Dual-energy computed tomography-based iodine concentration as a predictor of histopathological response to preoperative chemoradiotherapy for pancreatic cancer

Abstract

To explore predictors of the histopathological response to preoperative chemoradiotherapy (CRT) in patients with pancreatic cancer (PC) using dual-energy computed tomography-reconstructed images. This retrospective study divided 40 patients who had undergone preoperative CRT (50–60 Gy in 25 fractions) followed by surgical resection into two groups: the response group (Grades II, III and IV, evaluated from surgical specimens) and the nonresponse group (Grades Ia and Ib). The computed tomography number [in Hounsfield units (HUs)] and iodine concentration (IC) were measured at the locations of the aorta, PC and pancreatic parenchyma (PP) in the contrast-enhanced 4D dual-energy computed tomography images. Logistic regression analysis was performed to identify predictors of histopathological response. Univariate analysis did not reveal a significant relation between any parameter and patient characteristics or dosimetric parameters of the treatment plan. The HU and IC values in PP and the differences in HU and IC between the PP and PC (ΔHU and ΔIC, respectively) were significant predictors for distinguishing the response (n = 24) and nonresponse (n = 16) groups (P < 0.05). The IC in PP and ΔIC had a higher area under curve values [0.797 (95% confidence interval, 0.659–0.935) and 0.789 (0.650–0.928), respectively] than HU in PP and ΔHU [0.734 (0.580–0.889) and 0.721 (0.562–0.881), respectively]. The IC value could potentially be used for predicting the histopathological response in patients who have undergone preoperative CRT.

INTRODUCTION

Pancreatic cancer (PC), which is often diagnosed at an advanced stage, is a fatal malignancy with an increasing incidence. It is estimated that it will become the second most common cause of cancer-related deaths by 2030 [1]. PC’s etiology is not fully understood, and effective treatments are limited, resulting in a low survival rate [2]. Surgical tumor resection is considered the only curative option for patients with localized disease and could potentially provide long-term disease-free survival [3]. For resectable or borderline resectable PC, surgical rejection followed by adjuvant chemotherapy is considered the mainstream treatment [4]. However, standard treatment for PC has not been established. Neoadjuvant (preoperative) radiotherapy demonstrated the significant improvement in overall survival for patients with resectable PC compared with surgery alone or surgery with adjuvant radiotherapy [5]. The combination of preoperative chemotherapy with radiotherapy (CRT) aims to reduce tumor size and increase the probability of achieving negative margins during surgical resection, and is expected to improve treatment outcomes [6]. The Dutch Randomized Phase III PREOPANC Trial showed that the preoperative CRT significantly improved disease-free survival and locoregional failure-free interval compared with the immediate surgery group [7].

Histopathological responses have been associated with patient prognosis. Chatterjee et al. [8] divided patients with PC into two response groups: patients exhibiting a pathological complete response or minimal (<5%) residual tumor, and the remainder of the patients. Overall and disease-free periods of survival were significantly longer in the favorable (first-mentioned) response group. Jeon et al. [9] also demonstrated that pathological response could be a predictor of overall and disease-free survival after pancreatectomy. Hirata et al. [10] demonstrated that the dose intensity of radiation and gemcitabine (GEM) was significantly related to the pathological response of preoperative CRT for PC. Prediction of the histopathological response prior to surgery may be useful in determining the patient’s further treatment strategy, such as dose-escalation radiotherapy.

Dual-energy computed tomography (DECT) has recently become available in clinical practice and can reconstruct various types of images such as virtual monochromatic images (VMIs), effective atomic numbers and iodine density maps (IDMs) [11]. An IDM allows quantitative measurement of the amount of iodine in contrast-enhanced agents present in tissues and blood vessels. Noda et al. [12] analyzed the iodine concentration (IC) derived from DECT in patients who had undergone first-line chemotherapy for PC and found that the difference in IC for PC between the pancreatic parenchymatous and equilibrium phases was significantly higher in the favorable response group than in the nonfavorable response group. They concluded that the IC diagnostic factor was a potential biomarker for the assessment of chemotherapy response in patients with PC. To the best of our knowledge, no studies investigating the association between IC and histopathological responses after preoperative CRT have been published.

