- Email: [email protected]
Intraoperative Oncologic Staging and Outcomes for Lung Cancer Resection Vary by Surgeon Specialty
GENERAL THORACIC
Intraoperative Oncologic Staging and Outcomes for Lung Cancer Resection Vary by Surgeon Specialty Michelle C. Ellis, MD, Brian S. Diggs, PhD, John T. Vetto, MD, and Paul H. Schipper, MD Department of Surgery, Division of Cardiothoracic Surgery, and Division of Surgical Oncology, Oregon Health & Science University, Portland, Oregon
Background. In the United States the majority of lung cancer resections are performed by general surgeons, although surgeons specializing in thoracic surgery have demonstrated superior perioperative and long-term oncologic outcomes. Why these differences exist has not been well studied. We hypothesized that the completeness of intraoperative oncologic staging may explain some of these differences. Methods. The Nationwide Inpatient Sample (NIS) database was used to review 222,233 patients with primary lung cancer treated surgically with wedge resection, segmentectomy, lobectomy, or pneumonectomy from 1998 to 2007. Surgeons were classified as general thoracic surgeons if they performed greater than 75% general thoracic operations and less than 10% cardiac operations; they were classified as cardiac surgeons if they performed greater than 10% cardiac operations; they were classified as general surgeons if they performed less than 75% thoracic operations and less than 10% cardiac operations. The main outcome measure was the performance
of lymphadenectomy or mediastinoscopy during the same admission as the cancer resection. Results. The overall lymphadenectomy rate was 56% (n ⴝ 125,115) and was highest for general thoracic surgeons at 73% (n ⴝ 13,313), followed by 55% (n ⴝ 65,453) for general surgeons, and 54% (n ⴝ 46,349) for cardiac surgeons (p < 0.0001). General surgeons had a significantly higher risk for in-hospital mortality (odds ratio [OR], 1.47; confidence interval [CI], 1.14 to 1.90; p ⴝ 0.003) and postoperative complications (OR, 1.17; CI, 1.00 to 1.36; p ⴝ 0.043) compared with general thoracic surgeons. Conclusions. Surgeon specialty impacts the adequacy of oncologic staging in patients undergoing resection for primary lung cancer. Specifically, general thoracic surgeons performed intraoperative oncologic staging significantly more often than did their general surgeon and cardiac surgeon counterparts while achieving significantly lower in-hospital mortality and complication rates. (Ann Thorac Surg 2011;92:1958 – 64) © 2011 by The Society of Thoracic Surgeons
I
database, we sought to establish the frequency of lymphadenectomy or mediastinoscopy, or both, at the time of a lung cancer resection. Specifically, we looked for predictors of lymphadenectomy and trends over time.
n the United States, the majority of thoracic procedures are performed by general surgeons; however surgeons specializing in thoracic surgery have lower perioperative mortality and morbidity rates compared with general surgeons [1–3]. Patients who have lung resection for cancer performed by a board-certified cardiothoracic surgeon specializing in general thoracic surgery have a longer overall and cancer-specific 5-year survival [4]. Lymph node status is a main determinant of stage and prognosis in lung cancer, and a mediastinal lymphadenectomy has been recommended for all patients undergoing surgical resection for this disease [5– 8]. The current study sought to determine whether the completeness of intraoperative oncologic staging might explain the difference in surgical outcomes for lung cancer resections performed by various types of surgeon. Using the Nationwide Inpatient Sample (NIS)
Accepted for publication May 31, 2011. Presented at the Forty-seventh Annual Meeting of The Society of Thoracic Surgeons, San Diego, CA, Jan 31–Feb 2, 2011. Address correspondence to Dr Schipper, Division of Cardiothoracic Surgery, Oregon Health & Science University, Mail Code L353, 3181 SW Sam Jackson Park Rd, Portland, OR 97239; e-mail: [email protected].
© 2011 by The Society of Thoracic Surgeons Published by Elsevier Inc
Patients and Methods Design/Data Source We performed a retrospective review of the NIS database. The NIS is managed under the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality [9]. It is the largest all-payer database of national hospital discharges and contains a 20% stratified sample of nonfederal acute-care hospitals in the United States. Each record within the NIS database represents a single hospital discharge and contains information on demographic data, admission type, primary and secondary diagnoses and procedures, length of stay, hospital demographics, insurance payers, total hospital charges, and hospital transfer status. Statistical weights and other sampling variables are included that allow national estimates to be made from the sample. Institutional review board approval was waived for this study because of the deidentified nature of the data. 0003-4975/$36.00 doi:10.1016/j.athoracsur.2011.05.120
Inclusion Criteria We searched the NIS using the International Classification of Diseases, 9th revision, Clinical Modification (ICD9-CM) looking for the code for primary lung cancer (ICD-9-CM diagnosis code 162.⫻) in patients who were treated surgically from 1998 to 2007 as designated by an ICD-9-CM procedure code of wedge resection (32.9), segmentectomy (32.3⫻), lobectomy (32.4⫻), or pneumonectomy (32.5⫻). Primary outcome was the presence of lymphadenectomy (ICD-9-CM codes 40.11, 40.22, 40.29, 40.3, 40.50, or 40.59) or mediastinoscopy (ICD-9-CM code 34.22) during the same admission as the cancer resection. Secondary measures were patient demographics (sex, age on admission, insurance payer), hospital factors (rural versus urban, teaching status), and procedure factors (in-hospital mortality, complication rates). (The ICD-9-CM codes used for the definitions of cardiac and thoracic surgeons are available by contacting the authors.).
Classification of Surgeon Type Similar to our previous work [1], individual surgeons performing these procedures were divided into 3 groups based on the percentage of their total case volume spent doing general thoracic surgery, cardiac surgery, or any other type of surgery. A surgeon was considered a general thoracic surgeon if greater than 75% of total procedures performed were from ICD-9-CM codes considered to be general thoracic operations and no more than 10% of total procedures performed were from ICD9-CM codes considered to be cardiac operations. A surgeon was considered a cardiac surgeon if greater than 10% of procedures performed were cardiac in nature. A surgeon was considered a general surgeon if fewer than 75% of total procedures performed were general thoracic surgical procedures and fewer than 10% of total procedures performed were cardiac operations. Board certification and subspecialty training were not considered in these definitions, as this information was not available in the NIS. Also, because surgeon identifiers were not necessarily consistent from year to year, all surgeons were given year-specific identifiers and were classified based on that year’s case mix.a Lymphadenectomy rates by surgeon type were compared as a whole over the study period (1998 to 2007) as were year-by-year trends. Multivariate models were then constructed to determine predictors for lymphadenectomy, mediastinoscopy, inhospital mortality, and in-hospital morbidity.
Evaluation of Case Mix Definition Cutoffs A final multivariate model was constructed to evaluate the case mix definition of a general thoracic surgeon—in particular the cutoff of a 75% general thoracic case mix. Surgeons were divided into 6 groups. The first group contained the previous definition of a cardiac surgeon. The 6 remaining groups were defined based on the percentage of cases considered general thoracic surgical procedures: 0% to 20%, 21% to 40%, 41% to 60%, 61% to 80%, and 81% to 100%, while still maintaining that no
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
1959
more than 10% of total procedures performed were from ICD-9-CM codes considered cardiac operations.
Statistical Analysis All analyses used versions of statistical tests that take into account the hierarchical sampling structure of the NIS. Categorical comparisons were made using 2 tests. Multivariate analysis used binary logistic regression models. Multiple logistic regression models were created for each outcome, each with additional variables to assess the descriptive ability of the new variables and their effect on the already included variables. Surgeon volume was included as a logarithmically transformed variable to reflect the belief that as overall volume increases, improvements to outcome would come from increasing volume by some percentage, rather than increasing volume by some absolute number of cases. A significant difference identified in surgeon volume using this method would indicate that for each doubling of the dependent variable (surgeon volume) there was a difference in the desired outcome from the previous doubling. For example a surgeon performing 8 procedures was different in the examined outcome from the surgeon performing 4 procedures and the surgeon performing 16 procedures was different from the 1 performing 8 procedures. Statistical significance was taken as p ⬍ 0.05.
