Pancreatic ductal adenocarcinoma (PDAC) is characterized by a 5-year survival rate below 5 % [1]. The focus of research on the survival of patients with pancreatic cancer is dominated by distant metastases [1, 2], but the importance of local control is often forgotten [3, 4]. Uncontrolled local disease is the cause of death in about one third of patients with PDAC [5, 6]. The role of local control in survival was recently confirmed in a phase III randomized trial [7]. Furthermore, radiation dose escalation is possible [8, 9] and intensity modulated radiotherapy (IMRT) was shown to increase the 2-year local control rate from 38 [10, 11] to 59 % [1, 7, 11, 12, 13, 14, 15].

Lymphatic spread is an important prognostic factor of PDAC, and the presence of metastatic lymph nodes is a predictor of local failure [6]. Although some radiation oncologists include elective lymph nodes (eLNs) into the target volume [16, 17, 18, 19], others only irradiate the primary tumor (gross tumor volume; GTV) without the eLNs so as to increase the dose delivered to the GTV [10, 20]. This debate is very similar to the one for patients with non-small-cell lung cancer where a considerable dose is often delivered incidentally to the high-risk nodal volumes [21].

Several studies have described how to define the lymph node regions to be treated [16, 17, 19, 22], but there is no comparative literature on the extent to which specific eLNs are anatomically covered by the main PTVs to help choose the appropriate approach. Here, we aimed to quantify the percentage of eLNs included in the definitions of PTVs as delineated in the different contouring methods of four groups. For the Oxford (based on our previous work [16]) and Radiation Therapy Oncology Group (RTOG) [19] methods, the planning target volume (PTV) includes the GTV with margins plus additional eLNs, i.e., clinical target volume (CTV), while the SCALOP (Selective Chemoradiation in Advanced Localised Pancreatic Cancer trial; NCT01032057, clinicaltrials.gov) and the Michigan methods [10] only include the GTV without eLNs.

Patients and methods

We used the planning computer tomography sets (2.5-mm slides; performed with intravenous contrast in full exhale position) of 11 consecutive patients with locally advanced, inoperable PDAC of the pancreatic head who had undergone concurrent chemoradiation from January 2009 as part of the randomized phase II clinical trial ARC-II (EudraCT 2008-006302-42). Contouring was performed using the Eclipse system 8.9 (Varian, Palo Alto, CA, USA), in accordance with the ICRU 50/62 guidelines. The study was approved by the appropriate ethics committee and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

Oxford contouring method

The GTV, CTV, and PTV were outlined as previously described [16, 17]. The regional nodal areas, i.e., CTVs, included the areas based on vascular anatomy: the aorta and inferior vena cava (para; para-aortic, interaortocaval, and paracaval), the celiac trunk (CT), the superior mesenteric artery (SMA; superior mesenteric artery/vein distribution) and the portal vein (PV).

The CTV of the para-aortic nodes (CTV_para_ox) was created by outlining the aorta from the celiac trunk to the inferior mesenteric artery and expanding it by 10 mm in all directions (superior–inferior, anterior–posterior, lateral–lateral).

To create the CTV for the inferior vena cava (IVC; CTV_IVC_ox), the latter was outlined parallel to the aorta. A margin of 10 mm was added to the left and anteriorly following histopathological patterns of nodal disease to create the CTV_IVC_ox [23]. The CTV_para_ox and CTV_IVC_ox were merged to form the aortocaval lymph nodes (CTV_paraIVC_ox).

The Oxford CTV for the PV was defined similarly to the RTOG method [19], but on the right lateral side only to the level of the junction of the cystic duct with the common bile duct, which is the craniolateral edge of the lymphatic dissection of the hepatoduodenal ligament. As with the RTOG contouring method, a margin of 10 mm was added in all directions, except in the right lateral direction, to create the CTV (CTV_PV_ox) in order to protect the liver parenchyma. The CT and SMA and their corresponding CTVs (CTV_CT_ox and CTV_SMA_ox, respectively) were defined and contoured similarly to the RTOG protocol except that the posterior CTV margin was restricted to 5 mm within the anterior part of the aorta. The PTV_Oxford protocol incorporated all CTVs as previously described [11, 12]. We have included a detailed description of the step-by-step contouring of the PTV_Oxford protocol in the supplementary figures (Figs. 6, 7, 8, 9).

