Over 100,000 thoracotomies are performed annually in the United States alone, and each one of these procedures requires a careful preoperative assessment. Although for coronary artery bypass grafting, the most commonly performed cardiac operation, several large models of risk stratification have been developed and validated, allowing not only intra-institutional comparisons but also comparisons between entire countries, no such model exists for lung resections. The Society for Thoracic Surgery has been putting considerable effort into gathering at least a baseline set of data that would allow such modeling. The thoracic community hopes to be able to contribute in the near future towards such a database. At present, however, each potential patient for a lung resection has to be assessed for two key concepts: resectability, which is defined by the amount of lung tissue that can be safely removed without pulmonary insufficiency; and operability, which is defined as the ability of the patient to survive the procedure and any perioperative complications. Individual authors have identified a large number of variables that are associated with increased complications. Virtually all of these data were gathered in a retrospective fashion and were not randomized. A number of risk factors that have been identified warrant brief discussion, as follows.

Increased preoperative sputum production is associated with postoperative pulmonary complications. There are also some data suggesting that the presence of an upper respiratory infection can impair respiratory function [1, 2]. Perioperative antibiotic use is directed towards wound infections and not on a systematic basis against pulmonary complications. No trials have been conducted in recent years to evaluate the ability of routine antibiotic prophylaxis to protect against postoperative respiratory events. Treatment of suspected respiratory infection with antibiotics may decrease the risk of postoperative pulmonary complications, but the threat of multi-resistant organisms provides a strong incentive to limit the use of antibiotics to a setting in which they have proven efficacy. Surgery, therefore, should be delayed whenever possible to complete a course of antibiotics in patients diagnosed with pneumonia or bacterial upper respiratory infections in the preoperative period, and the same argument holds for patients with viral respiratory infections [2].

The influence of age is far less clear than in cardiac surgery where age has clearly been established as a risk factor. Patients over 70 years of age were thought to be at increased risk for thoracic surgery, but this could not be confirmed in more recent studies [3, 15].

Both body mass index greater than 30 kg/m2. and chronic obstructive pulmonary disease (COPD) have been identified as independent predictors of surgical site infections after cardiac surgery. Obesity is known to change pulmonary mechanics and predisposes to atelectasis and hypoxemia. The degree to which obesity, per se, contributes to pulmonary complications has not been clearly established. Hypoventilation syndromes and sleep disorders are especially common in obese patients with concomitant COPD, and are worse in the postoperative period [4]. Most of the time, however, it is not practical to delay a surgery for weeks or months while a patient attempts to lose weight. At present, increased monitoring is the only recommendation for avoiding alveolar-hypoventilation and atelectasis. Special care should be taken to assure that patients are alert enough to protect their airway, and non-invasive ventilation may be helpful when hypercapnia is present [4].

COPD and lung cancer are a common ground for malnutrition, and are associated with an increased risk for complications after lung resection. At present, there is insufficient data to justify perioperative total parenteral nutrition, because the results of clinical studies have been inconsistent. There are data from prospective randomized controlled trials that indicate oral nutritional supplements for as little as 8 weeks can result in weight gain and improved respiratory muscle strength [8]. It therefore appears reasonable to place a malnourished patient on an oral supplement while awaiting surgery [5].

Chronic smoking results in bronchial gland hyperplasia with increased mucous production and impaired ciliary function while it acutely raises carboxy hemoglobin levels. There are some data to suggest that a greater than 20 pack-year smoking history significantly increases a patient's surgical risk. It is difficult to separate smoking from the tobacco-related cardiopulmonary diseases, such as COPD and coronary artery disease. As far as smoking and surgery are concerned, unless a patient stops smoking weeks before surgery, there is no benefit derived, and there are actually data to suggest that the risk may be enhanced. In a randomized trial, patients who quit less than 8 weeks prior to coronary artery bypass grafting were actually found to have higher rates of respiratory complications than patients who continued to smoke; at the same time, patients who quit smoking for at least 8 weeks prior to surgery did better than patients who continued to smoke [7].

A vast body of literature suggests that chronic pulmonary disease is the most predictive characteristic of a patient undergoing thoracic surgery [9]. The degree of correlation, however, is still being debated. It is generally agreed that arterial blood gas analysis revealing hypercapnia (PaCO2 > 45 mm Hg) and marked hypoxemia (PaO2 < 50 mm Hg) is predictive of severe complications and death, although the supporting data are not uniform [9, 10]. There are strong arguments for treating asthma and COPD exacerbations before surgery. It has been documented that glucocorticoids decrease the severity and the duration of COPD exacerbations [11]. Although chronic steroid use has been associated with an increased risk of major gastrointestinal complications and infections, as well as delayed wound healing, there are no data to suggest that pulse therapy has the same deleterious effect. In addition, bronchodilators have been found to decrease the risk of postoperative respiratory complications. In combination, steroids and bronchodilators are a very effective means of eliminating the reversible component of bronchospasm [12].

In a landmark study published 40 years ago, Thoren demonstrated the superiority of chest physiotherapy initiated before surgery. That study was performed in a general surgical population [13]. He showed that he could lower respiratory complications by 50% in cholecystectomy patients. Chest physiotherapy includes deep breathing exercises, incentive spirometry, intermittent positive pressure breathing (IPPB), coughing and huffing maneuvers, and chest physical therapy with postural drainage. Since the time of Thoren's work, studies indicate that chest physiotherapy targeted towards high-risk patients, especially those with underlying pulmonary disease, shows some benefit [14]. It appears that deep breathing exercises, incentive spirometry, and IPPB are equally effective. Therefore, in today's cost-conscious setting, most institutions start with deep breathing exercises and incentive spirometry. It appears unnecessary to perform chest percussion in these patients, because it is no more effective than the other maneuvers described above, and seems to increase the need for analgesics. Whether or not pulmonary rehabilitation can contribute significantly to outcomes has recently gained some attention. It is a prerequisite for entry into the ongoing National Emphysema Treatment Trial for lung volume reduction. The data from this study are not yet complete.