Results Based on the selection criteria, 71,314 records were selected from the NIS, representing a national estimate of 352,046 cases. An anonymous surgeon identifier was missing in 129,813 cases; these were excluded resulting in a study population of 222,233. The excluded patients were similar demographically to the final study population. Of the 7,770 unique surgeon-year combinations, 61.7% (n ⫽ 4,791) were considered general surgeons, 2.9% (n ⫽ 227) were considered general thoracic surgeons, and 35.4% (n ⫽ 2,752) were considered cardiac surgeons. The demographics of the study population are listed in Table 1. Lung cancer resections were performed by general surgeons in 53.5% (n ⫽ 118,843) of patients, by thoracic surgeons in 8.2% (n ⫽ 18,284) of patients, and by cardiac surgeons in 38.3% (n ⫽ 85,106) of patients. Patients were similar with respect to age, sex, and procedure type performed among the surgeon types. Thoracic surgeons differed with respect to practice location, primary insurance payer, and patient income level (Table 1). The median number of general thoracic surgical procedures performed by general thoracic surgeons was 21 (interquartile range [IQR]; 4 to 85), for general surgeons the median number was 8 (IQR, 3 to 20), and for cardiac surgeons it was 23 (IQR, 12 to 39).
Lymphadenectomy and Mediastinoscopy Rates Lymphadenectomy was performed in 125,115 patients, representing an overall lymphadenectomy rate of 56.3%. Mediastinoscopy was performed in 24,083 patients and represented an overall rate of 10.8%. When lymphadenectomy rates were assessed by surgeon type, general
GENERAL THORACIC
Ann Thorac Surg 2011;92:1958 – 64
GENERAL THORACIC
1960
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
Table 1. Study Population Demographics Variable n Age (Years) (%) 18–50 51–70 71–80 81⫹ Sex (%) Male Female Procedure (%) Limited resection Lobectomy Pneumonectomy Location/status of hospital (%) Rural Urban nonteaching Urban teaching Primary payer (%) Medicare Private insurance Self-pay Other/unknown Income Quartiles (%) Lowest Second lowest Second highest Highest CS ⫽ cardiac surgeon; thoracic surgeon.
GS
GTS
CS
118,843 (53.5%)
18,284 (8.2%)
85,106 (38.3%)
7.6 52.1 33.7 6.6
8.3 53.6 31.4 6.6
7.2 53.7 33.2 5.9
51.7 48.3
50.5 49.5
54.0 46.0
13.3 77.8 8.9
12.9 78.7 8.4
8.9 81.3 9.8
Ann Thorac Surg 2011;92:1958 – 64
gical procedures performed was added to the multivariate model, the surgeon specialty definitions based on percentage of cases performed were no longer significant predictors of lymphadenectomy. However the volume of general thoracic surgical cases performed was a significant predictor of lymphadenectomy. For every doubling of general thoracic case volume there was a significant increase in the likelihood that a lymphadenectomy (odds ratio [OR], 1.28; 95% confidence interval [CI], 1.24 to 1.32; p ⬍ 0.001) and mediastinoscopy (OR, 1.19; CI, 1.11 to 1.28; p ⬍ 0.001) were performed. Conversely for every doubling of general surgery case volume there was a significant decrease in the likelihood a lymphadenectomy was performed (OR, 0.95; CI, 0.92 to 0.97; p ⬍ 0.001). Changes in cardiac surgery case volume did not significantly impact lymphadenectomy or mediastinoscopy rates in this multivariate model.
Mortality Rates 7.7 41.0 51.2
0.4 10.0 89.6
5.4 43.7 50.9
57.3 34.6 1.6 6.5
53.1 39.8 1.9 5.2
59.1 32.7 1.6 6.6
14.0 26.8 26.6 30.3
6.9 18.6 22.3 49.9
17.1 27.9 28.5 23.8
GS ⫽ general surgeon;
GTS ⫽ general
thoracic surgeons performed lymphadenectomy in 72.8% of patients compared with 55.1% and 54.5% for general surgeons and cardiac surgeons, respectively (p ⬍ 0.0001). Similarly mediastinoscopy rates were highest for general thoracic surgeons at 15.8%, followed by 10.9% and 9.6% for general and cardiac surgeons, respectively (p ⬍ 0.0001). A multivariate analysis was performed to determine predictors for lymphadenectomy. Patients less likely to have lymphadenectomy were those who were in the bottom 50% of income level, insured by Medicare as opposed to private insurance or self-pay, received their care at a rural or urban nonteaching hospital, or had resection performed by a cardiac or general surgeon. Specifically, a patient was more than twice as likely to have a lymphadenectomy performed if the lung cancer resection was completed by a general thoracic surgeon rather than a cardiac or general surgeon (Table 2). Patients who were more likely to undergo lymphadenectomy were women, those who had a lobectomy as opposed to pneumonectomy or limited lung resection, or those who had the procedure performed by a general thoracic surgeon. When volume of general thoracic sur-
When comparing the 3 surgeon types, lung cancer resections performed by general surgeons had a significantly higher risk for in-hospital mortality (OR, 1.55; CI, 1.20 to
Table 2. Predictors of Lymphadenectomy Variable Age (Years) 18–50 51–70 71–80 80⫹ Sex Male Female Surgeon type Thoracic surgeon General surgeon Cardiac surgeon Primary payer Medicare Private Insurance Self-pay Other/unknown Income level Lowest quartile Second lowest quartile Second highest quartile Highest quartile Procedure performed Limited lung resection Lobectomy Pneumonectomy Location/status of hospital Rural Urban nonteaching Urban teaching CI ⫽ confidence interval.
Odds Ratio (CI)
p Value
1.0 (Reference) 0.99 (0.92-1.07) 0.98 (0.90-1.08) 0.91 (0.81-1.03)
— 0.877 0.710 0.136
1.0 (Reference) 1.09 (1.05–1.14)
— ⬍0.001
1.0 (Reference) 0.47 (0.35–0.65) 0.47 (0.35–0.64)
— ⬍0.001 ⬍0.001
0.93 (0.87–0.99) 1.0 (Reference) 0.91 (0.73–1.14) 0.90 (0.81–1.00)
0.022 — 0.414 0.041
0.74 (0.66–0.83) 0.89 (0.81–0.97) 0.93 (0.86-1.00) 1.0 (Reference)
⬍0.001 0.009 0.054 —
1.0 (Reference) 1.38 (1.27–1.49) 0.97 (0.86–1.09)
— ⬍0.001 0.631
0.60 (0.44–0.81) 0.74 (0.66–0.83) 1.0 (Reference)
⬍0.001 ⬍0.001 —
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
Table 3. Predictors for In-Hospital Mortality Variable Age (Years) 18–50 51–70 71–80 80⫹ Sex Male Female Surgeon type Thoracic surgeon General surgeon Cardiac surgeon Primary payer Medicare Private Insurance Self-pay Other/unknown Income level Lowest quartile Second lowest quartile Second highest quartile Highest quartile Procedure performed Limited lung resection Lobectomy Pneumonectomy Location/status of hospital Rural Urban nonteaching Urban teaching
Odds Ratio (CI)
p Value
1.0 (Reference) 1.40 (1.05–1.87) 2.44 (1.80–3.29) 3.14 (2.24–4.39)
— 0.024 ⬍0.001 ⬍0.001
1.0 (Reference) 0.68 (0.60–0.76)
— ⬍0.001
1.0 (Reference) 1.79 (1.41–2.25) 1.50 (1.18–1.91)
— ⬍0.001 ⬍0.001
1.61 (1.37–1.89) 1.0 (Reference) 1.02 (0.59–1.77) 1.39 (1.07–1.81)
⬍0.001 — 0.945 0.013
1.28 (1.06–1.55) 1.25 (1.08–1.44) 1.10 (0.95–1.28) 1.0 (Reference)
0.010 0.003 0.207 —
1.0 (Reference) 1.08 (0.89–1.30) 3.65 (2.93–4.55)
— 0.456 ⬍0.001
0.93 (0.73–1.20) 1.07 (0.95–1.21) 1.0 (Reference)
0.602 0.257 —
1961
ative complications. Female sex was significantly protective against complications.
Trends Over Time
CI ⫽ confidence interval.
The study population was assessed as a whole and by surgeon type over the study period (1998 to 2007) to determine trends in patient demographics, lymphadenectomy and mediastinoscopy rates, in-hospital mortality, and complication rates. The study population as a whole remained relatively similar in terms of age, sex, insurance type, and location of care. The overall population was composed of a higher proportion of lowest income patients after 2002, but this was due to database adjustments in the definition of income quartiles. The overall in-hospital mortality remained relatively stable throughout the study period, whereas the overall lymphadenectomy rate rose steadily (Fig 1). When trends over time were assessed by surgeon type, significant changes in the lymphadenectomy rates were seen. Thoracic surgeon lymphadenectomy rates dipped in the midportion of the study period, whereas general and cardiac surgeon rates steadily increased (Fig 1).
