FDG-PET parameters predict for recurrence in anal cancer – results from a prospective, multicentre clinical trial

Patients with histologically confirmed non-metastatic squamous cell carcinoma of the anal canal were enrolled into a prospective study at three Australian centres investigating the role of multi-parametric magnetic resonance imaging (MRI) as a biomarker in AC. The protocol has been previously published [15]. Patients were staged according to the AJCC Cancer Staging Manual, Seventh Edition. p16 testing was performed on all tumours. Informed consent was obtained from all participants. The study was approved by the Hunter New England Human Research Ethics Committee (reference: HREC/12/HNE/408) and prospectively registered (ACTRN12614001219673).

Patients were treated with Intensity Modulated Radiotherapy (IMRT) or Volumetric Modulated Arc Therapy (VMAT). Prescription dose, contouring, and radiotherapy planning were mandated as per the Australasian Gastrointestinal Trials Group (AGITG) guidelines [16]. Radiotherapy doses ranged from 50.4-54Gy, depending on tumour stage. Continuous infusional 5-Fluorouracil (5-FU) was delivered during the first and fourth weeks of radiation treatment at a dose of 800-1000 mg/m²/day over 5 days. Mitomycin-C at 10 mg/kg was delivered on day one only.

FDG-PET/CT was performed at the time of staging (pre-CRT) and 12 weeks post-CRT. Images were acquired on a GE 690, GE 710 or Siemens Biograph mCT PET/CT. Patients were fasted for a minimum of 6 h prior to scanning to ensure a blood glucose level of < 10 mmol/L. Scanning was performed 1 h following FDG injection and extended from vertex of scalp to mid-thigh.

PET images were analysed utilising Mirada Medical workstation [Mirada Medical Ltd., Oxford, UK; software version Mirada XD3 (64-bit)]. A computer-generated volume of interest (VOI) was snapped around the primary tumour only (not involved lymph nodes) using six SUV thresholds from which seven PET parameters were extracted (Table 1).

Between January 2013 and January 2017, 19 patients were enrolled on the prospective study and underwent FDG-PET/CT imaging pre- and post- CRT. Patient and tumour characteristics are displayed in Table 2. All tumours were p16 positive.

Medium follow-up was 15.8 months (range: 8.2–54.3). Three patients (15.8%) developed a local recurrence, one of whom also experienced regional and distant recurrence. Two patients developed distant metastatic disease without local recurrence for a total of five patients (26.3%) with any recurrence. At last follow-up, one patient had died of disease.

Post-CRT PET parameters extracted from a VOI bounded by an SUV of 3 demonstrated the most promise (Table 3). The median SUV was the only statistically significant parameter for local recurrence (p < 0.01) and showed excellent discrimination (ROC AUC 1.00). All patients who had a median SUV greater than 3.4 experienced a local relapse. Median SUV did not reach statistical significance for detection of any recurrence (p = 0.07) and showed only fair discrimination (ROC AUC 0.75). The mean SUV at this threshold did not quite reach significance for prediction of local recurrence (p = 0.06) but displayed excellent discrimination (ROC AUC 0.91). For prediction of any recurrence, the only parameter reaching statistical significance was the MTV bounded by 41% maximum SUV on the pre-CRT PET (p = 0.03) which also showed excellent discrimination (ROC AUC 0.89). None of the pre-CRT PET scan parameters correlated with local recurrence. On LASSO regression, the only parameter retained was the SUVmax on the post-CRT PET. Figure 1 shows example FDG-PET/CT imaging in a patient who did (a-b), and a patient who did not (c-d), subsequently experience a local recurrence.

SUVmax is the most commonly used PET parameter in clinical practice. It is simple and quick to measure but is subject to noise because its value is taken from a single voxel. Alternatively, SUV mean, median or standard deviation within a VOI bounded by a threshold SUV may be more robust [17]. SUV thresholds for bounding VOI may be absolute (e.g. 3, 5, or 10) or relative to the SUVmax of the tumour. Another PET parameter in use is the peak SUV which is the average SUV contained within a small VOI approximating 1 cm³ that encompasses the voxels surrounding the SUVmax.

