Intraoperative Ultrasound Guidance in Breast-Conserving Surgery Improves Cosmetic Outcomes and Patient Satisfaction: Results of a Multicenter Randomized Controlled Trial (COBALT)

The COBALT trial was performed in accordance with the Declaration of Helsinki, guidelines for Good Clinical Practice, and the CONSORT statement.23 Central and local independent medical Ethics Review Boards of the participating hospitals approved the study protocol (registered at http://​www.​TrialRegister.​nl, number NTR2579).24

The COBALT trial was a comparative, two-arm, parallel group, randomized controlled trial (RCT) undertaken in six hospitals in The Netherlands between October 2010 and March 2012. Eligible patients were randomized in a 1:1 ratio to either USS or PGS.25 Patients eligible to participate in the COBALT trial were diagnosed with an early stage (T1–T2, N0–N1) palpable invasive breast cancer and were scheduled to undergo BCS. Women with preoperatively diagnosed ductal carcinoma in situ, multifocal disease, a history of neoadjuvant therapy, previous surgical treatment, or radiation therapy of the affected breast were excluded from the study, while patients who had invasive carcinoma with limited or extensive carcinoma in situ diagnosed postoperatively by the pathologist were not excluded.

Although the definition of tumor-free margins used in the current study was ‘no tumor cells at the margin’, the aim of both USS and PGS for all surgeons was to achieve complete tumor removal with surgically feasible healthy tissue margins of up to 1 cm. The closure of the lumpectomy cavity did not involve oncoplastic surgery techniques, thus allowing seroma formation.

Excess volume resection was defined as a calculated resection ratio (CRR) and calculated by dividing the volume of each specimen by the optimal resection volume, which is the spherical tumor volume plus an arbitrarily chosen, surgically feasible, tumor-free resection margin of 1 cm.7,26 In an ideal situation, the specimen volume is identical to the optimal resection volume and the CRR equals 1.0.

Follow-up was performed at 3, 6, and 12 months after initial surgery and consisted of digital photographs of the breasts and patient self-evaluation. Digital photographs consisted of standardized 4-viewpoint pictures (one frontal, two oblique, and one lateral picture of the breasts, from neck to waist), and were taken during regular outpatient clinic visits using a digital camera of at least 5 megapixel resolution.

The COBALT trial used subjective (panel- and self-evaluation) and objective [Breast Cancer Conservative Treatment cosmetic results (BCCT.core) software] cosmetic outcome evaluation techniques. In all these methods, cosmetic outcome of the treated breast was compared with the untreated breast and scored using the 4-point Likert scale, which classifies outcomes as excellent, good, fair, or poor. ‘Excellent’ meant identical to the untreated breast and ‘poor’ indicated a marked difference with the untreated breast.27 Patients who underwent mastectomy following primary surgery were scored as ‘poor’ on all three evaluation methods.

Patient, tumor, and treatment characteristics of both types of surgery were compared using the Chi squared test, Fisher’s exact test, or the two-independent sample t test. Intraclass correlation coefficient (ICC) was calculated to obtain interobserver reliability between the three panel members. The estimated consistency of agreement between the three observers at each follow-up occasion was considered poor when ICC was <0.40, fair when 0.40 ≤ ICC < 0.60, good when 0.60 ≤ ICC < 0.75, and excellent when 0.75 ≤ ICC < 1.00.29

All panel members’ scores were combined for each patient separately at a given follow-up time point by finding the mean overall cosmetic outcome score and rounding it to the nearest integer. A multilevel, mixed effect, proportional odds model for ordinal responses was used to model the odds of having a worse cosmetic outcome. Since the outcome was assessed at three different time points by three different methods of evaluation, each patient contributed nine outcome measurements, and the patient was considered as a random effect in the model. Candidate explanatory variables were study center, type of surgery, follow-up occasion, method of evaluation, T-stadium, body mass index (BMI; kg/m²), age, re-excision (yes/no), and boost (yes/no). These variables were considered as fixed effects in the model, and only those that were significant were retained in the final model. The effect of specimen volume on cosmetic outcome was investigated separately after excluding women who had undergone mastectomy after the primary surgery. The volume measurements were categorized into two groups (volume of more or less than 40 cc), and were based on the median volume. During revision of pathology reports, four specimens were identified with suspicious differences between volume and weight, and one specimen lacked volume measurement. Volumes of these specimens were therefore imputed using specimen weight (twice measured: in the operation theater and by the pathologist). A separate model was run to investigate patient satisfaction with the appearance of their breast. The proportionality assumptions for all models were checked and fulfilled.

A brief overview of patient and tumor characteristics is provided in Table 1. A complete overview of initial oncological results can be found in the study by Krekel et al.8 Mean specimen weight, volume, and CRR were significantly smaller in the USS group than in the PGS group (weight: 38 vs. 52 g; volume: 38 vs. 53 cc; and CRR: 1.0 vs. 1.7; all p < 0.01). Additional therapy was more frequently required in the PGS group than in the USS group (27 % vs. 11 %, p = 0.038) and five patients in the PGS group with tumor-involved margins underwent a mastectomy. Total missing data rates were 14 % (18 patients) at 3 and 6 months, and 12 % (14 patients) at 1 year. One patient refused to participate in follow-up, one patient had major complications due to chemotherapy and could not participate in follow-up, one patient refused follow-up after 3 months, and one patient was excluded from follow-up because she refused radiotherapy. The remaining missing data points per patient were due to the inability to attend the outpatient clinic visit despite several appointments.

The estimated consistency of agreement between the three observers was fair for responses on the first follow-up occasion (ICC 0.59) and good for the next two occasions (ICC 0.69 at 6 months; ICC 0.68 at 12 months).

Ultrasound-guided surgery resulted in better cosmetic outcomes than PGS—20 % versus 14 %, respectively, of the overall responses rated excellent. Only 6 % of the responses were poor with USS, compared to 13 % with PGS (Table 2). Overall, a patient had a 45 % reduced chance of a worse cosmetic outcome with USS compared with PGS (odds ratio [OR] 0.55, p = 0.067); the odds of having a worse cosmetic outcome became significantly greater by 12 months (OR 2.57, p < 0.001). When considering various evaluation methods, the odds of reporting a worse cosmetic outcome were lower for panel- and self-evaluation than for BCCT.core software. Furthermore, the odds of having a worse cosmetic outcome were 2.35-fold higher for T2 tumors than T1 tumors (p = 0.010). (Table 3)

To elaborate on our findings, in Table 4 we report the probabilities of scoring the breast as excellent, good, fair, or poor. For all three evaluation methods, the probability of an excellent or good score was clearly higher with USS than with PGS, and the cumulative probabilities of these responses were 70 versus 56 % for BCCT.core, 81 versus 69 % for panel evaluation, and 78 versus 66 % for patient self-evaluation, respectively.

The volume measurements were categorized into two groups (volume of more or less than 40 cc). Excision volumes >40 cc resulted in 2.78 higher odds of having worse outcome when compared with volumes ≤40 cc (OR 2.78, p = 0.002). When volume was included in the model as a covariate, the difference between the two types of surgery became nonsignificant (OR 1.18, p = 0.574). However, of the lumpectomies performed with USS, only 20 % had a volume larger than 40 cc, while the figure for PGS lumpectomies was 56 %. (Table 5)

After USS, 90 % of patients were very satisfied or satisfied with the appearance of their breasts compared to 80 % with PGS. The chances of being less satisfied were 86 % lower after USS than after PGS, controlled for follow-up time points, T-stadium, and BMI (OR 0.14, p = 0.006). (Table 2)