The aim of this study was to explore the potential of dosimetric parameters, the CT number expressed in Hounsfield units (HU) and IC reconstructed using DECT as predictors of the histopathological response to preoperative CRT for patients with PC.

MATERIALS AND METHODS

Patients

This retrospective study was approved by our institutional review board (no. 22208). All patients provided written informed consent for the use of their data in clinical research before the administration of radiotherapy. From October 2017 to January 2019, 58 patients with resectable or borderline resectable PC who underwent preoperative CRT (with or without induction chemotherapy) and underwent contrast-enhanced 4D DECT (CE-4D-DECT) [13] were enrolled in the study.

Simulation and treatment planning

Induction chemotherapy with either GEM, GEM and nab-paclitaxel (NabPTX), or FOLFIRINOX was performed before simulation because the administration of induction chemotherapy may allow for the selection of patients who may truly benefit from subsequent locoregional treatment including CRT and surgery [14]. CE-4D-DECT was performed with each patient immobilized by evacuating the cushion to the supine position. A brief description of the CE-4D-DECT procedure follows. An intravenous CE agent was injected to achieve a dose of 450 mg of iodine per kg of body weight using a fixed 30-s injection time while performing the step-and-shoot scan using a fast kilovoltage-switching DECT system (Revolution HD, GE Medical Systems, Waukesha, WI, USA). The time delay of the DECT acquisition was individually targeted at the center of the tumor in the pancreatic parenchymal phase (40 s from the time of injection). For the scanning, a tube voltage of 80/140 kVp, tube current of 360 mA and rotation time of 0.5 s were used, and the acquired data were divided into 10 respiratory phase image sets using a workstation (Advantage Sim, GE Medical Systems). The VMI at 77 keV (expiration and inspiration phases), average intensity projection (generated by averaging pixel densities for all phase images) and the IDM at the expiration phase were reconstructed with a slice thickness of 2.5 mm.

Using a treatment planning system (Eclipse, Varian Medical Systems, Palo Alto, CA, USA), the high-dose clinical target volume was defined as the area comprising the roots of the celiac and superior mesenteric arteries. The elective dose CTV was composed of the primary pancreatic tumor [incorporating the gross tumor volume (GTV)] with a 5 mm margin, retroperitoneal soft tissue and surrounding regional nodal areas. The internal target volume (ITV) was generated by combining the CTVs associated with the expiration and inspiration phases of the VMI. Isotropic margins of 3 and 5 mm were added to the ITV to create a high-dose planning target volume (PTVHD) and elective dose PTV (PTVED), respectively [10, 15]. Volumetric modulated arc therapy plans using the simultaneous integrated boost technique were generated to deliver prescription doses of 60 and 50 Gy in 25 fractions for PTVHD and PTVED, respectively (Fig. 1a). The dosimetric parameters of D_1%, D_50% and D_99%, which indicate doses for 1, 50 and 99% of the volume of the GTV, were measured. During the radiotherapy course, all the patients received concurrent chemotherapy (GEM, GEM and NabPTX, or S-1). Patients who did not receive induction chemotherapy received S-1 or GEM + S-1.

Fig. 1.

(a) Volumetric modulated arc therapy planning using the simultaneous integrated boost technique, and (b) quantitative measurement on VMIs and IDMs for two patients.

Histopathological examination of resected specimens

Histopathological examination of 4 μm thick sections stained with hematoxylin and eosin was conducted by pathologists, and the histopathological response to the preoperative CRT was evaluated for all resected tumors. In accordance with the Seventh Edition of the General Rules for the Study of Pancreatic Cancer, by the Japan Pancreas Society, the histopathological response was classified according to the following five levels [16]: Grade Ia, estimated extent of residual tumor ≥ 90%; Grade Ib, 50% ≤ estimated extent of residual tumor < 90%; Grade II, 10% ≤ estimated extent of residual tumor < 50%; Grade III, estimated extent of residual tumor < 10%; Grade IV, complete response. In this study, patients were divided into either the response group (Grades II, III and IV) or the nonresponse group (Grades Ia and Ib).

