- Email: [email protected]
Survival after resection for lung cancer is the outcome that matters
The American Journal of Surgery 188 (2004) 598 – 602
Scientific paper
Survival after resection for lung cancer is the outcome that matters Michael F. Reed, M.D.a,b,*, Mark Molloy, M.D.a,b, Erica L. Dalton, M.Ed.a,b, John A. Howington, M.D.a,b a
Department of Surgery, Cincinnati Veterans Administration Medical Center, Cincinnati, OH, USA b University of Cincinnati College of Medicine, Cincinnati, OH, USA Manuscript received June 10, 2004; revised manuscript July 26, 2004
Presented at the 28th Annual Symposium of the Association of VA Surgeons, Richmond, Virginia, April 25 –27, 2004
Abstract Background: Lung cancer is the leading cause of cancer mortality in the United States. Stage-specific survival is well documented in national data sets; however, there remains limited recording of longitudinal survival in individual centers. Methods: The VistA Surgery Package was employed to list operations performed by the thoracic surgery service at one Veterans Administration (VA) Medical Center. Results: During a period of 107 months, 416 thoracic operations were performed, 211 of them for lung cancer. Stage distribution was 66% stage I, 18% stage II, 12% stage III, and 4% stage IV. During follow-up, 102 patients died, 57 of them from disease-specific causes. Median survival was 39 months for stage I. Disease-specific median survival was 83 months for stage I, and 5-year survival was 52% (72% for stage IA and 32% for stage IB). Conclusions: Pulmonary resection offers high disease-free survival for early-stage lung cancer. Decentralized hospital computer programming (DHCP) allows individual oncology programs to reliably measure survival. Use of this important outcome measure in quality improvement programs facilitates realistic counseling of patients and meaningful assessments of practice effectiveness. © 2004 Excerpta Medica Inc. All rights reserved. Keywords: Lung cancer; Outcomes; Surgery
Lung cancer causes more deaths in the United States than any other cancer [1]. It results in greater mortality than breast, colon, and prostate cancer combined. Currently, approximately 170,000 Americans are diagnosed annually with lung cancer [2]. Worldwide, ⬎1 million people die each year from lung cancer [3]. The overall 5-year survival rate after the diagnosis of lung cancer is a dismal 14% compared with 86% for breast cancer, 61% for colon cancer, and 96% for prostate cancer [2]. The culprit for this cancer epidemic is tobacco. Indeed, smoking is the principal cause of preventable mortality in the United States and accounting for 1 in 5 deaths. Unfortunately, even if every smoker successfully quit now, lung cancer would still occur for a number of decades. At present, the majority of new cases of lung cancer occur in former smokers. Thus, con-
* Corresponding author. Tel.: ⫹1-513-584-1387; fax: ⫹1-513-5841745. E-mail address: [email protected]
tinued aggressive approaches to prevention, surveillance, and treatment are critical at the national and international levels as well as at individual institutions. The treatment strategy employed for any cancer principally depends on the stage of the disease. In lung cancer, the mainstay of treatment for early-stage disease is surgery. Adjuvant approaches, notably chemotherapy and radiation therapy, constitute the therapeutic modalities—primarily palliative—for advanced-stage cancer. However, even patients who receive complete surgical resection experience a 5-year survival ⬍50%. Low survival rates despite aggressive treatment have contributed to a pervasiveness of nihilism in the management as well as research of lung cancer. Despite the dismal prognosis for most patients with lung cancer, incremental improvement in the management of the disease is taking place. On the epidemiologic front, substantial success has been achieved in smoking cessation, thus resulting in a decreasing number of American smokers. The thoracic surgical management of lung cancer continues to
0002-9610/04/$ – see front matter © 2004 Excerpta Medica Inc. All rights reserved. doi:10.1016/j.amjsurg.2004.07.037
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
improve. Minimally invasive techniques such as video-assisted thoracoscopic surgery (VATS) are now employed routinely to perform pulmonary resections, including lobectomy, with equivalent oncologic outcomes as the traditional open techniques but with decreased morbidity [4,5]. Chemotherapy has long played a central role in the treatment of unresectable lung cancer. Moreover, it has recently been demonstrated to improve survival when administered in an adjuvant manner after complete surgical resection [6]. Radiation therapy is used in the local treatment of unresectable, recurrent, or metastatic disease. Techniques to more accurately deliver therapeutic doses while minimizing toxicity exhibit considerable progress. Thus, it is incumbent on clinicians managing patients with lung cancer to carefully evaluate treatment results and constantly appraise current and novel therapeutic modalities. Limitations exist in determining oncologic success of treatment approaches. Clinical studies, of necessity, often involve small cohorts of highly selected patients. Extrapolation to the general population may not always prove consistent. Larger pools are therefore employed. But national data sets may incorporate a limited number of high-volume centers with institutional and patient characteristics that are not representative of other specific programs. Ideally, individual oncologic programs would measure critical outcomes, especially long-term survival. However, complexities inherent in determining long-term survival have contributed to the adoption of surrogate end points for assessment of treatment success and oncologic program quality. Operative mortality, wound infection rates, and other data points collected in important surveys geared toward surgical quality improvement, such as the National Surgical Quality Improvement Program (NSQIP), are critical. However, they do not address specific oncologic outcomes including long-term survival, disease recurrence, and quality of life. Such surveys may also be limited to a specific field, such as surgery, which is less relevant to the individual cancer patient in whom a multimodality approach incorporating numerous specialties is required. The goal of this study was to focus on long-term survival, an outcome central to assessing the success of treating any patient with cancer. We aimed to exploit the unique capabilities of the VA electronic medical record system to determine stage-specific survival in an individual cancer program and to use this critical metric as a measure of practice effectiveness.