Evaluation of Case Mix Definition Cutoffs A surgeon performing 61% to 80% of cases as general thoracic surgical procedures was just as likely to perform a lymphadenectomy as a surgeon with an 81% to 100% general thoracic case mix. Surgeons in each group— 0% to 20%, 21% to 40%, or 41% to 60% general thoracic cases—were significantly less likely to perform a lymphadenectomy compared with surgeons who performed 81% to 100% general thoracic procedures. The likelihood of performing lymphadenectomy at the time of lung cancer resection based on general thoracic surgery case mix is shown in Table 4.
2.01; p ⬍ 0.001) compared with those performed by general thoracic surgeons. Unadjusted in-hospital mortality rates for thoracic, cardiac, and general surgeons were 2.3%, 3.4%, and 4.0%, respectively. Age greater than 50 years, Medicare rather than private insurance as primary payer, a procedure of pneumonectomy compared with more limited lung resection, and a non– thoracic surgeons performing the operations were all significant predictors for increased mortality on multivariate analysis. Female sex was significantly protective against in-hospital mortality (Table 3).
Complication Rates When comparing the 3 surgeon types, lung cancer resections performed by general surgeons had a significantly higher risk for postoperative complications (OR, 1.16; CI, 1.00 to 1.35; p ⫽ 0.049) compared with those performed by thoracic surgeons. The overall complication rate for thoracic, cardiac, and general surgeons was 28.5%, 31.3%, and 33.5%, respectively. On multivariate analysis, age greater than 50 years and general surgeons performing the procedures were significant predictors for postoper-
Fig 1. Lymphadenectomy rates overall and by surgeon type over time.
GENERAL THORACIC
Ann Thorac Surg 2011;92:1958 – 64
GENERAL THORACIC
1962
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
Table 4. Likelihood of Lymphadenectomy by Case Mix Type of Case General Thoracic Cases (%) 0–20 21–40 41–60 61–80 81–100 Cardiac Surgeon Cases
Odds Ratio (95% CI)
p Value
0.29 (0.21–0.41) 0.56 (0.39–0.80) 0.64 (0.44–0.93) 0.89 (0.63–1.26) 1.0 (Reference) 0.44 (0.80–1.02)
⬍0.001 0.002 0.020 0.527 — ⬍0.001
CI ⫽ confidence interval.
Comment All surgeons wish to provide the best care for their patients, yet differences in outcomes between surgeon types exist. By examining these differences we hope to shed light on areas for improvement and education. The current study demonstrates that general surgeons perform the majority of lung cancer resections in the United States (53.5%) with significantly lower median thoracic surgical procedure case volumes compared with general thoracic and cardiac surgeons. In other words, more than half of the lung cancer resections in the United States are completed by a large number of general surgeons each doing a small number of cases (median of 8 per year). Significant differences in lymphadenectomy rates were found when surgeon types were compared, with general thoracic surgeons having a significantly higher overall lymphadenectomy rate (72.8%). Specifically, if a non– general thoracic surgeon performed the lung cancer resection, patients were half as likely to undergo lymphadenectomy. Similar trends in mediastinoscopy rates were also seen, with a general surgeon rate of 10.9% compared with 15.8% for general thoracic surgeons. The observed mediastinoscopy rate is likely lower than actual rates because surgeons may perform mediastinoscopy as an outpatient procedure, with patients returning at a later date for the definitive procedure. These separate admissions are not captured in an identifiable way by the NIS dataset. The lower lymphadenectomy and mediastinoscopy rates among general surgeons were associated with significantly higher morbidity and in-hospital mortality rates. Farjah and colleagues [4], using the Surveillance Epidemiology and End Results–Medicare database and surgeon practice type definitions based on board certification (not case mix), reported a lymphadenectomy rate with lung cancer resection of 33% for general thoracic surgeons, 22% for cardiac surgeons, and 11% for general surgeons. These rates are lower than we identified for the surgeon groups. This difference may represent the lower rate of lymphadenectomy seen in the Medicare population and in older patients, as our data suggest. It may also be a result of our study’s need to define general thoracic surgeons, cardiac surgeons, and general surgeons (because surgeons are anonymous in the NIS) by case mix rather than board certification. When annual thoracic surgical case volume was introduced into the
Ann Thorac Surg 2011;92:1958 – 64
multivariate model, the effect of case mix was no longer significant, whereas for every doubling of thoracic cases performed, a significant increase in the lymphadenectomy rate was seen. This suggests that the raw volume of thoracic cases performed may be more important than overall case mix. When the case mix cutoffs used in the current study to define a general thoracic surgeon versus a general surgeon were adjusted by 5% intervals (both upward to 95% general thoracic cases and downward to 5% general thoracic cases), a significantly higher lymphadenectomy rate was seen with each incremental increase in case mix volume up to 85%. Beyond the 85% case mix cutoff, lymphadenectomy was performed at equivalent rates despite increasing case mix. This finding formed the basis of the multivariate model evaluating the general thoracic surgical case mix definition. In this model surgeons with 61% to 80% of their practice in general thoracic cases had a similar likelihood of performing a lymphadenectomy as those with 81% to 100% of their practice in general thoracic cases. Surgeons in the highest group (81% to 100%) were significantly more likely to perform a lymphadenectomy than surgeons performing 0% to 20%, 21% to 40%, or 41% to 60% general thoracic surgical procedures. By making these groups more granular we are better able to see which surgeons are responsible for the higher lymphadenectomy rates; specifically, the greater percentage of general thoracic surgery cases performed by a surgeon, the more likely they are to perform a lymphadenectomy. The NIS dataset does not include information on board certification of surgeons, which limits our analysis to defining surgeon groups based on case mix and overall case volume. It is difficult to determine the interplay of case mix and surgeon volume in our multivariate models. A single surgeon is only capable of carrying out a fixed maximum number of cases in a given year. An increase in total volume of 1 type of case will lead to a decrease in another type. Necessarily, volume and case mix may at times measure the same variable. The conclusions of these 2 models are similar however: The more general thoracic surgery a surgeon performs (whether measured in percentage of surgery or raw volume) the more likely they are to perform a lymphadenectomy at the time of lung cancer resection. The steady rise of lymphadenectomy rates seen over the study period is reassuring. General and cardiac surgeons consistently increased their lymphadenectomy rates during this time. This will hopefully translate into improved oncologic outcomes. However the lymphadenectomy rates of general surgeons and cardiac surgeons are still significantly lower than those of general thoracic surgeons. Further work needs to be done to disseminate knowledge and possibly to create educational training sessions to allow surgeons to be facile at not only performing lung resection but also adequately staging the mediastinum. Alternatively these procedures could be channeled to centers and surgeons capable of providing the appropriate level of care. Our study has limitations common to those using large databases. Individual charts are not available for review, allowing errors in data collection and entry to go unde-
tected. The large number of patients included in the study should minimize the impact of these possible errors. There is a financial incentive for surgeons to code for lymphadenectomy performed during lung cancer resection, which should aid in capturing these events. Despite these limitations, an almost 20 percentage point difference in lymphadenectomy rate by surgeon type was identified. In summary, surgeon specialty impacts the adequacy of oncologic staging in patients undergoing resection for primary lung cancer. Specifically, general thoracic surgeons performed intraoperative oncologic staging significantly more often than did their general surgery and cardiac surgery counterparts and achieved significantly lower mortality and complication rates.
References 1. Schipper PH, Diggs BS, Ungerleider RM, Welke KF. The influence of surgeon specialty on outcomes in general thoracic surgery: a national sample 1996 to 2005. Ann Thorac Surg 2009;88:1566 –73. 2. Goodney PP, Lucas FL, Stukel TA, Birkmeyer JD. Surgeon specialty and operative mortality with lung resection. Ann Surg 2005;2411:179 – 84.
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
1963
3. Dimick JB, Goodney PP, Orringer MB, Birkmeyer JD. Specialty training and mortality after esophageal cancer resection. Ann Thorac Surg 2005;80:282– 6. 4. Farjah F, Flum DR, Varghese TK Jr, Symons RG, Wood DE. Surgeon specialty and long-term survival after pulmonary resection for lung cancer. Ann Thorac Surg 2009;87:995–1006. 5. Scott WJ, Howington J, Feigenberg S, Movsas B, Pisters K, American College of Chest Physicians. Treatment of nonsmall cell lung cancer stage I and stage II: ACCP evidencebased clinical practice guidelines (2nd edition). Chest 2007; 132(3 Suppl):234S– 42S. 6. Detterbeck FC, Jantz MA, Wallace M, Vansteenkiste J, Silvestri GA, American College of Chest Physicians. Invasive mediastinal staging of lung cancer: ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007;132(3 Suppl): 202S–20S. 7. Murthy SC, Reznik SI, Ogwudu UC, et al. Winning the battle, losing the war: the noncurative “curative” resection for stage I adenocarcinoma of the lung. Ann Thorac Surg 2010;90:1067–74. 8. Darling GE, Allen MS, Decker PA, et al. Number of lymph nodes harvested from a mediastinal lymphadenectomy: results of the randomized, prospective ACOSOG Z0030 trial. Chest 2011;139:1124 –9. 9. Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS). Available at: www.hcup-us.ahrq.gov/ nisoverview.jsp. Accessed June 23, 2011.
