- Email: [email protected]
Perfusion CT can predict tumoral grading of pancreatic adenocarcinoma
European Journal of Radiology 82 (2013) 227–233
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Perfusion CT can predict tumoral grading of pancreatic adenocarcinoma M. D’Onofrio a,∗ , A. Gallotti a , W. Mantovani b , S. Crosara a , E. Manfrin c , M. Falconi d , A. Ventriglia a , G.A. Zamboni a , R. Manfredi a , R. Pozzi Mucelli a a
Department of Radiology, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy Department of Medicine and Public Health, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy c Department of Pathology, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy d Department of Surgery, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy b
a r t i c l e
i n f o
Article history: Received 12 May 2012 Received in revised form 26 September 2012 Accepted 29 September 2012 Keywords: Perfusion CT Pancreatic adenocarcinoma Abdominal radiology Tumor grading
a b s t r a c t Objectives: To describe perfusion CT features of locally advanced pancreatic ductal adenocarcinoma and to evaluate correlation with tumor grading. Methods: Thirty-two patients with locally advanced pancreatic adenocarcinoma were included in this study. Lesions were evaluated by P-CT and biopsy after patient’s informed consent. P-CT parameters have been assessed on a large single and on 6 small intratumoral ROIs. Values obtained have been compared and related to the tumor grading using Mann–Whitney U test. Sensibility, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy in predicting tumor grading have been calculated for cut-off values chosen by using ROC curves. Results: Out of 32 lesions, 12 were classified as low grade and 20 as high grade. A statistically significant difference between high and low grade neoplasms were demonstrated for PEI and BV parameters. PEI and BV cut-off values were respectively 17.8 HU and 14.8 ml/100 g. PEI identified high grade neoplasms with a 65% sensitivity, 92% specificity, 93% PPV, 61% NPV and 75% accuracy. BV identified high grade neoplasms with a 80% sensitivity, 75% specificity, 84% PPV, 69% NPV, 78% accuracy. Considering both PEI and BV, P-CT identified high grade lesions with a 60% sensitivity, 100% specificity, 100% PPV, 60% NPV and 75% accuracy. Conclusions: PEI and BV perfusion CT parameters proved their efficiency in identifying high grade pancreatic adenocarcinoma. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Pancreatic ductal adenocarcinoma (ADK) is the most common primary malignancy of the pancreas and is associated with a very poor prognosis, being worldwide one of the leading cause of cancerrelated death [1,2]. Patients with pancreatic ADK have an overall 1-year and 5-year survival rate after diagnosis below 20% and below 5%, respectively [3]. Surgical resection still remains the only potentially curative treatment [3–5]. However, only 5–25% of patients with pancreatic cancer are candidates for radical resection due to the difficulty in achieving early diagnosis [6,7]. About 40–50% of patients have metastases and approximately 40% of patients have locally advanced disease at the time of diagnosis [3,8]. Moreover, despite the fact that a combined surgical-systemic approach may offer a longer life expectancy, the long-term outcome remains dis-
∗ Corresponding author. Tel.: +39 045 8124140; fax: +39 045 8277808. E-mail address: [email protected] (M. D’Onofrio). 0720-048X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2012.09.023
mal, with a mortality of 1–5% and an overall 5-year survival rate after pancreaticoduodenectomy between 15% and 20% and after distal pancreatectomy between 8% and 15% [4,5,9,10]. Some clinico-pathological features could be advocated as risk factors for an unfavorable prognosis, such as age, gender, serum CA 19.9 level, duration of disease-related symptoms, tumor size and grade, lymph node involvement, perineural and retroperitoneal invasion [4,11]. The high incidence of recurrence after surgery is also independently related to the presence of positive transection margins [4,5,10]. The early recurrence of pancreatic adenocarcinoma after resection characterizes the patient population reported in literature as “early death” (ED). In this patient population the resection can be considered unnecessary. The preoperative correct identification of this group of patients is very important to minimize unnecessary resections but remains difficult owing to the post-operative assessment of some factors such as tumor resection margins and grading [4]. Molecular and immunohistochemical approaches have been used to better explain the inherited biological aggressiveness of
228
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
human cancers and to suggest the development of new treatment strategies [12–14]. The tumor growth, grading and progression and its metastatic spread have been shown strictly dependent on tumor angiogenesis. Recently, quantitative imaging techniques based on functional and targeted information have been developed [15–17]. Perfusion CT (P-CT) is a new imaging technique able to provide qualitative and quantitative information on perfusion parameters of tissues, which have been demonstrated to be correlated with histological markers of angiogenesis. The excellent linear relationship between tissue attenuation and iodinated contrast agent concentration allows an objective quantification of the perfusion parameters correlated with the hemodynamic changes caused by new vessels [12,18,19]. Accurate imaging detection, characterization and staging of pancreatic cancer are of paramount importance to provide patients an adequate treatment strategy. Functional imaging from P-CT may add useful information on tumor aggressiveness affecting treatment strategy and patient management. To our knowledge, there have been no previous reports in the literature in which the P-CT parameters of pancreatic ADK and tumor grading were compared. The purpose of this study was to prospectively describe the perfusion CT features of locally advanced pancreatic ductal
adenocarcinomas and to assess whether these features correlate with the tumor grading at pathology. 2. Material and methods 2.1. Patients Institutional review board was obtained for this prospective study, conducted in accordance with the principles of Helsinki Declaration. Thirty-two patients (17 males, 15 female; mean age 66.2 years; range, 43–83 years) were included in the study population (Table 1). All patients were required to provide written informed consent before to be involved in this study. Between December 2010 and June 2011, all patients afferent to our Institution with a suspicion of locally advanced pancreatic cancer were recruited. Locally advanced pancreatic cancer was considered as non resectable tumor in absence of any distant metastasis [3]. The inclusion criterion was the presence of locally advanced pancreatic ductal adenocarcinoma in patients who did not undergo any previous neoadjuvant or palliative therapies. All pancreatic lesions were pathologically confirmed by fine needle biopsy performed the same day immediately after the P-CT. Patients were excluded
Table 1 Patients, tumors and perfusion CT parameters.
