Impact of experience in breast cancer surgery on survival: the role of quality of care in a registry-based cohort

Previous studies have suggested that the surgeon's experience in breast cancer surgery may affect patient survival. In this registry-based retrospective cohort study, we examined whether quality of care could partly explain this association. TABAN, François, et al. Impact of experience in breast cancer surgery on survival: the role of quality of care in a registry-based cohort. Swiss Medical Weekly, 2019, vol. 149, p. w14704 DOI : 10.4414/smw.2019.14704

However, why the surgeon's experience could influence breast cancer mortality remains unclear.The authors of two reviews highlighted weaknesses in previously published studies.Most studies were based on administrative data, did not adjust their analyses for differences in the patient characteristics and analysed overall mortality, known to be strongly influenced by patients' co-morbidities [20,21].Finally, they suggested that the observed differences in survival could be explained by differences in the quality of breast cancer management (not limited to the surgery).
Switzerland has one of the most expensive healthcare system worldwide [22,23] and the canton of Geneva is among those with the highest medical density, with approximately 5 physicians per 1000 inhabitants.In this canton, a large proportion of breast cancer surgery is performed in the private sector, sometimes by breast cancer surgeons who perform fewer than five breast cancer operations per year.However, Geneva provides some of the best quality of care for breast cancer in Switzerland [24] and has breast can-Author contributions FT contributed to the study concept, interpreted the data and revised the different versions of the manuscript; NE contributed to the study design, data cleaning and statistical analyses, interpreted the results and revised the final manuscript; ER contributed to the study concept and design, analysed and interpreted the data and reviewed the different versions of the manuscript; CR contributed to the analysis and interpretation of the results and revised the final manuscript ; GF was responsible for the quality control of data and algorithms, performed the statistical analyses and revised the final manuscript; SB contributed to the analysis and interpretation of the results and revised the final manuscript; EDM contributed to the data analyses and interpretation and to the final manuscript editing and revision; TGL contributed to the interpretation of the results and revised the final manuscript; CB contributed to the study concept and design, interpreted the data, drafted the first manuscript and revised the different versions of the manuscript.All authors have approved the final manuscript.cer survival rates that are among the highest in Switzerland and Europe [25][26][27].
The aim of this population-based retrospective cohort study was to investigate whether the association between surgeon's experience and breast cancer mortality was also true in this very specific context, even after adjustment for patient and tumour characteristics, and whether this association could be confounded by the quality of care provided.

Material and methods
This study is reported according to the RECORD extension [28] of the STROBE statement [29] for reporting observational studies using routinely collected health data.

Ethical approval
Formal ethical approval and patient consent for this study was not required.The Geneva Cancer Registry (GCR) has a general authorisation [30], to collect nominative data and to analyse the anonymised data.

Design and setting
This retrospective cohort study was based on data routinely collected by the GCR, which has recorded all incident cancers occurring in Geneva, Switzerland (approximately 480,000 inhabitants in 2014) since 1970.The data recorded include sociodemographic variables, tumour characteristics coded according to the International Classification of Diseases for Oncology (ICD-O) [31], stage at diagnosis (coded according to the Tumor, Node, Metastasis Classification of Malignant Tumors [32]), and treatment received within 6 months of diagnosis, including the identity of the physician in charge of the first treatment for the patient, for the private sector.

Outcomes
The primary outcome of this study was the 5-year breast cancer-specific survival of patients according to surgeon's experience, adjusted for variables known to influence survival, for patient and tumour characteristics that were significantly associated with the surgeon's experience or patient survival in the present cohort, and for quality-of-care indicators.

Surgeon experience
To define surgeon experience, we calculated for each surgeon the average number of breast cancer operations performed per year among the resident population.In order to avoid fluctuations due to various reasons (e.g., decreasing activity during the last years of the surgeon's working life), we considered only the 3 years, between 2000 and 2009, during which the surgeon performed the highest number of breast cancer operations.We then stratified the surgeons into three categories: those performing ≤5, 6-10 or >10 breast cancer operations per year.