Materials and Methods Operative records for all cases performed by the thoracic surgery service at the Cincinnati VA Medical Center between July 1, 1994, and May 30, 2003, were exported as a tab-delimited text file from the surgical package in VistA (U.S. Government) to a Microsoft Excel (Redmond, WA) spreadsheet. Variables included in this core dataset were
599
patient name, social security number, date of surgery, procedure performed, operative time, attending surgeon, and level of resident training. The electronic medical record for each patient was then accessed to obtain the date of patient birth, histology of the resected specimen, patient status (alive or dead), and date of death or last follow-up. Patients who underwent resection for benign disease or small-cell carcinoma, as well as those who underwent only staging procedures, were excluded from subsequent analysis. Pathologic stage was determined in each case according to American Joint Committee on Cancer (AJCC) guidelines [7]. Survival analyses were performed in StatView (SAS Institute, Inc., Cary, NC) using the Kaplan-Meier method [8].
Results Case ascertainment using the VistA Surgery Package was employed covering a 107-month period from July 1, 1994, through May 31, 2003. This identified 416 thoracic surgical procedures performed at the Cincinnati VA Medical Center. Of these, 211 (51%) were operations for the surgical treatment of lung cancer. Operations for lung cancer staging, such as mediastinoscopy and anterior mediastinotomy (Chamberlain procedure), were not included as surgical treatment operations. Likewise, cases carried out for management of unresectable disease, for instance pleurodesis for malignant pleural effusion and laser ablation of endobronchial tumors, were also excluded. Among the patients who underwent surgery for lung cancer, complete data was available for 197 (93%). Pulmonary resection for non– small cell lung carcinoma (NSCLC) with curative intent was performed on 178 (90%) of the group with complete data available. All further analyses were performed on the 178 individuals who had resection with curative intent. Patient demographics were log captured to Excel from the VistA surgical package. Further variables were collected from the VA electronic medical record (DHCP) and added to the registry. The gender distribution was 175 men and 3 women. The age range was 45 to 90 years (mean 68). The specific operative procedure performed was verified from VistA as well as review of the operative report in DHCP. Lobectomy or bilobectomy accounted for 127 cases (71%), and wedge resection was performed in 38 cases (21%). Pneumonectomy was required in 9 operations (5%). Four patients underwent associated chest wall resection (2%). Final pathology reports in DHCP were reviewed, and the histologic subtypes were squamous cell carcinoma in 86 patients (48%), adenocarcinoma in 80 patients (45%), largecell carcinoma in 5 patients (3%), and unspecified NSCLC in 7 patients (4%). Staging was based on the revised international TNM system [9] accepted by the AJCC [7] and the International Union Against Cancer (IUCC) [10]. The information used for assigning the stage was based on review
600
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
Fig. 3. Longitudinal disease-specific survival. Sixty-month Kaplan-Meier survival curves are plotted for the patients with known cause of death, and excluding patients who died of unknown cause or other non-lung cancer problems, to demonstrate cumulative disease-free survival by stage (I through IV). Fig. 1. Overall survival. The number of patients alive and dead is noted for the overall group as well as for each individual stage (I through IV).
of the pathology reports of the resection specimen and the operative note. The stage distribution was 118 stage I (66%), 31 stage II (18%), 22 stage III (12%), and 7 stage IV (4%). Within stage I, 69 were stage IA (pT1N0M0) and 48 were stage IB (pT2N0M0). Overall survival was determined for the 178 patients treated with pulmonary resection for lung cancer. The average duration of follow-up was 29 months (range 0 to 100). During the study period, 102 patients died (57%). Of the 102 deaths, 61 (60%) were from disease-specific causes; 16 (16%) were related to other causes; and 25 (25%) were from unknown causes (Fig. 1). Longitudinal overall survival was analyzed for all 178 patients who underwent pulmonary resection for curative intent using the method of Kaplan and Meier [8]. The cumulative survival by stage (I to IV) is demonstrated in 60-month survival curves (Fig. 2). The 5-year survival was 35% for stage I, 37% for stage II, 23% for stage III, and 21% for stage IV. These values represent total deaths from all causes. The median survival was 39 months for stage I, 44 months for stage II, 15 months for stage III, and 23 months for stage IV.