I
II
Gender Female Male Ageb Site Right pancreas Body-tail Size (mm)b PFb PF1 PF2 PF3 PF4 PF5 PF6 Mean PF 1–6 TTPb TTP1 TTP2 TTP3 TTP4 TTP5 TTP6 Mean TTP 1–6 PEIb PEI1 PEI2 PEI3 PEI4 PEI5 PEI6 Mean PEI 1–6 BVb BV1 BV2 BV3 BV4 BV5 BV6 Mean BV 1–6
High grade (n = 20)
Low grade (n = 12)
9 11 65 (43; 83)
6 6 69 (46; 79)
11 9 4.4 (2.5; 7)
7 5 4.2 (3; 5.2)
5.9 (1.5; 75.1) 7.9 (1.1; 67.9) 4.5 (0.7; 72.5) 5.7 (2.3; 66.7) 8.5 (1.3; 98.4) 6 (1.3; 119.1) 8.4 (2.1; 62.7) 11.4 (2.6; 48.6) 70 (6; 101) 59.5 (5; 101) 70 (11; 101) 70 (10; 101) 70 (5; 101) 70 (5; 101) 60 (10; 101) 65 (25.8; 101) 16.2 (7; 40.9) 15.7 (4; 43.7) 14.3 (1.3; 37.3) 18.4 (1.9; 52.9) 19.3 (4.4; 44.2) 14.8 (4.6; 44.2) 16.1 (8.6; 37.8) 17.7 (5; 42.8) 11.3 (1.3; 37.3) 9.5 (0; 39.7) 7.7 (0; 33.8) 11 (0; 35.9) 9.7 (1.7; 37.5) 8.2 (0; 37.5) 9.1 (0.4; 28.8) 9.3 (0.4; 34.1)
8.9 (3.8; 22.9) 10.4 (4.6; 28.1) 11.3 (2.1; 29.4) 9.1 (1.9; 27.4) 8.3 (3.9; 19.5) 11.4 (4.8; 30.2) 13.2 (6.2; 53.4) 10.9 (5.7; 26.6) 82.5 (30; 100) 85 (30; 100) 70 (30; 100) 70 (30; 100) 85 (30; 100) 70 (25; 100) 70 (25; 100) 80.6 (30; 95) 26.3 (15.8; 47) 28.1 (13.8; 56.3) 25.3 (11.1; 70) 27.2 (3.4; 55) 27.8 (10; 59.4) 28.4 (8.3; 64.6) 32.3 (22; 58) 28.2 (14.7; 54.1) 19 (9; 47.2) 22.3 (12; 48.7) 18 (1.3; 60.3) 17.1 (2.9; 49.7) 18.9 (4.7; 53.1) 19.4 (3.5; 55.8) 23.6 (10.1; 57.1) 17.2 (7.8; 45.8)
p valuea 0.784c 0.239 0.854c 0.272 0.387 0.408 0.099 0.408 0.803 0.182 0.239 0.924 0.716 0.107 0.893 0.833 0.272 0.659 0.387 0.099 0.012 0.003 0.008 0.099 0.099 0.004 <0.0001 0.005 0.004 <0.0001 0.001 0.289 0.053 0.003 <0.0001 0.001
Total (n = 32) 15 17 66 (43; 83) 18 14 4.3 (2.5; 7) 8.2 (1.5; 75.1) 9.4 (1.1; 67.9) 7.6 (0.7; 72.5) 7.1 (1.9; 66.7) 8.5 (1.3; 98.4) 8 (1.3; 119.1) 11.4 (2.1; 62.7) 11 (2.6; 48.6) 70 (6; 101) 70 (5; 101) 70 (11; 101) 70 (10; 101) 70 (5; 101) 70 (5; 101) 70 (10; 101) 69.7 (25.8; 101) 19.9 (7; 47) 21.7 (4; 56.3) 21 (1.3; 70) 21.1 (1.9; 55) 24.7 (4.4; 59.4) 21.1 (4.6; 64.6) 21.6 (8.6; 58) 19.9 (5; 54.1) 13.4 (1.3; 47.2) 12.5 (0; 48.7) 11.8 (0; 60.3) 11.4 (0; 49.7) 12.3 (1.7; 53.1) 14.1 (0; 55.8) 11 (0.4; 57.1) 11.7 (0.4; 45.8)
In the 1st section of this table data regarding patients and tumors divided into high and low grade of malignancy are shown. In the 2nd section this table contains median values of P-CT parameters PF, TTP, PEI and BV measured on large single ROI and on the six small ROIs obtained for high and low grade of malignancy. a Mann–Whitney U test. b Median (range). c Fisher exact test.
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
in presence of resectable tumors at imaging examinations, if distant metastases were observed or if they had a cancer other than pancreatic adenocarcinoma. 2.2. Perfusion CT technique All perfusion CT examinations were performed at our Institution on a 64-row Multidetector CT scanner (Brilliance 64; Philips, Best, The Netherlands). After a minimum fast of 6 h, P-CT was performed with patients in supine position with their arms behind their head, during superficial and regular free-breath. An abdominal belt (Philips Medical Systems, Best, The Netherlands) was adjusted to reduce the artifacts caused by respiratory motion. An initial unenhanced wide coverage scan was acquired with a section thickness of 5 mm to correctly localize the tumor and select the volume on which perfusion scans had to be performed. The spatial coordinates were noted and used to plan the subsequent dynamic examination achieved with cine table position. P-CT consists of the acquisition of repeated CT scan of the volume being analyzed over time, during and after the administration of a bolus of iodinated contrast agent. High iodine concentration (370 mg/ml) contrast agents are preferred since there is a linear relationship between iodine concentration and tissue enhancement [16,17]. A total of 50 ml of contrast agent at an injection rate of 5 ml/s was administered according to the “short sharp bolus” technique through an antecubital vein, and followed by 40 ml of saline solution at the same flow rate [16]. CT scanning parameters included 120 kVp (for a slope method based scanning protocol, which is very sensitive to the image noise, a higher voltage has to be used than for a deconvolution method), tube current of 150 mAs, gantry rotation time of 0.5 s, increment 5 mm, 5-mm slice thickness and 100 mm in length. 64 slice CT scanners allow an axial coverage of about 40 mm. Using toggling table technique during dynamic images acquisition, axial coverage can be extended to 100 mm [19]. Scanning was initiated 12 s after starting contrast material injection and images were acquired at every 5 s for 40 s, and later at 50, 70, 90 and 120 s, for a total of 200 images. 2.3. Imaging and data analysis Image data were processed on a workstation (Extended Brilliance Workspace V4.5.2.40007, Philips, Best, The Netherlands) loaded with commercial perfusion CT software (CT Perfusion Software, Philips, Best, The Netherlands) based on slope method. Perfusion parameters for analyzing images acquired during the first pass were calculated with the use of dedicated semiautomatic software. An abdominal radiologist with 10 years of experience in pancreatic imaging, aware of the purpose of this study and kept blinded to the histopathological analysis, reviewed and interpreted all images. Tumor location and size were previously defined in each patient. The largest dimension was measured on axial scans and mean diameter calculated. The arterial input was measured by manually drawing a circular region of interest (ROI) within the abdominal aorta on a selected image, being careful to avoid any mural calcification. The arterial time/density curve and the end of the first pass phase were derived automatically by the software. To obtain P-CT parameters a single ROI was manually drawn within the tumor on the axial slice in which the tumor showed the maximum diameter being careful not to include the surrounding adipose tissue and adjacent normal vascular structures to avoid partial volume effect. Additionally, six smaller ROIs were manually drawn within the tumor, of which two in the central portion (called 1 and 2) and four in the periphery (called 3, 4, 5, 6). Qualitative information achieved by perfusion CT consists of analysis of color maps automatically generated by the software, able to provide an
229
immediate and panoramic representation of perfusion distribution within the tumor. Quantitative information consists of the analysis of P-CT parameters. The final numeric values expressed for each ROI are the mean values automatically calculated by the system from all measurements performed in each voxel included in an ROI. The following parameters have been calculated: • Perfusion (or blood flow, PF), expressed in ml/100 g of tissue/min, which consists of the ratio between the maximum slope of the time/density curve of tissue and the peak density reached by the aorta, selected as arterial input. • Time to Peak (TTP), expressed in seconds, which indicates the time interval between the arrival of contrast agent in the aorta (selected as arterial input) and the peak density in the tissue. It is a marker of perfusion pressure. • Peak Enhancement Intensity (PEI), expressed in Hounsfield Units, which indicates the peak attenuation reached by the tissue after contrast media injection. It is a marker of tissue blood volume. • Blood volume (BV), expressed in ml/100 g of tissue, which reflects the volume of blood flowing within all vessels in the tissue. It is a marker of tumor vascularity. 2.4. Pathological diagnosis In our Institution all patients with locally advanced pancreatic cancer usually undergo percutaneous ultrasound-guided fine needle biopsy to achieve a pathological confirmation prior to adjuvant therapies. 2.4.1. Biopsy procedure All patients included in the study population underwent biopsy under ultrasound guided percutaneous approach in the same day immediately after P-CT examination. Local anesthesia (lidocaine 1%) was administered in the abdominal wall at the chosen entry point. Multifrequency probes (ranging from 2.5 to 5 MHz) with lateral support were used during the aspiration biopsy, performed with fine needles (Menghini-type needles, 20–21-G caliper). Once the pancreatic tumor had been reached, the needle was gently inserted and moved with very small excursions. The unstructured material collected in the needle cavity was then placed on a glass slide. An adequate sample must contain no or little blood cells, which could interfere with analysis. To allow an optimal reading, the material was smeared into a thin layer and fixed with 95% alcohol to avoid cellular lyses and morphological artifacts. Immediate fixation and a modified Papanicolaou staining allowed to evaluate in real time the appropriateness of the sample. 2.4.2. Histopathological analysis Two independent pathologists respectively with 15 and 5 years of experience in pancreatic cytology, aware of the purpose of this study and kept blinded to the P-CT analysis, examined the glass slides randomly to make the grades as objective as possible. At low magnification the cellular density of the samples was evaluated followed by a specific characterization of both the cellular background and the diagnostic elements at higher magnification. The presence of confluent pathological cells and morphological features of nuclei and cytoplasm were accurately described. After having ensured a diagnosis of pancreatic ductal adenocarcinoma in all patient-population, tumors were classified in two different groups according to their grading, as follows: low grade, was defined as well- or moderate-differentiated lesions with low cytological atypia; high grade, was defined as poorly differentiated lesions with several cytological atypia.