Pulmonary function testing before lung resection is used to help assess the risk of perioperative morbidity and mortality. It also serves as the basis to estimate remaining lung function. Pulmonary function testing was a product of World War II, and over the past 50 years has further evolved. It initially relied on maximum breathing capacity (maximum voluntary ventilation; MVV), and now includes spirometry and a measurement of lung volume, diffusion capacity (diffusion capacity of the lung to carbon monoxide; DLCO) and arterial blood gases in conjunction with radionuclide scanning. None of the individual tests have been found to have a high correlation with outcomes, and only as a group have these tests been found to have a correlation.

One fundamental issue with using spirometry as a predictor is the question of what constitutes an acceptable endpoint. The crudest of all tests remains perioperative mortality, and it may be fair to use large government databases to get an idea of present outcomes, since individual institutional reports tend to report good results. For instance, data from all non-federated acute care hospitals in California between 1983 and 1986 showed rates of in-hospital deaths within 30 days of 10.1% for pneumonectomy, 3.4% for lobectomy, 2.9% for segmental resections, and 3.2% for wedge resections [15].

Unfortunately, no pulmonary function data are available. However, it was noted that age was a strong risk factor with the mortality for pneumonectomy increasing from 6.2% for patients less than 60 years of age to 12.5% for patients between ages 60 and 69, 18.8% for ages 70 to 79, and 29.2% for ages over 79. Lesser resections had a proportional increase also. The same trend was seen in national Medicare data between 1983 and 1985 [16]. This showed a 30-day mortality rate for pneumonectomy of 4.9% at ages 66 to 69, 11.7% for ages 70 to 74, and 26.5% for ages over 74. Mortality for lobectomy and lesser procedures increased from 5% to 8% over the same age ranges.

Miller reported on a series of 2,340 lung resections performed between 1974 and 1990 [17]. The 30-day hospital mortality rate was 5% for pneumonectomy. Candidates had to have an MVV of more than 50% predicted, DLCO of more than 50% predicted, ratio of forced expiratory volume in 1 s to forced vital capacity (FEV1/FVC) of more than 0.50, and predicted postoperative (PPO) FEV1 of more than 1 L. For lobectomy the criteria included an MVV of more than 40% and a PPO FEV1 of more than 1 with a mortality rate of 0.4% [17]. It is noteworthy that the derivation of accepted criteria is derived from a very small series, often of not more than a dozen patients. The agreed values are currently in flux with a tendency towards accepting lower numbers. The National Emphysema Treatment Trial, which includes lung volume reduction surgery, is currently a driving force to gather sophisticated data.

So far, some guidelines can be derived from the acquired experience:

1. All patients undergoing pulmonary resections that are not emergencies, even if they are very limited in scope, should have at least spirometry testing preoperatively.
2. This testing should include spirometry repeated after bronchodilator if the FEV1/FVC ratio is below normal, and also lung volume, MVV, diffusion capacity, and arterial blood gas measurement. These tests are especially helpful when pneumonectomy is considered.
3. A patient with an FEV1 of more than 2 L and a DLCO of more than 60% should tolerate a pneumonectomy and should not require any further tests. A patient with an FEV1 of more than 1.5 L and a DLCO of more than 50% should tolerate a lobectomy or a lesser procedure. However, if there is uncertainty that a lobectomy may be sufficient, further testing may be required.
4. If a patient fails the above criteria and a pneumonectomy is planned, an estimation of postoperative residual function should be performed. This currently appears to be most accurately done with a quantitative perfusion scan, with or without ventilation studies. For a lobectomy or lesser procedures, an estimate based on segment count is satisfactory in most cases. In borderline cases, the perfusion scan needs to be interpreted for regional abnormalities. The expectation is that in the early postoperative period, the loss will exceed the predictions, but 3 to 6 months after the lobectomy the FEV1 will tend to be higher than predicted [5]. A PPO FEV1 of more than 1 L or more than 40% of predicted normal and a PPO DLCO of more than 40% indicate that the patient is an appropriate candidate for the planned resection. Patients with a lower PPO FEV1, but still greater than 0.8 L or greater than 30% predicted, have higher risk of postoperative morbidity and more limiting exertional dyspnea. Generally, their mortality is slightly increased, but the increased mortality has been felt to be acceptable given their overall outlook.

Patients with lower predicted parameters appear to undergo resection only on a very individual basis and are candidates for exercise stress testing and assessment of pulmonary hemodynamics.

Cardiac risk assessment for lung resection surgery is limited by a lack of data pertaining specifically to this type of surgery. No prospective controlled data, in particular, are available. Use of the core recommendations of the American College of Cardiology/American Heart Association (ACC/AHA) guidelines are currently the best available method, not only in assessing the risk, but also how the data were gathered to derive the recommendations [18]. By that standard, there are no data from large perspective randomized studies (Level 1) available, but only from non-randomized cohort studies (Level 3), or consensus expert opinions (Level 5). The ACC/AHA guidelines state that preoperative testing is rarely indicated in patients who are symptom-free and have no deterioration in their functional status if they had a recent coronary evaluation within the last 2 years preceding the planned surgery [18]. Preoperative assessment should identify those patients who are deemed at high risk for cardiac complications based on clinical predictors. These predictors include, for example, recent myocardial infarction (within 1 month), or significant myocardial ischemia detected with non-invasive testing, decompensated congestive heart failure, symptomatic ventricular or uncontrolled supraventricular arrhythmias, and severe valvular heart disease. As outlined in the ACC/AHA guidelines, patients not deemed high-risk or emergencies can be categorized into either minor, intermediate, or major risk based on a combination of clinical predictors and planned surgical procedure. The functional capacity of an individual patient is helpful, especially in determining a risk for minor and intermediate patients, and is considered excellent if over 7 metabolic equivalents (METS), moderate if 4 to 7 METS, or poor if less than 4 METS [19].