Contouring of the peripancreatic lymph nodes

Although the Oxford method does not routinely involve contouring of the peripancreatic lymph nodes, for the purpose of the present study we decided to create the peripancreatic CTVs, because of their importance in the lymphatic drainage of PDAC. We first contoured the CTV_Pancreas that included the pancreas and the primary GTV to the sagittal plane through the left lateral aortic border; this plane constitutes the resection margin of a typical Whipple procedure. The CTV_Pancreas was expanded by 10 mm in the four corresponding directions (anterior, posterior, superior, inferior) to create each of the four peripancreatic nodal regions of interest (ROIs). Finally, the CTVPancreas was subtracted from each of the four ROIs to create the peripancreatic nodal CTVs surrounding the pancreas: anterior (CTV_AntPanc), posterior (CTV_PostPanc), superior (CTV_SupPanc), and inferior (CTV_InfPanc).

RTOG contouring method

The PTV (PTV_RTOG) and the CTVs for the vascular areas were contoured as described recently [19]. CTVs included the nodes for the aorta (CTV_para_RTOG), the CT (CTV_CT_RTOG), the SMA (CTV_SMA_RTOG), and the PV (CTV_PV_RTOG).

Michigan contouring method

The CTV consisted of the GTV plus 5 mm without including eLNs. The PTV (PTV_Michigan) consisted of CTV plus 5 mm in all directions [10, 20].

SCALOP contouring method

The PTV (PTV_SCALOP) was created by expanding the GTV by 20 mm in the superior–inferior direction and by 15 mm in the anterior–posterior and laterolateral direction (NCT01032057; clinicaltrials.gov). Similarly to the Michigan method, no eLNs were included in the outlined PTV.

Calculation of anatomical coverage of CTVs by the four PTVs

To calculate the percentage of CTV coverage by each of the four PTVs (Oxford, RTOG, SCALOP, Michigan), the volume of PTV that overlapped with the CTV (PTV∩CTV) was defined as shown in Fig. 1. Following that, the percentage of CTV coverage by each of the four PTVs was calculated by the ratio of (PTV∩CTV)/CTV. All contoured volumes were calculated in cm3.

Fig. 1
figure 1

Simplified illustration of calculation of the CTV (blue oval) coverage by the PTV (black oval). The volume of PTV that overlapped with the CTV (PTV ∩ CTV; red area) was contoured (left panel). To calculate the percentage of CTV coverage by each of the four PTVs, the ratio of (PTV ∩ CTV)/CTV was measured (right panel)

Full size image

Statistical analysis

The significance of differences between the means was measured with the Wilcoxon signed-rank test or ANOVA using the Bonferroni test with GraphPad Prism 4.0. The individual values from 11 patients were expressed as a scatter plot ± standard deviation (SD). A value of p < 0.05 was considered significant.

Results

Variability of PTV size and anatomical coverage of peripancreatic lymph node areas

We contoured the pancreas, the GTV, the peripancreatic CTVs, and the PTVs as shown in Fig. 2 a. There was no significant difference between PTV_Oxford and PTV_RTOG, while both PTVs were significantly larger than those of PTV_SCALOP and PTV_Michigan. The volumes of PTV_SCALOP and PTV_Michigan were likewise comparable (Fig. 2 b).

Fig. 2
figure 2

Anatomical coverage of peripancreatic lymph nodes. a Axial images demonstrating the pancreas, the GTV, the peripancreatic node CTVs— Anterior (CTV_AntPanc), Posterior (CTV_PostPanc), Superior (CTV_SupPanc), and Inferior (CTV_InfPanc)—and the four PTVs, PTV_Oxford, PTV_RTOG, PTV_SCALOP, and PTV_Michigan. The four arrows point to the four peripancreatic CTVs. Contoured structures and arrows are color-coded as indicated. b Total size of the PTV. c, d, e, f Comparison of the percentage of CTV coverage by the PTVs for each of the four peripancreatic node regions. Ao aorta, ns not significant. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

The following was observed regarding coverage of the high-risk peripancreatic CTVs by the four PTVs (Fig. 2 c, d, e, f):

  1. 1.

