© 2002 American Thoracic Society
Lung Cancer—Where Are We Today?Current Advances in Staging and Nonsurgical TreatmentDepartment of Respiratory Medicine, University College, London Hospitals National Health Service Trust, London, United Kingdom Correspondence and requests for reprints should be addressed to S. G. Spiro, M.D., F.R.C.P., Department of Thoracic Medicine, The Middlesex Hospital, Mortimer Street, London W1N 8AA, UK. E-mail: stephen.spiro{at}uclh.org
Lung cancer remains the commonest cause of cancer death in both men and women in the developed world, although mortality rates for men are dropping. Spiral computed tomography (CT) of the chest in middle-aged, smoking subjects may identify two to four times more lung cancers than a chest X-ray, with more than 70% of tumors being Stage I. The incidence of benign nodules is high, making interpretation difficult. Randomized controlled trials are required to determine whether spiral CT detects lung cancer early enough to improve mortality. Preoperative staging has relied on CT scans, but positron emission tomography scanning has greater sensitivity, specificity, and accuracy than CT and is recommended as the final confirmatory investigation when the CT shows resectable disease. In locally advanced non–small cell lung cancer, there is a small advantage for the addition of chemotherapy to radiotherapy, but no advantage for postoperative radiotherapy. Chemotherapy gives no benefit when given as neoadjuvant or adjuvant treatment around surgery. In advanced disease, newer cytotoxic agents confer a small survival advantage over older combinations, but the advantage in median survival over best supportive care remains a few months with modest improvements in quality of life. Survival with small cell lung cancer has shown little increase over the last 15 years despite multiple attempts to manipulate the timing, dose intensity of chemotherapy, and the potential of radiotherapy. Novel therapies are urgently needed for all cell types of lung cancer.
Key Words: chemotherapy • lung cancer • radiotherapy • screening • staging
Abstract Epidemiology Screening Chest X-ray and Sputum Cytology Spiral Computed Tomography Biological Screening Tools Staging Tests: an Update Who Sees the Patient? Computerized Tomography of the Chest Magnetic Resonance Imaging Positron Emission Tomography Endoscopic Biopsy Techniques The Search for Extrathoracic Metastasis Refining of the Staging Classification in an Attempt to Increase Resectability Advances in Radiotherapy in Non–Small Cell Lung Cancer Radical Radiotherapy for Stage I and II Disease Postoperative Radiotherapy Radical Radiotherapy for Stage IIIA and IIIB Disease Palliative Radiotherapy Interventional Bronchoscopy and Brachytherapy Chemotherapy for Non–Small Cell Lung Cancer Neoadjuvant Chemotherapy Adjuvant Chemotherapy and Surgery Chemotherapy and Radiotherapy in Locally Advanced Disease Chemotherapy in Advanced Disease Newer Chemotherapy Combinations Small Cell Lung Cancer Chemotherapy Treatment of Limited Disease Extensive Disease Elderly Patients Prophylactic Cranial Irradiation Detection of Early Lung Cancer and Photodynamic Therapy The Future Conclusion Lung cancer is one of the most important diseases in respiratory medicine. Worldwide, it is the commonest cancer in men, virtually the commonest in women, and has a greater total incidence than that of colorectal, cervical, and breast cancer combined. In 2001, lung cancer will have caused more than 1 million deaths worldwide and this global incidence is rising at 0.5% per annum. The etiology of the great majority of lung cancers has been known for nearly 50 years (1), but we have failed to make serious inroads into the powerbase of the tobacco industry. In the Western world, although the incidence of lung cancer in men has fallen over the last 20 years, a similar decline among female smokers is not yet evident, and adolescents are smoking in increasing numbers. Global cigarette sales are rising steadily with the ruthless pursuit of new conquests by the tobacco industry in Asia, China, and South America, some of the poorest countries in the world that cannot afford the cost of tobacco-related diseases. In China in 1998, one in four smokers died from tobacco-related causes, and 0.6 million deaths in 1990 were tobacco related, a figure that rose to 0.8 million in 2000 (2). Lung cancer is a disease for which there is no established screening, which presents late in its course, and has a median survival of 6–12 months from the time of diagnosis with an overall 5-year survival of 5–10%; and yet the major cause of this disease is clearly understood. Communities and countries addressing a smoking ban would probably achieve far more in the long term than we currently are with all our available treatments. Although surgery offers the best chance of cure in lung cancer, particularly in the case of non–small cell lung cancer, only a small proportion of patients are ever suitable for curative resection and the majority must rely on nonsurgical and adjuvant therapies. This review focuses on the role of chemotherapy and radiotherapy in both small cell lung cancer (SCLC) and non–small cell lung cancer (NSCLC). In addition, the potential role of screening for lung cancer and advances in staging tests are discussed.
At the end of the twentieth century, lung cancer had become one of the world's leading causes of preventable deaths. By 1950, case-control epidemiologic studies showed that cigarettes were strongly associated with the risk of lung cancer (3, 4). In 1962, the Royal College of Physicians in London intervened in a public health matter for the first time since 1725 and published a compelling document supporting the evidence that smoking caused lung cancer (5). Worldwide it is estimated that 47–52% of men and 10–12% of women smoke. Compared with women, men started smoking younger, smoked more and for a longer duration, inhaled more deeply, and bought cigarettes with a higher tar content (6, 7). Women took up smoking in the United States and Western Europe during the second World War. Recent case-control studies have shown female smokers to have a higher relative risk of lung cancer than males, after adjusting for age and average daily consumption. The incidence of lung cancer shows marked geographical variation, and is most common in developed countries and less so in developing countries, for example, those of Africa and South America (8). The low rates in these countries will inevitably rise to match those of the developed world. Within countries, lung cancer incidence among men considerably exceeds that of women, but the highest rates occur in the same regions for both sexes (Table 1) . Only 5–10% of all lung cancers are diagnosed in patients under the age of 50 years, with adenocarcinoma and a positive family history being common in these cases.