Although a patient who is deemed to be in a low-risk category should be able to proceed to surgery without further testing because further testing is not cost-effective, this approach may be modified by a high-risk surgical procedure. The intermediate risk group is the most troublesome to evaluate. It is also the one into which a considerable number of thoracic surgical patients fall. No definitive randomized studies for further testing in thoracic surgical patients, however, exist.

An attempt to refine the ACC/AHA guidelines by creating a cardiopulmonary risk index did not add much [19]. The intent was to assess both pulmonary and cardiac risk as a combined endpoint. From a cardiac perspective, however, it was not an adequate measure to assess the risk in thoracic surgery patients [20]. The argument that cardiac risk should be looked at independently stems from the fact that thoracic surgical patients possess the same risk factors as patients presenting for primary cardiac surgery. Pre-existing coronary artery disease, especially with a history of myocardial infarction, is a considerable risk factor. The incidence of reinfarction in patients with prior myocardial infarctions after major surgical procedures ranges from 3% to 17% with a subsequent mortality rate reported between 28% and 69% [21]. In addition, postoperative cardiac complications, predominantly myocardial infarction or congestive heart failure, are the most common cause of deaths [22, 23].

The coronary artery surgery study (CASS) database contains data regarding the cardiac risk before thoracic surgery [24]. Thoracic surgery, together with head and neck surgery, abdominal surgery, and vascular surgery, was identified as a group with a higher risk of perioperative death or myocardial infarction in patients with known coronary artery disease. For the 165 thoracic surgical patients analyzed, there was a trend towards benefiting from surgical coronary revascularization versus medical management, although it did not reach statistical significance. That database also yielded information indicating that patients who had coronary artery bypass surgery within 5 years prior to lung surgery were relatively protected from postoperative myocardial infarctions or deaths after non-cardiac surgery [24].

Since the significance of silent coronary disease in an asymptomatic patient seems to be rather small, the value of either medical or surgical interventions for that alone is questionable at best, considering that the risk of coronary artery bypass surgery when added to the risk of a subsequent thoracic procedure may actually be greater than the risk of the thoracic operation by itself. There seems little justification to perform a coronary operation just to get a patient through the following thoracic operation. However, if the long-term outlook is influenced by the degree of coronary disease, then revascularization has an appropriate role. The role of prophylactic coronary bypass surgery to specifically lower thoracic surgical risk is, at best, controversial, and not supported by any prospective randomized data [23].

The place of angioplasty is relatively new and unexplored in that setting. A Mayo Clinic series reported a perioperative myocardial infarction rate of 5.6% and a mortality rate of 1.9% in high-risk patients who underwent angioplasty on average 9 days before their planned non-cardiac surgery [25]. Given the potent antiplatelet agents leading to the feared bleeding diathesis, the optimal timing for surgery after an angioplasty is currently not known. Combined operations have been offered to patients who have symptomatic coronary disease and lung cancer. Again, no randomized data are available, but overall, stage for stage, patients who undergo combined procedures seem to have a worse survival rate than that of patients who have surgery for the same-stage lung cancer only [76]. Although technically feasible, most authors advocate a staged approach with coronary artery bypass grafting surgery preceding the lung resection [77].

Commonly used non-invasive testing available for cardiac risk assessment include echocardiography, radionuclide angiography, and exercise and non-exercise stress testing. Echocardiography is helpful when decreased left ventricular ejection fraction is suspected, as it has been shown that ejection fractions less than 35% represent an increased risk for postoperative congestive heart failure. Therefore, patients with uncontrolled and unknown left ventricular systolic or diastolic function benefit from this modality. Routine ventricular function assessment is not indicated, and radionuclide angiography or echocardiography is not predictive of perioperative ischemic events [26]. Exercise stress testing is inexpensive and provides information on functional capacity. The problem with it is that a significant number of thoracic surgical patients will be unable to attain target heart rates. A negative predictive value of a negative exercise stress test is reported to be over 90% for patients who undergo major surgery [26]. The alternative is a myocardial perfusion study using thallium combined with dipyridamole or adenosine. These studies are found to have a negative predictive value of over 95%, both in general surgical populations and in vascular surgery patients [27]. The problem from a thoracic point of view is that the dipyridamole and adenosine should be avoided in patients with clinical bronchospasm because these agents may induce bronchospasm and respiratory arrest. In addition, the use of theophylline or caffeine products should be avoided before testing with dipyridamole or adenosine because of interference with the test. Hence, dobutamine echocardiography is often preferred. Available data would suggest that the negative and positive predictive values are similar to that of thallium scintigraphy [28]. There are also data on the efficacy of dobutamine stress echocardiography in patients undergoing lung volume reduction surgery, the results of which were similar to the previously mentioned tests in patients having non-cardiac surgical procedures [29].

Cardiac considerations include concerns other than just the evaluation for ischemia. Overall, the incidence of postoperative cardiovascular complications is 25% to 30% after thoracic surgery, with the most common complications being arrhythmias, predominantly atrial fibrillation. Other issues include cardiogenic pulmonary edema, pulmonary hypertension, and core pulmonale, as well as valvular heart disease. Postoperative heart failure is especially feared in patients after pneumonectomy or lung volume reduction surgery. If not screened, patients for these procedures have faired poorly, and preoperative echocardiographic evaluation of the right heart function and pressures is feasible and may avoid right heart catheterization [30-32].

