Oesophageal IGRT considerations for SBRT of LA-NSCLC: barium-enhanced CBCT and interfraction motion

Up to 70% of patients with potentially curable locally-advanced non-small cell lung cancer (NSCLC) do not receive radical treatment and as many as 36% receive no treatment at all [1‐4]. Many factors play into why this might be the case, but concerns regarding competing comorbidities, oesophageal toxicity and the logistic requirements of attending daily treatment for 6 weeks are foremost. With increasing access to stereotactic body radiotherapy (SBRT), there is an ever-increasing proportion of early stage NSCLC patients receiving radical treatment [5‐7]. SBRT may have a similar impact if applied in the locally advanced setting. The safety of a hypofractionated regime of 15 × 4 Gy-fractions has been demonstrated [8‐10] and is now being compared to a conventional 30-fraction regime in a phase III trial [11]. We are currently investigating the extent to which hypofractionation can be safely achieved (ACTRN12619001186145) [12]. To accomplish this requires the most accurate image guidance possible to account for the daily position of organs-at-risk.

One of the key dose-limiting organs-at-risk when irradiating locally-advanced NSCLC is the oesophagus. Conventional fractionations with concurrent chemotherapy can result in ≥ Gr. 3 oesophagitis rates of 5–18% [13, 14], whilst hypofractionated regimes pose a greater risk of severe and even fatal reaction [10, 15]. The day-to-day position of the oesophagus can be inconsistent and is not fully accounted for during simulation and treatment planning [16‐18]. Additionally, with standard cone-beam computed tomographic (CBCT) imaging, the visibility of the oesophagus is poor due its small size and similar radiographic density to the adjacent tissues. With a narrow therapeutic index, the oesophagus represents a significant obstacle in utilising SBRT for targets within the mediastinum [19, 20].

We have previously investigated the use of oral contrast with thoracic CBCT and compared 50 ml each of Gastrografin and Barium Sulfate [21]. Barium allowed better visualisation of the oesophagus on CBCT, however its density and volume led to artifacts which potentially impaired IGRT for other structures including the tumour itself. Qiu et al., using an even larger volume of barium, quantified interfractional oesophageal movement in relation to bony anatomy, and found this to be as high as 36 mm [18]. The aim of this work was to determine the optimal volume of barium required to maintain oesophageal visibility and minimise imaging artifacts. Additionally, we aimed to quantify the interfraction motion of the oesophagus relative to the tumour, an essential IGRT consideration in applying SBRT for locally-advanced NSCLC.

We have demonstrated that a small amount of barium is sufficient in improving the interobserver contouring reproducibility of the oesophagus on CBCT imaging with no improvement with increasing dose. Barium improved the contouring reproducibility metrics from baseline for each patient. The amount of improvement the contrast provided varied between patients and was less dose dependent than it was patient dependent. Improvement correlated with the baseline measurement, indicating the poorer the interobserver contouring reproducibility without contrast, the greater the improvement the contrast provided. From these results we infer that the improved precision in observer contouring corresponds to improved oesophageal visibility due to barium administration. The only dose level that showed a significant improvement over the others was 10 mL, although this was found in the Kappa and Dice metrics and not the HD. Previous reports have shown that the Dice and HD do not correlate and instead complement one another [29], highlighting the importance of using a combination of metrics in studies like this [24].

There is growing evidence to show that hypofractionated, SBRT-like treatments may play a significant role in patients with locally-advanced NSCLC who are ineligible for concurrent chemoradiotherapy [8‐11, 30]. These regimes can cause severe toxicity to central organs like the oesophagus. The oesophagus is mobile and its position is influenced by both normal physiological functions (peristalsis, swallowing, respiration, cardiac rhythm) and tumour-related changes (tumour regression/progression, atelectasis, plural effusion) [16, 31]. Its position cannot be inferred by bony landmarks, nor assumed to be consistent over a course of radiotherapy spanning weeks, hence monitoring its daily position is paramount. We found the median interfraction oesophagus HD was 4 mm compared to the planning CT with 14% measuring over 10 mm. The magnitude of displacement could not be predicted by anatomical level, primary tumour laterality, overlap of the PTV with the oesophagus, nor by involvement of specific nodal stations. No patients in this study had significant tumour changes that required re-simulation, however reports have shown this can occur in 20% of locally-advanced lung cancer patients with significant risk to mediastinal structures if not actioned upon [32]. Figure 5 demonstrates a patient from this study with extreme variation in oesophageal position from as early as week 2 through to week 6. It could be argued that the introduction of an oral contrast agent to the oesophageal lumen may induce oesophageal motion through peristalsis. Given the small volume of barium used, the rapid nature of oesophageal transit times (seconds) [33] and the timeframe from administration to CBCT (minutes) this is unlikely to be a significant consideration. We also found the oesophageal volume decreased over the treatment course in most patients. We speculate that this is due to treatment related oesophagitis, causing constriction and luminal narrowing [34]. This then results in less barium accumulating in the lumen and as the external boundary of the lumen can be difficult to see on CBCT, the contoured volumes have shrunk.