The mortality rates for lung cancer closely parallel the incidence rates because of poor survival. Age-adjusted mortality rates increase exponentially until the age of 80 years in men and 70 years in women and then decline. In the United States, lung cancer accounts for 28% of all cancer deaths each year. Whereas it was responsible for 3% of all female cancer deaths in 1950, it accounted for 24% in 1995. The age-adjusted lung cancer death rate passed that of breast cancer among white women in the United States in 1986, and among black women in 1990 (9). However, there is mounting evidence that at least in the developed world death rates from lung cancer may have peaked. Between 1973 and 1994, the incidence of lung cancer in the United States for those over 65 years of age increased by 220% in women, but fell by 18% in men (10). For those under 65 years of age, the incidence of lung cancer increased by 58% in women and fell by 16% in men over the same period (10), and for those younger than 45 years old, age-adjusted incidence and mortality rates from lung cancer fell in both sexes (more so for men) with a projection that, in the United States, overall incidence of and mortality from the disease may begin a decline for both sexes at the beginning of the new millennium (11). Although lung cancer incidence has fallen in the United States, it remains the leading cause of cancer deaths worldwide, with a global incidence that continues to rise. There is also concern in the United States that the incidence of the disease may start to increase again as a result of increasing tobacco consumption (12). In addition, shifts have occurred in the incidence rates of the different histologic subtypes of lung cancer, with adenocarcinoma surpassing squamous cell tumors as the most frequent type in both white and black Americans.
Chest X-Ray and Sputum Cytology There has been much interest in the idea of screening to detect presymptomatic lung cancers, when presumably they would be at an earlier and more curable stage of their growth. In the 1970s there was a major effort, using chest X-ray and sputum cytology, to assess the prevalence of tumors and demonstrate that early detection would enhance survival and ultimately decrease mortality. The Mayo Lung Project (13), the Czechoslovakian Screening Study (14), and similar trials at Johns Hopkins Hospital (15) and Memorial Sloan-Kettering Hospital (16) all enrolled male smokers and variously compared annual or more frequent chest X-rays with or without additional sputum cytologic evaluation against a control group who had an initial or annual chest X-ray only. All these studies identified more tumors in the screened than control groups. The tumors were smaller, of a lower (more favorable) stage, and the resection rate and 5-year survival rates were better. However, overall mortality was not improved. Although all these were randomized controlled trials, only the Mayo Lung Project had a true control group that was unscreened. However, this study lacked power from the outset, with less than 20% power to detect a 10% benefit in lung cancer mortality and a 55% power to detect a 20% benefit. Moreover, there was further contamination as 55% of the control group had a chest X-ray in the previous year and 73% had a chest X-ray during the last 2 years. Compliance was also a significant problem. Interestingly, the screening seemed rather ineffectual, as the incidence of new tumors provided 206 new cancers, of which only 45 (22%) were resectable, compared with 60% of the prevalence tumors at baseline. One of the problems with these studies may have been the choice of mortality from lung cancer as the end point. It has been argued that all-cause mortality (17) or survival from lung cancer may be less biased end points (18). Indeed, Strauss has performed an important and impressive reanalysis of the Mayo Lung Cohort data. This analysis has shown that although the incidence of lung cancer was higher in the screened group, the survival of patients with lung cancer was also much higher in this group when compared with the control group; and this increased survival was directly related to tumor resection and therefore not due to overdiagnosis of "pseudo-disease" (18). Strauss argues that the randomization procedure for the Mayo Lung Project was suboptimal, because of unidentified confounding variables (18). So that although we understand a certain amount about the etiology of lung cancer, we still cannot accurately distinguish those 16% of male and 9% of female life-long smokers who will develop the disease from their fellow smokers who will not. Strauss finally lays to rest the years of debate around the Mayo Lung Project and explains the findings without having to resort to the counterintuitive concept of overdiagnosis; screening is worth doing because more resectable cases are picked up and more patients are cured (18). Another problem highlighted from the Mayo study is that of identifying early tumors on the chest X-ray, as 90% of peripheral and 75% of perihilar tumors were visible in retrospect on previous films (19). Quekel and coworkers more recently also reported a 19% miss rate of peripheral tumors, with an average size of 16 mm (20).