Postoperative arrhythmias, especially supraventricular arrhythmias, are very common. In a recent prospective cohort study, male sex, age over 70 years, significant valvular disease, history of supraventricular arrhythmia, asthma, congestive heart failure, and premature atrial contractions on preoperative EKG, American Society of Anesthesiology Class III or IV, and intrathoracic procedure (odds ratio, 9.2) all were identified as risk factors [33].

Given the high prevalence, prophylaxis has been practiced for many years. The use of digoxin for prevention of atrial arrhythmias is widespread. Despite its popularity, the use of digoxin for this purpose is questionable, and has not been supported by prospective trials. In addition, the use of calcium channel blockers for prophylaxis has not resulted in any reduction of atrial arrhythmias. Amiodarone is to be avoided because of the increased incidence of adult respiratory distress syndrome in that setting. As a consequence, no recommendation for prophylaxis of postoperative atrial arrhythmias can be made at present. Although the use of beta blockers in patients with COPD may worsen bronchospasm, there are data to support that in high-risk patients, when started preoperatively, it results in a 15% reduction in the combined endpoint of myocardial infarction, unstable angina, congestive heart failure, revascularization, and death at 6 months [34].

When Sir Peter Goldstraw was asked at the turn of the millennium to summarize oncologic principles guiding surgery for lung cancer, he came up with nine succinct points [35], which deserve to be repeated here:

1. Diagnosis must precede treatment. The spectrum of diagnostic tools available to establish a preoperative diagnosis has considerably widened, and will do so in the future. Rigid bronchoscopy has been supplemented by flexible bronchoscopy, mediastinoscopy, and video-assisted thoracoscopy, as well as positron-emission tomography (PET) scanning. Nevertheless, a surgeon still encounters lesions where no histologic preoperative diagnosis has been established, but which are highly suggestive for cancer, and which will require a thoracotomy. In these circumstances, an intraoperative effort to establish a tissue diagnosis prior to undertaking a large pulmonary resection should be made.
2. Surgery remains the only reliable prospect for cure. Having said that, this statement is qualified by the following:
3. To benefit from surgical cure, the patient must survive the operation. It is especially in this area that a national database will be extremely helpful, as it will provide for performance guidelines. On one hand, mortality figures should be kept within acceptable guidelines. On the other hand, in order to do so, patients with poorer performance status should not be excluded. Therefore, risk adjustment in large data sets will be helpful.
4. The prospects for cure depend on stage. It has been said by cynics that the only progress in lung cancer in the last 50 years has come from improved staging. Not only has the system been improved and agreed upon internationally, but also the tools in order to classify patients have made great strides. (Both of these topics are covered in detail elsewhere in this issue.)
5. All preoperative staging is inaccurate. This statement has to be taken with a caveat and implies that a systematic nodal dissection is performed. Goldstraw made this statement based on his own work in which he suggests that preoperative cTNM stage [clinical-diagnostic staging of cancer]. is accurate in less than 50% of cases [36]. To improve this percentage, the topics of pleural lavage cytology and sentinel node biopsy, as practiced in breast cancer surgery, are being actively explored.
6. Complete resection is the only operation to have an impact on survival. Surgery should have a good prospect of resulting in a complete resection. Exploratory thoracotomy, i.e., an opening and closing operation, carries a considerable mortality reported to be 3% to 4%. As is the case with appendectomy, on the other hand, if no exploratory thoracotomies are performed, the approach would be considered too conservative, whereas if too many are performed, the approach would be too aggressive. It appears that this number should not exceed 5% [37].
7. There is no survival advantage to resection greater than that which removes all macroscopic and microscopic disease. Here, the American College of Surgeons Oncology Group (ACOSOG) trial of en-bloc lymphadenectomy versus systematic lymph node dissection will hopefully help and answer the question of whether lymphadenectomy, in addition to staging, has a therapeutic role. Despite the fact that a randomized trial has shown a three-fold local recurrence rate in patients undergoing less than a lobectomy, this issue is not entirely settled, especially in the current era of neoadjuvant protocols [38].
8. Distant relapse remains the major obstacle to surgical cure. Despite PET scanning, recurrence at a distant site remains the most common mechanism of relapse after complete resection; hence, the interest in neoadjuvant protocols with the intent to decrease that number from two-thirds [39].
9. The surgeon has an important role for patients with inoperable disease. Malignant effusions of the pleural space, and of the pericardial space, as well as interventional airway management are areas where surgeons can contribute significantly to the quality of life in a patient who otherwise is unresectable.

Staging lung cancer is of utmost importance as attempts to determine the extent of the disease and subsequently stratifies patients in order to allow determination of prognosis and ultimately therapeutic options, and therefore in and of itself may have made the biggest contribution, not only to the understanding, but also to the change in outcome seen over the last twenty-five years. During that time period, we have seen a trend towards less and less invasive technology. Subsequently, a brief review of surgical options are presented, acknowledging that they all are preceded in general by varying imaging techniques. Thoracoscopic techniques the last twenty five years has seen an explosive technology for what is called minimally invasive procedures. Especially for staging purposes, thoracoscopic techniques appear very useful. Generally, two questions need to be answered:

1. A histologic diagnosis of the primary tumor needs to be obtained.
2. The second question often is regarding nodal involvement, especially N2 disease.