    Comparison of the percentage of CTV coverage between (a) PTV_Oxford and PTV_RTOG (both approximately 80 %) and (b) between PTV_SCALOP and PTV_Michigan (approximately 60 % and 50 %, respectively) revealed no statistically significant difference for any of the four peripancreatic CTVs. The CTV of the posterior peripancreatic region that has the highest risk of positive nodes was included close to 100 % in both the Oxford and RTOG PTVs.

  2. 2.

    Both the Oxford and RTOG PTVs contained statistically more of all four peripancreatic CTVs compared to PTV_Michigan.

  3. 3.

    Comparison of the percentage of CTV coverage between PTV_Oxford or PTV_RTOG and PTV_SCALOP showed a significant difference only for CTV_PostPanc and CTV_InfPanc.

Variability of aortic CTV and anatomical coverage of para-aortic lymph node areas

The aorta and the para-aortic CTVs were contoured as shown in Fig. 3 a. Comparison of the CTVs showed a significant difference between CTV_para_ox and CTV_para_RTOG (Fig. 3 b). The following was noted regarding the coverage of para-aortic CTVs by the four PTVs (Fig. 3 c, d):

  1. 1.

    Comparison of the percentage of CTV coverage between PTV_Oxford and PTV_RTOG revealed no significant difference for either CTV_para_ox or CTV_para_RTOG.

  2. 2.

    When we compared the percentage of CTV coverage between PTV_SCALOP and PTV_Michigan, there was only a marginally significant difference for both CTV_para_ox and CTV_para_RTOG.

  3. 3.

    As expected, comparing the percentage of CTV coverage between PTV_Oxford or between PTV_RTOG with PTV_Michigan and PTV_SCALOP resulted in a significant difference, for both CTV_para_ox and for CTV_para_RTOG.

 

Fig. 3
figure 3

Anatomical coverage of the lymph nodes of the aorta. a Axial images demonstrating the aorta, the CTV for para-aortic lymphatics contoured based on the Oxford (CTV_para_ox) and RTOG (CTV_para_RTOG) methods, the inferior vena cava (IVC) and its nodes based on the Oxford (CTV_IVC_ox) protocol and the four PTVs: PTV_Oxford, PTV_RTOG, PTV_SCALOP, and PTV_Michigan. Contoured structures are color-coded as indicated. b CTV sizes for the para-aortic areas. c, d Comparison of the percentage of CTV coverage by the PTVs for the para-aortic areas. ns not significant. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

Variability of celiac trunk and superior mesenteric artery CTVs and their anatomical coverage

The CT, the SMA, and their corresponding CTVs were delineated as shown in Fig. 5 a (only the CTVs for RTOG are shown here). As expected, comparison of the CTVs showed no significant difference for the CT (CTV_CT_ox vs. CTV_CT_RTOG) and SMA (CTV_SMA_ox vs. CTV_SMA_RTOG) (Fig. 4 b, c). The following was observed regarding the coverage of the CT and SMA CTVs by the four respective PTVs (Fig. 4 d, e, f, g):

  1. 1.

    Comparison of the percentage of CTV coverage between PTV_Oxford and PTV_RTOG revealed no statistically significant difference, neither for CTV_CT_ox, CTV_CT_RTOG, CTV_SMA_ox nor for CTV_SMA_RTOG.

  2. 2.

    Comparison of the percentage of CTV coverage between PTV_SCALOP and PTV_Michigan showed marginally significant difference only for CTV_CT_ox and CTV_SMA_RTOG but not for the other CTVs, again with SCALOP tending to cover more of the CTVs owing to the expansion rules of the GTV.

  3. 3.

    Comparison of the percentage of CTV coverage between PTV_Oxford or PTV_RTOG and PTV_Michigan revealed a significant difference for the CTVs of both the Oxford and RTOG methods.

