 It appears that broader application of CT (across many centres and more generally to patients) yielded worse results, despite major technological improvements (rapid, helical, higher resolution scanners). Similarly, the early studies of PET in NSCLC demonstrated sensitivity and specificity for mediastinal staging of over 90%,¹³ while the more recent results show these rates to be around 70–80% despite advancing technology.^11,14 Is the study by Tournoy¹ showing that we have reached the limits of what PET imaging can give us with regard to mediastinal staging?

Invasive staging tests have also undergone an evolution. Needle aspiration has gone from a “blind” transcarinal aspiration to real-time image-guided aspirations via either transoesophageal ultrasound or endobronchial ultrasound. The data from these newer interventions appear to be much better than the older transbronchial needle technique.^15,16 Mediastinoscopy has evolved from a simple lighted tube to video-mediastinoscopy and video-mediastinoscopic lymph-adenectomy, both of which have been reported to have better reliability than simple mediastinoscopy.^17–19 We will have to see whether the excellent results reported in the initial studies hold up as the technology disseminates.

So where does this leave us? Again, it is worth revisiting the evolution of CT imaging in lung cancer. There is no doubt that CT has become an integral and, for all practical purposes, an essential part of the assessment of a patient with lung cancer. This is because CT is invaluable in guiding the evaluation and choice of treatment in patients with NSCLC and not because it obviates the need for invasive mediastinal staging tests, as was originally anticipated. Chest CT scanning helps guide the need for extrathoracic imaging in that approximately 30% of patients with enlarged media-stinal nodes will have detectable distant metastases despite having no signs of symptoms of metastases,^7,20 whereas <5% will have distant metastases if they are asymptomatic and have a stage I tumour by CT scanning.^2,8,9,21,22 Patients who have a normal mediastinum by CT scanning but evidence of N1 node enlargement need invasive staging (because of a 20–25% incidence of mediastinal node involvement), whereas those with a peripheral stage I tumour have a very low incidence of N2,3 disease (<10%).^2,21 Furthermore, CT tells us how much lung would have to be removed if resection is to be done, and allows treatment planning for radiotherapy.

Perhaps we are not asking the right questions of PET imaging. The goal of the pretreatment evaluation of the patient is to predict the biological behaviour of the tumour and to allow us to choose a treatment strategy that has the best chance of being effective. Many studies have shown that PET intensity correlates with the propensity of tumours to reappear and spread.²³ Studies are beginning to address whether PET intensity can guide the need for invasive mediastinal staging.²⁴ The change in PET intensity a few days after a dose of chemotherapy is an excellent predictor of the response to that regimen.²³ Perhaps the study by Tournoy is telling us that the progress we can achieve through better PET technology and better science is limited if we are asking the same questions. Perhaps our conceptual framework of how we use tests to achieve better outcomes for patients needs to evolve and change.