Includes 89 448 patients across New York State from January 1, 2003, to December 31, 2013. The analysis by patient age used Cochran-Armitage tests.
Includes 87 449 patients across New York State from January 1, 2003, to October 1, 2013.
Includes 2299 surgeons treating 87 449 patients across New York State from January 1, 2003, to October 1, 2013. CL indicates confidence limits.
eTable 1. Coding Algorithms Used for Patient Selection and Variable Creation
eTable 2. Physician and Hospital Characteristics and the Choice of Reexcision Procedure 90 Days After Breast Conservation Surgery
eFigure. Patient Selection
Isaacs AJ, Gemignani ML, Pusic A, Sedrakyan A. Association of Breast Conservation Surgery for Cancer With 90-Day Reoperation Rates in New York State. JAMA Surg. 2016;151(7):648-655. doi:10.1001/jamasurg.2015.5535
For early-stage breast cancer, breast conservation surgery (BCS) is a conservative option for women and involves removing the tumor with a margin of surrounding breast tissue. If margins are not tumor free, patients undergo additional surgery to avoid local recurrence.
To investigate the use of BCS in New York State and to determine rates of reoperation, procedure choice, and the effect of surgeon experience on the odds of a reoperation 90 days after BCS.
Design, Setting, and Participants
A population-based sample of 89 448 women undergoing primary BCS for cancer were selected and examined from January 1, 2003, to December 31, 2013, in New York State mandatory reporting databases. All hospitals and ambulatory surgery centers in New York State were included. Data were analyzed from December 15, 2014, to November 1, 2015.
Main Outcomes and Measures
Rate of reoperations within 90 days of the initial BCS procedure.
During the study period, 89 448 women 20 years or older (mean [SD] age, 61.7 [13.7] years) underwent primary BCS. In 2013, 1416 women in New York aged 20 to 49 years underwent BCS compared with 3068 women aged 50 to 64 years and 3644 women 65 years or older. These numbers represent a significant decrease from 1960 women younger than 50 years in 2003 who underwent BCS (P < .001 for trend) but little change from the 2899 women aged 50 to 64 years and 3270 women 65 years or older who underwent BCS in 2003. Mean overall rate of 90-day reoperation was 30.9% (27 010 of 87 499 patients) and decreased over time from 39.5% (6630 of 16 805 patients) in 2003 to 2004 to 23.1% (5148 of 22 286 patients) in 2011 to 2013. Rates of reoperation were highest in women aged 20 to 49 years (37.7% [6990 of 18 524]) and lowest in women 65 years or older (26.3% [9656 of 36 691]) (P < .001 for trend). Over time, more patients underwent BCS as a subsequent procedure, from 4237 of 6630 patients (63.9%) in 2003 to 2004 to 4258 of 5148 (82.7%) in 2011 to 2013 (P < .001 for trend). Among the 19 466 women who underwent BCS as a second procedure, 2429 (12.5%) required a third intervention (2.7% of all women included). Significant surgeon-level variation was found in the data; 90-day rates of reoperations by surgeon ranged from 0% to 100%. Low-volume surgeons (<14 cases per year) had an unadjusted rate of 35.2% compared with 29.6% in middle-volume (14-33 cases per year) and 27.5% in high-volume (≥34 cases per year) surgeons. The difference persisted in adjusted analyses (odds ratio for low-volume surgeons, 1.49 [95% CI, 1.19-1.87]; for middle-volume surgeons, 1.20 [95% CI, 0.93-1.56]) compared with high-volume surgeons (used as the reference category).
Conclusions and Relevance
Use of BCS has decreased overall, most steeply in younger women. Nearly 1 in 4 women underwent a reoperation within 90 days of BCS across New York State from 2011 to 2013, compared with 2 in 5 from 2003 to 2004. Rates vary significantly by surgeon, and initial BCS performed by high-volume surgeons was associated with a 33% lower risk for a reoperation.