Indicators of quality of care
State-of-the-art breast cancer management was defined according to the quality indicators described by the European Society of Breast Cancer Specialists (EUSOMA) [33].We selected nine indicators for which information was available from the GCR database: (1) reported hormone receptor immune-activity, tumour size, and grading; (2) histological assessment before surgery; (3) a single operation for the primary tumour (excluding reconstruction); (4) sentinel lymph node excision for clinically negative axillae; (5) ≥10 lymph nodes removed when axillary dissection performed; (6) breast-conserving surgery for tumours ≤3 cm; (7) radiotherapy if indicated (after breast-conserving surgery if no metastasis or after mastectomy for pT3 or pT4 or positive margin or ≥pN2a); (8) endocrine therapy for oestrogen-receptor positive tumours; and (9) chemotherapy for oestrogen-receptor negative tumours >1 cm or with a positive lymph node (we also considered an age of ≤35 years as an indication for chemotherapy, according to the 2003 Saint Gallen Consensus [34]).We added two additional criteria, not included in EUSOMA: (10) axillary lymph node dissection if clinical involvement or positive sentinel lymph node biopsy and (11) presence of negative margins after the last surgery.Each indicator was scored 1 when correctly performed or 0 if not correctly performed, and was omitted from the score if not applicable to the patient.
For each patient, we calculated the proportion of pertinent indicators correctly fulfilled, as explained in detail in a previous study [35].This overall quality-of-care score was categorised as <75%, 75-90% or 90-100% of the items fulfilled.
In order to allow indirect comparison with the public sector, we additionally present the data reported in a previously published study on a similar cohort from a public breast cancer unit [35].

Breast cancer-specific mortality
The GCR performs active follow-up yearly, by linking the GCR files with those of the Cantonal Population Office.The cause of death is provided by the Federal Office for Statistics, and coded according to the International Statistical Classification of Diseases and Related Health Problems [29].The exact cause of death is confirmed by a physician at the GCR after consulting clinical records and/or inquiring of the patient's physician.

Surgeon experience
The patient and tumour characteristics, and the 11 individual items included in the quality-of-care score were reported according to the surgeon's experience as numbers (percentages) or means (95% confidence intervals [CIs]), and compared with a χ 2 test or analysis of variance (ANOVA), as appropriate, to identify the variables significantly associated with surgeon experience.5-year survival All patients were followed up from the date of confirmation of a breast cancer diagnosis until 31 December 2014, death or the date of loss to follow-up, whichever occurred first.Only deaths from breast cancer were considered.Each variable (patient and tumour characteristics) was included in a univariate Cox regression model to identify those significantly associated with 5-year breast cancer-specific mortality.
The crude association between surgeon experience and 5-year breast cancer-specific survival was examined graphically with Kaplan-Meier curves, and a Cox regression model was constructed to report the hazard ratios (HRs) and 95% CIs for comparison of breast cancer mortality in patients treated by surgeons performing 6-10 and >10 breast cancer operations/year with those treated by surgeons performing ≤5 operations/year (baseline).

Multivariate model
We used a multivariate Cox regression model including all variables known to be strongly associated with breast cancer-specific mortality (age, tumour stage, grade, and oestrogen and progesterone receptor status), and the patient or tumour characteristics that were shown to be associated with either 5-year breast cancer-specific survival or with surgeon experience in the univariate analyses.A variable was then dropped from the model if it were not significantly associated with the outcome, did not contribute significantly to the fit of the model to the data, established with a likelihood ratio test comparing the model including the variable with one excluding it (p >0.1), or did not act as a confounder, evident as a change in HR of >10%.Finally, we quantified the impact of the quality of care by introducing the quality-of-care score into the last multivariate model.
Missing data for different variables were retained in the models as a category labelled "unknown".We considered differences as statistically significant at p <0.05; all p-values reported are two-sided.The proportional hazard assumption was assessed graphically.

Description of the cohort
During the study period, 88 surgeons operated on 1489 breast cancer patients.Most (n = 67) surgeons were gynaecologists; 18 were thoracic and 3 were plastic surgeons.A total of 651 breast cancer patients (44%) were operated on by 5 surgeons who performed >10 operations/year, 434 (29%) by 12 surgeons who performed 6-10, and 404 (27%) by 71 surgeons who performed ≤5.Among the latter group, 37 (9%) women were operated on by one of the 25 surgeons who performed ≤1 breast cancer intervention per year.

Patient and tumour characteristics
During the study period, the patients recruited by the surgeons performing >10 operations/year increased from 39 to 50%, whereas recruitment by the surgeons performing ≤5 decreased from 31 to 21% (p <0.001).Compared with the patients treated by the latter, those treated by surgeons performing >10 operations/year were more often of a higher socioeconomic status and were less frequently born in Southern Europe (table 1).Although there were differences in the proportion of "unknown data" for some variables, tumour characteristics did not differ significantly across the surgeon groups (table 2).In comparison with previously published data from the public sector [35] patients tended to be younger, from higher socio-economic status and more often born in Switzerland.The tumour characteristics, however, were similar (tables 1 and 2).