In both instances, thoracoscopy has been found to be helpful. One has to be reminded, however, that successful thoracoscopy requires that the involved lung can be collapsed, i.e., that there is no significant pleural disease present and that the patient has enough reserve to be ventilated with one lung only.

Mediastinoscopy is not a very taxing procedure for the patient, and is often performed on an outpatient basis. The enthusiasm for a mediastinoscopy varies from institution to institution. The reader, however, should be reminded that in a CT scan, which shows enlarged mediastinal lymph nodes greater than 1 cm, the accuracy of that scan is only about 70%. Cervical mediastinoscopy allows access to the paratracheal lymph nodes, subcarinal lymph nodes and tracheobronchial angle lymph nodes, especially on the right. Because of the aortic arch, the para-aortic and aortapulmonary window lymph nodes are not accessible by this technique.

In order to assess the aortopulmonary window, especially in left upper lobe tumors where the aortic arch is in the way of traditional cervical mediastinoscopy, the surgeon has several options:

Traditionally, a Chamberlain procedure has been used. In this, the cartilaginous portion of the second or third rib is removed, and through an extrathoracic approach, the aortopulmonary window is explored and biopsied. Ginsberg has advocated an extended cervical mediastinoscopy in which, through a traditional cervical incision the dissection plane is carried anterior to the big vessel, the anterior mediastinal lymph nodes, and often the sub-aortic nodes are biopsied.

Thoracoscopy also can be done to assess the aortopulmonary window. An interesting question in times where FDG-PET is available is if invasive staging of the mediastinum is still necessary to the same degree. A recently published paper in that regard suggests that PET may reduce the necessity for invasive staging when the primary lesion's standardized uptake value is less than 2.5 and the mediastinum is PET-negative. In that study, invasive staging could have been reduced by approximately 12%.

According to the above principles, stage I lung cancer lends itself best to surgical resection, as by definition, macro- and micro-scopic disease should be completely excisable. Why, then, is it that the 5-year survival rates are disappointingly low, ranging from 57% to 67% [40]? It has been known for a long time that occult metastases are the culprit, and that many patients who seemingly have had completely resections suffer from systemic disease [42, 43]. Mountain and others have determined prognosticators for outcome in stage I lung cancer, and these include the size of the primary tumor, the type of histology, and the primary location, either central or involving the visceral pleura, as indicators of a worse outcome, i.e., higher likelihood of metastatic disease [41]. Knowing these results would lead one to believe that bigger surgical procedures may improve results. Just the contrary development has been observed over the past quarter century. In recent years in particular, surgeons have supported the development of thoracoscopic anatomic dissections using a small incision without spreading the ribs [44]. The claim is that comparable morbidity and mortality can be achieved with shorter hospital stays and more comfort to the patient [45].

This view is still not held by the majority of surgeons, who believe that more exposure, especially for palpation, which is deemed an essential part, is being sacrificed in thoracoscopic surgery. In addition, the technique is clearly not suitable for many cases with adhesions and scarring. Technical advances in the more "traditional approach" have also been achieved. With the routine use of epidural pain control, muscle-sparing incisions, and axillary thoracotomies, postoperative pain control, loss of pulmonary function, and short hospital stays are not significantly different from thoracoscopic approaches.

Given the overall disappointing 5-year survival statistics, a number of investigators have looked at chemotherapy in the adjuvant setting. The data of the individual studies are not easily interpretable, but the non-small-cell lung cancer collaborative group performed a meta-analysis from 52 randomized trials in 1995 [46]. When given cisplatin-based chemotherapy, a 5% absolute survival benefit at 5 years was found. The confidence intervals were between 1% and 10%, which still results in an absolute benefit. On the contrary, when using long-term alkylating agents, an increase in mortality of 5% at 5 years was found, therefore favoring surgery alone. For other regimens, no conclusion could be drawn. Based on intriguing reports from Japan, there is currently an interest in examining the use of oral fluorouracil derivatives [47, 48]. In light of two Japanese studies, which were well designed and have been scrutinized heavily, ACOSOG is embarking on a randomized study of postoperative tegafur and uracil combined with leucovorin (Orzel). The tegafur and uracil administered in the Japanese studies were well tolerated, and appeared to inhibit recurrences and prolong survival.

Preoperative chemotherapy has been intensively studied, especially in stage IIIA and IIIB non-small-cell lung cancer. The goal of these studies was primarily focused on shrinking the tumor size, thus allowing complete resection of initially questionable or unresectable tumors, and only a secondary goal was dealing with micrometastases.

The French Cooperative Oncology Group concluded that, after an analysis of relapse patterns, preoperative chemotherapy appears to act more on micrometastases than on local control, and they designed the first randomized study of preoperative chemotherapy for stage IB, II, and IIIA patients, which was presented at the American Society of Clinical Oncology meeting in May 1999 [49]. This study included 373 patients. The regimen included mitomycin, ifosfamide, and cisplatin at 3-week intervals. Initially two cycles were given, and to patients who responded, two additional postoperative cycles were given. The median survival was extended from 26 to 36 months, and appeared not to be statistically significant. In addition, there was no difference in local recurrence rate. Distant metastases, however, occurred significantly less in the chemotherapy arm. Criticisms of the study have been that the surgery-only patients faired less well than patients in other published series.

Ongoing studies of most United States and European oncology groups are including early stage tumors, as in the eagerly- awaited Southwest Oncology Group Trial 9901 [50].

