Endobiliary Stent Position Changes during External-beam Radiotherapy.

PURPOSE: Endobiliary stents can be used as surrogates for pancreatic localization when using cone-beam computed tomography (CBCT) during external-beam radiotherapy (EBRT). This work reports on interfraction stent position changes during EBRT for locally advanced pancreatic cancer (LAPC).
MATERIALS AND METHODS: Six patients with endobiliary stents who underwent EBRT for LAPC were assessed. Measurements from the most superior aspect of the stent (sup stent) and the most inferior aspect of the stent (inf stent) to the most inferior, posterior aspect of the L1 vertebra central spinous process were determined from daily treatment CBCTs and compared with those determined from the planning computed tomography (CT) scan. Changes in stent-L1 measurements were interpreted as changes in relative stent position.
RESULTS: Three patients showed mean interfraction stent position changes of ≥1 cm when treatment measurements were compared with planning measurements. The sup stent for patient A moved to the right (2.66 ± 2.77 cm) and inferiorly (3.0 ± 3.12 cm), and the inf stent moved to the right (1.92 ± 2.02 cm) inferiorly (3.23 ± 3.34 cm) and posteriorly (1.41 ± 1.43 cm). The inf stent for patient B moved superiorly (2.23 ± 0.49 cm) and posteriorly (1.72 ± 0.59 cm). The sup and inf stent for patient F moved inferiorly (0.98 ± 0.35 cm and 1.21 ± 0.38 cm, respectively). The remaining three patients C, D, and E showed interfraction position changes of <1 cm.
CONCLUSION: Endobiliary stent migration and deformation were observed in a small subset of patients. Further investigation is required before confirming their use as surrogates for LAPC target localization during image-guided EBRT.

Introduction

The prognosis for patients diagnosed with pancreatic cancer is poor. The 5-year survival rate in England for men and women diagnosed with pancreatic cancer is 3.9% and 4.4%, respectively [1]. For those with pancreatic tumours deemed surgically unresectable with no distant metastases, their disease is classified as locally advanced pancreatic cancer (LAPC) [2]. Chemotherapy alone or chemoradiotherapy may be used to treat patients with LAPC [3]. It is possible to treat LAPC with either conformal radiotherapy or intensity-modulated radiotherapy [4] though the presence of several critical organs near the pancreas (eg, spinal cord, kidneys, liver, and bowel) makes the planning and delivery of radiotherapy complex, especially if nodal volume inclusion creates a larger planning target volume (PTV) [5].

Image-guided radiotherapy (IGRT) can be used to ensure treatment accuracy during radiotherapy treatments and limit toxicity to organs at risk. While IGRT using megavolt (MV) or kilovolt (kV) planar images can enhance accuracy, these images provide mostly bony anatomy information with minimal soft tissue definition [6]. Using bony anatomy to align organs such as the pancreas has proven to be a poor predictor of pancreatic location [7]. Cone-beam computed tomography (CBCT) can provide additional soft tissue definition, but delineating the pancreas and pancreatic tumours on CBCT remains difficult. The lack of soft tissue definition is likely caused by abdominal organ mobility [8]. Although Feng et al [9] found that the abdominal wall and diaphragm were unsuitable surrogates for pancreatic tumour position, this has not excluded the use of other internal structures as surrogates. Studies have sought to use fiducial markers [10], surgical clips [11], electromagnetic transponders [12], and endobiliary stents [13] as pancreatic tumour surrogates to assist with image matching.

Endobiliary stents are commonly implanted in pancreatic cancer patients as a palliative measure to relieve symptoms associated with biliary strictures such as jaundice [14], thereby aiding chemoradiotherapy tolerability [2]. As the common biliary duct runs through the pancreatic head, using an endobiliary stent as a surrogate for pancreatic localization can serve as a potential alternative to the implantation of fiducial markers, which are placed percutaneously under radiographic guidance, intraoperatively, or using endoscopic ultrasound [15]. Fiducial marker insertion for IGRT use would be an additional procedure for patients with endobiliary stents already inserted for palliative purposes. It is also unknown how fiducial marker visibility may be affected if a stent is in place as well. Endobiliary stents can be visualized on CBCT, and it has been suggested that they can be used as a surrogate when determining pancreatic tumour positions [13]. Two studies have reported on CBCT-determined interfraction motion of endobiliary stents during external-beam radiotherapy (EBRT) and found minimal stent position changes (<1 cm) with no indication of stent migration [11], [13]. In contrast, Johanson et al [16] reported stent migration rates of 4.9% proximally and 5.9% distally in the general patient population. The work presented here seeks to report on the stability of endobiliary stents in patients who received chemoradiotherapy at our centre.