Cancer of the female breast is the most common cancer and the principal cause of cancer-related death among women globally.1 In the United States, an estimated 232 670 new cases of breast cancer would be diagnosed in 2014.2 For early-stage breast cancer, the most common treatments are mastectomy and breast conservation surgery (BCS) both of which are usually combined with adjuvant radiotherapy. Total mastectomy is a well-established treatment option with some cosmetic limitations that may influence the patient’s quality of life.3 Breast conservation surgery is a conservative option for women and involves removing the tumor with a margin of surrounding tumor-free breast tissue. Multiple randomized clinical trials comparing BCS with mastectomy have reported similar survival with a slightly higher risk for local recurrence compared with mastectomy.4- 6 Breast conservation surgery also may be associated with a more positive body image compared with mastectomy.3,7
However, conducting BCS requires expertise and, in many instances of nonpalpable cancers, image-guided localization. The main goal of the surgery is complete removal of the affected tissue while achieving an acceptable cosmetic result. Intraoperative pathologic assessment of the margins around the tumor have limited accuracy.8 If final pathologic results for resected tissue show that margins are involved, patients undergo additional surgery to avoid local recurrence.9- 12 Patients have the option of mastectomy or additional BCS at this point. Repeated excisions are stressful for patients, may lead to poor cosmetic outcomes,13 have been associated with increased risk for cancer recurrence,14,15 and add substantial costs to care.16- 18
Patient, hospital, and surgeon factors might be related to occurrence of reoperations. Some studies have identified that, in addition to patient age,11,19 breast size, surgical biopsy, and histologic findings18 are associated with a reexcision. Significant hospital level variation in reoperation rates have been observed in some studies and were not explained by patient characteristics.12,17 Surgeon volume of surgical procedures has been found to be associated with patient outcomes for many operations20- 22 and may be an important factor for reoperation.
From a policy perspective, rates of repeated lumpectomies might be considered a measure of quality.23 However, major concerns remain that injudicious quality measure will lead to surgical decisions to perform larger lumpectomies than needed, to ignore the pathologic margin, or to recommend mastectomy instead of BCS to avoid being penalized for the reoperation rate. We sought to inform these discussions by performing an investigative study of the use of BCS in New York State over time. We determine reoperation occurrence, the choice of procedure for the reoperation, and the effect of physician experience and patient characteristics on the odds of reoperation 90 days after BCS.
Question What patient- and surgeon-level variables have an effect on rates of reoperations after breast conservation surgery (BCS) in New York State?
Findings Among 89 448 women 20 years or older who underwent primary BCS for cancer from 2003 to 2013, the mean rate of reoperations within 90 days was 30.9%. Rates decreased over time and varied by surgeon, and BCS performed by high-volume surgeons was associated with a 33% lower risk for a reoperation.
Meaning Nearly 1 in 4 women will undergo a repeated operation 90 days after BCS, but rates vary widely by surgeon.
The New York Statewide Planning and Cooperation System (SPARCS) is a collection of patient-level details from every hospital discharge and ambulatory surgery encounter in New York State. Its current format contains patient demographics, diagnoses, procedures, and discharge status.24 All female patients 20 years or older who underwent BCS for breast cancer as determined by codes from the Current Procedural Terminology or the International Classification of Diseases, Ninth Revision (ICD-9), from January 1, 2003, to December 31, 2013, were selected. These years were chosen to reflect modern practice from available data at the time of analysis. Patients for whom BCS was not the primary procedure, who were not residents of New York State, who underwent BCS or mastectomy in the year before the incident procedure, or who had a known personal history of breast cancer were excluded. This study was approved by the institutional review board of Weill Cornell Medical College. Informed consent was waived because data were already collected for billing purposes.
Patient age and race, payer, and procedure setting were available in the SPARCS database. Patient family history of breast cancer, personal history of cancer, diagnosis of carcinoma in situ, and the presence of coronary artery disease were determined using ICD-9 coding algorithms (described in eTable 1 in the Supplement). Additional comorbidities of cerebrovascular disease, hypertension, hypothyroidism, obesity, chronic pulmonary disease, and depression were developed and validated using the methods of Elixhauser et al25 for administrative data. We also included a validated score created by van Walraven et al26 using the published measures of Elixhauser et al25 to quantify the overall comorbidity burden.
The SPARCS data include the license number of the surgeon who performed the primary procedure for each encounter, which allowed us to calculate the mean annual volume of BCS for each surgeon. Volume was broken into tertiles as suggested by Birkmeyer and colleagues20 in their landmark surgeon volume outcome study. In addition, we looked up the date of medical school graduation for each licensed surgeon in the database of New York State Office of Professionals27 to determine their years of experience. This linkage was possible for 75% of the available data. Surgeon experience was also described in tertiles.
For each patient, we looked within 90 days of the incident BCS for a repeated BCS or a mastectomy. Because laterality was provided on 26.6% of cases only, we did not consider any mismatched cases to be reoperations for patients for whom laterality was coded on the incident procedure and the reoperation. This mismatch only occurred in 0.16% of patients. We then repeated the method to look for a third excision surgery within 90 days of the second BCS.