DISCUSSION DR BILL PUTNAM (Nashville, TN): Dr Mathisen, Dr Reed. The authors present a thoughtful examination of the role of specialty training in the performance of lymphadenectomy for lung cancer. Like the sentinel paper of Birkmeyer on decreasing mortality with larger surgeon volumes, Dr Ellis and colleagues confirmed that with increased expertise, surgical treatment and mortality improves. The title suggests that completeness of lymphadenectomy was evaluated; however the authors analyze the presence or absence of lymphadenectomy, also a significant question. The American College of Surgeons Oncology Group trial Z30 evaluated survival following proscribed systematic mediastinal lymph node sampling in early stage disease compared to complete lymphadenectomy. In this ACOSOG study, operative mortality and long-term survival is similar and objectively defined the minimum intraoperative staging required. For my first question, operative mortality was different between surgeon groups, but quite low, less than 4% for all. Were there quality indicators such as length of stay or 1-year mortality that could be shared with us from your review of the data and how these may differ between these surgeon groups? DR ELLIS: I think that this data allows us to go in many directions. We decided to focus on lymphadenectomy rates, and as far as other factors such as length of stay or location of care, were not analyzed to determine whether or not these impacted outcomes. The NIS database only contains inpatient information and so 1-year mortality rates cannot be determined. DR PUTNAM: I was encouraged to see that over the period of the study, the lymphadenectomy rate increased for all groups, from about 45% to 68% over the study period, an increase of about 50%. So it does appear that something is working to the patient’s and to the surgeon’s advantage. The authors also note the necessary access to care, so critically important to our
patients, particularly in rural or urban nonteaching hospitals. Economic status and social constraints may limit these patients’ choice for location of care. While centers of excellence refine the optimal clinical practice, the ability to access that care can be limited for significant portions of the population. General surgeons and cardiac-focused surgeons may be tasked to care for a wide variety of patient problems, which may require skills beyond their usual level of practice. So for my second question, how do you propose we balance the need for quality of care and the challenges of access, particularly in remote rural or urban centers? What is the role of the general and the cardiac-focused surgeon in providing that care? Thank you for your detailed and informative presentation. DR ELLIS: I think you raise a very important issue, that is access to care. As our study demonstrated, over 60% of lung cancer resections are performed by general surgeons, with only 3% being completed by thoracic surgeons. I think the next step now that we know that these differences in lymphadenectomy rates exist among surgeon types is to find a way to ensure that quality care is delivered to all of our patients. This will need to be addressed on a national level. If the STS and the ACS decide to create centers of excellence where thoracic surgeons perform the majority of lung cancer resections, we need to target the pulmonologists and medical oncologists and encourage them to refer patients to thoracic surgeons. Alternatively, if we want to work toward improving lymphadenectomy rates in general, we need to disseminate knowledge and raise awareness in surgeons performing lung resections in these smaller communities regarding the importance and necessity of performing mediastinal staging. Whether this is through weekend seminars or mentorship programs, we need to get the word out that this is important in the care of our patients. Thank you.
GENERAL THORACIC
Ann Thorac Surg 2011;92:1958 – 64
GENERAL THORACIC
1964
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
DR RICHARD J. SHEMIN (Los Angeles, CA): My question is about your methodology and the choice to define a cardiac surgeon as surgeons who only do 10% cardiac cases? DR ELLIS: That’s correct. DR SHEMIN: Work that was done by the Dartmouth Atlas group and the STS in the year 2000 looked at cardiac surgeons who performed lung surgery. We found that cardiac surgeons did about 75% cardiac cases and only about 25% general thoracic cases and at low volume. Workforce studies, even up until 2010, continue to show that very significant numbers of adult cardiac surgeons perform general thoracic surgery. So I just wondered why you picked that very low bar to define the cardiac group and whether it is actually accurate. DR ELLIS: Thank you. We chose the 10% cutoff for cardiac cases to be inclusive rather than exclusive. Our rationale behind this was that hospitals are not going to let surgeons perform cardiactype surgeries unless they are a board-certified cardiac surgeons, and so a case mix of 10% or greater would have captured all of those cardiac surgeons. DR JOHN R. BENFIELD (Los Angeles, CA): Your wellpresented paper adds to previous evidence that general thoracic surgeons certified by the American Board of Thoracic Surgery (ABTS) do a better job operating on lung cancer patients than general surgeons who are neither ABTS certified nor subspecialized in general thoracic surgery. Previous evidence included the work Carolyn Reed, one of our moderators today. She had surveyed relevant data in South Carolina to which I referred in my STS Presidential Address (Benfield JR, Metamorphosis. (Ann Thorac Surg 1996;61:1045–1050). My proposal of 1996 remains the same today. I suggest that we recognize the reality that some general surgeons do general thoracic surgery. Some of my best general surgical residents are among them. However these excellent people are practicing what they learned from me more than 15 years ago. I have never seen any of them at a thoracic surgery meeting, and they are not practicing modern thoracic oncology. That is not apparent to their referring physicians and not good for their patients. We cannot stop what is ongoing, but we can perhaps make it better. My proposal, challenging as it would be, and anachronistic as it might sound, is that the STS develop a mechanism for taking in selected general surgeons who are doing a significant volume of general thoracic surgery at a reasonably acceptable level. I would welcome them into our educational environment with a form of limited STS membership. The goal would be to facilitate improvement in their level of care for patients who need thoracic surgery. Such limited STS membership should be time limited. I believe this would result in improved thoracic surgical care for patients and gradual disappearance of incompletely educated surgeons who dabble in general thoracic surgery.
Ann Thorac Surg 2011;92:1958 – 64
I thank you for your important contribution. DR ROHIT SHAHANI (Poughkeepsie, NY): Thank you for that paper. My question to you is, can you give us some numbers? You only gave us percentages, and with only 3% being what you qualified as true thoracic surgeons, that means we are making a generalization from your paper based on 97% of the work being done by non–thoracic surgeons. Could you give us an absolute number of the number of cases done by each specialty. DR ELLIS: There were 222,233 patients in our study. Using the case mix cutoffs I mentioned to define our surgeon groups, 61.7% were considered general surgeons, 2.9% thoracic surgeons, and 35.4% cardiac surgeons. Lung cancer resections were performed by general surgeons in 53.5% (n ⫽ 118,843) of patients, by thoracic surgeons in 8.2% (n ⫽ 18,284) of patients, and by cardiac surgeons in 38.3% (n ⫽ 85,106). Thoracic surgeons made up a small percentage of surgeons performing lung cancer resections but had a higher median volume compared with general surgeons. DR DANIEL RAYMOND (Rochester, NY): Could you just comment on your inclusion in the lymphadenectomy group? Did it require specific zones or did you just say if anyone got a lymph node that is a lymphadenectomy and it counted? Also the number of lymph nodes or the weight of lymph nodes, did you look at those numbers, some way of standardizing? DR ELLIS: We looked at whether there was a presence or absence of lymphadenectomy. Unfortunately the NIS database does not include information regarding nodal level or number of nodes resected and so we are left with a gross measure of lymphadenectomy. This makes our findings even more concerning, at this very low threshold there are significant discrepancies in lymphadenectomy rates by surgeon type. DR ANTHONY P. FURNARY (Portland, OR): Don’t you think it would be more accurate instead of using a percentage of volume to use actual volume of thoracic cases performed? As an example, if a general surgeon does 80 thoracic cases a year and 100 general surgery cases a year, his outcomes might be equivalent to those of a thoracic surgeon as defined by your study. Similarly, a cardiac surgeon who does 120 lobes a year and does 100 hearts a year might have similar outcomes to a thoracic surgeon who only does 23 cases a year. DR ELLIS: Thank you for your question. We actually did look at that, and when we looked at the general thoracic case volume, for every doubling there was an increased likelihood of lymphadenectomy performed, and this was linear all the way up to above 100 cases a year. And so I think the bottom line is the more thoracic surgery you do, the better you do it and the more likely you are to do it completely with a lymphadenectomy.