Patients and Methods

Patients

The patients included in this retrospective study had previously provided informed consent for treatment on a research ethics board–approved phase II trial conducted in accordance with the Helsinki Declaration of 1964 and its later amendments, examining the effects of a novel radiosensitizer nelfinavir in combination with chemoradiotherapy in patients with LAPC between October 2010 and September 2012. All trial patients with an endobiliary stent implanted at the time of their planning computed tomography (CT) scan who underwent daily CBCTs during their radiotherapy treatments were included in this analysis. Six patients satisfied the inclusion criteria.

Radiotherapy Planning

Patients received 59.4 Gy in 33 fractions prescribed with concurrent chemoradiation and nelfinavir. Treatment was planned in the supine position with patients' hands above their head, arms and head supported by a vacuum bag, and knees supported by a Kneefix (Civco Medical Systems, Kalona, IA). Planning CTs were captured during exhale breath-hold with a 16-slice helical CT scanner (GE Optima kV120, Smart mA, 0.25-cm slices; GE Healthcare, Little Chalfont, UK). At planning, patients also underwent fluoroscopy using a conventional simulator (Varian Medical Systems, Palo Alto, CA) or a 4-dimensional CT (4DCT) using the CT scanner with the stent visualized for breathing motion assessment. Breathing motion data collected during planning was used to establish a stent internal target volume contour for use in image matching. Gross tumour volumes (GTVs) and clinical tumour volumes (CTVs) were contoured on the planning CT by a radiation oncologist using commercially available radiotherapy contouring software (Varian Eclipse 10.0; Varian Medical Systems, Palo Alto, CA). Phase 1 consisted of 50.4 Gy prescribed to PTV1 (= PTV_Oxford using the Oxford contouring method as previously described by Fokas et al [5]), and phase 2 consisted of 9 Gy prescribed to PTV2 (= GTV + 1 cm superiorly, 2 cm inferiorly, and 1.5 cm circumferentially). For each phase, 3-dimensional conformal radiotherapy plans were delivered using a combination of 6-MV and 15-MV beam energies. Beam arrangements were individualized depending on tumour location in relation to normal tissue with up to 7 coplanar and noncoplanar beam angles.

Daily Image Guidance

Daily treatment CBCTs (21EX Clinac 110 kV, 20 mA, 20 ms, 360°/min, Varian Medical Systems) were captured during free breathing to assess patient position and tumour displacement. The treating radiation therapists compared the daily treatment CBCTs with the planning CT online using commercially available radiotherapy imaging software (Varian On-Board Imager, Varian Medical Systems) to determine displacement prior to treatment. As tumour definition was difficult to assess on the treatment CBCTs, the endobiliary stent, bony anatomy, and adjacent soft tissue structures (ie, organs and vasculature) were used as surrogates for the tumour. If displacement was greater than 0.3 cm in the right-left, superior-inferior, or anterior-posterior direction or if rotation was greater than 3°, then corrective action was taken and a repeat CBCT was acquired. If displacement was less than 0.3 cm and if rotation was less than 3°, then treatment proceeded.

Definition of Stent Characteristics

For this retrospective study, three points were identified on all available planning CTs and treatment CBCT scans for each patient: the most superior aspect of the stent (sup stent); the most inferior aspect of the stent (inf stent); and the most inferior, posterior aspect of the L1 vertebra central spinous process (L1). Using contouring software, a digital marker was placed at these points for each planning CT and treatment CBCT (Figure 1). A specific set of x, y, and z coordinates were associated with each marker; x corresponded to the right(+)/left(−) direction, y corresponded to the superior(+)/inferior(−) direction, and z corresponded to the anterior(+)/posterior(−) direction. All observations were made by a single radiation therapist with several years of experience in CBCT assessment. To determine intrauser variability, the stent and L1 marker placements were repeated on three separate occasions for the first five fractions.