Data were analyzed from December 15, 2014, to November 1, 2015. Raw counts, percentages, and means and SDs were presented for descriptive analyses. We examined changes over time and by patient age graphically and verified these data using Cochran-Armitage tests. For adjusted odds of reoperation within 90 days, we used hierarchical logistic regression models. Models were fit using adaptive quadrature in the generalized linear mixed model (GLIMMIX) procedure with a random physician-level intercept to account for clustering of patients within surgeons. Unadjusted and multivariable models are presented for each patient and physician factor. We used χ2 tests to examine factors associated with the choice of the repeated procedure, and all P values are 2-sided. To examine variation in physician reintervention rates, we used funnel plots. A small amount (<5%) of data were missing in hospital volume and payer variables, and a relatively large amount of data was missing for surgeon experience. Missing values were excluded in pairwise univariate analyses, and all cases with missing data were excluded from the multivariable model. Sensitivity analyses were performed for excluding cases with missing data from all analyses, which yielded similar estimates. All analyses were completed using SAS software (version 9.3; SAS Institute Inc).
From January 1, 2003, to December 31, 2013, 89 448 women 20 years or older underwent primary BCS for breast cancer across New York State (eFigure in the Supplement). The mean (SD) age was 61.7 (13.7) years, with 37 581 women (42.0%) 65 years or older and 18 857 women (21.1%) aged 20 to 49 years. Most patients were white (65 208 [75.5%]), had commercial insurance (44 837 [50.2%]), and underwent surgery in an outpatient setting (80 970 [90.5%]) (Table 1).
The use of BCS in New York State has changed over time. Use appeared to have peaked in 2004 with more than 8500 cases and decreased to a mean of 8078 in 2011 to 2013. In 2003, 1960 women with breast cancer aged 20 to 49 years, 2899 aged 50 to 64 years, and 3270 65 years or older underwent BCS. Although the use varied slightly over time for women 50 to 65 years and 65 years or older, it decreased significantly in women aged 20 to 49 years (P < .001 for trend), so that in 2013 only 1416 underwent BCS (compared with 3068 aged 50-64 years and 3644 65 years or older) (Figure 1).
Owing to insufficient follow-up among patients undergoing BCS after October 1, 2013, 87 499 patients contributed to the 90-day analysis of reoperation rates. The mean 90-day reoperation rate was 30.9% (27 010 of 87 499 patients) for the entire study period and decreased significantly over time, from 39.5% (6630 of 16 805 patients) in 2003 to 2004 to 23.1% (5148 of 22 286 patients) in 2011 to 2013 (P < .001 for trend) (Figure 2A). Reoperation rates were highest in women aged 20 to 49 years (37.7% [6990 of 18 524]) and lowest in the women 65 years or older (26.3% [9656 of 36 691]) (P < .001 for trend) (Figure 2B).
Over time, more patients underwent BCS as a repeated procedure, from 4237 of 6630 patients (63.9%) in 2003 to 2004 to 4258 of 5148 patients (82.7%) in 2011 to 2013 (P < .001 for trend) (Figure 2C). The women aged 50 to 64 years were more likely to undergo BCS than mastectomy as a repeated procedure (7714 of 10 364 [74.4%]) compared with patients aged 20 to 49 years (4860 of 6990 [69.5%]) and 65 years or older (71.4%) (P = .054 for trend) (Figure 2D). Women who were white (13 900 of 18 909 [73.5%] vs 5009 of 7206 [69.5%]) and who had commercial insurance (10 985 of 19 466 [56.4%] vs 3895 of 7544 [51.7%]), low comorbidity scores (4446 of 19 466 [22.8%] vs 1039 of 7544 [13.8%]), or in situ disease (4193 of 19 466 [21.5%] vs 1430 of 7544 [19.0%]) were also more likely to undergo BCS than mastectomy as a repeated procedure (eTable 2 in the Supplement). Of the 19 466 patients who underwent a second BCS, 2429 (12.5%) required a third intervention (2.7% of women undergoing a primary BCS).
We found a significant surgeon-level variation in the data, with rates of reoperations varying from 0% to 100%. Although the mean rate of reoperations was 30.8%, 454 of 2299 surgeons (19.7%) had rates higher than 50%. Of the 2299 surgeons who performed at least 1 primary BCS, 299 (13.0%) had reoperation rates above the 95% confidence limits for reoperation, including 140 surgeons (6.1%) above the 99.8% limits. On the other side, 139 surgeons (6.0%) had reoperation rates below the 95% confidence limits, and 67 (2.9%) had reoperation rates below the 99.8% limits (Figure 3).