230
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
2.5. Statistical analysis Distribution of continuous variables was reported as mean (or median in event of abnormal distribution) and range (minimum–maximum values). Categorical variables were presented as numbers and percentages. The comparison between the low grade and the high grade tumor groups was carried out using Student’s t test or the Mann–Whitney U test for continuous variables. Categorical data were compared by using the chi-square test and the Fisher exact test when necessary. Diagnostic accuracy was assessed as the area under the receiver-operating characteristic (ROC) plot. Diagnosis obtained by pathologic examination was the external gold standard. Sensitivity, specificity, positive and negative predictive values (and their 95% Confidence Interval) for each significant P-CT parameter in the prediction of tumor grade were also calculated by using cut-off values chosen on the basis of ROC curves. A p value less than or equal to 0.05 was considered statistically significant. 3. Results Perfusion CT examination and biopsy under ultrasound guided percutaneous approach were possible in all cases. 3.1. Pathology Of the 32 pancreatic ductal adenocarcinomas (mean diameter, 45.9 mm; range, 30–70 mm), 6 (18.8%) lesions were located in the head, 7 (21.8%) in the uncinate process, 5 (15.7%) in the neck, 14 (43.7%) in the body-tail of the gland. According to the pathological analysis, 12 (37.5%) were classified as low grade ADK and 20 (62.5%) as high grade (Table 1). No significant difference was observed between the mean diameter of low grade and high grade tumors (mean size, 42.6 mm; range, 30–52 mm; vs mean size, 46.4 mm; range, 25–70 mm; p = 0.272). 3.2. Perfusion CT parameters Median Perfusion (PF) value resulting from the measurements performed on the large ROI was 5.9 ml/100 g of tissue/min (range 1.5–75.1) in reference to high grade neoplasms and 8.9 ml/100 g of tissue/min (range 3.8–22.9) in reference to low grade neoplasms; median PF value resulting from the mean value of the measurements performed on the 6 small ROIs was 6.8 ml/100 g of tissue/min (range 2.6–48.6) in reference to high grade neoplasms and 10.9 ml/100 g of tissue/min (range 5.7–26.6) in reference to low grade neoplasms. There was no statistically significant difference between high grade and low grade neoplasms in terms of median PF values (Table 1) resulting both from
measurements performed on the large ROI (p = 0.387) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.924). Median Time To Peak (TTP) value resulting from the measurements performed on the large ROI was 70 s (range 6–101) in reference to high grade neoplasms and 82.5 s (range 30–100) in reference to low grade neoplasms; median TTP value resulting from the mean value of the measurements performed on the 6 small ROIs was 65 s (range 25.8–101) in reference to high grade neoplasms and 80.6 s (range 30–95) in reference to low grade neoplasms. There was no statistically significant difference between high grade and low grade neoplasms in terms of median TTP values (Table 1) resulting both from measurements performed on the large ROI (p = 0.716) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.099). Median Peak Enhancement Intensity (PEI) value resulting from the measurements performed on the large ROI was 16.2 HU (range 7–40.9) in reference to high grade neoplasms and 26.3 HU (range 15.8–47) in reference to low grade neoplasms; median PEI value resulting from the mean value of the measurements performed on the 6 small ROIs was 17.7 HU (range 5–42.8) in reference to high grade neoplasms and 28.2 HU (range 14.7–54.1) in reference to low grade neoplasms. There was statistically significant difference at Mann–Whitney U-test between high grade and low grade neoplasms in terms of median PEI values (Table 1) resulting both from measurements performed on the large ROI (p = 0.012) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.005). Median Blood Volume (BV) value resulting from the measurements performed on the large ROI was 11.3 ml/100 g of tissue (range 1.3–37.3) in reference to high grade neoplasms and 19 ml/100 g of tissue (range 9–47.2) in reference to low grade neoplasms; median BV value resulting from the mean value of the measurements performed on the 6 small ROIs was 9.3 ml/100 g of tissue (range 0.4–34.1) in reference to high grade neoplasms and 17.2 ml/100 g of tissue (range 7.8–45.8) in reference to low grade neoplasms. There was statistically significant difference at Mann–Whitney U-test between high grade (Fig. 1) and low grade (Fig. 2) neoplasms in terms of median BV values (Table 1) resulting both from measurements performed on the large ROI (p = 0.004) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.001). 3.3. Perfusion CT – statistical analysis Given these P-CT results, only PEI and BV parameters measured on the large ROI were taken into account, since the measurements performed on the large ROI and the mean value of the measurements performed on the 6 small ROIs have been proved to be statistically significant. The choice of the large ROI has been
Fig. 1. CT axial scan (left) showing a hypodense mass of the pancreatic body-tail (arrow). BV color-map (right) obtained during same scan, shows hypoperfusion of the lesion (arrow).
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
231
Fig. 2. CT axial scan (left) showing a hypodense mass of the head-uncinate process of the pancreas (arrow). BV color-map (right) obtained during same scan shows well perfused lesion (arrow).
