Imaging in the diagnosis and treatment of non-small cell lung cancer

INTRODUCTION

Despite international efforts to reduce tobacco smoking, lung cancer remains the leading cause of cancer-related death in both men and women in most developed economies.¹ Non-small cell lung cancer (NSCLC) accounts for the majority of all new cases. Diagnosis of NSCLC can result either from evaluation of suspicious symptoms or incidentally subsequent to an imaging finding. Early clinical diagnosis of lung cancer is difficult due to a frequent lack of symptoms. Even when symptoms are present they are not specific for cancer. Accordingly, by the time of clinical diagnosis most patients have loco-regionally advanced disease or systemic metastases. However, these are often undetected by conventional staging procedures and only become evident in the weeks or months after attempts at curative treatment.² Although more sensitive than clinical diagnosis, early radiological diagnosis is also hampered by lack of specificity.³ This means that many unnecessary biopsy procedures are performed to exclude malignancy in lung nodules identified in asymptomatic individuals.

Once there is a firm histopathological diagnosis of NSCLC, staging is critical for treatment selection and prognostic stratification. Although CT scanning with contrast remains the routine staging modality, it has significant limitations. In an attempt to overcome these limitations, an array of other non-invasive and minimally invasive diagnostic procedures has been developed to supplement CT for definition of the extent of disease. Options available for assessing thoracic disease extent include bronchoscopy, which can be combined with endobronchial ultrasound (EBUS), endoscopic ultrasound (EUS) via the oesophageus, and various techniques for accessing and pathologically sampling thoracic structures. Extra-thoracic disease detection techniques include magnetic resonance imaging (MRI), radionuclide bone scanning and ultrasound. PET with the glucose analogue, fluorine-18 fluorodeoxyglucose (FDG) can provide staging of both intrathoracic and extra-thoracic sites. Appropriate selection of which of these modalities should be used and the order in which they should be applied, requires considerable experience and insight into the diagnostic process. Working in a multidisciplinary environment, we have found that the choice of the most appropriate staging investigations is often difficult for clinicians.

We believe that the evaluation of known or suspected NSCLC should be highly individualized and based on Bayesian principles of disease probabilities. The choice of investigation is also critically dependent on the expertise available in a particular healthcare environment and the therapeutic options that are applicable, based on both patient and institutional characteristics. Within this framework, it is, however, important to detail some guiding principles.

APPLICATION OF BAYESIAN PRINCIPLES TO STAGING OF LUNG CANCER

The positive and negative predictive values of all diagnostic procedures are functions of both the intrinsic technical performance of the test and the prevalence of the disease being evaluated.⁴ The negative predictive value of a test for any given site of disease will tend to be lower when patients have advanced disease and the likelihood of additional, small metastases is high. Conversely, the positive predictive value will tend to be compromised in populations with a low likelihood of disease. Consequently, it is important to use tests with a high sensitivity in populations with a high a priori likelihood of disease and tests with high specificity for diseases with a low prevalence of disease.

For lung cancer, the challenge is not only to establish the diagnosis of malignancy, but also to exclude metastatic spread. Because the latter may involve a single cell in any tissue of the body, there is a seemingly insurmountable challenge for all diagnostic techniques with respect to sensitivity. However, most treatment decisions are currently based on the presence of discernible collections of metastatic cells. Because lung cancer has a particular predilection to metastasize and granulomatous diseases are relatively common in the thorax, tests with both high sensitivity and specificity are required for staging. Of the available techniques, it is likely in most clinical scenarios that PET/CT offers the best diagnostic performance characteristics for initial staging. Although currently controversial, we believe that FDG PET/CT could be justified as the initial and sometimes the only staging investigation.

IMPORTANCE OF CLINICAL MANAGEMENT OPTIONS TO DIAGNOSTIC CHOICE

The diagnostic and staging process is directed at informing management decisions. Therefore, the requirements with respect to diagnostic performance of the tests chosen for staging purposes will also vary depending on the therapeutic options available. Each of the available diagnostic staging tools has its own strengths and weaknesses and the aim is to match the strengths with the clinical question. Some specific clinical scenarios can now be considered in these terms.