Disease-specific survival was then determined to assess the mortality rate due to lung cancer in this cohort of patients who had undergone surgery for cure. The 60-month survival curves separated by stage are shown in Fig. 3. Patients who were alive with disease (n ⫽ 3) were excluded as were those who died from unknown causes (n ⫽ 25) and those who died from other causes (n ⫽ 16), among whom the presence or absence of recurrent lung cancer could not be assessed. The disease-specific 5-year survival for the 134 patients either alive without disease (NED) or dead from lung cancer was 52% for stage I, 48% for stage II, 33% for stage III, and 0% for stage IV. The median survival was 83 months for stage I, 44 months for stage II, 14 months for stage II, and 23 months for stage IV. We then focused on patients with stage I NSCLC who represented those with the best prognosis after surgical resection. Of the 118 stage I cases, 69 were stage IA and 48 were stage IB (one could not be accurately determined A or B). The total 5-year survival was 43% for stage IA and 24% for stage IB. However, disease-specific 5-year survival was much higher in stage IA at 72%, whereas disease-specific 5-year survival was 32% for stage IB. The median survival was 45 months for stage IA and 28 months for stage IB. However, the disease-specific median survival for stage IA had not yet been reached, whereas it was 31 months for stage IB.
Comments
Fig. 2. Longitudinal overall survival. Sixty-month Kaplan-Meier survival curves are plotted for the entire cohort to demonstrate cumulative survival by stage (I through IV) incorporating all causes of mortality.
Information gathered from the VA electronic medical record system is readily amenable to analysis to evaluate important outcome measures at each individual institution. We employed the VistA Surgical Package and DHCP records to determine the demographics and survival rates of patients treated surgically for lung cancer. Nearly all patients were male, which is not unexpected in the VA system in light of its gender mix and the older age of most patients
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
with lung cancer. Unfortunately, new diagnoses of lung cancer in the United States are now nearly equally divided between men and women [2]. The VA system can therefore be expected to treat increasing numbers of women with lung cancer. The wide age range of patients, from 45 to 90 years, is representative of the general population of NSCLC patients. In fact, it is not uncommon to encounter individuals ⬍40 years with lung cancer. The mean age of 68 years is typical for NSCLC. However, many studies favor enrollment of younger patients, excluding those with significant comorbidities and advanced age, and the mean age in those groups is thus lower. In this study, the goal was to accurately represent the entire spectrum of veterans with NSCLC to provide realistic prognostic advice to them. The main operative procedure performed was lobectomy, with the lobectomy-and-bilobectomy group accounting for 71% of cases. This is the optimal management for resection of primary NSCLC, assuming that a complete resection can be achieved by this approach [11]. But wedge resection, including segmentectomy, was used in 21% of cases. Although this approach is a suboptimal oncologic operation, resulting in increased local recurrence rates and decreased survival, it is necessary in certain patients. We reserve wedge resection with curative intent for those with significant pulmonary disease caused, for example, by chronic pulmonary disease or previous lung resection. Pneumonectomy was required in 5% of cases. We attempt to avoid pneumonectomy whenever possible in favor of parenchymal-sparing sleeve lobectomy, but occasional large, central lesions with main bronchial involvement require pneumonectomy. A small number of cases (2%) also required enbloc chest wall resection to achieve negative margins. The type of incision, for instance anterior thoracotomy or muscle-sparing procedure, the need for rib resection, and VATS approach were not recorded in the registry, in part because of less frequent use of minimally invasive approaches in the early period of the study. More recently, general thoracic surgical staff with increased experience in VATS procedures performed all pulmonary resections. Indeed, nearly all recent wedge resections, and most lobectomies for T1 and peripheral T2 lesions, have been performed by way of a VATS approach. All oncologic principles employed in an open lobectomy by way of thoracotomy are followed in performing a VATS lobectomy. Several studies have verified equal oncologic outcomes using VATS lobectomy, with the advantage of diminished morbidity as well as decreased length of stay and duration of chest tube drainage [4,5]. The histologic subtypes were of a typical distribution: 48% squamous cell carcinoma and 45% adenocarcinoma. Immunohistochemical studies— especially using the markers TTF-1, CK7, and CK20 —were used frequently to improve verification of the tumor as a lung primary and to assure it was not a metastasis. The overall survival after curative-intent pulmonary re-
601
section for NSCLC remains disappointing nationwide and in this study where ⬎50% of patients died during follow-up (Fig. 1). Of the 102 deaths, 60% were from NSCLC; 16% were from other causes; and 25% were from unknown cause but without documentation in the electronic medical record of disease recurrence. It would appear that in the VA population, lung cancer is a marker of increased risk of death, even from other causes. Interestingly, when we analyzed longitudinal overall survival by stage, the separation of the curves was limited and had overlap of stages I and II (Fig. 2). The 5-year survival of stages I and II were similar (35% and 37%, respectively) and was slightly lower for stages III (23%) and IV (21%). Median survival followed a comparable pattern with similar stage I and II survival and lower values for stages III and IV. The lack of separation of stages I and II, as well as the number of non–lung cancer deaths, led us to look at patients alive without disease and dead due to lung cancer, namely, longitudinal disease-specific survival. This showed the expected increased separation of stages (Fig. 3). The 5-year disease-specific survival was similar for stages I and II (52% and 48%, respectively) and lower for stage III (33%). Likewise, the median survival exhibited more separation of stages at 83 months for