Stage II lung cancer comprises a relatively small number of cases, and accounts for 12% to 19% of non-small-cell lung cancers. The 5-year survival rate ranges from 24% to 55%. This reaches from completely resected T1N1 lesions for which survival is 35% to 55%, to T2N1 lesions for which survival is 25% to 44.7%, to T3N0 lesions for which survival is 38% to 42.9%. This group contains many patients who were clinically staged as stage I patients. In the Mountain series of 5,319 patients, clinical N1 disease was observed in 279 [40]. However, N1 disease was found histologically in 364 patients, which represents a 30% upstaging. The problem can be reduced to a simply diagnostic one: one aspect thereof is the inability to accurately stage the mediastinum, the second aspect involves distant micrometastases.

Currently, stage II lung cancer remains largely a pathologic diagnosis, and preoperatively can only be suspected from high resolution computed tomography (CT) scans, which, at best, have a sensitivity and specificity of 63% and 80%, respectively, for the detection of mediastinal lymph nodes [51]. Metastases to mediastinal lymph nodes by the fluorodeoxy glucose (FDG) PET scanning were found with a sensitivity of 88% and a specificity of 93% [51]. The combination of CT and PET has been reported to improve the accuracy to 94% with a positive predictive value of 93% and a negative predictive value of 95% [53]. The problem of exact anatomic correlation, however, is a vexing one. Hilar uptake may be hidden behind the tumor, or already may be a sign of positive mediastinal lymph nodes. Recently, the argument has been made that among patients with suspected stage II disease, a mediastinoscopy should be performed in those cases in which there is absence of concurrence between PET and CT scans. That argument is based on reports that patients who are found to have N2 disease on mediastinoscopy rarely benefit from primary surgical resection [54]. These patients may benefit from being involved in an induction chemotherapy protocol [55, 56].

The principle of complete resection remains the surgical goal. This surgery should include a systematic mediastinal lymph node dissection of paratracheal, subcarinal, and paraesophageal nodes. Completely resected patients with N2 disease found incidentally at surgery following a negative mediastinoscopy have been reported to have a 41% 5-year survival versus 15% [54]. Whether en-bloc mediastinal lymph node dissection versus traditional mediastinal dissection yields better survival is currently under investigation. The present author believes that video-assisted thoracic surgery in stage II lung cancer is controversial at best. So much of the outcome hinges upon a complete resection and adequate staging that any sacrifice in exposure to achieve this goal is questionable.

Adjuvant chemotherapy, particularly in this group of patients, is a very hot topic. Reference is made to the previously described meta-analysis [46]. Since this study has shown a clear advantage of a cisplatin-based regimen, several newer agents have been found to be active against lung cancer. Several large international trials are underway to settle this question. The international adjuvant lung cancer trial has accrued two-thirds of the goal of 3,300 patients, and is to close within a year. The Medical Research Council of the United Kingdom is accruing patients with early lung cancer for its planned 10,000 early- stage completely resected patient lung trial in which cisplatin- based chemotherapy will be compared to observation, and in North America, an inter-group trial, together with the National Cancer Institute of Canada, is accruing patients to a trial randomizing observation to 15 weeks of vinorelbine and cisplatin chemotherapy.

In the neoadjuvant setting, again the only completed trial is that of the French Cooperative Oncology Group, described earlier [49]. Awaiting maturation is the trial conducted by the Bimodality Lung Oncology Team, known as BLOT. In this pilot study, 94 patients in stages IA, IB, and IIB determined by mediastinoscopy were treated with two cycles of neoadjuvant paclitaxel and carboplatin. A clinical response was seen in 54%; in 4% it was complete, and 82% of the patients underwent a complete resection [57]. At this stage, neoadjuvant chemotherapy remains experimental, while its effectiveness, safety, and tolerability have made it very attractive.

Forty percent of patients of the 80% of all lung cancer patients who present with non-small-cell disease will have locally advanced disease. Traditionally, these patients have been treated with surgical resections, radiotherapy, and chemotherapy, as well as various combinations of these treatment modalities. Few patients will actually be cured by surgery alone, and multi- modality treatment is rapidly gaining acceptance in this sub- group. The definition of resectable or, on the contrary, unresectable, especially in the era of multi-modality therapy, keeps constantly changing. Even after the revision of the staging system in 1997, stage III disease remains a very disparate group of patients [40].

Staging is all-important, and the surgeon contributes here primarily by performing mediastinoscopy. Some surgeons have advocated routine mediastinoscopy, but most prefer a selective approach. Generally, this procedure is used in the presence of superior sulcus tumors, and if there is evidence of mediastinal lymph nodes of greater than 1 cm for large central tumors.

An interesting question is, how does mediastinoscopy fit into a treatment plan where neoadjuvant therapy is considered? The arguments for the timing go both ways: some argue that neoadjuvant therapy should be reserved for patients with proven N2 disease, therefore recommending a mediastinoscopy prior to neoadjuvant therapy. Others argue that a patient who has positive N2 nodes after receiving neoadjuvant therapy is unlikely to benefit from surgery; therefore, mediastinoscopy has its greatest impact at that juncture in the treatment plan. A few surgeons do not hesitate to use mediastinoscopy twice in the management of patients. Randomized data at present are not available.

Given the observation that incidentally found N2 disease in resected specimens when it was not suspected preoperatively leads to a much better survival than N2 disease found either by mediastinoscopy or an agreeing combination of PET/CT, the present author's bias is to perform the mediastinoscopy after neoadjuvant therapy. Despite these efforts at improvement in determining N2 involvement prior to a resection, there will remain a group of patients who either at resection or subsequently in their pathologic staging will have N2 disease, and then will require adjuvant therapy.