ASSESSMENT OF SOLITARY PULMONARY NODULES

A solitary pulmonary nodule (SPN) is a radiological diagnosis and is usually applied to lesions measuring less than 3 cm in largest dimension identified incidentally on CXR or CT. For these nodules, the objective is to exclude a malignant basis. Although serial imaging by CXR or CT is frequently used to exclude progressive growth due to malignancy, the opportunity to detect and treat lung cancers at the earliest possible time, and thereby maximize the chance of cure, may be lost by following an observational policy. In addition, such a strategy can generate considerable patient anxiety and may have medico-legal consequences. As a result, many patients undergo aggressive attempts to secure a histopathological diagnosis. The procedures used to obtain this tissue are variably invasive and add considerably to the overall health expenditure.

The probability of malignancy of an SPN can be ascertained imperfectly by a combination of clinical and radiological features.⁵ The most reliable CT features of benign nodules are either a benign pattern of calcification or stability for 2 years on serial imaging.⁶ The negative predictive value for malignancy of these benign CT features is high but is exhibited by only a minority of nodules detected.⁷ Other morphological features of benign and malignant nodules overlap considerably, especially in patients with emphysema,⁸ and require further assessment or biopsy to exclude malignancy.

Problematically, modern multidetector CT scanning detects many more lung nodules than previously appreciated.⁹ Biopsy of all these nodules would add significant economic burden and potential morbidity. Because the majority of such lesions are benign, serial CT scanning has been advocated as a pragmatic compromise by the American College of Chest Physicians.¹⁰ The Fleischner Society published their guidelines for management of small pulmonary nodules in 2005, taking into consideration the size of the nodules, the cancer risk profile of the patient and certain morphological features of the nodules.¹¹ There has also been a recommendation suggesting that longer intervals should be used for follow up of small lung nodules based on statistical considerations related to the reproducibility of calliper measurements for small, as opposed to large lesions.¹² However, these recommendations do not incorporate any consideration regarding the potential biological aggressiveness of tumours and the clinical motivation to detect small but biologically aggressive tumours that have not yet spread.

Given the limitations and overlap of morphologic features in the differentiation of benign and malignant nodules on a single CT study, and the logistical, theoretical and medico-legal issues related to delaying positive diagnoses by requiring serial imaging, there is growing interest in tests that could provide a more rapid but non-invasive SPN characterization. These include more sophisticated radiological techniques using modern CT or MRI technology, such as dynamic contrast-enhanced ‘perfusion’ imaging. An overall accuracy of 92%, sensitivity of 94% and specificity of 90%, for distinguishing benign from malignant nodules using a combined analysis of wash-in and washout characteristics on multidetector CT has been reported recently.¹³ Similar results were also reported by using dynamic contrast enhanced MRI.¹⁴ Perfusion imaging is, however, still a work in progress and limited to academic or research institutions. Furthermore, these techniques generally have a highly restricted survey volume, limiting the capability for assessing potential remote sites of disease.

PET with FDG is now recognized to be a useful technique for characterizing SPN. Although false negatives can occur due to either partial volume effects, related to small tumour size or significant respiratory blurring of activity, or to low intrinsic FDG avidity of some subtypes of malignancy, a negative PET scan allows for a conservative approach.¹⁵ The significance of a positive FDG PET is somewhat dependent on the prevalence of granulomatous lung disease in the population from which the patient is drawn. In most developed countries visually increased FDG uptake compared to mediastinal blood is most often due to lung cancer and warrants biopsy.³ In addition, by potentially detecting previously unrecognized metastatic disease it is possible to select sites for biopsy confirmation, and to select and plan therapy once lung cancer is pathologically confirmed.¹⁶ There is evidence that use of FDG PET is cost-effective in a number of different healthcare environments.^3,17,18 Furthermore, with modern combined PET/CT scanners it would be feasible to perform dynamic contrast CT evaluation after demonstrating FDG avidity in an SPN in order to further characterize the nature of the lesion.¹⁹

In patients in whom an SPN could represent a metastatic lesion, EBUS can help in differentiating lung metastases from primary lung cancer by obtaining a specific biopsy diagnosis and thereby obviating the need for an invasive staging procedure. Similarly, peripheral nodules in patients with very emphysematous lungs with an accompanying high risk of pneumothorax from CT-guided fine needle aspiration (FNA) are amenable to transbronchial biopsy via EBUS.

