Study outline
This was a non-randomised prospective single centre study in patients with distal oesophageal carcinoma suitable for treatment with radiotherapy or concurrent chemo-radiotherapy (ClinicalTrials.gov identifier – NCT02285660, Registered 21/10/2014). Trial participants underwent standard-of-care imaging including 3D FDG PET-CT, endoscopic ultrasound and radiotherapy planning 4D CT. A trial-specific contrast-enhanced 4D FDG PET-CT in the treatment position with limited coverage of the lower thorax and upper abdomen was also performed. Subsequent treatment was not affected by the trial-specific PET-CT and was delivered according to institutional clinical protocols.
Patient selection and recruitment
Inclusion criteria were as follows: Age ≥ 18 years; World Health Organization (WHO) performance status 0–2; histologically proven distal oesophageal carcinoma; clinical decision made to proceed with radiotherapy +/− concurrent chemotherapy; measurable primary tumour +/− loco-regional metastatic lymph nodes on standard-of-care imaging; able to provide fully informed written consent; able to lie flat for 1 h; not pregnant or breast feeding. Female patients of childbearing potential agreed to use effective contraception, were surgically sterile, or were post-menopausal.
Exclusion criteria included: poorly controlled diabetes; renal impairment with estimated glomerular filtration rate < 30 mL/min; allergy to iodinated contrast media.
The study was approved by the regional Research Ethics Committee (Approval Reference 11/YH/0213). All patients provided informed written consent prior to trial entry. A total of 15 patients were recruited between October 2011 and July 2014. All patients provided informed written consent prior to study entry.
PET-CT technique
4D PET-CT scans were performed using a 64-slice GE Discovery 690 PET-CT scanner (GE Healthcare, Amersham, UK) using Real-time Position Management (RPM) respiratory-gating hardware (Varian Medical Systems Inc., Palo Alto, CA, USA), a flat couch top and laser alignment. Patients were scanned supine in the treatment position; immobilised with a wing board and both arms raised above their heads. Serum blood glucose was checked routinely and imaging was not performed in patients with a blood glucose level of >10 mmol/L. Patients fasted for at least 6 h prior to intravenous injection of 400 MBq of 18Fluorine-FDG. Patients were positioned on the hard-top couch with laser alignment to skin tattoos marked during standard-of-care radiotherapy planning CT 45 min’ post injection. A non-contrast CT of the treatment area (lower thorax/upper abdomen) was obtained using standard settings: 120 kV, variable mA (min 50 max 500, noise index 22.6), tube rotation time 0.4 s, pitch 1.984 with a 2.5 mm slice reconstruction. Static PET acquisition from mid thorax to upper abdomen was then performed scanning in a cranial direction with 23 slices (50% overlap) acquired over 1 min. 60-min after tracer injection a 4D respiratory-gated PET acquisition commenced scanning in a cranial direction over the same volume (20-min acquisition). Breaths per minute (BPM) were monitored during this acquisition and average breathing period in seconds (60/average BPM) was calculated. The cine acquisition parameters for 4D CT were based on average breathing period determined from the 4D PET acquisition. The 4D CT component was obtained 35 s following a bolus of 100 ml of iodinated contrast (Niopam 300, Bracco Ltd., High Wycombe, UK) injected at 2 ml/s using the following settings; 120 kV, 150 mA, tube rotation 0.4 s per rotation, pitch 1.984, 40 mm detector coverage (centred over the tumour) with a 2.5 mm helical thickness (16 images per rotation). Cine duration varied for individual patients (product of average breathing time and scanner rotation time, 0.4 s). PET images were reconstructed using a standard ordered subset expectation maximization (OSEM) algorithm with CT for attenuation correction. Both non-attenuation corrected and attenuation corrected datasets were reconstructed.
The 4D CT and 4D PET data was divided into 10 phase bins. Post-processing was used to generate an averaged 3D PET from the 4D PET scan. No co-registration was necessary as the PET and CT components were inherently registered.
Contouring
Contouring was performed using specialized software (RTx, Mirada Medical, Oxford UK). PET and CT images were displayed using preset window levels and/or colour scale per a standardized institutional protocol (Fig. 1). Patients were contoured per the National Cancer Research Institute (NCRI) UK NeoSCOPE trial protocol [18]. The tumour length and outlines were derived from the diagnostic imaging which included an Endoscopic Ultrasound, contrast-enhanced CT and 3D PET-CT. The longest tumour dimension was used for outlining. The maximal length of the tumour with specific reference to an anatomical structure e.g. carina or superior aortic arch was defined on all imaging techniques. This enabled the oncologist and radiologist to identify the superior and inferior extent of the diseased oesophagus in relation to these structures; thereby allowing them to outline this segment of the entire circumference of oesophagus to be outlined. All the 4D CT and PET scans included at least one of the reference structures (e.g. superior aortic arch or carina) and therefore the target volumes were produced consistently on each of these datasets with reference to these structures [2]. Local nodal involvement was included in the target volume but more distant nodes were not. An experienced radiation oncologist contoured all target volumes with access to clinical details and standard-of-care imaging; PET-derived contours were generated by the same radiation oncologist contouring with an experienced dual-certified Nuclear Medicine Physician/Radiologist. GTVs were delineated on i) 10-phase planning 4D CT, ii) 10-phase 4D CT inherently co-registered to 3D PET-CT acquired at the same scan session and iii) 4D PET-CT. CTV was delineated and trimmed to anatomical boundaries (vertebrae, pericardium, pleura). 4D datasets from each series were used to generate an internal target volume (ITV) encompassing effects of physiological motion on the CTV. Expansion to PTV was the ITV of each series with a 5-mm margin in all directions. (Fig. 2). A minimum interval of 2 weeks was specified between delineation using each different methodology for each patient, to minimize any potential for intra-observer recall.
Positional analysis
Five positional metrics were used to compare target volumes, calculated using ImSimQA software (v3.1.5, Oncology Systems Limited, Shrewsbury, UK): Dice index (DICE); sensitivity index (Se.Idx), inclusiveness index (Inc.Idx), centre of gravity distance (CGD) and mean distance to conformity (MDC). The Dice index produces output values ranging from 0 and 1 where 0 represents two contours with no overlap and 1 represents two contours that are perfectly overlapping [19]. The Se.Idx calculates the overlapping volume between a contour and a reference contour as a percentage of the volume of the reference volume. The Inc.Idx is the probability that a voxel of a contour is really a voxel of a reference contour. CGD is the distance between the geometric centres of two contours [20]. MDC is the mean of the distances between contours averaged over all positions not within the overlapping contour [20].
Statistical analysis
Descriptive statistics (median, range) and Mann-Whitney test were used to assess for statistically significant differences between tumour volumes and lengths. Non-parametric analysis of variance (ANOVA) was used to assess the statistical significance of positional metrics between different imaging techniques (Friedman test). Stastistical analysis was performed using IBM SPSS Statistics (Version 22, IBM Corp, Amonk, NY, USA). A p-value <0.05 was taken as evidence of statistical significance.