Quality-of-care indicators and overall score
Table 3 presents the 11 quality indicators according to surgeon experience.Significant differences across the categories of surgeon experience were observed for histological assessment before surgery, sentinel lymph node procedure (when indicated), and ≥10 lymph nodes removed during axillary dissection.The mean overall quality indicator score was high in all groups (above 82%), but was higher in the women treated by surgeons performing >10 operations/year; 50.5% of their patients benefited from >90% of pertinent items fulfilled.This proportion was 47.2% in patients treated by surgeons performing 6-10 operations/year and 34.7% for those treated by surgeons performing ≤5 (p <0.001).Most of the EUSOMA minimum requirements were reached except for sentinel lymph node excision (if indicated) and the number of lymph nodes removed, which did not reach 90% and 95%, respectively, in any group.Also, the administration of chemotherapy when indicated failed to reach the 80% required by EUSOMA.Interestingly, the public sector also failed to reach these standards (table 3).

Five-year breast cancer-specific survival
The 1489 patients represented a total of 7046.9 personyears of follow-up.
Variables significantly associated with 5-year breast cancer-specific survival in the univariate analyses included age, socioeconomic status, method of breast cancer detection, familial risk of breast cancer, stage, lymph node invasion, grade, histology, and oestrogen and progesterone receptor and HER2 status.

Multivariate analyses
The variables retained in the final Cox model were age, socioeconomic status, stage, lymph node invasion, grade, histology, and oestrogen and progesterone receptor status.
A second model was constructed that included the qualityof-care score.
In the crude analysis, the patients operated on by surgeons performing >10 operations/year presented with 66% lower breast cancer-specific mortality than those treated by surgeons performing ≤5 (HR 0.34, 95% CI 0.17-0.67;p = 0.002).Adjustment for patient and tumour characteristics reduced the strength of the association (HR adj-1 0.45, 95% CI 0.21-0.94;p = 0.034).In the final model, additional adjustment for quality of care further decreased the strength of the association (HR adj-2 0.51, 95% CI 0.24-1.08;p = 0.078) and failed to reach statistical significance.The crude HR comparing the patients treated by surgeons performing 6-10 operations/year with those treated by surgeons performing ≤5 followed the same trend.(table 4)

Discussion
This study confirms previously reported findings, highlights new ones and generates several questions.First, as previously reported, we found a statistically significant crude association between high surgeon experience, and improved breast cancer-specific survival in their patients.Second, this study demonstrates that the strength of this association decreases after adjustment for patient and tumour characteristics, and further decreases after adjustment for measurable indicators of quality of care, suggesting that these factors may, at least partly, explain the previously reported differences in survival.Third, the quality of care provided in the private sector for breast cancer is good and comparable to that of the public breast cancer unit, although some EUSOMA targets were not reached.Fourth, in Geneva, the surgeon's experience may not impact on the 5-year breast cancer-specific survival of patients operated on in the private sector.Finally, other factors reflecting the quality of care should be investigated as they may further decrease the reported association.

What was already known on the topic?
In crude analyses, patients operated on by surgeons performing >10 operations/year had a lower risk of death as a consequence of their breast cancer than patients operated on by surgeons with less experience, which is consistent with the findings of other researchers.Based on 12 of 63 studies published between 1990 and 2010, Gooiker et al. [19] reported that the pooled survival advantage conferred by high-volume surgeons was around 20% (range 10-39%).Other studies have also reported an association between hospital volume and breast cancer survival [19].
In particular, Skinner et al. reported that breast cancer pa-tients operated on by low-volume surgeons in high-volume hospitals had similar outcomes to those of breast cancer patients operated on by high-volume surgeons in low-volume hospitals [4].
Our study also confirms that patients as well as treatments may differ according to surgeon experience [36].Although better care has generally been observed among breast cancer patients treated by high-volume surgeons, none of the previously published studies have used the EUSOMA criteria to assess the quality of care received.As reported in previous studies, we found that surgeons who performed >10 operations/year more frequently performed a histological assessment before surgery [37,38], removed sentinel lymph nodes when indicated [39][40][41][42][43], removed an adequate number of axillary lymph nodes when performing axillary clearance [44,45], and referred their patients for adjuvant radiotherapy when indicated [5,45,46].

What does this study add?
What our study newly highlights is that the unexplained association between surgeon experience and breast cancer survival may be partly explained by patient and tumour characteristics, but also by the quality of care provided.In fact, taking these variables into account decreases the association between surgeon experience and breast cancer sur-  vival.This is an important finding, since our study highlights that better survival after breast cancer may be due to better quality of care, and not to the surgeon's technical ability.