THE STAGING OF CLINICALLY APPARENT LUNG CANCER

Because lung cancer is often asymptomatic in its early stages, patients presenting clinically usually have relatively advanced loco-regional disease. In these patients, the key diagnostic objective is to evaluate the extent of disease, because this determines the most appropriate form of treatment and also the prognosis for survival. The extent of disease is now classified based on characteristics of the primary tumour (T-stage), regional nodes (N-stage) and distant metastases (M-stage). The general strategy utilized for staging cancer has been to first characterize the primary tumour and subsequently to evaluate for regional nodal and distant metastatic disease. However, in lung cancer, a critical factor in broadly defining treatment options is the exclusion of distant metastases. In patients with symptoms or biochemical abnormalities suggesting metastatic disease, contrast CT, MRI, ultrasound or bone scanning may be the simplest and most cost-effective means of confirmation. However, with its ability to conveniently survey the entire body and the high contrast generally observed between tumour deposits and normal tissues, FDG PET is capable of detecting disease in adrenal glands, liver and other organs that may appear normal on conventional staging modalities. Existing evidence strongly indicates that PET is the most accurate non-invasive technique currently available for detection of distant metastases in lung cancer. We have demonstrated that consistent with Bayesian principles, the rate of PET-detected metastasis increases significantly with increasing pre-PET stage²⁰ and is particularly evident in patients with locally advanced disease being considered for radiotherapy.²¹ Our centre has also previously shown that PET evidence of distant metastasis, even if unsupported by other evidence, is powerfully associated with subsequent progression of metastatic disease and death.²² Furthermore, in a recent study, the use of dual modality PET/CT staging was shown to be more accurate for detection of distant metastasis in NSCLC than either PET or CT as single modalities.²³ Accurate localization of FDG avid regions on fused PET/CT images reduces the risk of false-positive interpretations of physiological phenomena such as uptake in bowel or metabolically active brown fat. Accordingly, previous estimates of the impact of PET on the management of lung cancer are likely to be surpassed in current clinical practice. A further advantage of PET/CT is to select the most convenient, accessible site for biopsy confirmation of systemic metastases and may even obviate the need for invasive procedures such as mediastinoscopy.

In the absence of systemic metastases, the status of mediastinal lymph nodes becomes the next critical factor in therapeutic selection. Survival in NSCLC is also significantly correlated with lymph node stage and drops precipitously when mediastinal nodes contain tumour. CT has long been the standard non-invasive method for detecting intrathoracic lymph node metastasis in NSCLC. Lymph node size is usually used for characterizing nodes as being likely benign or malignant. The accuracy of FDG PET in staging intrathoracic lymph nodes has been directly estimated in numerous clinico-pathological studies. In all of these studies, including a meta-analysis,²⁴ PET has been shown to be more accurate than CT for staging the mediastinum. Importantly, the status of lymph node involvement as defined by PET appears to be prognostically significant. It has been reported that survival is more strongly correlated with PET-stage than CT-based stage in a large group of patients, who were mostly surgical candidates.²⁵ We have also shown a strong association between PET stage and survival in cohorts of patients including both surgical and radical radiotherapy candidates.²² The best non-invasive results have been obtained by correlating the results of both PET and CT images and combined PET/CT scanners are likely to become the standard non-invasive tool for mediastinal staging. Preliminary results support this view.²⁶ Because FDG PET/CT is clearly superior to diagnostic CT for staging both M and N stage, one could make a case for using this as the primary staging tool in patients with a high likelihood of lung cancer based on clinical presentation (Fig. 1).

USE OF EUS TO ASSESS AND BIOPSY MEDIASTINAL LYMPH NODES

When no definite metastatic disease is apparent on either CT scan or PET scan, EUS can be used to further stage the mediastinum. EUS is a combination of endoscopy and ultra-sonography. EUS is performed with echo-endoscopes, which are endoscopes with a small ultrasound transducer mounted at the tip. A linear echo-endoscope can be used for tissue-sampling using FNA of para-oesophageal lymph nodes with the FNA needle visualized in real time. Consequently, one is able to obtain samples not only from large lymph nodes, but also small lymph nodes, as well as lung masses. The main prerequisite is that the targeted lesion is in close proximity to the oesophagus.