Image analysis

Based on the VMI during the expiration phase, a radiation oncologist placed as large a region of interest (ROI) as possible within the GTV using the Eclipse treatment planning system. The ROIs were positioned in three consecutive slices at maximum (Fig. 1b). Areas with large vessels and artifacts from bile duct stents were excluded from the ROIs. In addition, a medical physicist positioned the three consecutive ROIs in the pancreatic parenchyma (PP) and aorta in close proximity to the ROI placed within the GTV. Because the amount of contrast uptake in the PP varies from place to place, ROI was placed in a homogeneous region without large vessels and artifact. In addition, measurements on the caudal (pancreatic tail) side of the pancreas from the PC were avoided as much as possible because of the possibility of blood flow obstruction by the tumor. Subsequently, the ROIs were copied to the IDM at the same respiratory phase image. The mean values of the CT number and IC in the VMI and IDM were measured for each patient. The respective values of ΔHU and ΔIC were determined as the differences in the HU and IC values, respectively, between the PC and PP.

Statistical analysis

All statistical analyses were performed using the SPSS software (version 27, IBM Corp., Armonk, NY, USA), and P < 0.05 was considered statistically significant. A logistic regression model was used for univariate analysis to determine the odds ratios for a set of potential predictor variables: age, gender, implementation of induction chemotherapy, type of concurrent chemotherapy, tumor diameter, serum CA19-9 value before preoperative CRT, HU value in the aorta, PC and PP, IC in the aorta, PC and PP, and ΔHU, ΔIC, D_1%, D_50% and D_99% between the response and nonresponse groups. Multivariate logistic regression analysis using parameters that were statistically significant (P < 0.05) in the univariate analysis was performed to distinguish between the two groups. Receiver operating characteristic (ROC) curves were used to determine the optimal cutoff values for distinguishing the two groups and to estimate the corresponding sensitivities and specificities.

RESULTS

Thirteen patients out of 58 could not undergo surgical resection, and a total of 45 patient underwent surgical resection after the CRT (Fig. 2). Among them, several patients were excluded for the following reasons. For three patients, the GTV was too small to secure a quantitative value because of the inaccuracies associated with small volumes [17]; for one patient, the GTV was excluded from the irradiated volume because of a dose constraint of OAR; for one further patient, the radiotherapy treatment course could not be completed. Finally, 40 patients were analyzed in this retrospective study as shown in Table 1. Twenty-four of these patients were assigned to the response group (19, 3 and 2 patients were allocated to Grades II, III and IV, respectively), and 16 were assigned to the nonresponse group (all 16 patients were allocated to Grade Ib). Comparison of patient characteristics between the two groups is listed in Table 2. No significant differences were observed with respect to age, gender, receipt of induction chemotherapy, type of concurrent chemotherapy, tumor diameter or CA19–9 levels before preoperative CRT between the two groups (P > 0.05).

Fig. 2.

Flowchart of this study. Meaning of acronyms: PTV = planning target volume.

Table 1.

Patient characteristics enrolled in this study

Characteristics (n = 40)
Age (y)	Mean ± SD	66 ± 8
Gender (n)	Male/female	25/15
Induction chemotherapy (n)	GEM + NabPTX	23
	GEM only	1
	FOLFILINOX	2
	None	14
Concurrent chemotherapy (n)	GEM + NabPTX	19
	GEM + S-1	1
	GEM	3
	S-1	17
Treatment response (n)	Grade Ib	16
	Grade II	19
	Grade III	3
	Grade IV	2
Tumor diameter (n)	<10 mm	0
	10 mm ≤ <20 mm	14
	20 mm ≤ <30 mm	19
	30 mm ≤	7
Photon energy (n)	6 MV	40
Width of MLC (n)	2.5 mm	38
	5 mm	2
Prescription dose for PTVHD (n)	60 Gy	40
Prescription dose for PTVED (n)	50 Gy	40

Table 2.