stage I, at 44 months for stage II, and at 15 months for stage III. Stage I is subdivided into IA and IB depending on T status (T1 and T2, respectively). Because the survival for stage I was limited to 52% even when considered in a disease-specific manner, the difference between IA and IB was then evaluated. The overall 5-year survival was 43% and 24% for stages IA and IB with median survivals of 45 months and 28 months, respectively. The disease-specific 5-year survival, however, was quite different at 72% and 32%, suggesting that although the non–lung cancer deaths occurred among all stages, early-stage disease, especially stage IA, had the expected diminished risk of subsequent lung cancer death. Disease-specific median survival was also quite dissimilar within stage I, with median survival not yet reached in IA but 31 months for IB. Thus, pulmonary resection for NSCLC offers high disease-specific survival for very early-stage disease. Moreover, there is a reasonable difference in survival between stage IA and IB, demonstrating that the T status carries important prognostic implications. An important recent report indicates improvement in survival when adjuvant platinum-based chemotherapy is administered after complete resection of NSCLC [6]. The benefit was demonstrated in each T and N designation. The benefit in stage IA, T1N0M0, was not specifically addressed. Our results, with a dramatically lower diseasespecific survival for stage IB compared with stage IA, would favor consideration of adjuvant therapy for IB. The much-improved prognosis for patients with IA disease might argue that the benefits of adjuvant chemotherapy would be less in this group of patients and that its use should be carefully weighed against the toxicity of treatment. This study focused on measuring a critical oncologic
602
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
outcome (i.e., survival), in lung cancer patients after surgical resection at a single VA Medical Center. The information was readily gathered from the electronic medical record that is in operation at all VA institutions nationwide. This system is a valuable resource that can be exploited to allow individual programs, both surgical and nonsurgical, to analyze outcomes that are most relevant to specific diseases. National programs such as the NSQIP offer valuable information geared toward quality improvement in a general sense, but they do not address outcomes that are of concern for each individual disease or medical subspecialty. Published studies of outcomes after treatment for cancer or other diseases typically involve selected patients. Inclusion of patients with recent separate cancers, patients at either extreme of the age spectrum, and patients with significant medical comorbidities often makes accurate conclusions difficult. However, these patients are not excluded from treatment in nonstudy settings. Moreover, referral centers that enroll people in trials may already be biased by the self-selection of patients who are able to travel and interested in participation in protocols. Individual programs can therefore more accurately measure critical outcomes by using data from their own institution. Using the electronic medical record, VA institutions are uniquely positioned to perform these critical self-assessments. With high-volume cancer diagnoses, information gathered from DHCP allows an individual VA oncology program to reliably measure the probability of survival. Although lung cancer is a common disease amenable to this analysis because of its high incidence, a variety of other cancer diagnoses treated surgically could also be assessed. More importantly, a multidisciplinary approach within an individual VA would allow nonsurgical patients, for instance, those with inoperable disease or those with prohibitive medical comorbidities, to also be included in outcome measurements. Furthermore, the outcome after management of diagnoses other that cancer should also be assessed at individual centers using a similar approach. This report suggests that pulmonary resection for earlystage NSCLC at one VA Medical Center offers high disease-specific survival. Importantly, survival for stage IA is
far better than for stage IB NSCLC. The lower survival for resected patients with stages IB and higher supports the use of adjuvant chemotherapy in those case, whereas the better disease-specific survival for patients with resected stage IA suggests that adjuvant therapy might provide a limited survival benefit. In a manner similar to this study, information gathered from DHCP might allow any individual VA oncology program to reliably measure the probability of survival for high-volume cancer diagnoses. Routine use of this important outcome measure in quality improvement programs facilitates realistic counseling of patients and meaningful assessments of practice effectiveness.
References [1] Estimated New Cancer Cases and Deaths by Sex for all Sites: United States Cancer Facts and Figures—1999. Atlanta, GA: American Cancer Society, 1999. [2] Jemal A, Thomas A, Murray T, Thun M. Cancer statistics 2002. CA Cancer J Clin 2002;52:23– 47. [3] Carney DN. Lung cancer–Time to move on from chemotherapy. N Engl J Med 2002;346:126 – 8. [4] Daniels LJ, Balderson SS, Onaitis MW, D’Amico TA. Thoracoscopic lobectomy: A safe and effective strategy for patients with stage I lung cancer. Ann Thorac Surg 2002;74:860 – 4. [5] McKenna RJ Jr., Wolf RK, Brenner M, et al. Is lobectomy by video-assisted thoracic surgery an adequate cancer operation? Ann Thorac Surg 1998;66:1903–7. [6] The International Adjuvant Lung Cancer Trial Collaborative Group. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med 2004;350:351– 60. [7] American Joint Committee on Cancer. Lung. In: Fleming ID, Cooper JS, Hensen DE, et al., editors. Cancer Staging Manual. Philadelphia, PA: Lippincott-Raven; 1997:127–37. [8] Kaplan EL, Meier P. Non-parametric estimation from incomplete observation. J Am Stat Assoc 1958;53:457– 81. [9] Mountain C. Revisions in the International System for Staging Lung Cancer. Chest 1997;111:1710 –7. [10] International Union Against Cancer: Lung and pleural tumors. In: Hermanek P, Hutter RVP, Sobin L, editors. TNM Atlas (Illustrated Guide to the TNM/pTNM Classification). New York, NY: Wiley-Liss; 1997:153–71. [11] Ginsberg RJ, Rubinstein LV. Randomized trial of lobectomy versus limited resection for T1 N0 non-small cell lung cancer. Lung Cancer Study Group. Ann Thorac Surg 1995;60:615–23.
