Contralateral prophylactic mastectomy (CPM) for treatment of breast cancer has increased markedly over the last decade in the wake of greater patient awareness of the procedure. Today about 20% of patients receive CPM, representing about half of those who get any mastectomy, and rates vary markedly by region and by age.^1–3 Surgeons play a dominant role in advising patients newly diagnosed with breast cancer regarding their loco-regional treatment decisions.¹ Virtually all patients with curable disease see a surgeon and most patients are treated by the first surgeon they see. About two thirds of patients diagnosed with breast cancer receive a recommendation for initial breast conserving surgery (BCS) or mastectomy from their surgeon, and virtually all receive the recommended treatment.⁴ Similarly, when a surgeon recommends against CPM, few patients undergo the procedure.¹

Consensus statements by professional associations reinforce that CPM should be considered in patients at higher risk for 2^nd primary breast cancer but discouraged in those who are at average risk.^5–7 Despite this, only about a third of patients at average risk of a 2^nd primary who desired CPM reported a recommendation from their surgeon discouraging it.^1,8 Patients with the same risk of contralateral breast cancer may take in very different information and recommendations regarding CPM depending on which surgeon they see. However, no study has been published that examines the influence of the attending surgeon on variations in receipt of CPM in the community. Do patients with the same attributes get different treatment depending on which surgeon they see? What surgeon attitudes influence this potential variability? To address these questions, we used information from a large diverse contemporary sample of patients newly diagnosed with breast cancer and their attending surgeons to examine the influence of individual surgeon on the receipt of CPM.

Table 1 shows the distribution of patient (1a) and surgeon (1b) characteristics. The patient mean age was 62 (SD 11); 28% had an increased risk of 2^nd primary cancer, and 2% reported a BRCA mutation on testing; 16% of the total sample received CPM. Half of surgeons (52%) practiced for 20 years or more; almost one-third (30%) reported that they treated more than 50 cases of new breast cancer per year; and one quarter were female (25%).

Figure 1 shows the frequency of responses for the individual items in each surgeon scale. For the “favor initial breast conservation scale”, there was strong consensus favoring BCS (96%, probably or definitely, lower right figure) in the hypothetical case with no obvious contraindications to breast conservation. On the other hand, very few surgeons favored CPM (upper left figure, 96% probably or definitely not recommend), and a less unanimous but still substantial proportion would recommend against it (76%, upper right figure). There was wide variability with regard to the items that comprised the “reluctance to perform CPM scale.” Common reasons for performing CPM if requested were 1) to give patients peace of mind, 2) avoid patient conflict, and 3) improve cosmetic outcomes. Less common were to 1) avoid surveillance, 2) reduce recurrence, and 3) improve long-term quality of life. Least frequently endorsed were to avoid losing the patient or to improve survival.

Figure 2 shows the lack of a strong relationship between the two scales as demonstrated by the broad distribution of points representing individual surgeons in the scatterplot and the correlation of .30. There were a large number of surgeons who were relatively high on one scale and much lower on the other. There is some asymmetry in that while there are a reasonable number of surgeons who both favor breast conservation and are reluctant to do CPM, there is a much smaller number who are low on both scales. The two histograms show the distribution of surgeons across each scale on the respective axes.

Figure 3 shows the results of the three successive multilevel models including baseline patient variables, surgeon variables and site. The base model 1 (blue line, open circle) included elevated risk of a 2^nd primary breast cancer (high risk), BRCA mutation, and age. Mutation status had a dominant effect with an odds ratio of about 6 (6.2 in model 1 to 5.8 in model 3) for CPM receipt when the women reports being mutation positive, and this estimate was omitted from the figure so that the other odds ratios are better compared (see appendix table 1 for the full table of model estimates). However, <2% of the women reported being mutation positive (Table 1).