Intraoperative Oncologic Staging and Outcomes for Lung Cancer Resection Vary by Surgeon Specialty Michelle C. Ellis, MD, Brian S. Diggs, PhD, John T. Vetto, MD, and Paul H. Schipper, MD Department of Surgery, Division of Cardiothoracic Surgery, and Division of Surgical Oncology, Oregon Health & Science University, Portland, Oregon
Background. In the United States the majority of lung cancer resections are performed by general surgeons, although surgeons specializing in thoracic surgery have demonstrated superior perioperative and long-term oncologic outcomes. Why these differences exist has not been well studied. We hypothesized that the completeness of intraoperative oncologic staging may explain some of these differences. Methods. The Nationwide Inpatient Sample (NIS) database was used to review 222,233 patients with primary lung cancer treated surgically with wedge resection, segmentectomy, lobectomy, or pneumonectomy from 1998 to 2007. Surgeons were classified as general thoracic surgeons if they performed greater than 75% general thoracic operations and less than 10% cardiac operations; they were classified as cardiac surgeons if they performed greater than 10% cardiac operations; they were classified as general surgeons if they performed less than 75% thoracic operations and less than 10% cardiac operations. The main outcome measure was the performance
of lymphadenectomy or mediastinoscopy during the same admission as the cancer resection. Results. The overall lymphadenectomy rate was 56% (n ⴝ 125,115) and was highest for general thoracic surgeons at 73% (n ⴝ 13,313), followed by 55% (n ⴝ 65,453) for general surgeons, and 54% (n ⴝ 46,349) for cardiac surgeons (p < 0.0001). General surgeons had a significantly higher risk for in-hospital mortality (odds ratio [OR], 1.47; confidence interval [CI], 1.14 to 1.90; p ⴝ 0.003) and postoperative complications (OR, 1.17; CI, 1.00 to 1.36; p ⴝ 0.043) compared with general thoracic surgeons. Conclusions. Surgeon specialty impacts the adequacy of oncologic staging in patients undergoing resection for primary lung cancer. Specifically, general thoracic surgeons performed intraoperative oncologic staging significantly more often than did their general surgeon and cardiac surgeon counterparts while achieving significantly lower in-hospital mortality and complication rates. (Ann Thorac Surg 2011;92:1958 – 64) © 2011 by The Society of Thoracic Surgeons
I
database, we sought to establish the frequency of lymphadenectomy or mediastinoscopy, or both, at the time of a lung cancer resection. Specifically, we looked for predictors of lymphadenectomy and trends over time.
n the United States, the majority of thoracic procedures are performed by general surgeons; however surgeons specializing in thoracic surgery have lower perioperative mortality and morbidity rates compared with general surgeons [1–3]. Patients who have lung resection for cancer performed by a board-certified cardiothoracic surgeon specializing in general thoracic surgery have a longer overall and cancer-specific 5-year survival [4]. Lymph node status is a main determinant of stage and prognosis in lung cancer, and a mediastinal lymphadenectomy has been recommended for all patients undergoing surgical resection for this disease [5– 8]. The current study sought to determine whether the completeness of intraoperative oncologic staging might explain the difference in surgical outcomes for lung cancer resections performed by various types of surgeon. Using the Nationwide Inpatient Sample (NIS)
Accepted for publication May 31, 2011. Presented at the Forty-seventh Annual Meeting of The Society of Thoracic Surgeons, San Diego, CA, Jan 31–Feb 2, 2011. Address correspondence to Dr Schipper, Division of Cardiothoracic Surgery, Oregon Health & Science University, Mail Code L353, 3181 SW Sam Jackson Park Rd, Portland, OR 97239; e-mail: [email protected].
© 2011 by The Society of Thoracic Surgeons Published by Elsevier Inc
Patients and Methods Design/Data Source We performed a retrospective review of the NIS database. The NIS is managed under the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality [9]. It is the largest all-payer database of national hospital discharges and contains a 20% stratified sample of nonfederal acute-care hospitals in the United States. Each record within the NIS database represents a single hospital discharge and contains information on demographic data, admission type, primary and secondary diagnoses and procedures, length of stay, hospital demographics, insurance payers, total hospital charges, and hospital transfer status. Statistical weights and other sampling variables are included that allow national estimates to be made from the sample. Institutional review board approval was waived for this study because of the deidentified nature of the data. 0003-4975/$36.00 doi:10.1016/j.athoracsur.2011.05.120
Inclusion Criteria We searched the NIS using the International Classification of Diseases, 9th revision, Clinical Modification (ICD9-CM) looking for the code for primary lung cancer (ICD-9-CM diagnosis code 162.⫻) in patients who were treated surgically from 1998 to 2007 as designated by an ICD-9-CM procedure code of wedge resection (32.9), segmentectomy (32.3⫻), lobectomy (32.4⫻), or pneumonectomy (32.5⫻). Primary outcome was the presence of lymphadenectomy (ICD-9-CM codes 40.11, 40.22, 40.29, 40.3, 40.50, or 40.59) or mediastinoscopy (ICD-9-CM code 34.22) during the same admission as the cancer resection. Secondary measures were patient demographics (sex, age on admission, insurance payer), hospital factors (rural versus urban, teaching status), and procedure factors (in-hospital mortality, complication rates). (The ICD-9-CM codes used for the definitions of cardiac and thoracic surgeons are available by contacting the authors.).
Classification of Surgeon Type Similar to our previous work [1], individual surgeons performing these procedures were divided into 3 groups based on the percentage of their total case volume spent doing general thoracic surgery, cardiac surgery, or any other type of surgery. A surgeon was considered a general thoracic surgeon if greater than 75% of total procedures performed were from ICD-9-CM codes considered to be general thoracic operations and no more than 10% of total procedures performed were from ICD9-CM codes considered to be cardiac operations. A surgeon was considered a cardiac surgeon if greater than 10% of procedures performed were cardiac in nature. A surgeon was considered a general surgeon if fewer than 75% of total procedures performed were general thoracic surgical procedures and fewer than 10% of total procedures performed were cardiac operations. Board certification and subspecialty training were not considered in these definitions, as this information was not available in the NIS. Also, because surgeon identifiers were not necessarily consistent from year to year, all surgeons were given year-specific identifiers and were classified based on that year’s case mix.a Lymphadenectomy rates by surgeon type were compared as a whole over the study period (1998 to 2007) as were year-by-year trends. Multivariate models were then constructed to determine predictors for lymphadenectomy, mediastinoscopy, inhospital mortality, and in-hospital morbidity.
Evaluation of Case Mix Definition Cutoffs A final multivariate model was constructed to evaluate the case mix definition of a general thoracic surgeon—in particular the cutoff of a 75% general thoracic case mix. Surgeons were divided into 6 groups. The first group contained the previous definition of a cardiac surgeon. The 6 remaining groups were defined based on the percentage of cases considered general thoracic surgical procedures: 0% to 20%, 21% to 40%, 41% to 60%, 61% to 80%, and 81% to 100%, while still maintaining that no
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
1959
more than 10% of total procedures performed were from ICD-9-CM codes considered cardiac operations.
Statistical Analysis All analyses used versions of statistical tests that take into account the hierarchical sampling structure of the NIS. Categorical comparisons were made using 2 tests. Multivariate analysis used binary logistic regression models. Multiple logistic regression models were created for each outcome, each with additional variables to assess the descriptive ability of the new variables and their effect on the already included variables. Surgeon volume was included as a logarithmically transformed variable to reflect the belief that as overall volume increases, improvements to outcome would come from increasing volume by some percentage, rather than increasing volume by some absolute number of cases. A significant difference identified in surgeon volume using this method would indicate that for each doubling of the dependent variable (surgeon volume) there was a difference in the desired outcome from the previous doubling. For example a surgeon performing 8 procedures was different in the examined outcome from the surgeon performing 4 procedures and the surgeon performing 16 procedures was different from the 1 performing 8 procedures. Statistical significance was taken as p ⬍ 0.05.