Scientific paper
Survival after resection for lung cancer is the outcome that matters Michael F. Reed, M.D.a,b,*, Mark Molloy, M.D.a,b, Erica L. Dalton, M.Ed.a,b, John A. Howington, M.D.a,b a
Department of Surgery, Cincinnati Veterans Administration Medical Center, Cincinnati, OH, USA b University of Cincinnati College of Medicine, Cincinnati, OH, USA Manuscript received June 10, 2004; revised manuscript July 26, 2004
Presented at the 28th Annual Symposium of the Association of VA Surgeons, Richmond, Virginia, April 25 –27, 2004
Abstract Background: Lung cancer is the leading cause of cancer mortality in the United States. Stage-specific survival is well documented in national data sets; however, there remains limited recording of longitudinal survival in individual centers. Methods: The VistA Surgery Package was employed to list operations performed by the thoracic surgery service at one Veterans Administration (VA) Medical Center. Results: During a period of 107 months, 416 thoracic operations were performed, 211 of them for lung cancer. Stage distribution was 66% stage I, 18% stage II, 12% stage III, and 4% stage IV. During follow-up, 102 patients died, 57 of them from disease-specific causes. Median survival was 39 months for stage I. Disease-specific median survival was 83 months for stage I, and 5-year survival was 52% (72% for stage IA and 32% for stage IB). Conclusions: Pulmonary resection offers high disease-free survival for early-stage lung cancer. Decentralized hospital computer programming (DHCP) allows individual oncology programs to reliably measure survival. Use of this important outcome measure in quality improvement programs facilitates realistic counseling of patients and meaningful assessments of practice effectiveness. © 2004 Excerpta Medica Inc. All rights reserved. Keywords: Lung cancer; Outcomes; Surgery
Lung cancer causes more deaths in the United States than any other cancer [1]. It results in greater mortality than breast, colon, and prostate cancer combined. Currently, approximately 170,000 Americans are diagnosed annually with lung cancer [2]. Worldwide, ⬎1 million people die each year from lung cancer [3]. The overall 5-year survival rate after the diagnosis of lung cancer is a dismal 14% compared with 86% for breast cancer, 61% for colon cancer, and 96% for prostate cancer [2]. The culprit for this cancer epidemic is tobacco. Indeed, smoking is the principal cause of preventable mortality in the United States and accounting for 1 in 5 deaths. Unfortunately, even if every smoker successfully quit now, lung cancer would still occur for a number of decades. At present, the majority of new cases of lung cancer occur in former smokers. Thus, con-
* Corresponding author. Tel.: ⫹1-513-584-1387; fax: ⫹1-513-5841745. E-mail address: [email protected]
tinued aggressive approaches to prevention, surveillance, and treatment are critical at the national and international levels as well as at individual institutions. The treatment strategy employed for any cancer principally depends on the stage of the disease. In lung cancer, the mainstay of treatment for early-stage disease is surgery. Adjuvant approaches, notably chemotherapy and radiation therapy, constitute the therapeutic modalities—primarily palliative—for advanced-stage cancer. However, even patients who receive complete surgical resection experience a 5-year survival ⬍50%. Low survival rates despite aggressive treatment have contributed to a pervasiveness of nihilism in the management as well as research of lung cancer. Despite the dismal prognosis for most patients with lung cancer, incremental improvement in the management of the disease is taking place. On the epidemiologic front, substantial success has been achieved in smoking cessation, thus resulting in a decreasing number of American smokers. The thoracic surgical management of lung cancer continues to
0002-9610/04/$ – see front matter © 2004 Excerpta Medica Inc. All rights reserved. doi:10.1016/j.amjsurg.2004.07.037
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
improve. Minimally invasive techniques such as video-assisted thoracoscopic surgery (VATS) are now employed routinely to perform pulmonary resections, including lobectomy, with equivalent oncologic outcomes as the traditional open techniques but with decreased morbidity [4,5]. Chemotherapy has long played a central role in the treatment of unresectable lung cancer. Moreover, it has recently been demonstrated to improve survival when administered in an adjuvant manner after complete surgical resection [6]. Radiation therapy is used in the local treatment of unresectable, recurrent, or metastatic disease. Techniques to more accurately deliver therapeutic doses while minimizing toxicity exhibit considerable progress. Thus, it is incumbent on clinicians managing patients with lung cancer to carefully evaluate treatment results and constantly appraise current and novel therapeutic modalities. Limitations exist in determining oncologic success of treatment approaches. Clinical studies, of necessity, often involve small cohorts of highly selected patients. Extrapolation to the general population may not always prove consistent. Larger pools are therefore employed. But national data sets may incorporate a limited number of high-volume centers with institutional and patient characteristics that are not representative of other specific programs. Ideally, individual oncologic programs would measure critical outcomes, especially long-term survival. However, complexities inherent in determining long-term survival have contributed to the adoption of surrogate end points for assessment of treatment success and oncologic program quality. Operative mortality, wound infection rates, and other data points collected in important surveys geared toward surgical quality improvement, such as the National Surgical Quality Improvement Program (NSQIP), are critical. However, they do not address specific oncologic outcomes including long-term survival, disease recurrence, and quality of life. Such surveys may also be limited to a specific field, such as surgery, which is less relevant to the individual cancer patient in whom a multimodality approach incorporating numerous specialties is required. The goal of this study was to focus on long-term survival, an outcome central to assessing the success of treating any patient with cancer. We aimed to exploit the unique capabilities of the VA electronic medical record system to determine stage-specific survival in an individual cancer program and to use this critical metric as a measure of practice effectiveness.