Despite half a century of experience with radiation therapy, there appears to be no clear-cut consensus as to its overall benefit. The only agreed upon indication is the presence of microscopic residual disease at the resection margin. A greatly debated study was performed by the Lung Cancer Study Group consisting of a phase III trial with patients who had N1 and N2 disease and squamous cell carcinoma of the lung [58]. Patients were randomized between observation and mediastinal irradiation to 50 Gy in 5 weeks. Squamous cell carcinoma was chosen because of the alleged higher local failure rate and the lesser likelihood of distant disease. Two hundred thirty patients were entered. In the non-radiation group, 18% of local recurrences were seen, compared to 1% with radiation. This study has been subsequently criticized for non-uniform staging, and for a lack of demonstration of a survival benefit, as overall failure was not one of the endpoints in the study. Nevertheless, several other studies have subsequently confirmed that local control can be achieved with radiation, but that this does not translate into a survival benefit.

Early trials have consistently not shown a survival benefit, but some have actually demonstrated a detrimental effect on survival, probably due to cardiopulmonary toxicity. This latter effect has been borne out in a meta-analysis of 2,128 patients, published in 1998, and including nine randomized trials of postoperative radiation [59]. The overall risk ratio was found to be 1.21 (p = 0.001) for death. The recommendation of the authors was to abstain from postoperative mediastinal radiation. Criticisms regarding the choice of studies involved for this meta-analysis included that one-quarter of the patients had N1 disease, for whom it is known that they do not benefit from irradiation. Thus, they were just exposed to the detrimental effects. Further, included in the analysis were studies in which large daily doses were administered up to 3 Gy. It is well known that large daily fractions predispose to cardiopulmonary side effects.

What has been said about adjuvant chemotherapy in the previous paragraphs concerning Stage I and II disease applies also to stage III disease. The previously discussed meta-analysis, which showed a 5% overall survival advantage with cisplatin-based chemotherapy included IIIA disease [46]. A vexing problem that has not gone away, despite improved anti-emetics and growth factor support for bone marrow toxicity, is that only about 60% of the total planned dose of chemotherapy is given. The change from pulse therapy to prolonged continuous low-dose administration of a chemotherapeutic agent is again one of the very attractive features of the Japanese and the ACOSOG Orzel trials. From a surgeon's perspective, neoadjuvant therapy carries the perceived problem of increased morbidity and mortality associated with the subsequent lung resection. Sporadic reported studies claiming that this is true, and others that do not find any change in outcome, are all small and non-randomized. From a surgical standpoint, this demands an even more careful assessment of the patient's overall ability to tolerate surgery. This assessment may very well require repetition of some components of a workup, which may have been performed months ago prior to embarking on the neoadjuvant protocol.

Of particular concern in this setting are radiation pneumonias, which may require perioperative steroid administration, increased technical difficulties from scarring fibrosis, fusion of anatomical planes, and a tendency towards increased bleeding. Bronchial stump dehiscence is also of particular concern, and consideration to muscle flaps should be given. Specific toxicity issues from chemotherapy warrant special attention, especially pulmonary toxicity resulting in adult respiratory distress syndrome of agents such as mitomycin C, which may require pretreatment during surgery with steroids and avoidance of high inspired-oxygen fraction. Despite the surgeon's aim to keep the administration of fluids to a minimum, this may not be possible due to renal toxicity. Cardiac toxicity can occur from either chemotherapeutic agents or radiation.

In summary, unfortunately only about 10% to 20% of the entire population with stage III disease presents with unsuspected nodal metastases at the time of thoracotomy. Preoperatively, they have normal-sized ipsilateral nodes, which contain microscopic disease at the time of surgery. This finding places these patients in the most favorable category of N2 node-positive patients with a 5-year survival of 30% with surgery alone, and these are the patients who benefit from adjuvant chemotherapy [60].

In the neoadjuvant setting, the literature is replete with reports on phase II trials; some that show a benefit, and others that do not [62]. There simply is not yet enough solid phase III data available to make a solid recommendation, although from a practical standpoint, almost every institution offers its patients some form of preoperative chemoprophylaxis. Preoperative radiation has been compared to neoadjuvant chemotherapy without finding any advantage [61].

For the unresectable stage III patient, it appears that a number of phase III trials using higher dose cisplatin-based chemotherapy in conjunction with radiotherapy led to improved survivals as compared to radiation therapy (RT) alone. This came about after a randomized trial in 1990 demonstrated the benefit of sequential addition of chemotherapy to radiation therapy [64]. This finding translated into a 5-year survival advantage that was just twice as large as RT alone. The current debate in this area is whether concurrent chemotherapy plus RT should be used versus sequential RT and chemotherapy. Better staging in the IIIB group allows for identifying patients without N2 or N3 disease of good performance status who can tolerate an often-complex local resection, and the possibility of a 30% 5-year survival.

A number of studies, mostly from single institutions, have clearly demonstrated that nodal status is the most important determinant of outcome in patients with chest wall invasion. This is perhaps most glaringly demonstrated by the Memorial series reported in 1999, in which of 334 patients, only 175 had a complete resection [65]. The 159 patients who had no resection or an incomplete resection were all dead at 5 years. More important, however, for patients with N0 disease, 5-year survival was 50%, with N1 disease 27%, and with N2 disease 15%.

Several of these reported institutional studies included patients who received either preoperative or postoperative radiation, and no consensus can be reached as to the impact of that modality. From the classic study by Shaw and Paulson in 1961, however, we know that preoperative radiation followed by surgical resection in the setting of Pancoast tumor prevents local recurrence [66]. Maybe because, traditionally, radiation therapy has been used in this setting, current interest centers around combined induction chemoradiation protocols for this group of patients, such as in the phase II Southwest Oncology Group Trial 9416, in which two cycles of concurrent etoposide and cisplatin, along with 5 weeks of radiotherapy to 4500 cGy were used. The results are expected imminently.