Most surgeons (2087 of 2299 [90.8%]) performed a mean of less than 14 primary cases per year, whereas a relatively small percentage of surgeons (56 of 2299 [2.4%]) performed a mean of more than 33 cases per year, which accounts for 33.1% of women treated. Surgeons who performed less than 14 BCS procedures had a mean unadjusted reoperation rate of 35.2% compared with 29.6% in surgeons performing 14 to 33 BCS procedures and 27.5% in surgeons performing 34 or more BCS procedures. Higher volume was independently associated with reoperation rates, and the difference persisted in multivariable analyses (odds ratio [OR] for low volume, 1.49 [95% CI, 1.19-1. 87]; for middle volume, 1.20 [95% CI, 0.93-1.56]). Time since medical school graduation (duration of career) was associated with reoperation, but the difference disappeared after adjusting for surgeon volume and other patient factors (Table 2).
In adjusted analyses, only being younger than 65 years (OR for <50 years, 1.47 [95% CI, 1.37-1.58]; for 50-64 years, 1.21 [95% CI, 1.13-1.29]), nonwhite race (OR for black race, 1.07 [95% CI, 1.00-1.15]; for other race, 1.13 [95% CI, 1.05-1.20]), and carcinoma in situ (OR, 1.57 [95% CI, 1.49-1.65]) were associated with a higher risk for reoperation. Patients who did not have a high comorbidity score (OR for low comorbidity score, 0.73 [95% CI, 0.68-0.77]; for a medium comorbidity score, 0.78 [95% CI, 0.75-0.82]), who had noncommercial insurance (OR for Medicaid, 0.88 [95% CI, 0.81-0.95]; for Medicare, 0.85 [95% CI, 0.80-0.91]), and who had inpatient procedures (OR, 0.36 [95% CI, 0.33-0.39]) were less likely to undergo a repeated excision.
In this study, we found a steady decrease in rates of reoperation over time; however, nearly 1 in 4 women still underwent at least 1 additional surgical procedure for breast cancer. Further, we found that individual surgeons’ reoperation rates varied widely, and many surgeons had rates much higher or lower than expected. Finally, we documented a clear surgeon volume-outcome effect and increasing changes in selecting BCS as a reoperation procedure.
To our knowledge, this study is the first to examine reoperation rates in the Unites States at a population level using a cohort that includes certified and noncertified cancer centers. The reoperation rates found in this study are consistent with those reported by Jeevan and colleagues,28 who found that nearly 1 in 5 women underwent a second operation after BCS in the UK National Health Service. Wilke and colleagues19 examined the US National Cancer Database from 2004 to 2010 and found a mean reoperation rate of 23.6% in the country, with a small decrease in rates over time (2.5% for noninvasive and 2.2% for invasive cancers). They also observed a decreased rate of reoperations with increased patient age, which is aligned with our results.19 We believe that the findings of reduced occurrence of reoperations are encouraging and imply improvements in training and patient selection for BCS.
The reduction in the use of BCS as a primary cancer treatment corresponds to an increase in mastectomy rates during the last decade, especially among women younger than 50 years.29 Further, an increasing incidence in the United States of bilateral mastectomy, particularly contralateral prophylactic mastectomies, occurred during the same period as our study.30- 32 The reduction in the use of BCS in women younger than 50 years and a corresponding reduction in overall reoperation rates over time implies that surgeons may be selecting more appropriate patients for the procedures.
Fisher and colleagues33 also noted rising mastectomy rates and confirmed in a survey of 332 mastectomy recipients from 2006 to 2010 that fear of cancer recurrence was the biggest motivator for declining BCS. In addition, they found no difference in the reported reasons by age group. Addressing the motivations of the women who undergo BCS is beyond the scope of this study.
Our finding related to the effect of surgeon volume on reoperation is novel and unprecedented. Limited prior evidence on the topic was not conclusive. Lovrics and colleagues34 reviewed 489 patients referred for radiotherapy in Hamilton, Ontario, Canada, for predictors of positive margins after BCS and found that surgeons with higher volumes resected a larger amount of tissue and had lower rates of positive margins; however, these differences disappeared after case-mix adjustment. McCahill and colleagues17 also considered surgeon volume in their investigation of predictors of reexcisions within 311 patients with positive postoperative margins and 1909 patients with negative postoperative margins and found no association in either group. These studies, however, included a relatively small number of surgeons and centers. Large national and regional observational studies are the main tools for evaluation of patient selection, practice, and physician patterns in surgery,35 and our study is unique from this perspective. Furthermore, the variation of reoperation rates from 0% to 100% is stunning when we consider that about 20% of surgeons have a reoperation rate of more than 50%. A large number of surgeons had much lower than expected rates of second operations, which also requires reflection and future research. Until recently, no consensus existed about when to perform a reexcision related to the width of surgical margins.9,11,36 The Society of Surgical Oncology and American Society for Radiation Oncology have now recommended that ink-positive margins on invasive tumors and carcinoma in situ undergo reexcision; however, patients with negative but perceived close margins are not recommended for a second surgery.37 Assessment of how the consensus statements will affect reoperations in the future may further reduce reoperation rates.