Materials and Methods Operative records for all cases performed by the thoracic surgery service at the Cincinnati VA Medical Center between July 1, 1994, and May 30, 2003, were exported as a tab-delimited text file from the surgical package in VistA (U.S. Government) to a Microsoft Excel (Redmond, WA) spreadsheet. Variables included in this core dataset were
599
patient name, social security number, date of surgery, procedure performed, operative time, attending surgeon, and level of resident training. The electronic medical record for each patient was then accessed to obtain the date of patient birth, histology of the resected specimen, patient status (alive or dead), and date of death or last follow-up. Patients who underwent resection for benign disease or small-cell carcinoma, as well as those who underwent only staging procedures, were excluded from subsequent analysis. Pathologic stage was determined in each case according to American Joint Committee on Cancer (AJCC) guidelines [7]. Survival analyses were performed in StatView (SAS Institute, Inc., Cary, NC) using the Kaplan-Meier method [8].
Results Case ascertainment using the VistA Surgery Package was employed covering a 107-month period from July 1, 1994, through May 31, 2003. This identified 416 thoracic surgical procedures performed at the Cincinnati VA Medical Center. Of these, 211 (51%) were operations for the surgical treatment of lung cancer. Operations for lung cancer staging, such as mediastinoscopy and anterior mediastinotomy (Chamberlain procedure), were not included as surgical treatment operations. Likewise, cases carried out for management of unresectable disease, for instance pleurodesis for malignant pleural effusion and laser ablation of endobronchial tumors, were also excluded. Among the patients who underwent surgery for lung cancer, complete data was available for 197 (93%). Pulmonary resection for non– small cell lung carcinoma (NSCLC) with curative intent was performed on 178 (90%) of the group with complete data available. All further analyses were performed on the 178 individuals who had resection with curative intent. Patient demographics were log captured to Excel from the VistA surgical package. Further variables were collected from the VA electronic medical record (DHCP) and added to the registry. The gender distribution was 175 men and 3 women. The age range was 45 to 90 years (mean 68). The specific operative procedure performed was verified from VistA as well as review of the operative report in DHCP. Lobectomy or bilobectomy accounted for 127 cases (71%), and wedge resection was performed in 38 cases (21%). Pneumonectomy was required in 9 operations (5%). Four patients underwent associated chest wall resection (2%). Final pathology reports in DHCP were reviewed, and the histologic subtypes were squamous cell carcinoma in 86 patients (48%), adenocarcinoma in 80 patients (45%), largecell carcinoma in 5 patients (3%), and unspecified NSCLC in 7 patients (4%). Staging was based on the revised international TNM system [9] accepted by the AJCC [7] and the International Union Against Cancer (IUCC) [10]. The information used for assigning the stage was based on review
600
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
Fig. 3. Longitudinal disease-specific survival. Sixty-month Kaplan-Meier survival curves are plotted for the patients with known cause of death, and excluding patients who died of unknown cause or other non-lung cancer problems, to demonstrate cumulative disease-free survival by stage (I through IV). Fig. 1. Overall survival. The number of patients alive and dead is noted for the overall group as well as for each individual stage (I through IV).
of the pathology reports of the resection specimen and the operative note. The stage distribution was 118 stage I (66%), 31 stage II (18%), 22 stage III (12%), and 7 stage IV (4%). Within stage I, 69 were stage IA (pT1N0M0) and 48 were stage IB (pT2N0M0). Overall survival was determined for the 178 patients treated with pulmonary resection for lung cancer. The average duration of follow-up was 29 months (range 0 to 100). During the study period, 102 patients died (57%). Of the 102 deaths, 61 (60%) were from disease-specific causes; 16 (16%) were related to other causes; and 25 (25%) were from unknown causes (Fig. 1). Longitudinal overall survival was analyzed for all 178 patients who underwent pulmonary resection for curative intent using the method of Kaplan and Meier [8]. The cumulative survival by stage (I to IV) is demonstrated in 60-month survival curves (Fig. 2). The 5-year survival was 35% for stage I, 37% for stage II, 23% for stage III, and 21% for stage IV. These values represent total deaths from all causes. The median survival was 39 months for stage I, 44 months for stage II, 15 months for stage III, and 23 months for stage IV.
Disease-specific survival was then determined to assess the mortality rate due to lung cancer in this cohort of patients who had undergone surgery for cure. The 60-month survival curves separated by stage are shown in Fig. 3. Patients who were alive with disease (n ⫽ 3) were excluded as were those who died from unknown causes (n ⫽ 25) and those who died from other causes (n ⫽ 16), among whom the presence or absence of recurrent lung cancer could not be assessed. The disease-specific 5-year survival for the 134 patients either alive without disease (NED) or dead from lung cancer was 52% for stage I, 48% for stage II, 33% for stage III, and 0% for stage IV. The median survival was 83 months for stage I, 44 months for stage II, 14 months for stage II, and 23 months for stage IV. We then focused on patients with stage I NSCLC who represented those with the best prognosis after surgical resection. Of the 118 stage I cases, 69 were stage IA and 48 were stage IB (one could not be accurately determined A or B). The total 5-year survival was 43% for stage IA and 24% for stage IB. However, disease-specific 5-year survival was much higher in stage IA at 72%, whereas disease-specific 5-year survival was 32% for stage IB. The median survival was 45 months for stage IA and 28 months for stage IB. However, the disease-specific median survival for stage IA had not yet been reached, whereas it was 31 months for stage IB.