Table 2 Cut-off values for PEI perfusional CT parameter. Criterion
Sensitivity
95% CI
Specificity
95% CI
<7 <=14.9 <=15.8 <=17.8* <=18.7 <=21 <=23.1 <=23.5 <=27.9 <=29.5 <=30.3 <=31.7 <=40.9 <=47
0.00 45.00 50.00 65.00 65.00 70.00 70.00 75.00 75.00 85.00 85.00 90.00 100.00 100.00
0.0–16.8 23.1–68.5 27.2–72.8 40.8–84.6 40.8–84.6 45.7–88.1 45.7–88.1 50.9–91.3 50.9–91.3 62.1–96.8 62.1–96.8 68.3–98.8 83.2–100.0 83.2–100.0
100.00 100.00 91.67 91.67 75.00 75.00 66.67 66.67 33.33 33.33 25.00 16.67 16.67 0.00
73.5–100.0 73.5–100.0 61.5–99.8 61.5–99.8 42.8–94.5 42.8–94.5 34.9–90.1 34.9–90.1 9.9–65.1 9.9–65.1 5.5–57.2 2.1–48.4 2.1–48.4 0.0–26.5
Using ROC curves cut-off value for PEI P-CT parameter with the best performance in terms of sensitivity and specificity in the characterization of high grade and low grade neoplasms. Fig. 3. ROC curve for PEI parameter. Area under the curve (AUC) is 0.767.
suggested by the advantages in terms of time consuming in data analysis. ROC curve for PEI parameter in shown in Fig. 3 with an area under the curve (AUC) of 0.767. ROC curve for BV parameter in shown in Fig. 4 with an area under the curve (AUC) of 0.798. Using these ROC curves, the cut-off values for PEI and BV perfusion parameters with the best performance in terms of sensitivity
and specificity in the characterization of high grade or low grade neoplasms have been chosen. PEI cut-off value has been set at 17.8 HU with sensitivity and specificity of 65% and 91.67% respectively (Table 2). BV cut-off value has been set at 14.8 ml/100 g with sensitivity and specificity of 80% and 75% respectively (Table 3). PEI perfusion parameter with a cut-off of 17.8 HU can identify pancreatic high grade adenocarcinomas with a 65% sensitivity (95% CI = 0.441–0.859), a 91.7% specificity (95% CI = 0.760–1.073), a positive predictive value of 92.9%, a negative predictive value of
Table 3 Cut-off values for BV perfusional CT parameter.
Fig. 4. ROC curve for BV parameter. Area under the curve (AUC) is 0.798.
Criterion
Sensitivity
95% CI
Specificity
95% CI
< 1.3 <=8.5 <=9 <=10 <=10.4 <=12.3 <=14.5 <=14.8* <=15.7 <=18.5 <=19.2 <=26.5 <=28.3 <=37.3 <=47.2
0 35 35 40 45 70 70 80 80 85 85 90 90 100 100
0.0–16.8 15.4–59.2 15.4–59.2 19.1–63.9 23.1–68.5 45.7–88.1 45.7–88.1 56.3–94.3 56.3–94.3 62.1–96.8 62.1–96.8 68.3–98.8 68.3–98.8 83.2–100.0 83.2–100.0
100 100 91.67 91.67 83.33 83.33 75 75 58.33 58.33 41.67 41.67 25 25 0
73.5–100.0 73.5–100.0 61.5–99.8 61.5–99.8 51.6–97.9 51.6–97.9 42.8–94.5 42.8–94.5 27.7–84.8 27.7–84.8 15.2–72.3 15.2–72.3 5.5–57.2 5.5–57.2 0.0–26.5
Using ROC curves cut-off value for BV P-CT parameter with the best performance in terms of sensitivity and specificity in the characterization of high grade and low grade neoplasms.
232
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
61.1% and a 75% accuracy. BV perfusion parameter with a cut-off of 14.8 ml/100 g can identify pancreatic high grade adenocarcinomas with a 80% sensitivity (95% CI = 0.625–0.975), a 75% specificity (95% CI = 0.505–0.995), a positive predictive value of 84.3%, a negative predictive value of 69.2% and a 78.1% accuracy. Considering both PEI and BV perfusion parameter with the related cut-off values, it is possible to characterize pancreatic ductal high grade adenocarcinomas with a 60% sensitivity (95% CI = 0.385–0.815), a 100% specificity, a positive predictive value of 100%, a negative predictive value of 60% and a 75% accuracy.
4. Discussion Perfusion CT (P-CT) is an imaging technique able to quantify the tissue enhancement after contrast material administration. Main P-CT fields of application are Neuroimaging (in diagnosing stroke and choosing its proper treatment) and Oncologic Imaging in terms of tumor assessment, characterization and staging, angiogenesis assessment, response to therapy prediction and response to therapy monitoring [20]. Many other applications are however emerging also in non oncologic fields such the assessment of changes in organ (liver, pancreas and spleen) perfusion in cirrhotic patients [21]. Regarding the CT perfusion technique applied in this study, although patient should not move at all during image acquisition, the movement provides less impact on image quality by using CT scanners based on slope method than other processing methods as reported in literature [17]. A shallow breathing during images acquisition is considered acceptable and patients must be instructed not to breathe deeply when feeling the hot flush due to the contrast agent. Pancreatic ductal adenocarcinoma is the most common pancreatic primary malignancy accounting for 80% of pancreatic malignancies. The main aims of imaging consist of tumor detection and characterization, playing an important role in disease management. CT study represents the modality of choice in diagnosis confirmation, providing accurate staging. In more than 95% of the patients pancreatic ductal adenocarcinoma is diagnosed at an advanced stage, in presence of locally advanced disease (vascular or perineural invasion) or distant metastases. Prognosis and therapeutic approaches depend on lesion resectability at the moment of clinical presentation. As reported by Barugola et al., up to 30% of resected patients die of disease within 1 year after surgery [4]. In this subgroup recurrence is very early, and survival is similar to that observed in patients with advanced disease undergoing antitumoral therapies alone. These so-called “early deaths” (ED) after resection for curative intent may be attributed either to inadequate pre- or intra-operative staging or to a particularly aggressive behavior of the disease. Thus, the ability to identify patients at high risk of ED before surgery is an important clinical goal to circumvent a demanding surgical intervention. According to some authors symptoms lasting >40 days, CA 19–9 >200 U/ml and G2 pathological grading are easy-to-obtain preoperative parameters in order to identify patients with a disease not suitable for front-up surgery, even if deemed resectable by high quality imaging [4,11]. According to Barugola et al. [4] those patients with level of CA 19–9 >200 U/ml and symptoms lasting >40 days should undergo fine needle biopsy: if pathology report shows a poorly differentiated or anaplastic tumor, a non-surgical therapeutical approach should be carried out. After neoadjuvant therapy (or, if necessary, palliative treatment), in case of disease stability or down-staging, patients could undergo surgical intervention. The present study aims to demonstrate P-CT ability in direct and preoperative characterization of high grade pancreatic ductal