Comparison of patient characteristics between the response and nonresponse group

Characteristics		Response (n = 24)	Nonresponse (n = 16)	OR (95% CI)	P-value
Age (y), mean ± SD		66 ± 9	66 ± 8	0.992 (0.920–1.069)	0.830
Gender (n), male/female		17/7	8/8	2.429 (0.651–9.066)	0.187
Induction chemotherapy (n)				0.758 (0.198–2.898)	0.685
Yes		15	11
None		9	5
Concurrent chemotherapy (n)				0.455 (0.121–1.711)	0.244
GEM only, GEM + NabPTX or GEM + S-1		12	11
S-1 only		12	5
Tumor diameter (mm), mean ± SD		22.9 ± 6.8	22.6 ± 4.9	1.082 (0.282–3.061)	0.882
CA19–9 before chemoradiotherapy (U/ml), mean ± SD		75 ± 145	121 ± 220	0.999 (0.995–1.002)	0.442

Table 3 summarizes the comparison of quantitative values for the response and nonresponse groups. Univariate analyses revealed comparable values (P > 0.05) for HU (aorta and PC), IC (aorta and PC) and dosimetric parameters (D_1%, D_50% and D_99%). The HU and IC in PP in the non-response group were significantly higher (88.4 ± 15.5 HU and 32.5 ± 5.4 mg/ml, respectively) than those in the response group (70.3 ± 23.2 HU and 24.5 ± 7.4 mg/ml). In addition, the values of ΔHU (32.3 ± 14.2 HU) and ΔIC (14.4 ± 5.3 mg/ml) in the nonresponse group were higher than those in the response group (18.2 ± 20.0 HU and 8.2 ± 6.1 mg/ml for ΔHU and ΔIC, respectively). The histopathological response significantly correlated with HU in PP (P = 0.021), ΔHU (P = 0.03), IC in PP (0.005) and ΔIC (P = 0.007), with OR values [95% confidence interval (CI)] of 0.954 (0.917–0.993), 0.959 (0.923–0.996), 0.825 (0.723–0.942) and 0.840 (0.739–0.954), respectively, from univariate analysis. However, multivariate analysis did not indicate the significant correlation (P > 0.05).

Table 3.

Comparison of quantitative values between the response and nonresponse group

Parameters		Univariable						Multivariable
		Response		Nonresponse		OR (95% CI)	P-value	OR (95% CI)	P-value
		Mean	SD	Mean	SD
CT number (HU)	Aorta	222.5	55.3	227.6	42.5	0.998 (0.986–1.011)	0.754	NA
	PC	52.1	15.1	56.1	9.2	0.977 (0.928–1.028)	0.363	NA
	PP	70.3	23.2	88.4	15.5	0.954 (0.917–0.993)	0.021	1.043 (0.923–1.179)	0.498
	ΔHU	18.2	20.0	32.3	14.2	0.959 (0.923–0.996)	0.030	1.038 (0.919–1.172)	0.552
IC (mg/ml)	Aorta	95.3	27.4	98.4	20.2	0.995 (0.970–1.021)	0.700	NA
	PC	16.3	5.9	18.1	3.9	0.934 (0.825–1.058)	0.281	NA
	PP	24.5	7.4	32.5	5.4	0.825 (0.723–0.942)	0.005	0.767 (0.560–1.051)	0.099
	ΔIC	8.2	6.1	14.4	5.3	0.840 (0.739–0.954)	0.007	0.829 (0.605–1.138)	0.246
Dose for GTV (Gy)	D1%	58.5	3.3	57.0	3.9	1.124 (0.940–1.344)	0.201	NA
	D50%	53.3	3.3	51.8	2.6	1.195 (0.939–1.520)	0.147	NA
	D99%	49.4	1.4	48.6	1.5	1.480 (0.916–2.391)	0.110	NA

Table 4 lists the cutoff values, sensitivities, specificities and area under the curve (AUC) for distinguishing the two groups using quantitative parameters that were statistically significant in the univariate analysis. The highest sensitivity of 87.5%, with a cutoff value of 17.6 HU, was obtained using ΔHU. In contrast, the specificity of 83.3% was highest for IC in PP, with a cutoff value of 30.3 mg/ml. Overall, IC in PP and ΔIC had a higher AUC [0.797 (95% CI; 0.659–0.935) and 0.789 (0.650–0.928), respectively] than HU in PP and ΔHU [0.734 (0.580–0.889) and 0.721 (0.562–0.881), respectively].