Comments
Fig. 2. Longitudinal overall survival. Sixty-month Kaplan-Meier survival curves are plotted for the entire cohort to demonstrate cumulative survival by stage (I through IV) incorporating all causes of mortality.
Information gathered from the VA electronic medical record system is readily amenable to analysis to evaluate important outcome measures at each individual institution. We employed the VistA Surgical Package and DHCP records to determine the demographics and survival rates of patients treated surgically for lung cancer. Nearly all patients were male, which is not unexpected in the VA system in light of its gender mix and the older age of most patients
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
with lung cancer. Unfortunately, new diagnoses of lung cancer in the United States are now nearly equally divided between men and women [2]. The VA system can therefore be expected to treat increasing numbers of women with lung cancer. The wide age range of patients, from 45 to 90 years, is representative of the general population of NSCLC patients. In fact, it is not uncommon to encounter individuals ⬍40 years with lung cancer. The mean age of 68 years is typical for NSCLC. However, many studies favor enrollment of younger patients, excluding those with significant comorbidities and advanced age, and the mean age in those groups is thus lower. In this study, the goal was to accurately represent the entire spectrum of veterans with NSCLC to provide realistic prognostic advice to them. The main operative procedure performed was lobectomy, with the lobectomy-and-bilobectomy group accounting for 71% of cases. This is the optimal management for resection of primary NSCLC, assuming that a complete resection can be achieved by this approach [11]. But wedge resection, including segmentectomy, was used in 21% of cases. Although this approach is a suboptimal oncologic operation, resulting in increased local recurrence rates and decreased survival, it is necessary in certain patients. We reserve wedge resection with curative intent for those with significant pulmonary disease caused, for example, by chronic pulmonary disease or previous lung resection. Pneumonectomy was required in 5% of cases. We attempt to avoid pneumonectomy whenever possible in favor of parenchymal-sparing sleeve lobectomy, but occasional large, central lesions with main bronchial involvement require pneumonectomy. A small number of cases (2%) also required enbloc chest wall resection to achieve negative margins. The type of incision, for instance anterior thoracotomy or muscle-sparing procedure, the need for rib resection, and VATS approach were not recorded in the registry, in part because of less frequent use of minimally invasive approaches in the early period of the study. More recently, general thoracic surgical staff with increased experience in VATS procedures performed all pulmonary resections. Indeed, nearly all recent wedge resections, and most lobectomies for T1 and peripheral T2 lesions, have been performed by way of a VATS approach. All oncologic principles employed in an open lobectomy by way of thoracotomy are followed in performing a VATS lobectomy. Several studies have verified equal oncologic outcomes using VATS lobectomy, with the advantage of diminished morbidity as well as decreased length of stay and duration of chest tube drainage [4,5]. The histologic subtypes were of a typical distribution: 48% squamous cell carcinoma and 45% adenocarcinoma. Immunohistochemical studies— especially using the markers TTF-1, CK7, and CK20 —were used frequently to improve verification of the tumor as a lung primary and to assure it was not a metastasis. The overall survival after curative-intent pulmonary re-
601
section for NSCLC remains disappointing nationwide and in this study where ⬎50% of patients died during follow-up (Fig. 1). Of the 102 deaths, 60% were from NSCLC; 16% were from other causes; and 25% were from unknown cause but without documentation in the electronic medical record of disease recurrence. It would appear that in the VA population, lung cancer is a marker of increased risk of death, even from other causes. Interestingly, when we analyzed longitudinal overall survival by stage, the separation of the curves was limited and had overlap of stages I and II (Fig. 2). The 5-year survival of stages I and II were similar (35% and 37%, respectively) and was slightly lower for stages III (23%) and IV (21%). Median survival followed a comparable pattern with similar stage I and II survival and lower values for stages III and IV. The lack of separation of stages I and II, as well as the number of non–lung cancer deaths, led us to look at patients alive without disease and dead due to lung cancer, namely, longitudinal disease-specific survival. This showed the expected increased separation of stages (Fig. 3). The 5-year disease-specific survival was similar for stages I and II (52% and 48%, respectively) and lower for stage III (33%). Likewise, the median survival exhibited more separation of stages at 83 months for stage I, at 44 months for stage II, and at 15 months for stage III. Stage I is subdivided into IA and IB depending on T status (T1 and T2, respectively). Because the survival for stage I was limited to 52% even when considered in a disease-specific manner, the difference between IA and IB was then evaluated. The overall 5-year survival was 43% and 24% for stages IA and IB with median survivals of 45 months and 28 months, respectively. The disease-specific 5-year survival, however, was quite different at 72% and 32%, suggesting that although the non–lung cancer deaths occurred among all stages, early-stage disease, especially stage IA, had the expected diminished risk of subsequent lung cancer death. Disease-specific