adenocarcinomas in order to implement the identification of patients at high risk of ED, to provide each patient with the best treatment strategy. Thirty-two patients with locally advanced pancreatic ductal adenocarcinoma underwent P-CT imaging study and percutaneous US-guided fine needle biopsy. A statistically significant difference between high grade and low grade neoplasms was found in terms of median PEI and BV values resulting both from measurements performed on the large ROI (p = 0.012 for PEI and p = 0.004 for BV) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.005 for PEI and p = 0.001 for BV). Using ROC curves the cut-off values for PEI and BV respectively with the best performance in terms of sensitivity and specificity in the characterization of high grade tumors have been chosen. PEI perfusional parameter with a cut-off of 17.8 HU can identify pancreatic high grade adenocarcinomas with a 75% accuracy. BV perfusional parameter with a cut-off of 14.8 ml/100 g can identify pancreatic high grade adenocarcinomas with a 78.1% accuracy. Considering both PEI and BV perfusional parameter with the related cut-off values, it is possible to characterize pancreatic ductal high grade adenocarcinomas with a 60% sensitivity, a 100% specificity, a positive predictive value of 100%, a negative predictive value of 60% and a 75% accuracy. According to our results, P-CT offers a significant contribution both to clinical and laboratory data in identifying high grade pancreatic ductal adenocarcinoma, and consequently patients at high risk of ED, more accurately than the suspicion given by duration of symptoms and CA 19–9 levels. In everyday clinical practice patients with pancreatic ductal adenocarcinoma classified as high grade at P-CT should undergo biopsy to achieve a grading confirmation even if resectable at imaging. Actually, standard of care for pancreatic lesions typical for ductal adenocarcinoma deemed resectable at imaging is surgical intervention, without pathological confirmation. Considering its accuracy values, P-CT could adequately select all patients with a high grade lesion and justify fine needle biopsy to confirm the grade itself, even in presence of a resectable disease at imaging. The limitations of this study are the relative small number of patients included and the analysis of a fine needle biopsy sample instead of the whole surgical specimen. Other prospective studies about resectable pancreatic lesions based on a larger population are needed. 5. Conclusion Perfusion CT can predict tumor grade of pancreatic adenocarcinoma. In particular, PEI and BV perfusion parameters proved their efficiency in identifying high grade pancreatic adenocarcinoma. References [1] Schima W, Ba-Ssalamah A, Kolblinger C, Kulinna-Cosentini C, Puespoek A, Gotzinger P. Pancreatic adenocarcinoma. European Radiology 2007;17(3):638–49. [2] D’Onofrio M, Gallotti A, Pozzi Mucelli R. Imaging techniques in pancreatic tumors. Expert Review of Medical Devices 2010;7(2):257–73. [3] Willett CG, Czito BG, Bendell JC, Ryan DP. Locally advanced pancreatic cancer. Journal of Clinical Oncology 2005;23(20):4538–44. [4] Barugola G, Partelli S, Marcucci S, et al. Resectable pancreatic cancer: who really benefits from resection? Annals of Surgical Oncology 2009;16(12):3316–22. [5] Barugola G, Partelli S, Crippa S, et al. Outcomes after resection of locally advanced or borderline resectable pancreatic cancer after neoadjuvant therapy. American Journal of Surgery 2012;203(2):132–9. [6] Zuckerman DS, Ryan DP. Adjuvant therapy for pancreatic cancer: a review. Cancer 2008;112(2):243–9. [7] Han SS, Jang JY, Kim SW, Kim WH, Lee KU, Park YH. Analysis of longterm survivors after surgical resection for pancreatic cancer. Pancreas 2006;32(3):271–5. [8] Czito BG, Willett CG, Clark JW, Fernandez Del Castillo C. Current perspectives on locally advanced pancreatic cancer. Oncology 2000;14(11):1535–45, discussion 46, 49–52.
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233 [9] Picozzi VJ, Pisters PW, Vickers SM, Strasberg SM. Strength of the evidence: adjuvant therapy for resected pancreatic cancer. Journal of Gastrointestinal Surgery 2008;12(4):657–61. [10] Barugola G, Falconi M, Bettini R, et al. The determinant factors of recurrence following resection for ductal pancreatic cancer. Journal of the Pancreas 2007;8(Suppl. 1):132–40. [11] Yeo CJ, Cameron JL. Pancreatic cancer. Current Problems in Surgery 1999;36(2):59–152. [12] Park MS, Klotz E, Kim MJ, et al. Perfusion CT: noninvasive surrogate marker for stratification of pancreatic cancer response to concurrent chemo- and radiation therapy. Radiology 2009;250(1):110–7. [13] Kuwahara K, Sasaki T, Kuwada Y, Murakami M, Yamasaki S, Chayama K. Expressions of angiogenic factors in pancreatic ductal carcinoma: a correlative study with clinicopathologic parameters and patient survival. Pancreas 2003;26(4):344–9. [14] Hattori Y, Gabata T, Zen Y, Mochizuki K, Kitagawa H, Matsui O. Poorly enhanced areas of pancreatic adenocarcinomas on late-phase dynamic computed tomography: comparison with pathological findings. Pancreas 2010;39(8):1263–70. [15] Wang ZQ, Li JS, Lu GM, Zhang XH, Chen ZQ, Meng K. Correlation of CT enhancement, tumor angiogenesis and pathologic grading
[16]
[17] [18]
[19] [20] [21]
233
of pancreatic carcinoma. World Journal of Gastroenterology 2003;9(9): 2100–4. Bellomi M, Viotti S, Preda L, D’Andrea G, Bonello L, Petralia G. Perfusion CT in solid body-tumours, Part II: clinical applications and future development. Radiologia Medica 2010;115(6):858–74. Kambadakone AR, Sahani DV. Body perfusion CT: technique, clinical applications, and advances. Radiologic Clinics of North America 2009;47(1):161–78. Kambadakone AR, Sharma A, Catalano OA, Hahn PF, Sahani DV. Protocol modifications for CT perfusion (CTp) examinations of abdomen-pelvic tumors: impact on radiation dose and data processing time. European Radiology 2011;21(6):1293–300. Miles KA. Perfusion CT for the assessment of tumour vascularity: which protocol? British Journal of Radiology 2003;76(Spec no. 1):S36–42. Miles KA. Perfusion imaging with computed tomography: brain and beyond. European Radiology 2006;16(Suppl. 7):M37–43. Motosugi U, Ichikawa T, Sou H, Morisaka H, Sano K, Araki T. Multi-organ perfusion CT in the abdomen using a 320-detector row CT scanner: Preliminary results of perfusion changes in the liver, spleen, and pancreas of cirrhotic patients. European Journal of Radiology 2012;81(10):2533–7.