Spiral Computed Tomography
These studies are all hypothesis generating, but it is too early to know whether detecting tumors that are in general smaller than when discovered on a chest radiograph will decrease mortality. All these studies have the in-built problems of lead time bias, length time bias, and overdiagnosis bias (27). Only large randomized controlled trials (RCTs) with a follow-up of 10 years or more and rigorous use of all-cause mortality as an end point (17) will answer this fundamental question. In addition to the prevalence data now available, the incidence data from the current uncontrolled studies will give valuable information as to how many of the smaller nodules (less than 10 mm in diameter) were, in fact, tumors and not identified as such during the initial CT screen. Although identifying nodules 10 mm in size or smaller gives a yield of cancers smaller in size than those discovered by conventional chest X-rays, these tumors will have undergone 25 to 30 volume doublings and will have a considerable propensity to form metastases (28). Furthermore, there are data to suggest that the relationship between tumor size, survival, and stage at presentation is not clear cut. One study of 510 patients found no statistical relationship between tumors of less than 3 cm and survival; patients with 3-cm masses had the same outcome as those with 1-cm nodules (29). In a related study of 620 patients there was no relationship between size of the primary tumor and stage at presentation. Patients with a 1-cm tumor had a similar stage distribution as those with 2- to 3-cm masses (30). Thus the biological behavior of tumors is variable and a fundamental part of the issue of long-term outcome. There is growing pressure to include low-dose helical CT in the armamentarium directed at finding lung cancer for good emotive (31) but not yet evidence-based reasons. Much work still needs to be done. The larger prevalence studies have been performed in countries where peripheral adenocarcinomas are commoner—the United States and Japan. This appears not to be the case in Europe and care must be taken when advocating a technique such as this more widely. The choice of population to screen will have a major effect on the prevalence of tumors found, as already clearly demonstrated in the data accumulated so far. Age, sex, smoking history, and the presence of airway obstruction are the major risk factors for the development of lung cancer. The issue of false-positive scans will need to be addressed. In the Japanese and Lung Cancer Action Project studies there were large numbers of subjects with noncalcified pulmonary nodules: 233 of the 1,000 in the Lung Cancer Action Project (24) and 66% in the Mayo study (26). The anxiety generated, the potential for overinvestigation, and the radiologic exposure these individuals receive suggest a need for further thought. It is also worth noting that in clinical practice, most lung cancers occur centrally and are diagnosed by bronchoscopy. These tumors hardly feature in the CT screening studies; only two central tumors were discovered in the Lung Cancer Action Project screen (24) and one in the study by Sone and coworkers, and that tumor was found by cytology, not CT (23). It would appear that central lung cancers are too aggressive to remain occult and produce symptoms leading to diagnosis before or between screening tests because of their situation in major airways. They behave in an entirely different way than intrapulmonary "nodule" or peripheral cancer. What of the cost in terms of machine time, scan interpretation, and resultant action? Attempts have been made to analyze the cost-effectiveness of screening for lung cancer, but such models make enormous assumptions and are probably premature (32). A proper screening program will require dedicated CT scanners, which may need to be mobile. In many countries there is already an unacceptable waiting time for staging CTs in patients known or suspected to have lung cancer, and the additional burden of screening is not possible. Once hundreds of scans are generated by screening, radiographers will have to be trained to report them and show only abnormal scans to radiologists for practical time reasons. Inevitably, further high-resolution CT scans will have to be performed on subjects with abnormalities and many will then require biopsies. Many will also need regular follow-up high-resolution scans for several years. The costs and logistics and possible long-term effects of the investigative irradiation are considerable. There is therefore no sensible alternative to embarking on carefully constructed RCTs in defined populations of sufficient numbers, members of which are followed for long enough to provide a clear answer about the potential of CT screening. Additional problems will occur if the technology of imaging moves ahead so fast that improvements will have to be incorporated into these prospective studies. For example, three-dimensional volumetric analysis of a nodule is already available and is more sensitive for showing size change than simple CT (33). Finally, will any control population accept an annual chest X-ray, or perhaps no screening chest X-ray, while being deprived of a chance for CT screening?
Biological Screening Tools Mao and coworkers (39) looked at early chromosomal and genetic alterations in lung epithelial cells and found that point mutations in the p53 and K-ras genes in sputum samples preceded the clinical diagnosis of lung cancer in one case by more than 1 year. Other groups have identified areas of genomic instability that cause microsatellite alterations that can act as clonal markers of early malignant disease (40).
It is beyond the remit of a single review to comprehensively summarize the current lung cancer staging literature, but newer techniques are becoming available and these, together with basic assessment of the patient, are discussed.
Who Sees the Patient? A similar review of the referral and treatment practice in a city in Yorkshire, England also found that almost half of patients with newly diagnosed lung cancer were not sent to a respiratory physician, and the treatment rates for surgery, radiotherapy, and chemotherapy for those patients were approximately half the rates for patients seen by a respiratory physician (42). Both studies reinforced the UK National Cancer Plan to identify a respiratory physician with an interest in lung cancer in every hospital to organize the care for patients with newly diagnosed lung cancer. It is probable that in an aging population referral patterns in other countries will be similar to those in the UK, with no exclusive referral pattern to a respiratory physician.
Computerized Tomography of the Chest Another study (47) suggests there are advantages if CT precedes bronchoscopy and the information from CT is used by the bronchoscopist. Costs were not greater, as the number of invasive tests was reduced. Of 171 patients suspected of having endobronchial cancer, 90 had a CT performed and reviewed before bronchoscopy. Six needed no further investigation because the CT was either normal, or consistent with benign disease or with widespread metastatic disease. Of the remainder, fiberoptic bronchoscopy was diagnostic in 50 of 68 (73%) compared with 44 of 81 patients (58%) who had a bronchoscopy first. Overall, a positive diagnosis was made after a single invasive test in 76% of the group having a CT first, and in 54% of the group that underwent bronchoscopy first. Only 7 of the CT-first group needed more than one invasive investigation, compared with 15 patients (18%) of the fiberoptic bronchoscopy-first group. The additional cost of a spiral CT in each patient was offset by the need for fewer invasive tests, even though they were more expensive. Because the majority of patients with lung cancer have a CT during their workup, it may be best done before fiberoptic bronchoscopy. The spiral CT, using a special staging technique, is the mainstay of staging in lung cancer. This involves an automated bolus injection of contrast 20–30 seconds before the scanning is initiated. This time interval allows optimal enhancement of the mediastinal blood vessels. A maximum slice thickness of 5 mm is used to prevent errors from partial volume effects. The new multislice CT systems allow the whole thorax to be scanned with 3-mm slices during a single breath hold. Despite advances in CT scanning technology, there remain important limitations for its use in staging, with preoperative predictions differing from operative staging in 35–45% of cases, with patients being both over- and understaged (48, 49). CT staging remains unsatisfactory for detecting hilar (N1) and mediastinal (N2 and N3) lymph node metastases, and for chest wall involvement (T3) or mediastinal invasion (T4), in which sensitivity and specificity can be less than 65% (50–53). These are critical areas that may make the difference between surgical and nonsurgical management decisions. One development has been single-photon emission CT in which technetium-99m-labeled tetrofosmin is taken up by lung cancers. In one study of 34 patients with lung cancer, CT when combined with single-photon emission CT gave a sensitivity of 94.7% and a specificity of 93.3% for the detection of mediastinal metastases; these levels of sensitivity and specificity were greater than those achieved with either technique alone (54). Dynamic expiratory CT scanning can be used to assess chest wall and mediastinal fixation by showing decreased mobility of the fixed tumor (55). Ultrasound may also be useful for chest wall assessment. In a series of 120 patients with contiguity between the tumor and the chest wall at CT but no definitive invasion (as diagnosed by bony erosion), 19 patients were judged to have invasive tumor on ultrasound with a sensitivity and specificity of 100 and 98%, respectively, when compared with operative findings (56). In the 1990s, many studies compared CT findings with the gold standard of mediastinoscopy or surgery. They showed that, regardless of the threshold size of lymph node chosen, CT findings in isolation could not be taken as clear evidence of malignant nodal involvement and about 20% of all nodes deemed malignant on CT criteria will be benign. Size alone cannot be an exclusion criterion and the clinician needs to prove by biopsy or resection that a node is indeed malignant. CT, however, continues to play an important and necessary part in the evaluation of patients with lung cancer, and its use is supported by the most recent American Thoracic Society/European Respiratory Society statement on pretreatment evaluation in NSCLC, in which CT is recommended for evaluation of mediastinal nodes in all patients with suspected NSCLC (57).
