A total of 596 patients were randomized, with 298 in each arm of the trial. Patient demographics and baseline characteristics are listed in Table 1. Patients underwent extensive imaging before surgery. The mean extent of disease was similar in the two groups. The main specimen volume was similar in both groups, reflecting no difference in surgical procedure before randomization.
The disposition of patients in both arms of the trial is shown in Fig. 2. In similar proportions of patients in both arms, the main lumpectomy specimen contained at least one positive margin (Fig. 2, phase I). In patients with positive margins on initial lumpectomy specimens, an average of two margins was involved, with no difference between the two arms. With reference to the patients with positive main specimen margins, surgeons correctly identified all positive margins on the main specimen and removed additional tissue from those involved margins (Fig. 2, phase II) in 33 of 147 cases (22 %) in the control arm, versus 101 of 163 (62 %) cases in the device arm (p < 0.0001). Patients for whom positive margins on the main specimen were not identified remained with positive final margins after the lumpectomy (Fig. 2, phase III, branches C1 and D1). Although the main specimen was cleared, some final margins were persistently positive because of disease identified at the edge of the additional tissue resected (phase III, branches C2 and D2) in 8 and 22 cases for the control and device arms, respectively. Interestingly, additional tissue was removed from the lumpectomy cavity in both arms in cases where the main specimen was found to have clear margins, resulting in positive final margins (phase III, branches C3 and D3) in 2 and 8 patients in the control and device arms, respectively.
Table 2 lists the patients’ final margin status after the primary lumpectomy procedure. In the control arm, 41.6 % (Fig. 2, branches C1, C2, and C3) of patients had positive margins compared with 30.9 % (Fig. 2, branches D1, D2, and D3) of patients in the device arm (p = 0.008), representing a 26 % reduction in the positive margin rate. Even though these patients had positive margins, surgeons determined that certain patients were not candidates for reexcision because the involved margins were recorded to be at skin or fascia. Excluding these patients, the significant difference in candidates for reexcision was maintained, favoring the device arm (p = 0.013). More patients in the control arm were candidates for reexcision because of positive margins originating from the main specimen. In contrast, there were more candidates for reexcision in the device arm on the basis of additional cavity shavings removed.
As shown in Table 2, 19.8 % of patients in the device arm underwent second procedures for reexcision of lumpectomy margins compared with 25.8 % of patients in the control arm, representing a 6 % absolute (23 % relative) reduction associated with MarginProbe use. The analysis of this difference also accounted for the small but statistically insignificant (prerandomization) difference between arms in the number of main lumpectomy specimens with positive margins (Fig. 2, phase I). With regard to reexcision procedures that were required because of positive margins originating from the main lumpectomy specimens (Fig. 2, branches C1 and D1), the control arm rate was 20.8 % compared with 10.0 % in the device arm, a 47 % reduction (p = 0.002).
To further evaluate device performance, the volume of tissue resected was analyzed. Both true-positive and false-positive device readings resulted in excision of additional breast tissue. Therefore, total volumes of excision were calculated across all surgeries (Table 3). As expected, the volume of tissue in main lumpectomy specimens was identical in the two arms. In the device arm, there was more tissue removed in the first surgical procedure, representing both true-positive and false-positive margin excisions. However, there was more tissue removed in reexcision procedures in the control arm. This led to an overall difference of 8.5 ml in tissue volume removed between the two study arms. When normalized to baseline breast volume, the difference between the arms was 2.6 %.
The performance of the MarginProbe in the provision of diagnostic information was also evaluated. The margin-level sensitivity of the device was 75.2 % (95 % CI: 69.4–81.0), with that of the control arm being 33.9 % (95 % CI: 27.6–40.2). False-negative rates were 24.8 and 66.1 % in the device and control arms, respectively. The increase in sensitivity in identification of positive margins by the device came at the expense of a reduction in margin-level specificity: 46.4 % (95 % CI: 42.9–49.9) device versus 83.4 % control (95 % CI: 81.0–85.8). False-positive rates were 53.6 and 16.6 %, in device and control arms, respectively.
Similar adverse event rates were observed in both groups: device, 6 events (2 %), and control, 5 events (2 %). Of these reports, only 1 event was possibly related to the study device (wound infection).