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Perfusion CT can predict tumoral grading of pancreatic adenocarcinoma M. D’Onofrio a,∗ , A. Gallotti a , W. Mantovani b , S. Crosara a , E. Manfrin c , M. Falconi d , A. Ventriglia a , G.A. Zamboni a , R. Manfredi a , R. Pozzi Mucelli a a
Department of Radiology, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy Department of Medicine and Public Health, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy c Department of Pathology, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy d Department of Surgery, University Hospital G.B. Rossi Piazzale L.A. Scuro 10, 37134 University of Verona, Verona, Italy b
a r t i c l e
i n f o
Article history: Received 12 May 2012 Received in revised form 26 September 2012 Accepted 29 September 2012 Keywords: Perfusion CT Pancreatic adenocarcinoma Abdominal radiology Tumor grading
a b s t r a c t Objectives: To describe perfusion CT features of locally advanced pancreatic ductal adenocarcinoma and to evaluate correlation with tumor grading. Methods: Thirty-two patients with locally advanced pancreatic adenocarcinoma were included in this study. Lesions were evaluated by P-CT and biopsy after patient’s informed consent. P-CT parameters have been assessed on a large single and on 6 small intratumoral ROIs. Values obtained have been compared and related to the tumor grading using Mann–Whitney U test. Sensibility, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy in predicting tumor grading have been calculated for cut-off values chosen by using ROC curves. Results: Out of 32 lesions, 12 were classified as low grade and 20 as high grade. A statistically significant difference between high and low grade neoplasms were demonstrated for PEI and BV parameters. PEI and BV cut-off values were respectively 17.8 HU and 14.8 ml/100 g. PEI identified high grade neoplasms with a 65% sensitivity, 92% specificity, 93% PPV, 61% NPV and 75% accuracy. BV identified high grade neoplasms with a 80% sensitivity, 75% specificity, 84% PPV, 69% NPV, 78% accuracy. Considering both PEI and BV, P-CT identified high grade lesions with a 60% sensitivity, 100% specificity, 100% PPV, 60% NPV and 75% accuracy. Conclusions: PEI and BV perfusion CT parameters proved their efficiency in identifying high grade pancreatic adenocarcinoma. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Pancreatic ductal adenocarcinoma (ADK) is the most common primary malignancy of the pancreas and is associated with a very poor prognosis, being worldwide one of the leading cause of cancerrelated death [1,2]. Patients with pancreatic ADK have an overall 1-year and 5-year survival rate after diagnosis below 20% and below 5%, respectively [3]. Surgical resection still remains the only potentially curative treatment [3–5]. However, only 5–25% of patients with pancreatic cancer are candidates for radical resection due to the difficulty in achieving early diagnosis [6,7]. About 40–50% of patients have metastases and approximately 40% of patients have locally advanced disease at the time of diagnosis [3,8]. Moreover, despite the fact that a combined surgical-systemic approach may offer a longer life expectancy, the long-term outcome remains dis-
∗ Corresponding author. Tel.: +39 045 8124140; fax: +39 045 8277808. E-mail address: [email protected] (M. D’Onofrio). 0720-048X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2012.09.023
mal, with a mortality of 1–5% and an overall 5-year survival rate after pancreaticoduodenectomy between 15% and 20% and after distal pancreatectomy between 8% and 15% [4,5,9,10]. Some clinico-pathological features could be advocated as risk factors for an unfavorable prognosis, such as age, gender, serum CA 19.9 level, duration of disease-related symptoms, tumor size and grade, lymph node involvement, perineural and retroperitoneal invasion [4,11]. The high incidence of recurrence after surgery is also independently related to the presence of positive transection margins [4,5,10]. The early recurrence of pancreatic adenocarcinoma after resection characterizes the patient population reported in literature as “early death” (ED). In this patient population the resection can be considered unnecessary. The preoperative correct identification of this group of patients is very important to minimize unnecessary resections but remains difficult owing to the post-operative assessment of some factors such as tumor resection margins and grading [4]. Molecular and immunohistochemical approaches have been used to better explain the inherited biological aggressiveness of
228
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
human cancers and to suggest the development of new treatment strategies [12–14]. The tumor growth, grading and progression and its metastatic spread have been shown strictly dependent on tumor angiogenesis. Recently, quantitative imaging techniques based on functional and targeted information have been developed [15–17]. Perfusion CT (P-CT) is a new imaging technique able to provide qualitative and quantitative information on perfusion parameters of tissues, which have been demonstrated to be correlated with histological markers of angiogenesis. The excellent linear relationship between tissue attenuation and iodinated contrast agent concentration allows an objective quantification of the perfusion parameters correlated with the hemodynamic changes caused by new vessels [12,18,19]. Accurate imaging detection, characterization and staging of pancreatic cancer are of paramount importance to provide patients an adequate treatment strategy. Functional imaging from P-CT may add useful information on tumor aggressiveness affecting treatment strategy and patient management. To our knowledge, there have been no previous reports in the literature in which the P-CT parameters of pancreatic ADK and tumor grading were compared. The purpose of this study was to prospectively describe the perfusion CT features of locally advanced pancreatic ductal
adenocarcinomas and to assess whether these features correlate with the tumor grading at pathology. 2. Material and methods 2.1. Patients Institutional review board was obtained for this prospective study, conducted in accordance with the principles of Helsinki Declaration. Thirty-two patients (17 males, 15 female; mean age 66.2 years; range, 43–83 years) were included in the study population (Table 1). All patients were required to provide written informed consent before to be involved in this study. Between December 2010 and June 2011, all patients afferent to our Institution with a suspicion of locally advanced pancreatic cancer were recruited. Locally advanced pancreatic cancer was considered as non resectable tumor in absence of any distant metastasis [3]. The inclusion criterion was the presence of locally advanced pancreatic ductal adenocarcinoma in patients who did not undergo any previous neoadjuvant or palliative therapies. All pancreatic lesions were pathologically confirmed by fine needle biopsy performed the same day immediately after the P-CT. Patients were excluded
Table 1 Patients, tumors and perfusion CT parameters.
I
II
Gender Female Male Ageb Site Right pancreas Body-tail Size (mm)b PFb PF1 PF2 PF3 PF4 PF5 PF6 Mean PF 1–6 TTPb TTP1 TTP2 TTP3 TTP4 TTP5 TTP6 Mean TTP 1–6 PEIb PEI1 PEI2 PEI3 PEI4 PEI5 PEI6 Mean PEI 1–6 BVb BV1 BV2 BV3 BV4 BV5 BV6 Mean BV 1–6
High grade (n = 20)
Low grade (n = 12)
9 11 65 (43; 83)
6 6 69 (46; 79)
11 9 4.4 (2.5; 7)
7 5 4.2 (3; 5.2)
5.9 (1.5; 75.1) 7.9 (1.1; 67.9) 4.5 (0.7; 72.5) 5.7 (2.3; 66.7) 8.5 (1.3; 98.4) 6 (1.3; 119.1) 8.4 (2.1; 62.7) 11.4 (2.6; 48.6) 70 (6; 101) 59.5 (5; 101) 70 (11; 101) 70 (10; 101) 70 (5; 101) 70 (5; 101) 60 (10; 101) 65 (25.8; 101) 16.2 (7; 40.9) 15.7 (4; 43.7) 14.3 (1.3; 37.3) 18.4 (1.9; 52.9) 19.3 (4.4; 44.2) 14.8 (4.6; 44.2) 16.1 (8.6; 37.8) 17.7 (5; 42.8) 11.3 (1.3; 37.3) 9.5 (0; 39.7) 7.7 (0; 33.8) 11 (0; 35.9) 9.7 (1.7; 37.5) 8.2 (0; 37.5) 9.1 (0.4; 28.8) 9.3 (0.4; 34.1)
8.9 (3.8; 22.9) 10.4 (4.6; 28.1) 11.3 (2.1; 29.4) 9.1 (1.9; 27.4) 8.3 (3.9; 19.5) 11.4 (4.8; 30.2) 13.2 (6.2; 53.4) 10.9 (5.7; 26.6) 82.5 (30; 100) 85 (30; 100) 70 (30; 100) 70 (30; 100) 85 (30; 100) 70 (25; 100) 70 (25; 100) 80.6 (30; 95) 26.3 (15.8; 47) 28.1 (13.8; 56.3) 25.3 (11.1; 70) 27.2 (3.4; 55) 27.8 (10; 59.4) 28.4 (8.3; 64.6) 32.3 (22; 58) 28.2 (14.7; 54.1) 19 (9; 47.2) 22.3 (12; 48.7) 18 (1.3; 60.3) 17.1 (2.9; 49.7) 18.9 (4.7; 53.1) 19.4 (3.5; 55.8) 23.6 (10.1; 57.1) 17.2 (7.8; 45.8)
p valuea 0.784c 0.239 0.854c 0.272 0.387 0.408 0.099 0.408 0.803 0.182 0.239 0.924 0.716 0.107 0.893 0.833 0.272 0.659 0.387 0.099 0.012 0.003 0.008 0.099 0.099 0.004 <0.0001 0.005 0.004 <0.0001 0.001 0.289 0.053 0.003 <0.0001 0.001
Total (n = 32) 15 17 66 (43; 83) 18 14 4.3 (2.5; 7) 8.2 (1.5; 75.1) 9.4 (1.1; 67.9) 7.6 (0.7; 72.5) 7.1 (1.9; 66.7) 8.5 (1.3; 98.4) 8 (1.3; 119.1) 11.4 (2.1; 62.7) 11 (2.6; 48.6) 70 (6; 101) 70 (5; 101) 70 (11; 101) 70 (10; 101) 70 (5; 101) 70 (5; 101) 70 (10; 101) 69.7 (25.8; 101) 19.9 (7; 47) 21.7 (4; 56.3) 21 (1.3; 70) 21.1 (1.9; 55) 24.7 (4.4; 59.4) 21.1 (4.6; 64.6) 21.6 (8.6; 58) 19.9 (5; 54.1) 13.4 (1.3; 47.2) 12.5 (0; 48.7) 11.8 (0; 60.3) 11.4 (0; 49.7) 12.3 (1.7; 53.1) 14.1 (0; 55.8) 11 (0.4; 57.1) 11.7 (0.4; 45.8)
In the 1st section of this table data regarding patients and tumors divided into high and low grade of malignancy are shown. In the 2nd section this table contains median values of P-CT parameters PF, TTP, PEI and BV measured on large single ROI and on the six small ROIs obtained for high and low grade of malignancy. a Mann–Whitney U test. b Median (range). c Fisher exact test.