Magnetic Resonance Imaging
Positron Emission Tomography As well as having a role in the evaluation of parenchymal nodules, PET is also valuable for evaluation of the mediastinum. However, image resolution of the current PET scanners is only 4–8 mm and requires complementary CT. The precise anatomical information from CT adds to the metabolic map of PET and helps distinguish, for example, N1 from N2 disease and central tumors from enlarged lymph nodes. A total of 29 published studies that examined the suitability of PET for the staging of NSCLC were reviewed by Laking and Price (69). A meta-analysis confirmed that PET is significantly more accurate than CT for detection of nodal mediastinal metastases, with a sensitivity and specificity of 79 and 91%, respectively, for PET versus 60 and 77%, respectively, for CT (70). The usefulness of the extra information gained from PET is itself dependent on the initial CT scan, so that PET has a sensitivity and specificity of 74 and 96%, respectively, for detecting metastasis in normal-sized mediastinal lymph nodes compared with 95 and 76%, respectively, when these lymph nodes are enlarged (71). It is important to remember this when drawing up clinical protocols or considering individual patients. False-positive mediastinal nodal scans occur in sarcoid and tuberculosis and other infections. Is PET sensitive and specific enough to replace mediastinoscopy and lymph node sampling before thoracotomy and prevent futile operations without denying surgery to appropriate candidates? Several studies have addressed this question. In one study of 100 patients, PET accurately staged NSCLC in 83% of cases compared with 65% by conventional imaging (thoracic CT, bone scintigraphy, and brain CT or MRI). PET identified 9 patients with metastases that were missed on conventional imaging whereas 10 patients thought to have metastases were shown not to by PET. PET was more sensitive than conventional imaging for bone, and adrenal metastases, but is inappropriate for the detection of brain metastases because of the high glucose uptake of the normal brain. The negative predictive value of PET for N2 disease was 96%, similar to that of mediastinoscopy, suggesting that patients with negative mediastinal PET could go straight to surgical resection of the primary tumor (72). In a comparison of PET with CT against the gold standard of mediastinal lymph node dissection in 102 patients with resectable NSLC (73), results were complicated by the high sensitivity (75%) and low specificity (66%) of CT scanning for detection of mediastinal metastases, but only PET results (91% sensitive and 86% specific) correlated with the histopathology of the mediastinal lymph nodes. PET altered the stage determined by conventional imaging in 62 patients (42 were upstaged and 20 were downstaged). However, PET was still wrong in 13 cases (conventional imaging was wrong in 32) and surgical staging was required for a definitive result (73). This emphasizes that no one with a positive PET scan should be denied surgery without positive histology (71). PET was actually of greatest value in 11 patients in whom distant metastases were found. However, in nine patients PET was falsely positive for distant metastases. In another study, treatment plans based on conventional staging were compared with those based on incorporation of PET. PET changed management for 40 of 153 patients (34 patients had their treatment changed from curative to palliative and 6 patients had their treatment changed from palliative to curative) and gave more accurate prognosis of individual patients (74). More recently, an attempt has been made by a group in The Netherlands to see whether patients actually benefit if PET is incorporated into the workup, and to address this question treatment plans based on conventional staging were compared with those based on incorporation of PET. In this PLUS (PET in Lung Cancer Staging) study 188 patients with NSCLC from 9 participating hospitals were randomized to conventional workup (CWU) or CWU plus PET. The primary outcome was the ability of PET to minimize futile thoracotomies. Eighteen patients in the CWU group and 32 in the CWU plus PET group did not have a thoracotomy. In the former group, 41% had a futile operation, as opposed to only 21% in the PET group (p < 0.003). Importantly, there was no decrease in justified surgery due to PET. Assessment of resectability by CT and PET was discordant in one-third of the cases, and PET was correct in two-thirds. PET was superior to CT in identifying the best mediastinoscopy site and in 10 cases only PET suggested the positive biopsy site. Overall one futile thoracotomy was avoided for every five PET scans (75). Who should have a PET scan? Despite the expense of PET scanning and its limited availability, cost–benefit analyses of published data, in both the United States (76) and Europe (71), have shown that it is cost-effective to carry out total body PET in patients with a negative mediastinal CT and an apparently resectable tumor as the cost is balanced by a better selection of patients for surgery. Patients with a positive mediastinal CT and no clinical suggestion of metastatic disease should go straight to mediastinoscopy. However, these recommendations remain impractical until there is better access to PET scanners and radiologists to interpret them. Ideally, because of the high negative predictive value, PET scanning should be performed in all those with no evidence of metastatic disease on CT who are considered for surgery; and, failing this, definitely in those preoperative patients with suspicious N2/N3 disease on CT scan.