median survival was also quite dissimilar within stage I, with median survival not yet reached in IA but 31 months for IB. Thus, pulmonary resection for NSCLC offers high disease-specific survival for very early-stage disease. Moreover, there is a reasonable difference in survival between stage IA and IB, demonstrating that the T status carries important prognostic implications. An important recent report indicates improvement in survival when adjuvant platinum-based chemotherapy is administered after complete resection of NSCLC [6]. The benefit was demonstrated in each T and N designation. The benefit in stage IA, T1N0M0, was not specifically addressed. Our results, with a dramatically lower diseasespecific survival for stage IB compared with stage IA, would favor consideration of adjuvant therapy for IB. The much-improved prognosis for patients with IA disease might argue that the benefits of adjuvant chemotherapy would be less in this group of patients and that its use should be carefully weighed against the toxicity of treatment. This study focused on measuring a critical oncologic
602
M. Reed et al. / The American Journal of Surgery 188 (2004) 598 – 602
outcome (i.e., survival), in lung cancer patients after surgical resection at a single VA Medical Center. The information was readily gathered from the electronic medical record that is in operation at all VA institutions nationwide. This system is a valuable resource that can be exploited to allow individual programs, both surgical and nonsurgical, to analyze outcomes that are most relevant to specific diseases. National programs such as the NSQIP offer valuable information geared toward quality improvement in a general sense, but they do not address outcomes that are of concern for each individual disease or medical subspecialty. Published studies of outcomes after treatment for cancer or other diseases typically involve selected patients. Inclusion of patients with recent separate cancers, patients at either extreme of the age spectrum, and patients with significant medical comorbidities often makes accurate conclusions difficult. However, these patients are not excluded from treatment in nonstudy settings. Moreover, referral centers that enroll people in trials may already be biased by the self-selection of patients who are able to travel and interested in participation in protocols. Individual programs can therefore more accurately measure critical outcomes by using data from their own institution. Using the electronic medical record, VA institutions are uniquely positioned to perform these critical self-assessments. With high-volume cancer diagnoses, information gathered from DHCP allows an individual VA oncology program to reliably measure the probability of survival. Although lung cancer is a common disease amenable to this analysis because of its high incidence, a variety of other cancer diagnoses treated surgically could also be assessed. More importantly, a multidisciplinary approach within an individual VA would allow nonsurgical patients, for instance, those with inoperable disease or those with prohibitive medical comorbidities, to also be included in outcome measurements. Furthermore, the outcome after management of diagnoses other that cancer should also be assessed at individual centers using a similar approach. This report suggests that pulmonary resection for earlystage NSCLC at one VA Medical Center offers high disease-specific survival. Importantly, survival for stage IA is
far better than for stage IB NSCLC. The lower survival for resected patients with stages IB and higher supports the use of adjuvant chemotherapy in those case, whereas the better disease-specific survival for patients with resected stage IA suggests that adjuvant therapy might provide a limited survival benefit. In a manner similar to this study, information gathered from DHCP might allow any individual VA oncology program to reliably measure the probability of survival for high-volume cancer diagnoses. Routine use of this important outcome measure in quality improvement programs facilitates realistic counseling of patients and meaningful assessments of practice effectiveness.
References [1] Estimated New Cancer Cases and Deaths by Sex for all Sites: United States Cancer Facts and Figures—1999. Atlanta, GA: American Cancer Society, 1999. [2] Jemal A, Thomas A, Murray T, Thun M. Cancer statistics 2002. CA Cancer J Clin 2002;52:23– 47. [3] Carney DN. Lung cancer–Time to move on from chemotherapy. N Engl J Med 2002;346:126 – 8. [4] Daniels LJ, Balderson SS, Onaitis MW, D’Amico TA. Thoracoscopic lobectomy: A safe and effective strategy for patients with stage I lung cancer. Ann Thorac Surg 2002;74:860 – 4. [5] McKenna RJ Jr., Wolf RK, Brenner M, et al. Is lobectomy by video-assisted thoracic surgery an adequate cancer operation? Ann Thorac Surg 1998;66:1903–7. [6] The International Adjuvant Lung Cancer Trial Collaborative Group. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med 2004;350:351– 60. [7] American Joint Committee on Cancer. Lung. In: Fleming ID, Cooper JS, Hensen DE, et al., editors. Cancer Staging Manual. Philadelphia, PA: Lippincott-Raven; 1997:127–37. [8] Kaplan EL, Meier P. Non-parametric estimation from incomplete observation. J Am Stat Assoc 1958;53:457– 81. [9] Mountain C. Revisions in the International System for Staging Lung Cancer. Chest 1997;111:1710 –7. [10] International Union Against Cancer: Lung and pleural tumors. In: Hermanek P, Hutter RVP, Sobin L, editors. TNM Atlas (Illustrated Guide to the TNM/pTNM Classification). New York, NY: Wiley-Liss; 1997:153–71. [11] Ginsberg RJ, Rubinstein LV. Randomized trial of lobectomy versus limited resection for T1 N0 non-small cell lung cancer. Lung Cancer Study Group. Ann Thorac Surg 1995;60:615–23.