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
in presence of resectable tumors at imaging examinations, if distant metastases were observed or if they had a cancer other than pancreatic adenocarcinoma. 2.2. Perfusion CT technique All perfusion CT examinations were performed at our Institution on a 64-row Multidetector CT scanner (Brilliance 64; Philips, Best, The Netherlands). After a minimum fast of 6 h, P-CT was performed with patients in supine position with their arms behind their head, during superficial and regular free-breath. An abdominal belt (Philips Medical Systems, Best, The Netherlands) was adjusted to reduce the artifacts caused by respiratory motion. An initial unenhanced wide coverage scan was acquired with a section thickness of 5 mm to correctly localize the tumor and select the volume on which perfusion scans had to be performed. The spatial coordinates were noted and used to plan the subsequent dynamic examination achieved with cine table position. P-CT consists of the acquisition of repeated CT scan of the volume being analyzed over time, during and after the administration of a bolus of iodinated contrast agent. High iodine concentration (370 mg/ml) contrast agents are preferred since there is a linear relationship between iodine concentration and tissue enhancement [16,17]. A total of 50 ml of contrast agent at an injection rate of 5 ml/s was administered according to the “short sharp bolus” technique through an antecubital vein, and followed by 40 ml of saline solution at the same flow rate [16]. CT scanning parameters included 120 kVp (for a slope method based scanning protocol, which is very sensitive to the image noise, a higher voltage has to be used than for a deconvolution method), tube current of 150 mAs, gantry rotation time of 0.5 s, increment 5 mm, 5-mm slice thickness and 100 mm in length. 64 slice CT scanners allow an axial coverage of about 40 mm. Using toggling table technique during dynamic images acquisition, axial coverage can be extended to 100 mm [19]. Scanning was initiated 12 s after starting contrast material injection and images were acquired at every 5 s for 40 s, and later at 50, 70, 90 and 120 s, for a total of 200 images. 2.3. Imaging and data analysis Image data were processed on a workstation (Extended Brilliance Workspace V4.5.2.40007, Philips, Best, The Netherlands) loaded with commercial perfusion CT software (CT Perfusion Software, Philips, Best, The Netherlands) based on slope method. Perfusion parameters for analyzing images acquired during the first pass were calculated with the use of dedicated semiautomatic software. An abdominal radiologist with 10 years of experience in pancreatic imaging, aware of the purpose of this study and kept blinded to the histopathological analysis, reviewed and interpreted all images. Tumor location and size were previously defined in each patient. The largest dimension was measured on axial scans and mean diameter calculated. The arterial input was measured by manually drawing a circular region of interest (ROI) within the abdominal aorta on a selected image, being careful to avoid any mural calcification. The arterial time/density curve and the end of the first pass phase were derived automatically by the software. To obtain P-CT parameters a single ROI was manually drawn within the tumor on the axial slice in which the tumor showed the maximum diameter being careful not to include the surrounding adipose tissue and adjacent normal vascular structures to avoid partial volume effect. Additionally, six smaller ROIs were manually drawn within the tumor, of which two in the central portion (called 1 and 2) and four in the periphery (called 3, 4, 5, 6). Qualitative information achieved by perfusion CT consists of analysis of color maps automatically generated by the software, able to provide an
229
immediate and panoramic representation of perfusion distribution within the tumor. Quantitative information consists of the analysis of P-CT parameters. The final numeric values expressed for each ROI are the mean values automatically calculated by the system from all measurements performed in each voxel included in an ROI. The following parameters have been calculated: • Perfusion (or blood flow, PF), expressed in ml/100 g of tissue/min, which consists of the ratio between the maximum slope of the time/density curve of tissue and the peak density reached by the aorta, selected as arterial input. • Time to Peak (TTP), expressed in seconds, which indicates the time interval between the arrival of contrast agent in the aorta (selected as arterial input) and the peak density in the tissue. It is a marker of perfusion pressure. • Peak Enhancement Intensity (PEI), expressed in Hounsfield Units, which indicates the peak attenuation reached by the tissue after contrast media injection. It is a marker of tissue blood volume. • Blood volume (BV), expressed in ml/100 g of tissue, which reflects the volume of blood flowing within all vessels in the tissue. It is a marker of tumor vascularity. 2.4. Pathological diagnosis In our Institution all patients with locally advanced pancreatic cancer usually undergo percutaneous ultrasound-guided fine needle biopsy to achieve a pathological confirmation prior to adjuvant therapies. 2.4.1. Biopsy procedure All patients included in the study population underwent biopsy under ultrasound guided percutaneous approach in the same day immediately after P-CT examination. Local anesthesia (lidocaine 1%) was administered in the abdominal wall at the chosen entry point. Multifrequency probes (ranging from 2.5 to 5 MHz) with lateral support were used during the aspiration biopsy, performed with fine needles (Menghini-type needles, 20–21-G caliper). Once the pancreatic tumor had been reached, the needle was gently inserted and moved with very small excursions. The unstructured material collected in the needle cavity was then placed on a glass slide. An adequate sample must contain no or little blood cells, which could interfere with analysis. To allow an optimal reading, the material was smeared into a thin layer and fixed with 95% alcohol to avoid cellular lyses and morphological artifacts. Immediate fixation and a modified Papanicolaou staining allowed to evaluate in real time the appropriateness of the sample. 2.4.2. Histopathological analysis Two independent pathologists respectively with 15 and 5 years of experience in pancreatic cytology, aware of the purpose of this study and kept blinded to the P-CT analysis, examined the glass slides randomly to make the grades as objective as possible. At low magnification the cellular density of the samples was evaluated followed by a specific characterization of both the cellular background and the diagnostic elements at higher magnification. The presence of confluent pathological cells and morphological features of nuclei and cytoplasm were accurately described. After having ensured a diagnosis of pancreatic ductal adenocarcinoma in all patient-population, tumors were classified in two different groups according to their grading, as follows: low grade, was defined as well- or moderate-differentiated lesions with low cytological atypia; high grade, was defined as poorly differentiated lesions with several cytological atypia.