Endoscopic Biopsy Techniques Another technique that is becoming increasingly important in the sampling of mediastinal, but not hilar, lymph nodes is transesophageal lymph node sampling under endoscopic ultrasound guidance (EUS) (82). This has the added advantage of avoided contamination of lymph node samples with malignant cells from the bronchial tree. EUS is a technique that has been in use for more than 10 years. It makes use of a modified endoscope with an ultrasound transducer at the tip and gives excellent views of the structures that lie adjacent to the gut lumen. EUS from the esophagus gives access to the subcarinal (Level 7), aortopulmonary (Level 5), and posterior (Levels 8 and 9) mediastinum and is able to resolve nodes as small as 3 mm. However, the views of the paratracheal and anterior mediastinal areas are limited by distortion caused by tracheal air. By using curved echo-endoscopes it is possible to perform fine needle aspiration (EUS-FNA) of abnormal subcarinal and aortopulmonary window nodes with negligible risk of infection or bleeding (83, 84). This had a sensitivity of 96% for malignancy in lymph nodes when bronchoscopy had been unhelpful (85, 86). Silvestri and coworkers looked at 27 patients with known or suspected lung cancer who underwent CT scan and EUS-FNA. They showed that EUS-FNA improved the sensitivity of CT scanning and granted access to lymph nodes not reached at mediastinoscopy (79). Wallace and coworkers studied 121 patients with lung cancer, using EUS-FNA of abnormal nodes. Of these, 97 had enlarged mediastinal lymph nodes and EUS-FNA confirmed malignancy in 75 (77%). In addition, 10 of 24 (42%) patients with normal mediastinal lymph nodes on CT had Stage III or IV disease on EUS-FNA (84), suggesting that it might be an even more powerful staging tool than mediastinoscopy (87). Only one study has directly compared mediastinoscopy (upper and anterior mediastinum) with EUS-FNA (subcarinal and posterior mediastinal lesions) and, although there were only a small number of patients, the suggestion is that the two techniques may prove complementary, with different lymph node stations targeted by the two techniques (88). More recently, Larsen and coworkers have looked at the effect of EUS-FNA on the management of 84 patients with mediastinal masses adjacent to the esophagus. Diagnosis was confirmed by thoracotomy, mediastinoscopy, or follow-up over at least 1 year. In 29 patients with known lung cancer who underwent mediastinal staging, EUS-FNA has a specificity of 100%, a sensitivity of 90%, a negative predictive value of 82%, and a positive predictive value of 100%. Similar figures applied for 50 patients with mediastinal masses but no obvious lung primary. The results from EUS-FNA provided a definite diagnosis and obviated the need for 28 mediastinoscopies and 18 thoracotomies. There were no complications from the procedure (89). So, what is the role of EUS-FNA in patient management? Harewood and coworkers have used models based on the medical literature to look at cost minimization in the accurate staging of patients with NSCLC and enlarged (greater than 1 cm) subcarinal lymph nodes on CT scan. The lowest cost workup was by initial EUS-FNA provided that the probability of subcarinal lymph nodes metastases was greater than 24%, assuming a sensitivity for EUS-FNA higher than 76% (90), in keeping with other studies (91). EUS-FNA may prove as valuable as or more so than mediastinoscopy, and ideally is the investigation of choice for diagnostic evaluation of CT-suspicious lymph nodes at Levels 5, 7, 8, and 9. However, because of the requirement for expensive equipment and a skilled endoscopist, EUS-FNA is available only in a small number of institutions. In addition, the role of EUS-FNA in the evaluation of patients with apparently resectable lung cancers and normal mediastinal CT scans is unknown, but there is some evidence that it might identify some of the 10% of patients with N2/N3 disease who are not picked up by CT scan or mediastinoscopy. There may be some advantages over PET scanning, which has a false-positive rate of up to 13%, although the possibility of overstaging with EUS-FNA has not really been addressed.
The Search for Extrathoracic Metastasis The presence of extrathoracic metastatic disease in NSCLC is dependent on the extent of intrathoracic involvement, that is, the worse the primary tumor and nodal involvement, the greater the likelihood of metastatic disease, whereas the incidence of silent metastases in Stage I disease is low (1%) (92, 93). Several studies, including two meta-analyses of the literature, have found distant metastases in only 2.5–5% of patients with potentially operable NSCLC despite normal clinical examination (94–97). The metastases most commonly affected brain, bone, liver, and adrenal glands in that order (95, 98). How best to identify these patients preoperatively and prevent a needless thoracotomy is not clear. The literature is divided, with some studies showing that screening all patients for extrathoracic metastases before thoracotomy is cost-effective (99) and others finding that this was not the case (92). It is now standard to include the adrenals and liver as part of a staging CT of the chest and upper abdomen (100).
Adrenals.
Brain. More recently, a multicenter, prospective randomized trial of 634 patients by the Canadian Oncology Group was designed to finally answer the question concerning whether to search for occult metastases in the asymptomatic patient with a resectable lung tumor and no clinical suggestion of extrathoracic spread (99). Although thoracotomy without recurrence occurred less often in patients who underwent full investigation (bone scintigraphy and CT of the head, thorax, and abdomen) as opposed to limited investigation (CT of the thorax with mediastinoscopy and other investigations as clinically indicated), the survival results were similar (99). In the meantime, we agree with the recommendations of Silvestri, that, before attempted resection, all patients should have a comprehensive clinical examination and even the subtlest of abnormalities should be investigated (108). Asymptomatic patients with Stage I disease should not be investigated further, but a routine search for metastases is recommended in any patient with known or suspected N2 disease (108).