Endoscopic ultrasound was initially used for FNA of mediastinal lesions/lymph nodes in the early 1990s.^27–31 Endoscopic approaches do not have a significant role in staging patients for remote metastatic disease metastasis, apart from some cases where an adrenal metastasis is suspected,^32–34 and therefore should be used primarily to exclude mediastinal involvement in patients in patients with negative or equivocal evidence of metastatic disease on FDG PET, or to pathologically confirm mediastinal involvement suspected on other imaging tests when this is required for treatment selection and planning. EUS can sample all lymph nodes next to the oesophagus, including posterior mediastinal, subcarinal, aorto-pulmonary window and lower para-oesophageal stations, but not stations anterior to the trachea. If EUS with FNA confirms mediastinal lymph node involvement, then mediastinoscopy or mediastinotomy can be avoided.³⁵ Most reports describe an accuracy of around 80–95%, including an Australian study of more than 50 patients with an overall sensitivity, specificity and accuracy were 93%, 100% and 95%.³⁶ Another major advantage of EUS is the ability to image and sample lymph nodes down to the size of 4–5 mm. This is important because it has been shown that 70% of mediastinal lymph node with tumour infiltration can be normal in size.³⁷ Studies have shown 12–25% of the patients with normal size mediastinal lymph nodes have malignant involvement on EUS-FNA.^38,39 There are encouraging data using EUS to detect micrometastatic involvement of lymph nodes.⁴⁰ Although the sensitivity of PET/CT for detection of disease in normal sized nodes is a clear advantage compared to CT, partial volume effects can limit its sensitivity for small volume metastases. Therefore, EUS may more definitively evaluate patients with negative or equivocal uptake in mediastinal nodes, particularly if the risk of nodal involvement is increased by the presence of a large primary. Like any other investigations, one of the shortcomings of EUS is that the negative predictive value is not ideal. It is reported to be around 77% compared to around 91% for mediastinoscopy.⁴¹ It should be noted that EUS-FNA has also been shown to be accurate in diagnosing inflammatory conditions.^42,43

While EUS has become widely used for staging both gastrointestinal and lung malignancies, the application of ultrasound technology to the bronchial tree has lagged behind. Although the first reports of EBUS were published in the early 1990s,⁴⁴ the technology is only recently gaining acceptance, and even then, in only some centres. The reasons for this are multiple. Technological limitations may have been a factor as flexible bronchoscopes are narrower in diameter than gastrointestinal endoscopes and require smaller ultrasound probes, which have only recently become widely available. Regardless of the past impediments, EBUS certainly has a role to play in the diagnosis and staging of lung cancer by allowing histological sampling of mediastinal lymph nodes not otherwise accessible by EUS or mediastinoscopy with a diagnostic sensitivity of 92% and specificity of 100%.⁴⁵ We believe that the utility of EUS and EBUS will be further enhanced by more routine use of combined PET/CT allowing anatomically accurate localization of nodal stations requiring biopsy. This might involve nodes with increased uptake, which are thought likely to be granulomatous based on the pattern of abnormality, or nodes that are highly suspicious of malignancy, which, if confirmed pathologically, would alter treatment choice. Alternatively it may be used to exclude small volume disease that is not apparent on PET but might be pertinent to treatment choice. There are several scenarios where this may be of clinical importance. These include the assessment of post treatment response following neo-adjuvant chemotherapy and radiation for stage IIIA NSCLC, particularly in the setting of prior mediastinoscopy or radiation, which render mediastinoscopy more complicated.

THE STAGING OF EARLY STAGE LUNG CANCER

Only when macroscopic systemic and nodal metastatic disease has been excluded with a reasonably high degree of certainty does detailed delineation of the anatomical relations of the primary tumour (T-stage) become important to therapeutic planning. Only with the very earliest detection and confirmation of NSCLC is the likelihood of metastatic disease sufficiently low to obviate staging of the draining hilar nodes as a minimum.⁴⁶ Because PET has limited spatial resolution and relatively poor definition of anatomical planes, FDG PET has a limited role in T-stage definition. However, in situations where the primary tumour abuts soft tissue of similar CT density to the primary, PET may provide better contrast resolution between tumoral and normal tissue. Examples include pleurally based tumours that are potentially invading the chest wall, and central tumours abutting the mediastinum. However, for practical purposes, CT and MRI remain the best methods for non-invasively determining T-stage in most patients, but both are clearly inferior to operative and pathological evaluation. In initial diagnosis EBUS has also been used to assess depth of invasion of the tracheo-bronchial tree in NSCLC. When combined with transbronchial needle aspiration (EBUS-TBNA) or transbronchial biopsy (EBUS-TBB), histological assessment of both the primary tumour and nodal stations becomes possible. The sensitivity, specificity and accuracy of EBUS with biopsy for peripheral lung cancers has been reported as 63–79%, 100% and 77–85%, respectively.^47,48 EBUS with biopsy is also uniquely suited to diagnosis of a proximal or hilar mass that does not have an endobronchial component. In addition to the sampling of mediastinal lymph nodes or masses, EUS can also directly sample accessible parenchymal lung masses with good success.⁴⁹ EUS and EBUS can thus complement or replace thoracoscopy and mediastinoscopy for staging primary tumours and lymph node stations that are otherwise difficult to access. Again, FDG PET/CT can potentially guide the endoscopist regarding the likely location and accessibility of lesions.