Results Based on the selection criteria, 71,314 records were selected from the NIS, representing a national estimate of 352,046 cases. An anonymous surgeon identifier was missing in 129,813 cases; these were excluded resulting in a study population of 222,233. The excluded patients were similar demographically to the final study population. Of the 7,770 unique surgeon-year combinations, 61.7% (n ⫽ 4,791) were considered general surgeons, 2.9% (n ⫽ 227) were considered general thoracic surgeons, and 35.4% (n ⫽ 2,752) were considered cardiac surgeons. The demographics of the study population are listed in Table 1. Lung cancer resections were performed by general surgeons in 53.5% (n ⫽ 118,843) of patients, by thoracic surgeons in 8.2% (n ⫽ 18,284) of patients, and by cardiac surgeons in 38.3% (n ⫽ 85,106) of patients. Patients were similar with respect to age, sex, and procedure type performed among the surgeon types. Thoracic surgeons differed with respect to practice location, primary insurance payer, and patient income level (Table 1). The median number of general thoracic surgical procedures performed by general thoracic surgeons was 21 (interquartile range [IQR]; 4 to 85), for general surgeons the median number was 8 (IQR, 3 to 20), and for cardiac surgeons it was 23 (IQR, 12 to 39).
Lymphadenectomy and Mediastinoscopy Rates Lymphadenectomy was performed in 125,115 patients, representing an overall lymphadenectomy rate of 56.3%. Mediastinoscopy was performed in 24,083 patients and represented an overall rate of 10.8%. When lymphadenectomy rates were assessed by surgeon type, general
GENERAL THORACIC
Ann Thorac Surg 2011;92:1958 – 64
GENERAL THORACIC
1960
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
Table 1. Study Population Demographics Variable n Age (Years) (%) 18–50 51–70 71–80 81⫹ Sex (%) Male Female Procedure (%) Limited resection Lobectomy Pneumonectomy Location/status of hospital (%) Rural Urban nonteaching Urban teaching Primary payer (%) Medicare Private insurance Self-pay Other/unknown Income Quartiles (%) Lowest Second lowest Second highest Highest CS ⫽ cardiac surgeon; thoracic surgeon.
GS
GTS
CS
118,843 (53.5%)
18,284 (8.2%)
85,106 (38.3%)
7.6 52.1 33.7 6.6
8.3 53.6 31.4 6.6
7.2 53.7 33.2 5.9
51.7 48.3
50.5 49.5
54.0 46.0
13.3 77.8 8.9
12.9 78.7 8.4
8.9 81.3 9.8
Ann Thorac Surg 2011;92:1958 – 64
gical procedures performed was added to the multivariate model, the surgeon specialty definitions based on percentage of cases performed were no longer significant predictors of lymphadenectomy. However the volume of general thoracic surgical cases performed was a significant predictor of lymphadenectomy. For every doubling of general thoracic case volume there was a significant increase in the likelihood that a lymphadenectomy (odds ratio [OR], 1.28; 95% confidence interval [CI], 1.24 to 1.32; p ⬍ 0.001) and mediastinoscopy (OR, 1.19; CI, 1.11 to 1.28; p ⬍ 0.001) were performed. Conversely for every doubling of general surgery case volume there was a significant decrease in the likelihood a lymphadenectomy was performed (OR, 0.95; CI, 0.92 to 0.97; p ⬍ 0.001). Changes in cardiac surgery case volume did not significantly impact lymphadenectomy or mediastinoscopy rates in this multivariate model.
Mortality Rates 7.7 41.0 51.2
0.4 10.0 89.6
5.4 43.7 50.9
57.3 34.6 1.6 6.5
53.1 39.8 1.9 5.2
59.1 32.7 1.6 6.6
14.0 26.8 26.6 30.3
6.9 18.6 22.3 49.9
17.1 27.9 28.5 23.8
GS ⫽ general surgeon;
GTS ⫽ general
thoracic surgeons performed lymphadenectomy in 72.8% of patients compared with 55.1% and 54.5% for general surgeons and cardiac surgeons, respectively (p ⬍ 0.0001). Similarly mediastinoscopy rates were highest for general thoracic surgeons at 15.8%, followed by 10.9% and 9.6% for general and cardiac surgeons, respectively (p ⬍ 0.0001). A multivariate analysis was performed to determine predictors for lymphadenectomy. Patients less likely to have lymphadenectomy were those who were in the bottom 50% of income level, insured by Medicare as opposed to private insurance or self-pay, received their care at a rural or urban nonteaching hospital, or had resection performed by a cardiac or general surgeon. Specifically, a patient was more than twice as likely to have a lymphadenectomy performed if the lung cancer resection was completed by a general thoracic surgeon rather than a cardiac or general surgeon (Table 2). Patients who were more likely to undergo lymphadenectomy were women, those who had a lobectomy as opposed to pneumonectomy or limited lung resection, or those who had the procedure performed by a general thoracic surgeon. When volume of general thoracic sur-
When comparing the 3 surgeon types, lung cancer resections performed by general surgeons had a significantly higher risk for in-hospital mortality (OR, 1.55; CI, 1.20 to
Table 2. Predictors of Lymphadenectomy Variable Age (Years) 18–50 51–70 71–80 80⫹ Sex Male Female Surgeon type Thoracic surgeon General surgeon Cardiac surgeon Primary payer Medicare Private Insurance Self-pay Other/unknown Income level Lowest quartile Second lowest quartile Second highest quartile Highest quartile Procedure performed Limited lung resection Lobectomy Pneumonectomy Location/status of hospital Rural Urban nonteaching Urban teaching CI ⫽ confidence interval.
Odds Ratio (CI)
p Value
1.0 (Reference) 0.99 (0.92-1.07) 0.98 (0.90-1.08) 0.91 (0.81-1.03)
— 0.877 0.710 0.136
1.0 (Reference) 1.09 (1.05–1.14)
— ⬍0.001
1.0 (Reference) 0.47 (0.35–0.65) 0.47 (0.35–0.64)
— ⬍0.001 ⬍0.001
0.93 (0.87–0.99) 1.0 (Reference) 0.91 (0.73–1.14) 0.90 (0.81–1.00)
0.022 — 0.414 0.041
0.74 (0.66–0.83) 0.89 (0.81–0.97) 0.93 (0.86-1.00) 1.0 (Reference)
⬍0.001 0.009 0.054 —
1.0 (Reference) 1.38 (1.27–1.49) 0.97 (0.86–1.09)
— ⬍0.001 0.631
0.60 (0.44–0.81) 0.74 (0.66–0.83) 1.0 (Reference)
⬍0.001 ⬍0.001 —
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
Table 3. Predictors for In-Hospital Mortality Variable Age (Years) 18–50 51–70 71–80 80⫹ Sex Male Female Surgeon type Thoracic surgeon General surgeon Cardiac surgeon Primary payer Medicare Private Insurance Self-pay Other/unknown Income level Lowest quartile Second lowest quartile Second highest quartile Highest quartile Procedure performed Limited lung resection Lobectomy Pneumonectomy Location/status of hospital Rural Urban nonteaching Urban teaching
Odds Ratio (CI)
p Value
1.0 (Reference) 1.40 (1.05–1.87) 2.44 (1.80–3.29) 3.14 (2.24–4.39)
— 0.024 ⬍0.001 ⬍0.001
1.0 (Reference) 0.68 (0.60–0.76)
— ⬍0.001
1.0 (Reference) 1.79 (1.41–2.25) 1.50 (1.18–1.91)
— ⬍0.001 ⬍0.001
1.61 (1.37–1.89) 1.0 (Reference) 1.02 (0.59–1.77) 1.39 (1.07–1.81)
⬍0.001 — 0.945 0.013
1.28 (1.06–1.55) 1.25 (1.08–1.44) 1.10 (0.95–1.28) 1.0 (Reference)
0.010 0.003 0.207 —
1.0 (Reference) 1.08 (0.89–1.30) 3.65 (2.93–4.55)
— 0.456 ⬍0.001
0.93 (0.73–1.20) 1.07 (0.95–1.21) 1.0 (Reference)
0.602 0.257 —
1961
ative complications. Female sex was significantly protective against complications.
Trends Over Time
CI ⫽ confidence interval.
The study population was assessed as a whole and by surgeon type over the study period (1998 to 2007) to determine trends in patient demographics, lymphadenectomy and mediastinoscopy rates, in-hospital mortality, and complication rates. The study population as a whole remained relatively similar in terms of age, sex, insurance type, and location of care. The overall population was composed of a higher proportion of lowest income patients after 2002, but this was due to database adjustments in the definition of income quartiles. The overall in-hospital mortality remained relatively stable throughout the study period, whereas the overall lymphadenectomy rate rose steadily (Fig 1). When trends over time were assessed by surgeon type, significant changes in the lymphadenectomy rates were seen. Thoracic surgeon lymphadenectomy rates dipped in the midportion of the study period, whereas general and cardiac surgeon rates steadily increased (Fig 1).