Figure 1

Placement of the sup stent measurement point on planning CT and day 10 CBCT to show stent position change. The far left (A1) and middle left (A2) images show axial views of the planning CT and day 10 CBCT, respectively. The top right (B1) and bottom right (B2) images show coronal views of the planning CT and day 10 CBCT, respectively. Digital markers were placed at the most superior end of the stent using contouring software. The green marker shows the sup stent position from the planning CT, and the red marker shows the sup stent position from the day 10 CBCT. Digital markers for the sup stent, inf stent, and L1 vertebra were placed on all available images, and the x, y, and z coordinates for these were used to determine the distances from the stent ends to the L1 vertebra.

Stent type (metal or plastic) was determined visually using the planning CT. Stent length was approximated by subtracting the inf stent y coordinate from the sup stent y coordinate from the planning CT. Respiratory stent motion was determined using the planning 4DCT or fluoroscopy. Virtual markers were placed at the sup stent on the images associated with maximum inhale and exhale, and the distance between these markers were measured in the superior-inferior direction to estimate the planning stent breathing motion.

Daily treatment stent-to-L1 (stent-L1) measurements were determined using the daily CBCTs taken immediately before treatment delivery. The distance between the sup stent and the L1 vertebra was calculated by subtracting the L1 coordinates from the sup stent coordinates (sup-L1). The distance between the inf stent and the L1 vertebra was calculated by subtracting the L1 coordinates from the inf stent coordinates (inf-L1). Changes in stent-L1 measurements were interpreted as stent position changes. Limited soft tissue definition on CBCT necessitated the comparison of surrogate motion with a bony match as in previous studies [7], [13]. Once the sup-L1 and inf-L1 measurements were obtained for each CBCT and planning CT, the planning stent-L1 measurements were then subtracted from the treatment stent-L1 measurements to determine the interfraction stent position relative to the baseline planning CT for both the sup stent and the inf stent. The mean, median, maximum, minimum, and interquartile range of interfraction stent motion was calculated using commercially available statistics software (IBM SPSS Statistics 20.0; IBM Corp, Armonk, NY).

Point-based analysis was performed to facilitate comparison with previous studies that examined interfraction stent position change using motion analysis of a single point (single end of stent [11] or centre of mass [13]). Because two points at opposing ends of the stent were used to determine position change for this study, any rotational error was captured as oppositional movement of the two stent ends. Also, since endobiliary stents are flexible structures making assessment of rotation alone difficult, using two points of measurement accounted for both potential rotation and potential deformation.

Results

Patient details can be found in Table 1. Six patients accrued from October 2010 to September 2012 were included in this study (A, B, C, D, E, and F) with a mean age of 63 years (range, 60–65 years) and a median GTV of 19.33 cm³ (range, 6.45 cm³–41.10 cm³). All LAPC masses were located in the pancreatic head.

Table 1

Patient Details Established from Planning

Patient	Sex	Age (y)	Tumour Location	Gross Tumour Volume (cm³)	Stent Type	Stent Length (cm)	Sup Stent to L1 Distance − Y Component (cm)	Planning Stent Sup-to-Inf Breathing Motion (cm)
A	Male	65	Head	27.78	Plastic	7.76	5.59	1.15
B	Female	64	Head	41.10	Metal	10.10	6.09	1.87
C	Female	60	Head	6.45	Plastic	6.81	6.41	1.41
D	Female	63	Head	28.62	Plastic	9.21	2.71	Not available
E.1	Female	66	Head	19.33	Plastic	8.95	6.96	0.92
E.2	Female	66	Head	Not contoured	Plastic	8.91	8.52	Not available
F	Female	62	Head	14.00	Metal	9.69	9.23	0.53

Inf, inferior; Sup, superior.