Refining of the Staging Classification in an Attempt to Increase Resectability What were the 1997 modifications and have they made a difference? The modifications included a regrouping of TNM (T, tumor; N, node; M, metastasis) subsets in Stages I, II, and IIIA (Table 2) , some minor changes to the TNM classification to clarify satellite lesions, and recommendations for the classification of mediastinal, hilar, and intrapulmonary lymph nodes, combining the features described by the American Joint Committee on Cancer and the American Thoracic Society.
Stage I has been divided into IA and IB to reflect the different 5-year survivals of 67 and 57% for pathologic stage (pStage) I and pStage II, respectively (109), and as an impetus to focus attention on the need to improve survival of patients with Stage IB disease, among whom the 5-year survival is 46–57%. Stage II includes T1 (IIA) and T2 (IIB) tumors with spread to the peribronchial, lobar, and hilar lymph nodes (N1), and like Stage I patients these patients should be considered for surgery. T3N0M0 was moved from IIIA to IIB to reflect the better prognosis compared with other Stage III disease presentations, and its similar prognosis to that of T2N1M0. Stage IIIA now includes mainly patients with N2 disease. This group remains extremely heterogeneous, with only a small minority considered resectable. In particular, there are differing prognoses for those patients with preoperatively diagnosed N2 disease (5-year survival of 9% for both pathologic [111] and clinically [112] diagnosed disease) as compared with "unforeseen" N2 disease (5-year survival of 24–34% [111, 112]). Stages IIIB and IV, which are rarely considered resectable, are unchanged apart from the clarification of satellite lesions. Those satellite lesions in the same lobe are now T4 (Stage IIIB), and those that are ipsilateral and in a different lobe or contralateral are now M1 (Stage IV). Stage IV includes any patient with distant metastases; however, the demarcation for supraclavicular (N3; Stage IIIB) versus cervical nodal metastases (M1; Stage IV) remains imprecise and if there is any doubt the patient should be assigned the better prognostic stage. Tracheobronchial lymph nodes are designated as intrapulmonary hilar lymph nodes (N1) instead of mediastinal (N2). (However, if a lymph node can be sampled at mediastinoscopy without creating a pneumothorax then it should be designated N2.) An extremely detailed single-institution staging analysis was made of 3,043 patients with primary carcinoma of the lung who underwent thoracotomy between 1961 and 1995. The aim of the study was to see how the new staging system stood up in practice (113). Patients were assigned a clinical stage (cStage) and a pathologic stage (with at least 100 patients in each stage) and were followed up for a mean of 116 months. All patients who underwent complete resections were staged by meticulous mediastinal dissection, rather than by mediastinal lymph node sampling. When survival curves were plotted for each pathologic stage against time, there were significant differences between the survival curves of all pathologic stages except for an overlap between pStage IB and pStage IIA. There was a much smaller difference between the survival curves for each clinical stage, emphasizing the superiority of accurate pathologic staging. The study was presented as an endorsement of the current staging with some recommendations: although patients designated as T3N0M0 had a good prognosis, those with tumors invading the chest wall, superior sulcus, diaphragm, and ribs had a poorer prognosis and should be reclassified as T4; patients with separate tumor nodules in a different lobe had a better prognosis than other patients with M1 disease and should be reclassified. However, Naruke and coworkers do not comment on the heterogeneity of Stage IIIA, which includes patients with N1 disease (5-year survival of 41.8%) and those with both bulky and "unforeseen" N2M0 disease (overall 5-year survival of 19.9%) (113). The survival rates of patients with T1N2 and T3N1 lung cancers are probably higher than those of patients with other subsets of Stage IIIA, but they are only considered as part of the larger groupings T1–2N2M0 and T3N1–2M0 and their better prognosis is masked (113). In two other prospective studies of 586 patients (114) and 2,361 patients (115), there was reasonable correlation between stage groupings and 5-year prognosis but again no significant difference in survival between Stage IB and IIA patients. There was again significant heterogeneity among those assigned to Stage IIIA, with 5-year survival spanning from 25 to 35% in patients with T3N1 disease (similar to Stage IIB survival of 27–33%) to 6–7% in those patients with T3N2 disease (114, 115), leading to the suggestion that T3N1M0 be moved to Stage IIB to reflect the better prognosis of this group (115). Another point from the study by Naruke and coworkers (113) is that we are still failing some patients who have the best chance of a surgical cure: 21% of pStage IA and 29.2% of cStage IA do not live for 5 years. This suggests that anatomical staging is either too inaccurate or is not the whole answer and raises a question concerning whether some patients have particularly aggressive tumors that could be identified preoperatively. One study of 1,020 cases of pStage IA and IB lung cancer showed more prognostic significance if pStage I is divided into 4 groups depending on tumor size (0–2, 2.1–4, 4.1–6, and 6.1–8 cm), but even the group that does best has only a 63% 5-year survival (116). This has raised interest in prospective and retrospective studies to look at molecular genetic markers, growth factors, receptors, and host and tumor factors relating to cell proliferation and angiogenesis. For example, D'Amico and colleagues looked at a series of 408 patients who underwent resection for pStage I NSCLC and found a higher association of recurrence and death with four molecular markers: p53, Factor VIII, Erb-b2, and CD44 (hazard ratio [HR], 1.4–1.68) (117). Molecular staging remains a research tool but undoubtedly will play a greater part in lung cancer management over the next 10–20 years, informing our treatment decisions. Although much progress has been made, further refinement of the classification of lung cancer is inevitable, and this will require the meticulous standardized collection of staging and survival data from individual patients. In particular, the new guidelines will have to incorporate some of the new methods of investigation now available. Probably the greatest advance has been the development of metabolic staging using PET. This provides a move away from the strictly anatomical imaging used to stage lung cancer and promises to refine the selection of patients for resection; a promise that remains to be definitively proven. Another advance may be the molecular staging of lung cancer in an attempt to classify lung cancers not just by cell type, but on the basis of genetic make-up and biochemical behavior to account for the differences in metastatic potential and sensitivity to chemoradiation that different tumors show. How will we measure our success? Good staging should result in better treatment choices and ultimately increased survival, with better quality of life, and a reduction in futile thoracotomies, noncurative operations, and ineffective chemoradiotherapy.