Evaluation of Case Mix Definition Cutoffs A surgeon performing 61% to 80% of cases as general thoracic surgical procedures was just as likely to perform a lymphadenectomy as a surgeon with an 81% to 100% general thoracic case mix. Surgeons in each group— 0% to 20%, 21% to 40%, or 41% to 60% general thoracic cases—were significantly less likely to perform a lymphadenectomy compared with surgeons who performed 81% to 100% general thoracic procedures. The likelihood of performing lymphadenectomy at the time of lung cancer resection based on general thoracic surgery case mix is shown in Table 4.
2.01; p ⬍ 0.001) compared with those performed by general thoracic surgeons. Unadjusted in-hospital mortality rates for thoracic, cardiac, and general surgeons were 2.3%, 3.4%, and 4.0%, respectively. Age greater than 50 years, Medicare rather than private insurance as primary payer, a procedure of pneumonectomy compared with more limited lung resection, and a non– thoracic surgeons performing the operations were all significant predictors for increased mortality on multivariate analysis. Female sex was significantly protective against in-hospital mortality (Table 3).
Complication Rates When comparing the 3 surgeon types, lung cancer resections performed by general surgeons had a significantly higher risk for postoperative complications (OR, 1.16; CI, 1.00 to 1.35; p ⫽ 0.049) compared with those performed by thoracic surgeons. The overall complication rate for thoracic, cardiac, and general surgeons was 28.5%, 31.3%, and 33.5%, respectively. On multivariate analysis, age greater than 50 years and general surgeons performing the procedures were significant predictors for postoper-
Fig 1. Lymphadenectomy rates overall and by surgeon type over time.
GENERAL THORACIC
Ann Thorac Surg 2011;92:1958 – 64
GENERAL THORACIC
1962
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
Table 4. Likelihood of Lymphadenectomy by Case Mix Type of Case General Thoracic Cases (%) 0–20 21–40 41–60 61–80 81–100 Cardiac Surgeon Cases
Odds Ratio (95% CI)
p Value
0.29 (0.21–0.41) 0.56 (0.39–0.80) 0.64 (0.44–0.93) 0.89 (0.63–1.26) 1.0 (Reference) 0.44 (0.80–1.02)
⬍0.001 0.002 0.020 0.527 — ⬍0.001
CI ⫽ confidence interval.
Comment All surgeons wish to provide the best care for their patients, yet differences in outcomes between surgeon types exist. By examining these differences we hope to shed light on areas for improvement and education. The current study demonstrates that general surgeons perform the majority of lung cancer resections in the United States (53.5%) with significantly lower median thoracic surgical procedure case volumes compared with general thoracic and cardiac surgeons. In other words, more than half of the lung cancer resections in the United States are completed by a large number of general surgeons each doing a small number of cases (median of 8 per year). Significant differences in lymphadenectomy rates were found when surgeon types were compared, with general thoracic surgeons having a significantly higher overall lymphadenectomy rate (72.8%). Specifically, if a non– general thoracic surgeon performed the lung cancer resection, patients were half as likely to undergo lymphadenectomy. Similar trends in mediastinoscopy rates were also seen, with a general surgeon rate of 10.9% compared with 15.8% for general thoracic surgeons. The observed mediastinoscopy rate is likely lower than actual rates because surgeons may perform mediastinoscopy as an outpatient procedure, with patients returning at a later date for the definitive procedure. These separate admissions are not captured in an identifiable way by the NIS dataset. The lower lymphadenectomy and mediastinoscopy rates among general surgeons were associated with significantly higher morbidity and in-hospital mortality rates. Farjah and colleagues [4], using the Surveillance Epidemiology and End Results–Medicare database and surgeon practice type definitions based on board certification (not case mix), reported a lymphadenectomy rate with lung cancer resection of 33% for general thoracic surgeons, 22% for cardiac surgeons, and 11% for general surgeons. These rates are lower than we identified for the surgeon groups. This difference may represent the lower rate of lymphadenectomy seen in the Medicare population and in older patients, as our data suggest. It may also be a result of our study’s need to define general thoracic surgeons, cardiac surgeons, and general surgeons (because surgeons are anonymous in the NIS) by case mix rather than board certification. When annual thoracic surgical case volume was introduced into the
Ann Thorac Surg 2011;92:1958 – 64
multivariate model, the effect of case mix was no longer significant, whereas for every doubling of thoracic cases performed, a significant increase in the lymphadenectomy rate was seen. This suggests that the raw volume of thoracic cases performed may be more important than overall case mix. When the case mix cutoffs used in the current study to define a general thoracic surgeon versus a general surgeon were adjusted by 5% intervals (both upward to 95% general thoracic cases and downward to 5% general thoracic cases), a significantly higher lymphadenectomy rate was seen with each incremental increase in case mix volume up to 85%. Beyond the 85% case mix cutoff, lymphadenectomy was performed at equivalent rates despite increasing case mix. This finding formed the basis of the multivariate model evaluating the general thoracic surgical case mix definition. In this model surgeons with 61% to 80% of their practice in general thoracic cases had a similar likelihood of performing a lymphadenectomy as those with 81% to 100% of their practice in general thoracic cases. Surgeons in the highest group (81% to 100%) were significantly more likely to perform a lymphadenectomy than surgeons performing 0% to 20%, 21% to 40%, or 41% to 60% general thoracic surgical procedures. By making these groups more granular we are better able to see which surgeons are responsible for the higher lymphadenectomy rates; specifically, the greater percentage of general thoracic surgery cases performed by a surgeon, the more likely they are to perform a lymphadenectomy. The NIS dataset does not include information on board certification of surgeons, which limits our analysis to defining surgeon groups based on case mix and overall case volume. It is difficult to determine the interplay of case mix and surgeon volume in our multivariate models. A single surgeon is only capable of carrying out a fixed maximum number of cases in a given year. An increase in total volume of 1 type of case will lead to a decrease in another type. Necessarily, volume and case mix may at times measure the same variable. The conclusions of these 2 models are similar however: The more general thoracic surgery a surgeon performs (whether measured in percentage of surgery or raw volume) the more likely they are to perform a lymphadenectomy at the time of lung cancer resection. The steady rise of lymphadenectomy rates seen over the study period is reassuring. General and cardiac surgeons consistently increased their lymphadenectomy rates during this time. This will hopefully translate into improved oncologic outcomes. However the lymphadenectomy rates of general surgeons and cardiac surgeons are still significantly lower than those of general thoracic surgeons. Further work needs to be done to disseminate knowledge and possibly to create educational training sessions to allow surgeons to be facile at not only performing lung resection but also adequately staging the mediastinum. Alternatively these procedures could be channeled to centers and surgeons capable of providing the appropriate level of care. Our study has limitations common to those using large databases. Individual charts are not available for review, allowing errors in data collection and entry to go unde-
tected. The large number of patients included in the study should minimize the impact of these possible errors. There is a financial incentive for surgeons to code for lymphadenectomy performed during lung cancer resection, which should aid in capturing these events. Despite these limitations, an almost 20 percentage point difference in lymphadenectomy rate by surgeon type was identified. In summary, surgeon specialty impacts the adequacy of oncologic staging in patients undergoing resection for primary lung cancer. Specifically, general thoracic surgeons performed intraoperative oncologic staging significantly more often than did their general surgery and cardiac surgery counterparts and achieved significantly lower mortality and complication rates.
References 1. Schipper PH, Diggs BS, Ungerleider RM, Welke KF. The influence of surgeon specialty on outcomes in general thoracic surgery: a national sample 1996 to 2005. Ann Thorac Surg 2009;88:1566 –73. 2. Goodney PP, Lucas FL, Stukel TA, Birkmeyer JD. Surgeon specialty and operative mortality with lung resection. Ann Surg 2005;2411:179 – 84.
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
1963
3. Dimick JB, Goodney PP, Orringer MB, Birkmeyer JD. Specialty training and mortality after esophageal cancer resection. Ann Thorac Surg 2005;80:282– 6. 4. Farjah F, Flum DR, Varghese TK Jr, Symons RG, Wood DE. Surgeon specialty and long-term survival after pulmonary resection for lung cancer. Ann Thorac Surg 2009;87:995–1006. 5. Scott WJ, Howington J, Feigenberg S, Movsas B, Pisters K, American College of Chest Physicians. Treatment of nonsmall cell lung cancer stage I and stage II: ACCP evidencebased clinical practice guidelines (2nd edition). Chest 2007; 132(3 Suppl):234S– 42S. 6. Detterbeck FC, Jantz MA, Wallace M, Vansteenkiste J, Silvestri GA, American College of Chest Physicians. Invasive mediastinal staging of lung cancer: ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007;132(3 Suppl): 202S–20S. 7. Murthy SC, Reznik SI, Ogwudu UC, et al. Winning the battle, losing the war: the noncurative “curative” resection for stage I adenocarcinoma of the lung. Ann Thorac Surg 2010;90:1067–74. 8. Darling GE, Allen MS, Decker PA, et al. Number of lymph nodes harvested from a mediastinal lymphadenectomy: results of the randomized, prospective ACOSOG Z0030 trial. Chest 2011;139:1124 –9. 9. Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS). Available at: www.hcup-us.ahrq.gov/ nisoverview.jsp. Accessed June 23, 2011.