IMPACT OF NON-INVASIVE STAGING ON PATIENT MANAGEMENT

There are multiple therapeutic options for the treatment of lung cancer, the choice of which is determined by the extent of disease. Because aggressive treatments aimed at improving the still disappointing cure rates come at considerable cost both to the healthcare economy and in terms of patient morbidity, their appropriate use relies on accurate definition of tumoral extent, particularly to avoid futile treatment. By demonstrating clinically occult disease, imaging has the potential to alter patient management.

In a prospective trial evaluating the impact of FDG PET on patient management that used a methodology of comparing a prospectively assigned treatment plan based on conventional staging information with that actually performed after PET information became available, we demonstrated that FDG PET substantially alters the management of patients with known or suspected lung cancer.⁵⁰ A randomized trial performed in the Netherlands evaluated the ability of PET to alter treatment in patients planned for surgery of lung cancer after conventional staging. In this study, the group that was randomized to have PET had a significant reduction in the ‘futile thoracotomy rate’⁵¹ and appeared to be highly cost-effective.⁵² Although the conclusions of a subsequent randomized trial performed in Australia⁵³ appear to be at odds with the Dutch study, the Australian study was highly biased by selection of patients with very early stage disease (92% had stage 1). Furthermore, review of the raw Australian data indicates that had the PET information been incorporated into decision-making, around 25% of patients could appropriately have had their management changed to avoid thoracotomy.

In patients being planned for radical chemoradiation, our group has demonstrated that FDG PET has a high impact on treatment selection and delivery, and more powerfully stratifies patient prognosis than conventional staging.⁵⁴ In our experience, approximately 30% of patients are spared futile radical radiotherapy because of either, detection of unexpected distant metastasis by PET, or of PET-detected intrathoracic disease that is too extensive for radical irradiation to be safely performed. A comparison of two prospective cohorts staged with and without PET found that the median estimated survival was almost double in PET staged patients, despite being well matched based on conventional prognostic indicators including conventional stage, Eastern Cooperative Oncology Group status and weight loss.⁵⁵ Preliminary data also indicate that the new combined PET/CT scanner may provide the most efficient and accurate means of integrating structural and molecular information into the staging process while facilitating more accurate treatment-planning.⁵⁶ Hence, the impact of this technology on clinical management may even be higher.

Endoscopic ultrasound has also been shown to change management in a number of studies,^57–60 mainly by avoiding mediastinoscopy. However, stage migration may also change the modality of treatment. In addition, because the procedure is performed in a day procedure setting, cost-effectiveness appears to be superior to that of mediastinoscopy.^61–65

CONCLUSIONS

FDG PET and, particularly, PET/CT provide significantly superior staging information than conventional methods in NSCLC. By providing more accurate information on the extent of disease, it is possible to identify patients with extensive, and therefore currently incurable, disease and spare them from futile and potentially toxic radical treatments.

More accurate staging of the mediastinum becomes a diagnostic goal in patients without systemic metastases on PET or other techniques. EUS or EBUS with ultrasound-guided biopsy offer minimally invasive approaches to several diagnostically challenging situations that would otherwise require invasive staging. These are generally safe procedures and in experienced hands have excellent diagnostic accuracy. Rather than competing with PET, we believe that they provide a complementary role and can best be deployed in the context of prior PET staging. In particular, PET/CT can be used to select the most appropriate nodal stations to sample with reference to the available treatment options and will often indicate the most appropriate tissue sampling technique to perform.

With the availability of a wide range of imaging techniques, few patients should reach the point of definitive treatment without clear definition of the extent of disease.