Our analysis of the effect of patient characteristics on reoperation rates also updates the results of prior investigations. Aziz and colleagues11 investigated positive margin status and reexcisions in 1430 women undergoing BCS from 1987 to 1997. They found that the odds of a positive margin were inversely proportional to patient age and diagnosis of carcinoma in situ. Wilke and colleagues19 also noted higher odds of a repeated operation with younger age and black race. Other studies investigating the role of race in breast cancer treatments and outcomes, however, have found that the effect disappears after adjusting for socioeconomic status,38,39 which we have not been able to include in our model. The decrease in reoperation rates with increased patient age and comorbidity may indicate that surgeons remove more tissue during an initial BCS to avoid a reoperation in higher-risk patients.
The extreme variability regarding reoperations among surgeons indicates a need for professional societies to provide guidelines and education to surgeons and trainees. The timing is right; the recent report by the American Society of Breast Surgeons Mastery of Breast Surgery (SM) Program12 examined surgeon characteristics and variation and found higher rates of reoperations among surgeons performing less than 10 lumpectomies during the 11-month study period. However, we caution against using reoperation rates as a quality metric just yet, because it may lead to unintended detrimental consequences, such as unnecessarily large excisions and poor cosmetic results. Emphasis should remain on performing the correct operation for the correct indication because future guidance is developed using surgeon- and patient-reported outcomes data.
A major limitation of this study is that tumor size, grade, and staging are unavailable in this data source, and therefore we could not adjust for these variables. These cancer variables have been shown in other studies17- 19 to have an important influence on positive margin rates and reoperation rates after BCS. Furthermore, pathologic information on positive margins and other clinical factors, such as use of magnetic resonance imaging, breast size, and tumor location, is unavailable. We believe, however, in the context of physician-level outcomes, that these factors will not bias the results; the recommendation for BCS will be made by surgeons based on their knowledge of the patients’ disease, and we have no reason to believe that unknowable patient characteristics will be unbalanced between surgeons. Lovrics and colleagues34 compared the tumor characteristics of patients undergoing BCS and found no significant differences among low-, middle-, or high-volume surgeons.
In addition to these limitations, this study does not measure outcomes such as patient satisfaction or cancer recurrence. Breast or surgical oncology fellowship training also could not be ascertained.
Nearly 1 in 4 women underwent a reoperation within 90 days of BCS across New York State from 2011 to 2013, which is reduced compared with 2 in 5 observed in 2003 to 2004. Reoperation rates vary significantly by surgeon from 0% to 100%, and initial BCS procedures performed by low-volume surgeons were associated with a 50% higher risk for a reoperation when compared with the highest-volume surgeons. After developing guidelines, quality metrics need to be developed and studies need to assess the effect of guidelines on rates of reoperations after BCS.
Correction: This article was corrected on March 23, 2016, to fix the key in Figure 1.
Corresponding Author: Art Sedrakyan, MD, PhD, Department of Healthcare Policy and Research, Weill Cornell Medical College, 402 E 67th St, Ste 223, New York, NY 10065 (firstname.lastname@example.org).
Accepted for Publication: December 1, 2015.
Published Online: February 17, 2016. doi:10.1001/jamasurg.2015.5535.
Author Contributions: Ms Isaacs had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: Isaacs, Gemignani, Sedrakyan.
Drafting of the manuscript: Isaacs, Gemignani, Sedrakyan.
Critical revision of the manuscript for important intellectual content: Isaacs, Pusic, Sedrakyan.
Statistical analysis: Isaacs, Sedrakyan.
Administrative, technical, or material support: Sedrakyan.
Study supervision: Sedrakyan.
Conflict of Interest Disclosures: Dr Sedrakyan reported receiving funding from the US Food and Drug Administration for establishing the MDEpiNet Science and Infrastructure Center, but no specific funding for this research project. Ms Isaacs reported working as a senior analyst at the Center. No other disclosures were reported.