Strengths and limitations
One of the strengths of our study is that we examined breast cancer-specific mortality and not overall mortality, which is influenced by patients' comorbidities.A second strength is that we adjusted our final survival model for all patient and tumour characteristics that are known to be associated with survival, or that were associated with surgeon experience or survival in our cohort.Finally, we used well-defined and recognised quality indicators to control for the impact of the quality of care, although we were unable to quantify all of these criteria based on registry data.The main limitation of this study is its observational nature.However, it is unlikely that a randomised clinical trial of this issue will ever be performed for practical and ethical reasons.Furthermore, we cannot exclude residual confounding by unrecorded variables.For example, the GCR collects information on patient characteristics and treatments, but does not collect detailed information regarding the specifics of the surgical procedures used.Also, the GCR records the name of the physician responsible for the first treatment administrated only, and for this reason, patients who had received neoadjuvant chemotherapy (about 6%) had to be excluded from our analyses.The surgeons' experience was probably underestimated in this study because we considered only the operations performed on breast cancer patients living in Geneva.Resident cancer patients represent 75% of all breast cancer patients treated in Geneva.However, we have no reason to believe that the proportion of nonresidents operated on differed according to surgeon experience, and we are quite confident that our categorisation is robust.The cut-offs used to define surgeon experience were lower than those used in most other studies [19], which reflects the reality of a city such as Geneva with both a high number of health providers in the private sector and a small population.Other studies have used various cut-off values to classify surgeon volume [19,47] and have shown a positive relationship between surgeon volume and breast cancer survival, independently of the cut-offs used.Also, this study focuses on breast cancer patients treated in the private sector; no extrapolation of our results to the public sector can be made, and we were unable to reproduce similar analyses for the public sector since the identity of the surgeons in university hospitals is unclear.However, indirect comparison with a public breast cancer unit during a similar time period showed comparable quality of care [35].We did not control for the potential impact of "hospital volume", but we are quite confident that, since there are only three private hospitals in Geneva, which are very similar in size, in their recruitment of breast cancer patients and in the quality of care they provide, this should not influence our re-  sults.Finally, some EUSOMA quality-of-care indicators were unavailable, some have changed in the latest version (i.e., recommendation on the number of lymph nodes to remove), the reasons why some procedures were performed remain unknown, and residual confounding is possible.For example, differences probably exist between surgeons in their access to multidisciplinary care.A multidisciplinary approach, which is now routinely available in specialist breast cancer units, could balance out any effect of the surgeon's experience on survival [48].At the time the breast cancer patients were enrolled in our study, a breast cancer network, SONGe (réseau de Sénologie et ONco-gynecologie Genevois), attracted some private professionals with a particular interest in the field of breast cancer care.Breast cancer surgeons affiliated to such a network may be more likely to work in a multidisciplinary context and probably have greater experience in breast cancer surgery than those who are not affiliated.Such a network could affect breast cancer-specific survival and could, once adjusted for, further decrease the strength of the association observed.

Conclusions
This study suggests that the previously reported association between surgeon experience and breast cancer mortality may be at least partly explained by patient selection and measurable indicators of quality of care.Further adjustment for variables reflecting quality of care, such as the degree of involvement in a breast cancer network with multidisciplinary meetings and co-operation, should be explored as they may confirm our findings by further decreasing the strength of the association.
BC = breast cancer; CI = confidence interval; HER2 = human epidermal growth factor receptor-2 a Surgeon's experience: mean annual new primary breast cancer (invasive or in situ) operations during the 3 years with the highest number of breast cancer intervention along the study period.b p-value of a χ 2 test c Data from Taban et al. 2013 [35] d Available since 2001 Original article Swiss Med Wkly.2019;149:w14704 Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table 1 :
Characteristics of breast cancer patients according to the surgeon's experience (Geneva Cancer Registry 2000-2009).
BC = breast cancer; CI = confidence interval a Surgeon's experience: mean annual new primary breast cancer (invasive or in situ) operations during the 3 years with the highest number of breast cancer intervention along the study period.b p-value of a χ 2 test c Data from Taban et al. 2013 [35] d p-value for ANOVA test Original article Swiss Med Wkly.2019;149:w14704 Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table 2 :
Characteristics of the tumours according to the surgeon's experience (Geneva Cancer Registry 2000-2009).

Table 3 :
[33]ity of diagnosis assessment and treatment according to the surgeon's experience(Geneva Cancer Registry, 2000-2009).CI = confidence interval; SD = standard deviation a Surgeon's experience: mean annual new primary breast cancer (invasive or in situ) operations during 3 years with the highest number of breast cancer intervention along the study period b p-value of the χ 2 test leaving non-pertinent out c Data from Taban et al. 2013[35]d Del Turco et al. 2010[33]e ANOVA Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.