DISCUSSION DR BILL PUTNAM (Nashville, TN): Dr Mathisen, Dr Reed. The authors present a thoughtful examination of the role of specialty training in the performance of lymphadenectomy for lung cancer. Like the sentinel paper of Birkmeyer on decreasing mortality with larger surgeon volumes, Dr Ellis and colleagues confirmed that with increased expertise, surgical treatment and mortality improves. The title suggests that completeness of lymphadenectomy was evaluated; however the authors analyze the presence or absence of lymphadenectomy, also a significant question. The American College of Surgeons Oncology Group trial Z30 evaluated survival following proscribed systematic mediastinal lymph node sampling in early stage disease compared to complete lymphadenectomy. In this ACOSOG study, operative mortality and long-term survival is similar and objectively defined the minimum intraoperative staging required. For my first question, operative mortality was different between surgeon groups, but quite low, less than 4% for all. Were there quality indicators such as length of stay or 1-year mortality that could be shared with us from your review of the data and how these may differ between these surgeon groups? DR ELLIS: I think that this data allows us to go in many directions. We decided to focus on lymphadenectomy rates, and as far as other factors such as length of stay or location of care, were not analyzed to determine whether or not these impacted outcomes. The NIS database only contains inpatient information and so 1-year mortality rates cannot be determined. DR PUTNAM: I was encouraged to see that over the period of the study, the lymphadenectomy rate increased for all groups, from about 45% to 68% over the study period, an increase of about 50%. So it does appear that something is working to the patient’s and to the surgeon’s advantage. The authors also note the necessary access to care, so critically important to our
patients, particularly in rural or urban nonteaching hospitals. Economic status and social constraints may limit these patients’ choice for location of care. While centers of excellence refine the optimal clinical practice, the ability to access that care can be limited for significant portions of the population. General surgeons and cardiac-focused surgeons may be tasked to care for a wide variety of patient problems, which may require skills beyond their usual level of practice. So for my second question, how do you propose we balance the need for quality of care and the challenges of access, particularly in remote rural or urban centers? What is the role of the general and the cardiac-focused surgeon in providing that care? Thank you for your detailed and informative presentation. DR ELLIS: I think you raise a very important issue, that is access to care. As our study demonstrated, over 60% of lung cancer resections are performed by general surgeons, with only 3% being completed by thoracic surgeons. I think the next step now that we know that these differences in lymphadenectomy rates exist among surgeon types is to find a way to ensure that quality care is delivered to all of our patients. This will need to be addressed on a national level. If the STS and the ACS decide to create centers of excellence where thoracic surgeons perform the majority of lung cancer resections, we need to target the pulmonologists and medical oncologists and encourage them to refer patients to thoracic surgeons. Alternatively, if we want to work toward improving lymphadenectomy rates in general, we need to disseminate knowledge and raise awareness in surgeons performing lung resections in these smaller communities regarding the importance and necessity of performing mediastinal staging. Whether this is through weekend seminars or mentorship programs, we need to get the word out that this is important in the care of our patients. Thank you.
GENERAL THORACIC
Ann Thorac Surg 2011;92:1958 – 64
GENERAL THORACIC
1964
ELLIS ET AL STAGING AND LUNG CANCER RESECTION OUTCOMES
DR RICHARD J. SHEMIN (Los Angeles, CA): My question is about your methodology and the choice to define a cardiac surgeon as surgeons who only do 10% cardiac cases? DR ELLIS: That’s correct. DR SHEMIN: Work that was done by the Dartmouth Atlas group and the STS in the year 2000 looked at cardiac surgeons who performed lung surgery. We found that cardiac surgeons did about 75% cardiac cases and only about 25% general thoracic cases and at low volume. Workforce studies, even up until 2010, continue to show that very significant numbers of adult cardiac surgeons perform general thoracic surgery. So I just wondered why you picked that very low bar to define the cardiac group and whether it is actually accurate. DR ELLIS: Thank you. We chose the 10% cutoff for cardiac cases to be inclusive rather than exclusive. Our rationale behind this was that hospitals are not going to let surgeons perform cardiactype surgeries unless they are a board-certified cardiac surgeons, and so a case mix of 10% or greater would have captured all of those cardiac surgeons. DR JOHN R. BENFIELD (Los Angeles, CA): Your wellpresented paper adds to previous evidence that general thoracic surgeons certified by the American Board of Thoracic Surgery (ABTS) do a better job operating on lung cancer patients than general surgeons who are neither ABTS certified nor subspecialized in general thoracic surgery. Previous evidence included the work Carolyn Reed, one of our moderators today. She had surveyed relevant data in South Carolina to which I referred in my STS Presidential Address (Benfield JR, Metamorphosis. (Ann Thorac Surg 1996;61:1045–1050). My proposal of 1996 remains the same today. I suggest that we recognize the reality that some general surgeons do general thoracic surgery. Some of my best general surgical residents are among them. However these excellent people are practicing what they learned from me more than 15 years ago. I have never seen any of them at a thoracic surgery meeting, and they are not practicing modern thoracic oncology. That is not apparent to their referring physicians and not good for their patients. We cannot stop what is ongoing, but we can perhaps make it better. My proposal, challenging as it would be, and anachronistic as it might sound, is that the STS develop a mechanism for taking in selected general surgeons who are doing a significant volume of general thoracic surgery at a reasonably acceptable level. I would welcome them into our educational environment with a form of limited STS membership. The goal would be to facilitate improvement in their level of care for patients who need thoracic surgery. Such limited STS membership should be time limited. I believe this would result in improved thoracic surgical care for patients and gradual disappearance of incompletely educated surgeons who dabble in general thoracic surgery.
Ann Thorac Surg 2011;92:1958 – 64
I thank you for your important contribution. DR ROHIT SHAHANI (Poughkeepsie, NY): Thank you for that paper. My question to you is, can you give us some numbers? You only gave us percentages, and with only 3% being what you qualified as true thoracic surgeons, that means we are making a generalization from your paper based on 97% of the work being done by non–thoracic surgeons. Could you give us an absolute number of the number of cases done by each specialty. DR ELLIS: There were 222,233 patients in our study. Using the case mix cutoffs I mentioned to define our surgeon groups, 61.7% were considered general surgeons, 2.9% thoracic surgeons, and 35.4% cardiac surgeons. Lung cancer resections were performed by general surgeons in 53.5% (n ⫽ 118,843) of patients, by thoracic surgeons in 8.2% (n ⫽ 18,284) of patients, and by cardiac surgeons in 38.3% (n ⫽ 85,106). Thoracic surgeons made up a small percentage of surgeons performing lung cancer resections but had a higher median volume compared with general surgeons. DR DANIEL RAYMOND (Rochester, NY): Could you just comment on your inclusion in the lymphadenectomy group? Did it require specific zones or did you just say if anyone got a lymph node that is a lymphadenectomy and it counted? Also the number of lymph nodes or the weight of lymph nodes, did you look at those numbers, some way of standardizing? DR ELLIS: We looked at whether there was a presence or absence of lymphadenectomy. Unfortunately the NIS database does not include information regarding nodal level or number of nodes resected and so we are left with a gross measure of lymphadenectomy. This makes our findings even more concerning, at this very low threshold there are significant discrepancies in lymphadenectomy rates by surgeon type. DR ANTHONY P. FURNARY (Portland, OR): Don’t you think it would be more accurate instead of using a percentage of volume to use actual volume of thoracic cases performed? As an example, if a general surgeon does 80 thoracic cases a year and 100 general surgery cases a year, his outcomes might be equivalent to those of a thoracic surgeon as defined by your study. Similarly, a cardiac surgeon who does 120 lobes a year and does 100 hearts a year might have similar outcomes to a thoracic surgeon who only does 23 cases a year. DR ELLIS: Thank you for your question. We actually did look at that, and when we looked at the general thoracic case volume, for every doubling there was an increased likelihood of lymphadenectomy performed, and this was linear all the way up to above 100 cases a year. And so I think the bottom line is the more thoracic surgery you do, the better you do it and the more likely you are to do it completely with a lymphadenectomy.