 

Fig. 4
figure 4

Anatomical coverage of the lymph nodes of the celiac trunk (CT) and superior mesenteric artery (SMA). a Axial images demonstrating the CTV for lymphatics of the CT and SMA, based on the RTOG method (CTV_CT_RTOG and CTV_SMA_RTOG, respectively), and the four PTVs: PTV_Oxford, PTV_RTOG, PTV_SCALOP, and PTV_Michigan. Contoured structures are color-coded as indicated. b, c CTV size for the CT and SMA lymph nodes contoured according to the Oxford and RTOG methods. Comparison of the percentage of CTV coverage by the PTVs for the CT (d, e) and SMA (f, g) areas contoured according to the Oxford and RTOG methods. Ao aorta, ns not significant. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image
Fig. 5
figure 5

Anatomical coverage of the lymph nodes of the portal vein (PV). a Axial images demonstrating the PV, the CTV for PV lymphatics contoured according to the Oxford (CTV_PV_ox) and RTOG (CTV_PV_RTOG) methods and the four PTVs: PTV_Oxford, PTV_RTOG, PTV_SCALOP, and PTV_Michigan. Contoured structures are color-coded as indicated. b CTV sizes for the PV lymph node areas. c, d Comparison of the percentage of CTV coverage by the PTVs for the PV areas. Ao aorta, ns not significant. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

Variability of portal vein CTVs and their anatomical coverage

We contoured the PV and the corresponding CTVs as shown in Fig. 5 a (only the CTVs for RTOG are shown here). Comparison of the CTV sizes showed no significant difference for the PV (CTV_PV_ox vs. CTV_PV_RTOG; Fig. 5 b). The following was observed concerning the coverage of the PV by the four PTVs (Fig. 5 c, d):

  1. 1.

    Comparison of the percentage of CTVs coverage between PTV_Oxford and PTV_RTOG revealed a significant difference for both CTV_PV_ox and CTV_PV_RTOG, which is attributable to a longer track of the vessels forming the CTV in the RTOG definition than in the Oxford definition.

  2. 2.

    Comparison of the percentage of PV CTV coverage between PTV_SCALOP and PTV_Michigan showed a significant difference for both CTV_PV_ox and CTV_PV_RTOG.

  3. 3.

    Comparison of the percentage of PV CTVs coverage between PTV_Oxford or PTV_RTOG and PTV_Michigan as well as PTV_SCALOP revealed a statistically significant difference.

Discussion

The inclusion of eLNs in the radiotherapy of PDAC remains debatable. Many nodal areas are included unintentionally into the PTVs because of their proximity to the pancreatic tumors. We intended to measure the degree of anatomical inclusion of eLN volumes according to four definitions of the target volumes in PDAC. As expected, the total volumes of the RTOG and the Oxford PTVs were significantly larger than their counterparts that did not include eLNs. This difference can have implications, since the tolerance of gemcitabine-based chemoradiotherapy is related to the size of the PTV [24] and hence radiation dose escalation will always have to exclude the eLNs [11]. On the other hand, the presence of metastatic lymph nodes was shown to be associated with local recurrence (HR 1.89) [6], although detailed studies are lacking.

We previously analyzed the patterns of regional lymphatic spread in PDAC [16] and showed that the posterior pancreaticoduodenal nodes were at the highest risk for metastasis (37 %), followed by the suprapancreatic head nodes (25 %), the infrapancreatic head nodes (24 %), the anterior pancreaticoduodenal nodes (23 %), nodes in the hepatoduodenal ligament (18 %), and the superior mesenteric nodes (10 %); all the other nodal areas had less than  10 % metastasis. This was recently confirmed by a pooled analysis by Sun et al. [22]. Para-aortic lymph node metastasis is probably underreported in both studies because these nodes are not routinely dissected. In the literature on radical lymph node dissection, a frequency of 15–18 % and a much higher rate of micrometastasis is reported [16].