Table 4.

Cutoff values, sensitivities, specificities and AUC for distinguishing response and nonresponse groups

Parameter	Cutoff value	Sensitivity (%)	Specificity (%)	AUC (95% CI)
PP (HU)	81.0 HU	81.3	58.3	0.734 (0.580–0.889)
ΔHU	17.6 HU	87.5	62.5	0.721 (0.562–0.881)
PP (IC)	30.3 mg/ml	68.8	83.3	0.797 (0.659–0.935)
ΔIC	10.6 mg/ml	75.0	75.0	0.789 (0.650–0.928)

DISCUSSION

In this study, we explored the potential of the IC measured using DECT for predicting the histopathological response to preoperative CRT in patients with PC. The IDMs derived from CE-DECT allow the amount of iodine present in blood vessels or tissues to be quantified, and the amount of iodine taken up by tissues can be used to evaluate hypoxic conditions. Hypoxia is a feature of the solid tumor microenvironment caused by insufficient and inhomogeneous vasculature, and a lack of blood supply disturbs tumor oxygenation. Under hypoxic conditions, an increase by a factor of ~2.5–3 in the radiation dose is required to obtain the same degree of damage as in well-oxygenated conditions [18]. Aoki et al. [19] measured the IC for primary lung cancer patients who underwent stereotactic radiotherapy and determined a median value of the average IC within the GTV of 1.86 mg/ml. In that study, the 2-year local control rate was significantly higher in the high-average IC group, and the authors concluded that IC might be used for a quantitative assessment of the radioresistance caused by the hypoxic tumor environment. The potential of the IC to predict treatment response has been reported for various tumor sites, such as head and neck, esophageal and cervical cancers [20–22].

PC is recognized as a severely hypoxic tumor, with pO₂ values ranging from 0 to 5.5 mm Hg, whereas the values for normal PP range from 9.3 to 92.7 mm Hg [23]. Delivering a dose with higher biological effectiveness is one option for overcoming this problem. Previous studies have reported that a higher biologically effective dose may improve local control and overall survival in patients with PC [24–26]. However, PC is often surrounded by radiosensitive OARs, such as the duodenum, bowel and stomach, and dose-escalating radiotherapy may cause adverse side effects. Thus, patient-specific treatment is required to predict treatment response and administer higher doses to those patients predicted to have a lower response. Visualizing and quantitatively assessing hypoxia in PC is a challenge for the prediction of treatment response. Metran-Nascente used the 2-nitroimidazole positron emission tomography (PET) tracer 18F-fluoroazomycin arabinoside and found that the degree of hypoxia varied among patients [27]. Dhani et al. [28] investigated surgical specimens from patients who received the preoperative 2-nitroimidazole tracer pimonidazole and demonstrated significant intra- and inter-tumoral heterogeneity of hypoxia.