Age was one of the strongest predictors after BRCA status, with a 50% reduction in odds of CPM for each decade increase in age. Overall, model 1 predicted CPM extremely well with an area under the curve (ROC) of 0.84 [0.83, 0.86] and the model explains 35% of the variability in the likelihood of CPM. Patient factors explained about 15% of the variability in the likelihood of CPM, but the surgeon identifier by itself explains even more, about 20% of the variance. The odds of a patient receiving CPM would increase 2.8-fold (95% CI 2.1,3.4) if she were to see a surgeon with a practice approach one standard deviation above a surgeon with the average CPM rate (independent of age, date of diagnosis, BRCA status and risk of recurrence). In model 2 (red line, solid circle) we added the 2 surgeon scale scores for “favor initial breast conservation” and “reluctance to perform CPM”. Both substantially decrease the odds of receipt of CPM (adjusted OR 0.7 and 0.6 per s.d. respectively) and explain 25% of the surgeon influence. Finally, in Model 3 (green line, cross) we added patient geographic site, which is highly correlated with receipt of CPM: patients in Los Angeles County are much less likely to have received CPM (OR 0.5) and the variable explains a quarter of the remaining surgeon effect.

Figure 4 shows the independent effects of both attitude scales on rates of CPM in the population in model 2. The probabilities estimated the effect of changing the scale values in the population averaging over all the other variables and remaining surgeon variation. The x axis represents the score for the reluctance to do CPM scale (from least to most reluctant surgeons). The modifiers “most, more, less, and least” refer to one or two standard deviations on either side of the mean (average) scale score in our population of surgeons. The y-axis shows the expected CPM rate. The vertical layered lines indicate the effect of different reluctance scale scores on the favor initial breast conservation scale (least, average, and most favoring surgeons). So for example, 13% was the average rate of CPM for a surgeon who had average scores for both scale scores (center point on the orange line). But there was a wide variation in CPM rates based upon variation in these scale levels. At the extreme, the rate of CPM was 34% for surgeons who least favored initial breast conservation and were least reluctant to perform CPM (upper left point). By contrast, the estimated rate was 4% for surgeons who most favored initial breast conservation and were most reluctant to perform CPM (lower right point).

In our study we observed a striking influence of the attending surgeon on whether or not a patient undergoes CPM after diagnosis of unilateral breast cancer. The individual attending surgeon explained a large amount (20%) of the overall variation in CPM use in this large diverse population-based patient sample, after adjusting for factors that determine the risk of 2^nd primary breast cancer and age, one of the major determinants of CPM. Surgeon attitudes about the approach to initial surgery or response to patient requests for CPM explained about a quarter of this surgeon effect. In a scenario of a typical patient with no contraindications to breast conservation and at average risk for a second primary cancer, a large majority of surgeons favored an initial breast conservation approach and a substantial majority would recommend against CPM. There was less consensus about the willingness to perform CPM if requested by the patient. We observed a range of reasons why a surgeon would be willing to perform CPM if asked: give peace of mind, yield better cosmetic outcomes, avoid conflict with patient, reduce need for surveillance, improve long-term quality of life, reduce recurrence of invasive disease, avoid losing patient to another surgeon, or improve survival (in order of endorsement). Both of the attitudes scales independently affected the likelihood that a patient would get CPM. This varied from 34% for the (relatively rare) attending surgeon who least favored initial beast conservation and was least reluctant to perform CPM if asked, to 4% for surgeon who strongly favored initial breast conservation and was most reluctant to perform CPM if asked.

We also examined pertinent patient factors. CPM was correlated with guideline-concordant clinical factors (elevated risk of genetic mutation or BRCA mutation on testing) and patient age, which may reflect physician and/or patient views that CPM is an increasingly more relevant alternative the younger the age of onset of breast cancer. Finally, the geographic location variable explained some of remaining surgeon variation, suggesting at least some regional surgeon peer effects and potentially a regional difference in patient population attitudes toward CPM.

Prior literature has demonstrated the marked increase in receipt of CPM after diagnosis of breast cancer and underlying factors driving the trend (dominantly greater patient awareness and interest for the procedure).^2,13–18 More recent research has shown that surgeon recommendation against CPM powerfully reduces receipt. But only about a one third of patients who consider CPM report that their surgeon recommended against it and one third reported no substantial discussion with their surgeon about it.^1,8 However, our study is the first to estimate how much the likelihood of a women receiving CPM varies across surgeon.

Aspects of the study merit comment. We used a large population based contemporary diverse patient sample with a high response rate. Virtually all patients identified their attending surgeon and surgeon survey response was high. The measures were highly relevant to clinical practice and the methods were appropriate to the research questions and study design. However, there were some weaknesses. Despite high survey response rates, there was inevitable decay in the sample given the requirement for completed surveys from both the surgeon and the patient. Finally, the findings were limited to large regions of the country.