Qiu et al. [18] have also recently reported on the use of barium with pre-treatment CBCTs for this purpose. They utilised 30 mL of barium before and a further 30 mL during CBCT acquisition and found oesophageal motion was greater in the left–right direction than the anterior–posterior direction, more prominent in the middle oesophagus and in patients with right sided disease or with bulky mediastinal lymph nodes. However, these positional variations were reported with reference to the bony landmarks. It is well understood that thoracic tumours move independent of bones [35, 36] and so in the context of hypofractionated SBRT treatments online soft tissue match to tumour is preferred [37]. We thus sought in this study to understand the relationship of the oesophageal position in relation to the tumour position over the course of treatment to help guide dosimetric decisions during treatment planning. The addition of barium to the planning and treatment of these patients can also help guide and verify the position of high dose gradients, for example when utilising a contralateral oesophageal-sparing technique [38, 39].

The consequence of barium-induced image artifact on CBCT has not been investigated until now. Artifact perception was not dose dependent as we had hypothesized, but instead influenced by individual patient and tumour anatomy. Proximity of organs/targets to the oesophagus had a greater impact than barium dose. Artifact influenced visibility of the central vessels more than other organs, which is acceptable given they are not as dose-restricting and their position is consistent. It remains of concern that some observers perceived the artifact to impede their ability to visualise the GTV. The more the PTV overlapped with the oesophagus, the greater the chance the barium influenced one's ability to clearly see the GTV on CBCT. However, in nearly all of these patients the oesophageal visibility was greatly improved by the barium, so a cost–benefit question arises in some patients between oesophageal and GTV visibility. Furthermore, the perception of image artifact is subjective and varies significantly between observers. In practice this subjectivity may be reduced because multiple staff are generally involved with online pre-treatment CBCT evaluation.

We found that in order to capture 95% of the oesophageal positional variations during treatment planning a planning risk volume (PRV) of 15 mm would be required. Qiu et al. [18] reported smaller 95% margins between 2.8 and 10.3 mm depending on location and direction, whilst Cohen et al. reported 8–12 mm although theirs reported from CBCTs of oesophageal cancers [40]. The limits of SBRT for locally-advanced NSCLC are not yet known and we are investigating this in a phase 1 dose escalation trial [12]. To incorporate a 15 mm PRV around the oesophagus will be impractical for most node positive NSCLC undergoing SBRT and a large PRV may jeopardise tumour control. Given a proportion of the shifts appeared random in nature (Fig. 4) the best solution may be online adaptive treatment strategies where the precise oesophageal position of the day is determined and the patient simply not treated if there is a significant shift into the high dose region. Additionally, with accurate visualisation and determination that a shift is more systematic, replanning or adaptive changes can be initiated.

This study has a number of limitations. The patient numbers are small and the results highlight the great inhomogeneity that this disease cohort presents. As the sample size per barium dose group was small, eliminating the influence of patient or disease variables on the results per dose group was impractical. Similarly, the number of treatment fractions, weekly CBCTs and patient numbers differed between groups, which may have influenced the results. Our finding of improved interobserver contouring reproducibility with barium demonstrates improved precision of oesophageal localisation. It does not necessarily, however demonstrate improved accuracy of oesophageal position. Considering the image quality of a linac CBCT (and not an MRI-linac for example) the true position of the oesophagus cannot be accurately determined. As with all radiotherapy studies drawing conclusion from human-generated contours, the resultant ground truth is dependent on the number and expertise of the observers. We intentionally chose both radiation oncologists and SBRT-trained therapists to participate as this reflects the true allocation of responsibility in the clinic. The study is further strengthened by the use of 6 observers to contour and assess image-artifact on each dataset, reducing the influence of investigator bias. Finally, in this study CBCTs were acquired weekly as per department protocol for conventionally fractionated radical treatment. Our results could have benefited had daily CBCTs been used. This study could be repeated in the future in patients receiving a hypofractionated course when daily CBCTs would be standard.

Small doses of barium delivered conveniently prior to patient setup greatly improve the interobserver contouring reproducibility of the oesophagus on CBCT and do not significantly degrade the image quality. Barium has helped show that the position of the oesophagus in relation to the tumour is variable and unpredictable over a course of treatment, requiring careful dosimetric consideration, close monitoring and ideally adaptive treatment strategies.