CBCTs were acquired for all patients for all 33 radiotherapy fractions with 198 CBCTs acquired immediately prior to treatment delivery. Three CBCTs were excluded because of software issues and a stent reinsertion, leaving 195 CBCTs suitable for stent position analysis. For patient A, CBCTs acquired on fractions 9 and 19 could not be analysed using the contouring software, leaving 31 fractions available for analysis. Six patients were included in the stent position analysis but assessed as seven (A, B, C, D, E.1, E.2, and F) because one patient required a stent reinsertion. Patient E required a stent reinsertion on fraction 15 as stent patency failure was suspected. Therefore, fraction 15 was omitted from analysis and fractions 1–14 (patient E.1) were analysed separately from 16–33 (patient E.2). The initial planning CT provided the baseline stent information for patient E.1, and a repeat CT scan taken at fraction 15 after stent reinsertion provided the baseline stent information for patient E.2 (Table 1).

Changes in Stent Position

Of the six patients assessed, three patients (A, B, and F) exhibited pronounced and sustained stent position changes as determined by examining the individual mean position change for each patient in each direction (Table 2) and by visualizing position change graphically over time (Figure 2). Patient A showed sustained stent position changes with the sup stent moving to the right (2.66 ± 2.77 cm) and inferiorly (−3.0 ± 3.12 cm) and the inf stent moving to the right (1.92 ± 2.02 cm) inferiorly (−3.23±3.34 cm) and posteriorly (−1.41 ± 1.43 cm). Patient B showed sustained stent position changes with the inf stent moving superiorly (2.23 ± 0.49 cm) and posteriorly (−1.72 ± 0.59 cm). Patient F showed sustained stent position changes with both the sup and inf stent moving inferiorly (−0.98 ± 0.35 cm and −1.21 ± 0.38 cm, respectively). The sustained position change was apparent from the start of treatment for patients B and F but did not become apparent in patient A until after fraction 5 (Figure 2). The degree of position change did not appear to increase with time. The free-breathing superior-inferior stent motion determined from planning for patients A, B, and F were 1.15 cm, 1.87 cm, and 0.53 cm, respectively (Table 1). Although the sup and inf stent for patient A and F moved similarly during treatment, only the inf stent for patient B showed sustained position changes (Figure 2, Figure 3). The population mean position change (μx,y,z) (ie, the mean of all patients' individual means of position change) for the sup stent was μx= 0.44 ± 1.09 cm, μy = −0.45 ± 1.26 cm, and μz = −0.09 ± 0.26 cm. The population mean position change for the inf stent for all patients was μx= 0.33 ± 0.84 cm, μy= −0.43 ± 1.65 cm, and μz = −0.31 ± 0.95 cm.

Table 2

Interfraction Stent Position Change

Patient		x (cm)		y (cm)		z (cm)
		Sup-L1	Inf-L1	Sup-L1	Inf-L1	Sup-L1	Inf-L1
A	Max	4.00	3.42	0.50	0.08	0.95	0.09
	Min	−0.49	−0.95	−4.16	−4.34	−0.92	−2.06
	Mean	2.66	1.92	−3.00	−3.23	0.43	−1.41
	SD	2.77	2.02	−3.12	−3.34	0.48	−1.43
B	Max	1.28	0.85	0.84	3.1	0.71	−0.73
	Min	−0.95	−2.12	−0.92	1.37	−0.96	−3.34
	Mean	−0.06	−0.61	−0.08	2.23	−0.37	−1.72
	SD	0.48	0.70	0.39	0.49	0.38	0.59
C	Max	0.26	0.83	1.61	0.75	0.66	0.94
	Min	−1.36	−0.69	0.01	−0.90	−0.53	−0.37
	Mean	−0.64	0.29	0.71	−0.10	−0.13	0.16
	SD	0.47	0.33	0.37	0.40	0.24	0.27
D	Max	0.94	0.72	1.25	1.14	0.27	0.98
	Min	−0.42	−0.37	−0.21	−0.27	−0.77	0.01
	Mean	0.23	0.20	0.54	0.42	−0.22	0.52
	SD	0.29	0.28	0.28	0.31	0.26	0.22
E.1	Max	1.76	0.68	0.60	0.34	0.17	−0.14
	Min	0.15	−1.50	−0.60	−1.74	−0.55	−1.15
	Mean	0.86	−0.30	0.04	−0.73	−0.22	−0.63
	SD	0.49	0.66	0.32	0.50	0.20	0.34
E.2	Max	0.26	1.40	0.12	−0.02	0.77	0.75
	Min	−0.56	0.08	−0.91	−0.84	−0.66	0.20
	Mean	−0.28	0.85	−0.41	−0.40	0.02	0.43
	SD	0.21	0.36	0.23	0.20	0.32	0.15
F	Max	0.83	0.91	−0.20	−0.30	0.52	1.18
	Min	−0.23	−0.95	−1.68	−1.92	−0.51	−0.20
	Mean	0.33	−0.04	−0.98	−1.21	−0.11	0.48
	SD	0.27	0.40	0.35	0.38	0.23	0.35
All patients
Mean of means		0.44	0.33	−0.45	−0.43	−0.09	−0.31
SD of means		1.09	0.84	1.26	1.66	0.26	0.95

