Introduction
Early stage breast cancer patients and their surgeons are confronted with the complex decision between breast conserving surgery (BCS) with radiotherapy or mastectomy (MST) with or without breast reconstruction. Both have similar overall survival [
1‐
3]. In the absence of an oncological contraindication for BCS, the treatment choice is a matter of expected cosmetic result that influences quality of life (QoL) [
4,
5]. The surgeon and patient discuss the expected cosmetic result of both BCS and MST, but preoperative prediction of the cosmetic result is typically based on informal assessment by the surgeon. An objective decision aid taking cosmesis and QoL into consideration does not exist.
It is generally believed that the benefit of BCS over MST depends on the cosmetic result after BCS. Good cosmesis after BCS has shown to yield a substantial QoL benefit over poor cosmesis [
4]. Therefore, it can be hypothesized that BCS is preferable when the chance of a good cosmetic result is high, but MST (with or without breast reconstruction) is preferable when BCS is unlikely to have a good cosmetic result. We have previously demonstrated that a good cosmetic result after BCS can be predicted preoperatively by tumor volume/breast volume ratio (TV/BV ratio) and tumor location [
6]. It remains unknown above what specific chance of good cosmetic result, BCS results in higher QoL over MST. If BCS is already performed for a low chance of good cosmesis, too many patients receive BCS increasing the incidence of poor cosmesis and reducing QoL. If BCS is only performed in those patients where good cosmesis is very likely and MST is performed otherwise, too many patients receive MST reducing QoL.
The aim was to guide decision making for patients with early stage breast cancer using a decision model that considers both predicted cosmetic result and QoL after BCS and MST. The treatment threshold—when to treat with BCS or MST—for optimal QoL is calculated. In clinical practice, this decision model could inform the treatment decision by weighing QoL of each treatment option.
Discussion
BCS is the treatment with superior QoL for breast cancer patients with a chance on good cosmetic result exceeding 36%. This 36% threshold is reached in case of an upper lateral, lower lateral, upper medial, lower medial, and central tumor location in combination with a tumor/breast volume ratio below 21.6, 4.1, 15.1, 3.2, and 14.7, respectively. In case of larger tumor/breast volume ratio’s, MST (with or without breast reconstruction) will result in higher QoL.
Due to the progress in surgical techniques and improving oncological outcomes, QoL and cosmesis have an increasingly prominent role in breast cancer treatment. Incorporating QoL and cosmesis in a treatment decision model comes with challenges due to their subjective entity. Cosmesis cannot be captured in clear-cut classifications and a (gold) standard for measuring cosmesis after BCS is lacking. Consequently, the validity of this decision model depends on the methodology of measuring and predicting cosmetic result and the definition of good and poor cosmesis. This is a weakness of our study since the validity of our definitions and self-developed Erasmus MC Panel questionnaire have not been tested extensively. The panel assessment is therefore study specific. How to measure cosmetic result after BCS is still a matter of debate, however, a multiple-person panel evaluation remains the most used and recommended [
18]. It does come with disadvantages as it remains a subjective method with low reproducibility results influenced by the type of observers (experts versus non-experts) and the scale used. Consensus increases if the outcome is dichotomized. It could be argued that cosmetic result should be based on a patient-reported outcome measure (PROM). However, PROM’s only deliver scores on a continuous scale and their predictive ability is low (e.g., TV/BV ratio was not associated with patient evaluation of cosmetic result) [
6]. The more efficient BCCT.core software for cosmetic result evaluation might become the future standard, but until today, agreement with panel evaluations differs [
19,
20]. The 12 patients with false positive result (i.e., model predicted a good cosmetic result but the panel evaluated cosmesis as poor) could be explained by the fact that 8 out of 12 patients had visible radiotherapy fibrosis or skin color changes and retractions around the scar which is not taken into consideration in the predicted cosmetic result. Another explanation could be that 3/12 (25%) of false positive patients were operated by a general oncological surgeon (i.e., not breast cancer specific surgeon) as compared to 8/69 (12%) of the total study population. This suggests that surgeon experience matters for achieving a good cosmetic result. The lack of other significant parameters, like radiotherapy boost, in the prediction model for cosmetic result was a disadvantage. Yet the model proved to discriminate well between poor and good cosmetic result (AUC of 0.83). As far as we know, other prediction models for cosmetic result or other decision models including cosmetic result and/or QoL are not available in the literature. Therefore, our models could not be compared or tested for external validity.