A comparison of the relative coverage of specific nodal areas according to the delineation criteria yields intriguing data. The posterior peripancreatic nodal region has a specifically important role as it presents the highest metastatic rate, it is intricately connected to the para-aortic region, and it is the typical site where microscopic disease (R1) is encountered after pancreaticoduodenectomy [22]. Interestingly, both the SCALOP and the Michigan PTVs covered this region by about 60 and 50 %, respectively, whereas anatomical coverage by the Oxford and RTOG PTVs was approximately 95 %. Notably, this region can be irradiated without any additional toxicity. The other peripancreatic regions were also involved to a large degree in the tumor-only concepts because of their proximity to the primary tumor. This could impact on the local control of patients treated according to the Michigan and SCALOP protocols. However, especially the superior and the inferior pancreaticoduodenal nodes as well as the right lateral nodes are difficult to cover with a high radiation dose because they are close to the duodenum.

The largest discrepancies in volume were found for the para-aortic and the interaortocaval nodes. The RTOG definition extends more to the right than the Oxford contouring, which could increase the dose distribution to the right kidney. Additionally, metastases to the right and behind the vena cava inferior are rare and therefore these regions can be spared [23]. Treatment of the para-aortic and interaortocaval nodes is especially debatable because according to the TNM classification these nodes are classified as distant metastasis if positive. However, a very rich lymphatic network connects them directly to the posterior pancreatic region, which is reflected in the high rate of para-aortic micrometastases [25], and hence the RTOG definition appears to be justified.

Regarding the PV area, the PTV_RTOG extends more into the hepatic hilum despite there being no significant difference in the total volumes with the PTV_Oxford. We avoid extending this volume too much to the right so as to spare the liver. Notably, treatment of this area adhering to strict expansion rules would lead to the entire circumference of the duodenum next to the hepatoduodenal ligament being included into the PTV. We advocate trimming this region or treating it with IMRT and a locally lower dose.

Finally, the superior mesenteric region was included to a significant degree in the SCALOP PTVs; however, as the superior mesenteric vessels are a frequent reason for unresectability, this region is often included into the PTV. Of note, Sun et al. found a higher rate of nodal disease in this region (16 %) than was found in our previous analysis (10 %) [16, 22]. However, their report confirmed the relatively low rate of metastasis in the CT region, which was included in the Oxford definition, mainly because of the proximity to the tumor PTV and the CTV of the hepatoduodenal ligament.

Practical conclusions can be drawn and a more relaxed position between the pro- and contra-eLN irradiation groups could be achieved by differentiating between nodal areas of“high risk” and“low risk”: The posterior peripancreatic region should always be included, especially in patients with borderline-resectable disease. Owing to the intricate connections with the para-aortic region and the relative ease of including this region without major consequences on the organs at risk, it is practical to include it between the superior and inferior border of the PTV of the primary tumor. On the other hand, inclusion of the nodes of the hepatoduodenal ligament often comes at the price of treatment of the entire circumference of the duodenum and therefore this option should be examined more critically. Under no circumstances should inclusion of this area compromise application of a high dose to the area of the tumor, especially because of the close contact with the duodenum. Moreover, the para-aortic regions beyond the level of the tumor PTV are of a lower order of importance in the ranking of eLNs.

Based on our experience, we currently advocate radiotherapy according to the Oxford contouring guidelines, since it is a feasible method that enables administration of a high dose (59.4 Gy) to the primary tumor region and it is associated with good tolerance. Radiotherapy planning should include three-dimensional conformal radiotherapy, and IMRT is preferred. An image-guiding method such as four-dimensional computed tomography should be applied where possible. The maximum PTV should not exceed 800 cc, and careful planning often results in a PTV lower than 600 cc at our institute [12].

Our study had some limitations. No differences in patient positioning and respiratory motion control were taken into account. Furthermore, our analysis consisted in only 11 patients and can therefore not be generalized, even if we analyzed the total number of approximately 600 structures in this study.

Conclusion

In conclusion, there are major differences in the definition of PTVs between the four protocols presented here, reflecting the controversy between the groups that are for and against inclusion of eLNs. A significant amount of eLNs are also targeted in the concepts aiming to treat the tumor only. A more differentiated view of the relative importance of specific known areas could contribute toward achieving more common ground between the two groups in the future, and hopefully toward increasing local control in patients with pancreatic cancer.