Volumetric PET parameters include the MTV, a measure of metabolic tumour volume above a threshold SUV, and total lesion glycolysis (TLG), which is the product of the mean SUV and MTV. Many feel that MTV and TLG better reflect metabolic tumour burden [17] and studies that have correlated MTV and TLG with prognosis [18‐20] have included both cervical and head and neck cancer treated with CRT [21, 22].

FDG-PET/CT is recommended for AC staging and radiotherapy planning [2, 4]. Retrospective studies have correlated staging PET parameters with outcome in AC. Kidd et al. reported that a higher SUVmax was associated with an increased risk of lymph node metastases and worse disease-free survival [10]. In a smaller study, Deantonio et al. also showed that higher SUVmax predicted for more advanced stage but were unable to demonstrate any correlation with survival [11].

Three other studies were unable to correlate SUVmax with outcome [12‐14]. They did however correlate MTV with outcome, although at different cut-off values. This is likely due to varying techniques used to calculate the MTV. Bazan et al. used a threshold of 50% SUVmax, included the primary and any nodal disease, and found a cut-point of 26 ml correlated with PFS [12]. Gauthe and colleagues also used a threshold of 50% SUVmax but only included the primary tumour. They found a cut-off of 7 ml correlated with overall survival [13]. Shali et al. included all metabolically active disease (primary, lymph nodes, and distant metastases) and used a threshold to define the volume of 30% of the SUVmax [14]. Not surprisingly, the MTV cut-off was higher (45 ml) than the others.

While these studies suffered from a retrospective design and heterogenous treatments, we too found that a volumetric parameter on the pre-treatment PET better correlated with recurrence than an SUV parameter. Our volume was bounded by 41% of the SUVmax and surrounded the primary tumour only. We chose to focus on the primary tumour as this is the most common site of recurrence [23‐25] and is more practical to measure in routine clinical practice. We also pre-determined which parameters to examine prior to performing our analysis in order to reduce the impact of multiple statistical testing.

Retrospective studies have investigated the predictive value of post-treatment PET scans in AC [5‐9]. While most have reported that a complete response on PET correlated with improved survival, all suffered from heterogeneous treatment, variable timing of PET imaging, and visual assessment only. Following definitive treatment, the timing of PET imaging in particular can have a significant impact on the sensitivity and specificity.

This was highlighted in a thorough prospective study by Mistrangelo et al. who performed PET imaging and anal biopsies at 1- and 3-months post completion of CRT on 53 patients [26]. Again, treatment techniques varied and imaging assessment was qualitative, but compared to 1 month, PET imaging at 3 months was found to have both improved sensitivity (66.6 vs. 100%) and specificity (92.5 vs. 97.4%) for detection of persistent disease.

The timing of post treatment PET imaging in the published retrospective studies ranged from 0.9–5.4 months [5], 2-7 months [6], 1-8 months [7], 1-8 months [8], and 1-6 months [9]. This significantly undermines the ability to draw robust conclusions from their data and limits its application to clinical practice.

Cardenas et al. is the only published study to perform quantitative analysis of post-treatment PET imaging [27]. Treatment was non-standardised, the only parameter assessed was SUVmax and, unfortunately, they fail to report a range for the timing of PET imaging. In addition, their follow-up ranged from 3.6–94.1 months, which is less than the median time to PET imaging (3.8 months). As such, our analysis is both novel and lacks some of the weaknesses present in the current literature.

While our study has many strengths, its major limitation is the small sample size which reflects the scarcity of AC. We also acknowledge that extracting seven PET parameters at six SUV thresholds generated a large amount of data which increased the likelihood of finding significant outcomes. However, such an exploratory analysis is useful for identifying signals which warrant further investigation in larger patient cohorts, which was the aim of this preliminary study.

That FDG-PET/CT imaging was performed on three different PET/CT systems is both a strength and weakness. While it introduces more variables, our results are likely more robust, they reflect real world practice, and are hence more applicable in the community. Future studies could consider including all metabolically active disease (primary and nodes) which may (or may not) better predict for any recurrence.

Finally, the emergence of digital PET/CT with enhanced image quality, greater sensitivity, increased signal to noise ratio, faster time-of-flight resolution, and improved quantitative accuracy means that PET/CT will represent an increasingly attractive biomarker in the future [28].