Inf, inferior; Max, maximum; Min, minimum; SD, standard deviation; Sup, superior.

Figure 2

Stent position changes as shown by the difference when the planned stent-L1 distance is subtracted from the daily treatment stent-L1 distance.

Figure 3

Interfraction stent position relative to the baseline planning CT stent position along the y-axis. Box plots with outliers represented by ⋆’s and ○’s, non-outlier maximums and minimums represented by whiskers, interquartile ranges represented by gray rectangles, and median values represented as horizontal lines intersecting the interquartile ranges.

The standard deviation of position change (σ) exhibited by the inf stent was greater than that exhibited by the sup stent (Table 2). For the y component of the stent-L1 distance, six patients (B, C, D, E.1, E.2, and F) had σinf stent> σsup stent. For the z component of the stent-L1 distance, five patients (A, B, C, E.1, and F) had σinf stent> σsup stent. Repeat stent measurements indicated minimal intrauser variability.

Discussion

This study shows that the position of endobiliary stents can vary greatly over the course of EBRT for LAPC in six patients. Utilizing endobiliary stents as imaging surrogates for pancreatic tumours should be done cautiously as these stents may lack positional stability during EBRT. In this study, three patients exhibited sustained stent position changes from their planning CT. We saw a greater rate of stent migration than previously published by Johanson et al [16], which may be evidence of distal stent migration only. The migration rate differential may be related to study population selection as Johanson's sample consisted of a generalized patient population that received endobiliary stents [16], whereas this study had pancreatic cancer patients that received stents prior to chemoradiotherapy.

It is unknown how the presence of malignancy, chemotherapy, or EBRT affected stent migration incidence in this study, but the development of radiotherapy-induced pancreatic fibrosis may play a role. While malignancy and chemotherapy may have influenced stent migration, Lofts et al [17] found that cancer patients on a chemo-regime of epirubicin, cisplatin, and 5-fluorouracil did not have a significant increase in stent blockage or shortening of stent patency duration. Significant interfraction deformation of the pancreatic head, duodenum, and stomach can occur during pancreatic head irradiation [18], and these interfraction anatomic changes may affect stent migration.

Stent migration could be attributed to pancreatic irradiation; however, other studies have found that endobiliary stents were stable during radiotherapy [13], [19]. Zhu et al [19] found no incidence of stent migration when they treated biliary malignancy with intraluminal radiotherapy, but this may be due to their use of a stent-in-stent configuration. Engineer et al [13] concluded that endobiliary stents were stable tumour surrogates during EBRT after they found population mean position changes of less than 0.4 cm in all directions during EBRT. We found similar population mean position changes when patients were analysed together (<0.5 cm), but when patients were examined individually, three patients exhibited individual mean interfraction motions of greater than 1 cm in at least one direction, indicating stent migration occurrence. This contradicts the findings of Whitfield et al [11] who reported that their stent patients had individual mean interfraction stent motions of less than 1 cm for all directions during EBRT. Previous studies that examined CBCT-determined interfraction stent motion during EBRT had small samples (three [11] and five patients [13]). Perhaps with larger study sizes, more cases of stent migration would have also presented themselves in these studies.