In general, there have been few advances in radiotherapy to improve the survival of patients with lung cancer. There is a probable dose–response effect for radical radiotherapy up to doses toward 70 grays (Gy), and standard doses can provide excellent palliation for some symptoms in patients with advanced disease. Various questions have been raised: can concurrent chemotherapy act synergistically with radiotherapy? Does the manner of dose administration matter (e.g., conventional daily doses versus accelerated regimens of more than one dose a day)? When part of a multimodal treatment, should radiotherapy be given concurrent with chemotherapy or sequentially? Most of these questions surround the use of radiotherapy in locally advanced inoperable Stage IIIA or Stage IIIB disease, but it also has been evaluated as an alternative to surgery and after successful resection (postoperative radiotherapy).
Radical Radiotherapy for Stage I and II Disease Using a complete response (CR) as a prerequisite for long-term survival or cure, the CR rates ranged from 38 to 46% (120–122), but declined with increasing initial tumor size. The best results occur in tumors less than 4 cm in diameter, for which CR rates range from 48 to 52% and local relapse rates are lower, than for larger tumors with CR rates of up to 20% and higher local relapse rates (123, 126, 127). Overall, 5-year survival varies among these studies from 6% (118) to 32% (120). Whereas tumor size and radiation dose appear to be prognostically important variables, there seems to be no effect of age or histologic cell type on survival. It is possible that modern treatment with CT planning and conformal treatment, in which the shape of the radiation beam is molded to the tumor, may produce better results, but there are as yet no data.
Postoperative Radiotherapy A study by Keller and coworkers (131) enrolled 488 patients who underwent resection of Stage II or Stage IIIA disease, and were then randomized to PORT alone (50.4 Gy) or PORT with four cycles of cisplatin and etoposide concurrent with 50.4 Gy. The median survival for patients undergoing PORT alone was 39 months, and the median survival for those receiving chemotherapy and radiotherapy was 38 months, that is, there was no additional advantage for chemotherapy.
Radical Radiotherapy for Stage IIIA and IIIB Disease It is beyond the remit of this review to discuss the basic principles of irradiation, fractionation, and field size. However, prognosis after radical radiotherapy depends on initial tumor size and nodal status, although most studies report an overall 3-year survival of 2–20% and a median survival of 8–12 months (133–136). Better survival is reported for higher doses, for example, the 3-year survival was 6% with 40 Gy, 10% after 50 Gy, and 15% after 60 Gy given in daily fractions of 2 Gy. The higher dose regimens also provided better, more durable local control (137). Dose intensity can also be increased by three-dimensional conformal radiotherapy, which restricts the dose to the tumor while protecting normal tissues (138) and is the result of better imaging technology using CT and MRI (139). These techniques continue to be evaluated. In addition to dose, studies have addressed the question of hyperfractionation (more than one dose per day). Hyperfractionation regimens use two or three fractions per day of 1–1.2 Gy separated by at least a 6-hour interval, while keeping the same total dose as the classic once-daily fraction regimens. This approach was originally investigated by the Radiation Therapy Oncology Group (RTOG). After a large Phase II trial to select the optimal radiation dose, the RTOG performed a three-armed trial: a daily schedule to 60 Gy, 69.6 Gy with two daily fractions each of 1.2 Gy, and a third arm of induction chemotherapy consisting of two cycles of cisplatin and vinblastine followed by 60 Gy in daily fractions for 6 weeks (140, 141). The 3-year survival rates were 6% after 60 Gy, 15% in the induction chemotherapy arm, and 13% after hyperfractionation. The results were not statistically significantly better for induction chemoradiation compared with radiation alone. The CHART regimen (continuous hyperfractionated accelerated radiotherapy) was developed in the UK. Treatment is given three times a day as 1.5-Gy fractions for 12 days including weekends to a total of 54 Gy and was compared with conventional daily radiotherapy to the same total dose (142). Overall, in 563 randomized patients, CHART demonstrated a 9% improvement in survival at 2 years (20 to 29%) and this was even more marked for squamous cell tumors (82% of all cases), for which there was a 34% reduction in the relative risk of death and an absolute improvement in survival at 2 years from 19 to 33%. Although severe dysphagia occurred more commonly in the CHART group, it was transient and manageable. CHART has provided a significant step forward in the methodology of administering radiotherapy, but for logistic reasons, mainly the inability to provide radiographer expertise at weekends, CHARTWEL (CHART with WeekEnd Leave) is being examined with and without adjuvant chemotherapy. American groups utilizing versions of CHART have also eliminated weekend treatments, but deliver three fractions of irradiation per day as HART (hyperfractionated accelerated radiation therapy). An Eastern Cooperative Oncology Group (ECOG) feasibility study using this regimen obtained a median survival of 13 months with acceptable toxicity (143).
Palliative Radiotherapy The most recent MRC palliation study (146) assessed patients with good performance status and compared 2 fractions, each separated by 1 week (total, 17 Gy), with 39 Gy given in 13 fractions. The study confirmed good palliation with a two-fraction regimen, but a small survival advantage for the higher dose regimen (3% at 2 years). This was not confirmed by an RTOG study, which showed no survival difference between 30 Gy in 10 fractions, 40 Gy in 20 fractions, and 40 Gy in 10 fractions (147).