Fig. 6
figure 6

Based on our treatment protocol (Oxford), patients with pancreatic cancer will receive a total dose of 50.4 Gy in 28 fractions (1.8 Gy/fraction) according to planning target volume Oxford (PTV_Oxford; shown below), followed by an additional dose of 9.0 Gy in 5 fractions according to planning target volume 2 (PTV2; boost dose). As a first step, we contour the gross tumor volume (GTV) and the PTV2. GTV includes the macroscopic pancreatic tumor visible on imaging. PTV2 = GTV + 20 mm in the craniocaudal direction and 15 mm in all other directions. If the planning computer tomography scan is taken in full exhale breath-hold, then PTV2 = GTV + 20 mm caudal, 10 mm cranial, and 15 mm in all other directions. The structures are color-coded as indicated

Full size image
Fig. 7
figure 7

As a second step, we contour the clinical target volumes (CTVs) for the aortocaval lymph nodes. We outline the aorta from the celiac trunk (include the celiac trunk anteriorly to its bifurcation) to the inferior mesenteric artery (for tumors of the lateral two thirds of the pancreatic tail, the inferior border is the lower border of the lower renal vein). Then we add a margin of + 15 mm in all directions except inferiorly (only 10 mm cranial if planning computer tomography scan taken in full exhale breath-hold). Where this volume extends into bone, manually edit this volume to include a maximum of 5–10 mm of bone. Label this as para-aortic lymph nodes (CTV_para_ox). Following this, we outline the inferior vena cava from the level of the inferior mesenteric artery (for tumors of the lateral two thirds of the pancreatic tail, the inferior border is the lower border of the lower renal vein) to the level of the celiac trunk. Then we add a margin + 15 mm in all directions except inferiorly and posteriorly (only 10 mm cranial if planning computer tomography scan taken in full exhale breath-hold). We subsequently edit this volume such that the right margin is 15 mm laterally to the right from the right edge of the anterior half of the IVC and 15 mm laterally to the right from the left edge of the posterior half of the IVC, aiming to exclude liver parenchyma. We label this as CTV_IVC_ox. We finally merge the two volumes to form the AortoCaval lymph nodes (CTV_paraIVC_ox). The structures are color-coded as indicated

Full size image
Fig. 8
figure 8

As a third step, we contour the clinical target volumes for the lymph nodes of the portal vein (PV), celiac trunk (CT), and superior mesenteric artery (SMA), as described in “Patients and methods.” The CTV_CT_ox and CTV_SMA_ox are contoured similarly to the RTOG guidelines. For that purpose, the CT and SMA are expanded by 10 mm in all directions. The posterior CTV margin is restricted to 5 mm within the anterior part of aorta. The Oxford CTV for PV (CTV_PV_ox) is contoured similarly to the RTOG method except that on the right lateral side CTV_PV_ox is delineated only to the level of the junction of the cystic duct with the common bile duct (described in [16]). We ensure enclosure of the common bile duct lymph nodes. The structures are color-coded as indicated. The orange arrows point to landmarks used to assist contouring

Full size image
Fig. 9
figure 9

As a fourth and final step, we create the PTV_Oxford by merging the following structures: PTV2 (boost region), elective nodal regions (depending on the site of the tumor in the pancreas, as described in [16]) and any enlarged lymph nodes. Note that the CTV_para_ox has been expanded by 5 mm before final creation of the PTV_Oxford. The structures are color-coded as indicated. The PTV_Oxford contour may then require minor manual editing to achieve a satisfactory elective nodal volume. We often “smooth out” the PTV_Oxford to avoid any “holes” within the contour. When necessary, we manually edit the PTV_Oxford to enable adequate sparing of at least one kidney. Depending on the localization of the primary tumor position, we edit the PTV_Oxford to enable some sparing of the right lateral wall of the duodenum (in conjunction with image guidance, such as 4D-computed tomography). Finally, we ensure that the PTV_Oxford encompasses (with a circumferential 1-cm margin) the superior mesenteric vein for 3 cm inferior to the confluence of the splenic vein with the superior mesenteric vein

Full size image