The more inferior stent position during treatment CBCTs may be due to the planning CTs being captured during exhale breath-hold and the daily CBCTs being captured during free-breathing; the stents appeared inferior on CBCTs because inhalation caused inferior pressure on the diaphragm and abdominal contents. However, not all patients showed more inferiorly positioned stents. Mean y values for patients C and D hovered closely around the planned value. The influence of variable inhale and exhale lung volumes on stent position as seen on free-breathing CBCTs is unknown. The majority of the free-breathing cycle is spent in exhale breath position [20], which suggests that stent appearance on free-breathing CBCT roughly correlates with its appearance on exhale breath-hold CT. The inferior stent position change exhibited by patients A and F was larger than the sup-to-inf breathing motion established during planning. Inhalation may not be the only contributing factor to an inferior position change; distal stent migration may have occurred. This cannot be stated conclusively because 4DCT-determined pancreatic tumour motion may not have good correlation with the range of motion demonstrated during radiotherapy [21], [22]. Other nonrespiratory physiological processes may impact on stent position. To distinguish between interfraction motion and intrafraction motion, more detailed treatment imaging such as 4-dimensional CBCT could be used to determine stent positions on respiratory-correlated CBCT projections [23]. Exhale breath-hold CBCTs could also be used to minimize the respiratory component of intrafraction motion and facilitate correspondence with the exhale breath-hold planning CT. Ideally, stent position would be compared to the pancreas to monitor migration occurrence, but limited soft tissue definition on CBCT necessitates the comparison of surrogate motion with a bony match as in previous studies [7], [13]. More accurate soft tissue imaging and further study would be required to determine whether change in stent position is representative of position change of the pancreatic head.

Most patients had greater standard deviation of position change exhibited by the inf stent than the sup stent. Greater freedom of movement was associated with the inf stent in the superior-inferior and anterior-posterior planes than the sup stent. Sup stent motion was likely restricted by the concentration, size, and rigidity of internal organs in the upper abdomen. Generally, both ends moved in the same direction.

In patient B, the magnitude and direction of position change did not correspond for the sup stent and inf stent. Stent deformation may have occurred. The cause of this is unknown but may be a result of bowel filling changes from planning to treatment or weight loss during treatment. The incidence of stent deformation reinforces the method of selecting two position points on the stent. Previous studies selected a single stent position point such as a single end of stent or centre of mass [11], [13] to assess interfraction stent motion and did not report on stent deformation. As two points at opposing ends of the stent were used to determine position change for this study, any rotational error was captured as oppositional movement of the two stent ends. Endobiliary stents are flexible structures, making the assessment of rotation alone difficult. By using two points of measurement, both potential rotation and potential deformation could be accounted for.

Because patients were referred from a number of centres, the patients had a variety of both metal and plastic stents in place. Although the variation in stent type did not affect visualization of the stent on CBCT nor did it affect the results, a greater number of patients with both metal and plastic stents are required to determine this statistically.

A limitation of this study was the small sample size, which did not allow for more robust statistical analysis. There was considerable diversity between the patients' results as shown by the large standard deviation of all patients' means. Without more statistical tests, it is difficult to say whether this diversity would also be found in a larger group of patients. Additional study patients would allow for the identification of possible trends with regards to stent migration as well as enable better extrapolation to the wider irradiated LAPC population. Other limitations of this study were that the method and timing of stent insertion were not controlled nor was the stent type. With patients being referred for radiotherapy after stent insertion, controlling the previously mentioned variables will require forethought and coordination. Careful, consistent stent selection and insertion could reduce or eliminate stent migration. Shortening the time frame from stent insertion to the start and completion of radiotherapy may also reduce stent migration during treatment. For those receiving short-course radiotherapy for pancreatic cancer [24], [25], [26], the position of the endobiliary stent may be stable during IGRT because of the shorter time span of treatment. For those patients who may undergo long-course radiotherapy with a pre-existing stent, it is worth investigating whether these patients may require an additional procedure to exchange for a more stable endobiliary stent or to insert additional fiducials to facilitate image guidance during EBRT.

Conclusion

Through examination of daily treatment images, large sustained endobiliary stent position changes were observed in a small subset of patients during EBRT for LAPC, indicating the occurrence of stent migration and deformation. Further study is needed to determine whether change in stent position is representative of position change of the pancreatic head and tumour, and whether stents are appropriate for use as tumour surrogates.