Interventional Bronchoscopy and Brachytherapy Brachytherapy seems to be used after endobronchial tumor clearance, but few studies have evaluated its benefit. One prospective RCT of neodymium:yttrium–aluminum–garnet (Nd:YAG) laser alone versus laser plus brachytherapy reported an additional 7-month symptom-free interval with both treatments and a reduced need for further endoscopic interventions (150). As a palliative tool, brachytherapy seems impressive with hemoptysis being controlled after one treatment in 70% of sufferers, pneumonia in 57% and cough and dyspnea in 30%. The main complication is massive hemoptysis which can occur in up to 9% of cases in larger series (151, 152).
As surgical resection or radical radiotherapy may cure only 10% of all patients with NSCLC, 90% will present with or develop advanced disease and ultimately die from their tumor. Although chemotherapy may be a logical approach, there is virtually no evidence that it can cure NSCLC. There is, however, increasing evidence that it can palliate and prolong life in some patients, but only for a few months. The monetary cost for this extension of life is high and increases with the development of newer, more expensive drugs, which may have some advantages in ease of administration and toxicity but only small gains in terms of prolonging median survival, with more patients alive at 1 year. A newer regimen, such as paclitaxel in advanced NSCLC, costs 12 times that of mitomycin, ifosfamide, and cisplatin (MIC) in the UK. It saves on inpatient administration time and may, in some studies, result in a greater number of patients being alive at 1 year. The cost effectiveness of these regimens is the subject of debate in the UK at present. The other costs of chemotherapy, that is, its toxicity and its potential detriment to quality of life, are even more important questions, the answers to which are only now beginning to emerge. Chemotherapy has been evaluated as neoadjuvant and adjuvant treatment around surgery, neoadjuvant and adjuvant around radiotherapy, and as primary treatment for advanced inoperable disease.
Neoadjuvant Chemotherapy There are only five published RCTs of neoadjuvant chemotherapy versus surgery alone; three were negative (153–155) and two were positive (156, 157) (Table 3) . A potential confounder in all these studies is the inconsistent addition of pre- and/or postoperative radiotherapy for some patients, usually if residual disease remained after resection.
Much of the data for the possible benefit of neoadjuvant chemotherapy is from Phase II trials, often with small numbers, and some trials use chemotherapy alone and others use both chemotherapy and radiotherapy preoperatively. The chemotherapy regimens vary, and the administration of radiotherapy also varies: preoperatively, intraoperatively, postoperatively, or not at all. Some studies accepted clinical staging and others documented only pathologic nodal staging. Of the Phase II studies giving chemotherapy followed by surgery (158–162), four used mitomycin, vinblastine, and cisplatin (MVC); one used cisplatin and 5-fluorouracil, and one used vinblastine and cisplatin. Some of these studies also gave irradiation, usually postoperatively, but details are scanty. Response rates to chemotherapy varied from 51 to 78% and resection rates were high: 51–68%. The overall perioperative mortality ranged from 0 to 17% and median survival ranged from 12 to 20 months. About 25% of all resected patients had a subsequent local relapse. Of interest, in a UK study (162), 45% of potential patients refused to enter such a study design, refusing chemotherapy. No clinical or pathologic complete responses were obtained in those given neoadjuvant treatment. A retrospective review of one institution's experience of the perioperative mortality for patients undergoing neoadjuvant chemotherapy and surgery (n = 76) compared with surgery alone (n = 259 patients) found no differences for mortality or morbidity based on clinical stage, postoperative stage, or extent of resection between these two groups of patients, which is reassuring (163). There are other Phase II studies in which neoadjuvant radiotherapy was given with chemotherapy (164–167). In essence, the resection rates after the combination treatment were a little higher than for chemotherapy alone (52–76%), but the median survival rates (13–22 months) were no better. The role of induction therapy has been addressed only in RCTs for early-stage or minimal N2 disease (negative CT of the mediastinum with either no or minimal N2 involvement at mediastinoscopy). These patients would be eligible for surgery alone, and have less bulky disease than those in the Phase II studies described above. The five studies are summarized in Table 3. The two positive studies by Roth and coworkers (156, 168) and Rosell and coworkers (157, 169) closed early because of disparity in survival between the two arms, a statistical hazard when applying early closure rules to small studies. However, the median survival differences have become smaller at 5 years (0 versus 17% in the study by Rosell and coworkers, and 15 versus 36% for Roth and coworkers), although still clearly favorable for the combined modality treatment. These studies have been criticized because the surgery-alone arms have fared badly, particularly in the study by Rosell and coworkers, in which there was a higher rate of tumor K-ras mutation and DNA aneuploidy, which are indicators of poor prognosis, than in the chemotherapy arm. The much larger French study just published (155) included 355 patients with clinical Stage I (except T1N0), II, and IIIA disease randomized to surgery or two courses of MIC followed by surgery with an option for two further courses postoperatively. Patients with T3 and N2 disease received postoperative irradiation. A pathologic CR was seen in 11% of patients receiving neoadjuvant chemotherapy, but the survival data for the entire study were not significantly different at 3 years (p < 0.15; Table 3). There was no difference in the postoperative mortality rates: 6.7% in the chemotherapy arm and 4.5% in the surgery-only arm. The median survival was 37 months with neoadjuvant chemotherapy and 26 months with surgery alone. There was a significant survival advantage with neoadjuvant chemotherapy for patients with N0 and N1 disease, but not for those with N2 disease. It is suggestive, therefore, that there may be an added advantage for chemotherapy for survival with Stage I and II NSCLC. However, there was no evidence of added benefit with chemotherapy for Stage IIIA disease. Thus, valuable information is emerging to define the role of neoadjuvant chemotherapy in patients with resectable disease. This population of NSCLC patients includes those most likely to respond to chemotherapy (good performance status, small-volume disease, and normal biochemical values) and, clearly, a small percentage improvement in overall survival data a |
TOP
ABSTRACT
CONTENTS