Trial set-up and patient population
This was a prospective, single arm, multi-centre feasibility study. All required regulatory approvals were obtained prior to patient recruitment. Between 2 February 2018 and 2 July 2019, women ≥18 years of age diagnosed with biopsy confirmed invasive breast cancer and scheduled to undergo primary WLE with or without sentinel lymph node biopsy (SLNB) or axillary lymph node dissection (ALND) were recruited at three study sites in Poland. Exclusion criteria were ductal carcinoma in situ (DCIS), pleomorphic lobular carcinoma in situ (LCIS), surgery or radiotherapy to the ipsilateral breast, impalpable lesions scheduled to have radio-guided occult lesion localization (ROLL), neoadjuvant chemotherapy, known hypersensitivity to 18F-FDG, pregnancy, lactation, an active history of other cancer or other health issues compromising study participation. Women of childbearing age were required to have a negative beta-HCG qualitative test result, a history of surgical sterilization or amenorrhea in the past 12 months.
Radiopharmaceutical administration
Patients scheduled for SLNB received up to 150 MBq technetium-99m-labelled nanocolloid (99mTc-nanocolloid) administered according to local practice on the morning of surgery. A potentially higher injected 99mTc-nanocolloid activity aimed to reduce the effect of 18F-FDG crosstalk due to down-scatter of 511 keV 18F ɣ-photons into the 99mTc ɣ-probe energy window. Following the 99mTc-nanocolloid injection all trial patients received an intravenous injection of 18F-FDG ≤5 MBq/kg, with a maximum of 300 MBq, 60-180 minutes prior to intraoperative FAR imaging. The administered activity was determined on a per site and per patient basis.
Surgery
Following induction of anaesthesia, patients due to undergo SLNB received an intraoperative blue dye injection as per local guidelines and SLNs were identified by using a hand-held ɣ-probe and blue discoloration of lymph nodes. In the majority of cases WLE was performed prior to the SLNB/ALND, to ensure a minimum signal intensity loss from radioactive decay of 18F-FDG in the time window between injection and FAR imaging. The WLE specimen was removed and sutures/surgical clips were placed on the specimen to record the anatomical orientation.
FAR imaging system
The LightPath® Imaging System (Lightpoint Medical Ltd., Chesham, UK), an in vitro diagnostic device, was used to visualize the location and distribution of 18F-FDG using FAR. This system is further described by Ciarrocchi et al. [13]. A 12 µm thick flexible scintillating film was used as an accessory to the LightPath® System (Fig. 1) and consisted of a multilayer sandwich construction as follows: 3 µm of mylar, 6 µm of P43 scintillating phosphor and 3 µm of mylar. The thinness of these layers made the scintillator insensitive to the 18F-FDG 511 keV ɣ-photons [10].
Images were acquired using a 300 seconds acquisition time and 8x8 pixel binning (pixel resolution 938 μm). These imaging settings were based on the results from the clinical 18F-FDG CLI study in breast cancer [6], and an in vitro FAR study performed with the LightPath® Imaging System [14].
To reject the discrete signals from high-energy annihilation ɣ-photons three emCCD frames were taken with an acquisition time of 100 s each. The emCCD frames were then combined into a single image by applying a spatial-temporal median filter (3 x 3 pixels x 3 time points). A Gaussian smoothing filter (σ = 3 pixels) was applied and the resultant image was scaled and translated for overlay display on top of the reference image. The resulting merged image was called the FAR image.
Intraoperative imaging
Following surgical excision and orientation, the WLE specimen was positioned in a disposable specimen tray, inserted in the imaging chamber and a reference image was acquired. A transparent plastic sheet was attached to the LightPath® monitor and based on the position of the specimen on the reference image the specimen contours and tumour margins were outlined with a marker pen (Fig. 2). The specimen was then draped with a 5 μm Mylar separator sheet and a flexible scintillator film and a FAR image acquired. The separator sheet acted as a protective layer to prevent contamination of the scintillator film from biological tissue and fluids. The aforementioned steps were repeated with the WLE specimen reorientated, placing the previously occult margins in the field of view.
All images were acquired within 60 – 180 minutes following 18F-FDG administration. Upon completion of FAR imaging the surgical specimens were sent for histopathology assessment.
The first study patient of each surgeon was considered a familiarization patient for the surgeon to become acquainted with FAR imaging and the intraoperative study procedures. These familiarization patients were excluded from the analysis dataset.
Radiation exposure
Prior to commencement of the study, staff at all study sites received radiation safety training as per local policies and procedures. At the start of each surgical procedure, designated staff members were issued with electronic personal radiation dose (EPD) monitors, worn in the anterior top pocket of the clothing. Ring dosimeters for the extremities were worn depending on local requirements. Thorough contamination monitoring of equipment, rooms, waste and staff was completed after each use, to ensure that potential radioactive contamination was identified and appropriately managed.
Histopathology
Histopathological analysis was performed according to local practice with reporting on tumour characteristics, invasive and in situ components and distance from tumour to all 6 margins (anterior, posterior, lateral, medial, superior and inferior). Local margin definitions were used as follows: a positive margin for invasive cancer was defined as tumour cells at the margin (= 0 mm) at site 1 and 2, and tumour cells <1 mm of margin at site 3. For DCIS the definition was 0 mm (site 1 and 2) and <2 mm (site 3). The histopathologists were blinded to the intraoperative FAR findings.
Image analysis
Following acquisition, the FAR image was assessed by the surgeon during surgery and postoperatively by three independent breast surgeons, the central readers (Central Read). It was left to the operating surgeon’s discretion to use the FAR images for clinical decision-making. The Central Read was performed post-operatively to provide a controlled and standardized analysis environment; central readers were blinded to the intraoperative FAR and postoperative histopathology results. The central readers visually analysed the FAR image in the LightPath® software (version 2.0.20) to identify areas of increased signal intensity called “hotspots”. A hotspot was classified as either a tumour hotspot or artefact. Each central reader analysed approximately one-third of the total number of images. A tumour hotspot was defined as a focal area of more than 1 mm in diameter that displayed an increased signal intensity over tissue background. The presence of a tumour hotspot indicated the tumour was close to the surface as the positrons from 18F-FDG travel only approximately 1 mm in tissue [11]. All readers were taught to recognize image artefacts as part of the formal training that took place prior to image analysis. A tumour hotspot on the intact WLE images resulted in a tumour margin being classified as positive on FAR.
The emCCD images were quantitatively analysed by an independent researcher/surgical fellow blinded to the histopathology results using OsiriX Lite version 11 (Pixmeo SARL, Geneva, Switzerland). Based on the findings from the Central Read, regions of interest (ROIs) were manually drawn to mark areas of increased signal intensity and background signal within and outside the specimen’s contours (Fig. 3). The location, mean, minimum and maximum radiance (photons/s/cm2/sr/MBq) were recorded for each ROI. Tumour-to-tissue background ratios (TBR) were calculated. The radiance was decay corrected to make the ROIs comparable between images taken at different time points.
Images acquired outside the 60 – 180 minutes time window or containing an image acquisition artefact were excluded from analysis.
Statistical analysis
The diagnostic accuracy of FAR was assessed by calculating the sensitivity, specificity, positive predicting value (PPV), negative predictive value (NPV), and overall accuracy. Subgroup analysis was performed to assess the effect of study site, tumour characteristics, time of imaging and 18F-FDG activity at time of imaging (i.e. decay corrected activity) on diagnostic accuracy. Thresholds for subgroup analysis were determined by means of ROC curves.