Tumors less than 2 cm from the carina, but sparing the carina itself, are obviously found in a very small group of patients, which falls very much within the realm of the discussion held for stage III patients without N2 or N3 disease. If this can be proven, these patients may become candidates for technically- challenging resections, as described by Dartwell and colleagues [67].

Only approximately 10% of patients who initially present with lung cancer have intracranial metastases, but approximately 27% of patients with N1 disease have cerebral metastases only [68]. Ninety-four percent of these metastases will be solitary. If left untreated, the patient dies within 1 month from symptomatic cerebral metastases [69]. Magilligan updated his initial report from 1976 after he first reported on combined resection of a synchronous solitary cerebral metastasis and primary non-small- cell lung cancer [70]. This retrospective series included 41 patients, and revealed a 1-year survival of 55%, 31% at 2 years, 21% at 5 years, and 15% at 10 years. Several retrospective series since then have confirmed Magilligan's experience, suggesting a 1-year survival rate of approximately 50%, and a 5-year survival rate ranging from 10% to 30%.

With the recent surge in interest in patients with end-stage emphysema, physicians have become alerted to the fact that approximately 15% of these patients will present with lung masses on CT scan [71]. Fifteen percent thereof, or 2% overall, will have lung cancer. These are clearly patients who fall below standard calculations of PPO FEV1, and therefore, by traditional criteria, have been deemed inoperable. With the extension of the selection criteria, as it has been applied to patients in whom lung volume reduction surgery is used, reports of lung cancer resection in this patient population have appeared [72, 73]. It is the present author's contention that lung volume reduction surgery has slowly evolved towards more lung being resected, and several institutions now perform near lobectomies routinely. It is therefore conceivable that in highly selected patients, the tumor may be within the part of the lung that was to be resected for emphysema, and therefore a lobectomy can be performed [78]. This is especially true for upper lobe lesions.

Whether non-anatomical resections in this patient population prolong life to the point that the patient succumbs to emphysema rather than cancer is unknown. A patient with an FEV of less than 30% has a 5-year life expectancy of 30% [78]. Given the recent warnings about lung volume reduction surgery and the published results favoring a lobectomy versus a limited resection, adherence to the oncological principles, as they are so aptly summarized by Goldstraw [35], should serve as guidance: the operation may be attempted if complete resection with complete staging can be achieved. Data on less than a lobectomy for these patients is missing, but may become available through the National Emphysema Treatment Trial.

Small-cell lung cancer seems to be diminishing, and now makes up less than 20% of all lung cancers. In most centers, small-cell lung cancers are treated by the medical oncologist. There are, however, a number of reports of studies in which small-cell lung cancers, mostly as peripheral nodules, were resected [74]. The most favorable results are reported in patients in whom the diagnosis was either made at surgery or on pathological specimen after resection. A 5-year survival rate of approximately 25% has been reported. The current recommendations are for four cycles of adjuvant chemotherapy based on a number of prospective and retrospective studies that show a benefit even in very early disease stages. Whether surgery should be added to chemotherapy and/or radiation combined chemotherapy has never been tested in a prospective fashion. According to the University of Toronto Lung Oncology Group, for very limited disease treated with surgery and then chemotherapy, similar results to non-small-cell lung cancer stage for stage can be expected [75].

Despite approximately 35,000 people undergoing surgical resection of non-small cell carcinoma for intended cure, recurrent disease will develop in 30 to 80% of those cases. In addition, a second primary lung cancer will occur with an incidence of 1 to 2% per patient year. The topic of follow-up subsequent to lung cancer resection is not at all covered in the 70's and 80's, and only in the last decade have people begun to report on follow-up programs. To date there is no standardized agreed upon follow-up program established. Maybe the most realistic data stems from a survey by Naunheim and coworkers who queried members of the Society of Thoracic Surgeons with 768 respondents. In that survey, it was established that within the initial five years following resection, the respondents performed on average eleven office visits with four complete blood counts and four liver function determinations, eleven chest radiographs, one chest CT, and one sputum cytology. The frequency of visits was greatest within the first two years, and then trailed off. The intensity of follow-up has been greatly debated and a recent study published by Walsh from MD Anderson Cancer Center reported in a retrospective follow-up of 358 patients that only 9% of all patients had a recurrence while they were free of symptoms, and of those, only one-third (2.9%) could be treated with curative intent. This means that overall, only 2% of all patients undergoing surveillance following lung cancer were able to be cured of recurrent cancer. The argument was made that intense frequent follow-up is not cost-effective and that less intense follow-up would not change the outcome at lower overall costs to patients in society. With the current interest in low density helical CAT scans as a screening tool, it would appear that the post-resection group is the ideal patient population for this new modality. The cost of a low density helical CAT scan has been significantly lowered over the past two years, and one current proposal for surveillance suggests to use a low density helical CAT scan twice in the first year, and subsequently once a year up until year five post-resection. Additional tests that may become of value and have been used in the past, but with newer ways of interpretation, may be more helpful, especially in this high-risk group consist of sputum cytology. At this point, the routine use of biologic markers in surveillance cannot be advocated on a routine basis. This technology, although promising, is not commonly available and remains rather expensive.

Surgery in the treatment of lung cancer remains a central pillar. In order to optimize surgical interventions, careful preoperative assessment is necessary. The main risks to be considered are pulmonary and cardiac status. After operability has been established, resctability is assessed by staging the patient. Surgery offers a variety of tools to accomplish complete staging prior to resection. Successful resection is accomplished if the cancer can be removed completely. In order to do so, a multidisciplinary approach is more and more evolving. In non-operative patients, surgical techniques have been developed to deal with airway obstructions and drain fluid and diffusions.