QoL is a complex and multifactorial outcome that is not captured by cosmesis and type of surgery only. For example, age, chemotherapy, and fear of recurrence have been shown to be associated with QoL [
21‐
23]. However, fear of recurrence was not considered here and no significant association was seen between QoL and age, stage, or chemotherapy. The lack of adjusted utilities for these factors was a drawback. Future studies should also address fear of recurrence. The mean utility found for poor and good cosmesis after BCS were not significantly different (
P = 0.055). No sample size calculation was performed for this study and it was not powered to show a significant QoL difference. The nature of our tertiary referral university hospital caused a small study population for BCS. This had multiple implications, like the lack of power to detect QoL differences and large (statistical) uncertainty of the utilities for BCS. To decrease this uncertainty, the mean utility for BCS in general measured by EQ-5D in 1050 patients from Freedman et al. was used [
11]. The results are, however, not directly generalizable beyond our study population and need to be validated. Due to the small sample size, the results of this study are preliminary. Work to validate the results in a larger study is underway.
It was assumed that QoL after BCS would be related to the cosmetic result. However, cosmesis was not significantly associated with EORTC QLQ C-30 and BR-23 scores, although absolute differences were seen [e.g., body image score of 87 vs. 73 (0–100) in good versus poor cosmesis,
P = 0.203] [
6]. This could be explained by the fact that all the questionnaires are general or disease-specific, not cosmetic result specific. The assessment of cosmesis after BCS by an independent panel, and the lack of a panel assessment after MST (with and without breast reconstruction), could also explain the lack of association with PROM’s. Studies with patients’ cosmetic evaluation, however, were significantly associated with QoL [
4,
5]. A larger study population is needed to draw final conclusions about whether EQ-5D is sensitive for breast cancer surgery outcome differences. Promising alternatives for future efficiency analysis are the recently published EORTC Quality of Life Utility Measure-Core 10 dimensions (QLU-C10D) [
24], EORTC Quality of Life for women undergoing breast reconstruction (QLQ-BRECON26) [
25], and the modernized EORTC BR23 module that is currently under development.
Another disadvantage was the complexity and use of different sources for generating the utilities. Namely, EORTC values with EQ-5D UK tariffs from our study population, EQ-5D values with unknown origin of tariffs from Freedman et al. and FACT values with EQ-5D US tariffs from Jagsi at el. The time of measurement was 38 months after surgery, 60 months after surgery, and 50 months after diagnosis, respectively. We did not correct for these time differences. Neither did we correct for the variety in study population size, but the standard deviations in the PSA show that the uncertainty for most BCS related parameters is larger when compared to the MST-related parameters caused by a smaller study population. The mastectomy group had more stage III patients (1 vs. 18%) who underwent less radiotherapy (100 vs. 30%), but this was not unexpected. An alternative for using multiple sources to define the utilities for each health state was not available since, as far as we know, no study has been published presenting EQ-5D derived QoL values for the different surgical treatments of breast cancer. We were the first to compare utilities between BCS and MST.
Whether the differences in utilities found between the treatment options are clinically meaningful can be questioned. The minimally important difference has been estimated to be 0.08 for UK-based EQ-5D scores and 0.06 for US-based EQ-5D scores by Pickard et al [
26]. We found a difference between good and poor cosmetic result after BCS of 0.065. However, Pickard et al. estimated the minimally important differences in advanced stage patients with cancers from all kinds of primary origin. Our study population consists of early stage breast cancer patients that already have a high baseline QoL (towards the maximum of one) which may impede finding large QoL differences. Other studies comparing QoL between treatment options showed comparable ranking order of the treatments to our decision model [
7,
27,
28]. Which is important since our decision model assumes BCS with good cosmesis is preferred over MST (with or without reconstruction) followed by BCS with poor cosmesis. Atisha et al. and Jagsi et al. found that satisfaction with breasts as measured by Breast-Q questionnaire was highest in patients receiving MST with autologous reconstruction and slightly decreased with BCS, followed by MST with implant reconstruction, and was the lowest with MST only [
7,
27]. Han et al. found that QoL and satisfaction as measured by EORTC QLQ-C30 & BR23 was higher for BCS as compared to MST or reconstructive surgery [
28]. These studies suggest that current QoL differences found are clinically meaningful. It has also been found that utility increases by 0.031 if the patient is given a treatment choice versus restricting choice to MST alone [
29]. Furthermore, if poor cosmesis is expected, other treatment alternatives are available, namely, neo-adjuvant chemotherapy and oncoplastic surgery including therapeutic mammoplasty. Utility studies are needed regarding these alternatives to enable implementation in a treatment decision model. This study should be interpreted as preliminary and a first step towards a more ideal treatment decision model.
The treatment decision model presented here is currently being studied in a randomized controlled trial in patients who are candidates for both BCS and MST to study the effectiveness in improving cosmetic result and QoL over the present situation [
30]. Here, TV/BV ratio is measured by ultrasound instead of MRI since it is less invasive, widely available, and more cost-effective. Validation of the volume measurements is currently awaiting publication. There is a fair amount of QoL benefit (0.02 per patient) that can be expected if more research is performed that decreases parameter uncertainty as shown by our value of information analysis. Currently, we are measuring utilities in a larger study population for all health states to reduce uncertainty and improve the decision model.