CT scans provide a higher spatial resolution (with a matrix size of 512 × 512) and can be performed at a lower cost than PET examinations. Cai et al. demonstrated that the ratio of HU values for PP and PC in portal-venous phase was significantly associated with progression-free survival after intraoperative radiotherapy and stated that hypo-attenuation in CT imaging may reflect decreased blood flow [29]. In a report by Noda et al. [12], the difference in IC between the pancreatic parenchymatous and equilibrium phases was superior to the HU value for assessing the hemodynamics of PC. Fujita et al. [30] calculated the extracellular volume (ECV) based on IC of the PC and aorta in the equilibrium phase and demonstrated the potential of the ECV for predicting response to preoperative neoadjuvant chemotherapy with an AUC of 0.798 (0.682–0.886). Our results support the outcomes of these studies by confirming that the treatment response of patients with PC may be predicted using CT images, while our quantitative values were obtained in the pancreatic parenchymal phase. The pancreatic parenchymal phase is useful for contouring in radiotherapy because of its ability to clearly delineate pancreatic tumors. Noid et al. calculated the ECV based on IC of PC and aorta at late-arterial-phase, and found that there was a linear correlation between the ECV and the change before and after CRT in CA19-9 [31]. The low ECV value may represent hypoxic and highly resistance to radiotherapy, while ECV calculation requires hematocrit values, which cannot be calculated by DECT alone. More simply, Koay et al. [32] demonstrated that ΔHU between PC and PP during the pancreatic parenchymal phase is associated with clinical outcomes, and patients with higher ΔHU resulted in the poorer distant metastasis-free survival and overall survival. In the report, high ΔHU tumors were more likely to contain several poor prognosis mutations in combination with KRAS, such as TP53, SMAD4 and PIK3CA. Another report by Koay also demonstrated that high ΔHU PCs had lower progression-free survival and overall survival for preoperative CRT, and the imaging-based biomarker has the potential for predicting treatment outcome [33]. We have demonstrated that IC in PP and ΔIC could be a predictive factor for histopathological response of preoperative CRT, and that the AUC was higher for IC in PP and ΔIC than for HU in PP and ΔHU. Iodine uptake in the tumor is lower than in the PP, therefore the differences between patients are small, and the amount taken up by the PP may be an important factor in determining ΔIC. The IDM eliminates the signal from the PC/PP and measures only the iodine density, which is thought to be a more accurate assessment than HU values. In pancreatic parenchymal phase images, quantitative values obtained by DECT may reflect tumor features rather than blood flow.

There are other imaging modalities for predicting treatment outcome for patients with PC. Hyun et al. [34] assessed intratumoral heterogeneity measured by 18F-fluorodeoxyglucose (FDG) PET texture analysis and demonstrated that first-order entropy is associated with a 2-year survival prediction [AUC of 0.720 (0.625–0.815)]. Xu et al. [35] showed that metabolic tumor volume (MTV) and total lesion glycolysis (TLG) obtained from preoperative 18F-FDG PET examination might serve as surrogate markers for prediction of outcome in patients with PC (AUC of 0.712 for overall survival and 0.697 for recurrence-free survival using MTV; AUC of 0.706 and 0.696 using TLG). Park et al. [36] demonstrated that the volume transfer constant derived from perfusion CT could distinguish patients who underwent concurrent CRT into response and nonresponse groups with 75.0% sensitivity and 90.0% specificity. In contrast, Koch et al. [37] found that apparent diffusion coefficients obtained from diffusion-weighted magnetic resonance imaging did not contribute to the prediction of treatment outcomes for patients with PC. Compared with these findings, our study demonstrated the favorable association between IC and histopathological response. In addition, the reconstructed images from DECT simulations, such as VMI, water density images and effective atomic numbers, have advantages in improving tumor visualization and the accuracy of dose calculation [13, 15, 38, 39]. Therefore, the use of DECT in radiotherapy is expected to increase.

This study had several limitations. First, the number of patients was limited; a larger number of cases should be examined to improve the reliability of the results. Second, IC was measured in the IDMs during the expiration phase of CE-4D-DECT, and the images included larger motion artifacts than the breath-holding images. Third, there are several image analysis techniques for exploring the predictive factors of histopathological response, such as histogram analysis [12] and radiomics analysis [40]. Because these techniques extract dozens to thousands of image features, the limited number of cases in this study did not allow for multivariate analysis using these features. Fourth, PC is depicted differently depending on the timing of imaging [41]; however, in this study, only the parenchymal phase of the pancreas was examined. Finally, this research utilized data from the past several years for analysis, but it did not include information on patients from more recent times. Consequently, it is important to note that the findings of this study may have limited direct applicability to the current patient population.

In conclusion, the IC at pancreatic parenchymal phase measured by DECT may reflect features of PC, and the IC in PP and the difference in IC between PP and PC has the potential for predicting histopathological responses in patients who undergo preoperative CRT.

CONFLICT OF INTEREST

The authors declare no conflicts of interest in relation to this study.

FUNDING

This study was supported by a JSPS KAKENHI Grant (Grant-in-Aid for Scientific Research (C) 21 K07742).

DATA AVAILABILITY

The original contributions presented in the study are included in the article. Further inquiries can be directed to the corresponding author.