230
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
2.5. Statistical analysis Distribution of continuous variables was reported as mean (or median in event of abnormal distribution) and range (minimum–maximum values). Categorical variables were presented as numbers and percentages. The comparison between the low grade and the high grade tumor groups was carried out using Student’s t test or the Mann–Whitney U test for continuous variables. Categorical data were compared by using the chi-square test and the Fisher exact test when necessary. Diagnostic accuracy was assessed as the area under the receiver-operating characteristic (ROC) plot. Diagnosis obtained by pathologic examination was the external gold standard. Sensitivity, specificity, positive and negative predictive values (and their 95% Confidence Interval) for each significant P-CT parameter in the prediction of tumor grade were also calculated by using cut-off values chosen on the basis of ROC curves. A p value less than or equal to 0.05 was considered statistically significant. 3. Results Perfusion CT examination and biopsy under ultrasound guided percutaneous approach were possible in all cases. 3.1. Pathology Of the 32 pancreatic ductal adenocarcinomas (mean diameter, 45.9 mm; range, 30–70 mm), 6 (18.8%) lesions were located in the head, 7 (21.8%) in the uncinate process, 5 (15.7%) in the neck, 14 (43.7%) in the body-tail of the gland. According to the pathological analysis, 12 (37.5%) were classified as low grade ADK and 20 (62.5%) as high grade (Table 1). No significant difference was observed between the mean diameter of low grade and high grade tumors (mean size, 42.6 mm; range, 30–52 mm; vs mean size, 46.4 mm; range, 25–70 mm; p = 0.272). 3.2. Perfusion CT parameters Median Perfusion (PF) value resulting from the measurements performed on the large ROI was 5.9 ml/100 g of tissue/min (range 1.5–75.1) in reference to high grade neoplasms and 8.9 ml/100 g of tissue/min (range 3.8–22.9) in reference to low grade neoplasms; median PF value resulting from the mean value of the measurements performed on the 6 small ROIs was 6.8 ml/100 g of tissue/min (range 2.6–48.6) in reference to high grade neoplasms and 10.9 ml/100 g of tissue/min (range 5.7–26.6) in reference to low grade neoplasms. There was no statistically significant difference between high grade and low grade neoplasms in terms of median PF values (Table 1) resulting both from
measurements performed on the large ROI (p = 0.387) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.924). Median Time To Peak (TTP) value resulting from the measurements performed on the large ROI was 70 s (range 6–101) in reference to high grade neoplasms and 82.5 s (range 30–100) in reference to low grade neoplasms; median TTP value resulting from the mean value of the measurements performed on the 6 small ROIs was 65 s (range 25.8–101) in reference to high grade neoplasms and 80.6 s (range 30–95) in reference to low grade neoplasms. There was no statistically significant difference between high grade and low grade neoplasms in terms of median TTP values (Table 1) resulting both from measurements performed on the large ROI (p = 0.716) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.099). Median Peak Enhancement Intensity (PEI) value resulting from the measurements performed on the large ROI was 16.2 HU (range 7–40.9) in reference to high grade neoplasms and 26.3 HU (range 15.8–47) in reference to low grade neoplasms; median PEI value resulting from the mean value of the measurements performed on the 6 small ROIs was 17.7 HU (range 5–42.8) in reference to high grade neoplasms and 28.2 HU (range 14.7–54.1) in reference to low grade neoplasms. There was statistically significant difference at Mann–Whitney U-test between high grade and low grade neoplasms in terms of median PEI values (Table 1) resulting both from measurements performed on the large ROI (p = 0.012) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.005). Median Blood Volume (BV) value resulting from the measurements performed on the large ROI was 11.3 ml/100 g of tissue (range 1.3–37.3) in reference to high grade neoplasms and 19 ml/100 g of tissue (range 9–47.2) in reference to low grade neoplasms; median BV value resulting from the mean value of the measurements performed on the 6 small ROIs was 9.3 ml/100 g of tissue (range 0.4–34.1) in reference to high grade neoplasms and 17.2 ml/100 g of tissue (range 7.8–45.8) in reference to low grade neoplasms. There was statistically significant difference at Mann–Whitney U-test between high grade (Fig. 1) and low grade (Fig. 2) neoplasms in terms of median BV values (Table 1) resulting both from measurements performed on the large ROI (p = 0.004) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.001). 3.3. Perfusion CT – statistical analysis Given these P-CT results, only PEI and BV parameters measured on the large ROI were taken into account, since the measurements performed on the large ROI and the mean value of the measurements performed on the 6 small ROIs have been proved to be statistically significant. The choice of the large ROI has been
Fig. 1. CT axial scan (left) showing a hypodense mass of the pancreatic body-tail (arrow). BV color-map (right) obtained during same scan, shows hypoperfusion of the lesion (arrow).
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
231
Fig. 2. CT axial scan (left) showing a hypodense mass of the head-uncinate process of the pancreas (arrow). BV color-map (right) obtained during same scan shows well perfused lesion (arrow).
Table 2 Cut-off values for PEI perfusional CT parameter. Criterion
Sensitivity
95% CI
Specificity
95% CI
<7 <=14.9 <=15.8 <=17.8* <=18.7 <=21 <=23.1 <=23.5 <=27.9 <=29.5 <=30.3 <=31.7 <=40.9 <=47
0.00 45.00 50.00 65.00 65.00 70.00 70.00 75.00 75.00 85.00 85.00 90.00 100.00 100.00
0.0–16.8 23.1–68.5 27.2–72.8 40.8–84.6 40.8–84.6 45.7–88.1 45.7–88.1 50.9–91.3 50.9–91.3 62.1–96.8 62.1–96.8 68.3–98.8 83.2–100.0 83.2–100.0
100.00 100.00 91.67 91.67 75.00 75.00 66.67 66.67 33.33 33.33 25.00 16.67 16.67 0.00
73.5–100.0 73.5–100.0 61.5–99.8 61.5–99.8 42.8–94.5 42.8–94.5 34.9–90.1 34.9–90.1 9.9–65.1 9.9–65.1 5.5–57.2 2.1–48.4 2.1–48.4 0.0–26.5
Using ROC curves cut-off value for PEI P-CT parameter with the best performance in terms of sensitivity and specificity in the characterization of high grade and low grade neoplasms. Fig. 3. ROC curve for PEI parameter. Area under the curve (AUC) is 0.767.
suggested by the advantages in terms of time consuming in data analysis. ROC curve for PEI parameter in shown in Fig. 3 with an area under the curve (AUC) of 0.767. ROC curve for BV parameter in shown in Fig. 4 with an area under the curve (AUC) of 0.798. Using these ROC curves, the cut-off values for PEI and BV perfusion parameters with the best performance in terms of sensitivity
and specificity in the characterization of high grade or low grade neoplasms have been chosen. PEI cut-off value has been set at 17.8 HU with sensitivity and specificity of 65% and 91.67% respectively (Table 2). BV cut-off value has been set at 14.8 ml/100 g with sensitivity and specificity of 80% and 75% respectively (Table 3). PEI perfusion parameter with a cut-off of 17.8 HU can identify pancreatic high grade adenocarcinomas with a 65% sensitivity (95% CI = 0.441–0.859), a 91.7% specificity (95% CI = 0.760–1.073), a positive predictive value of 92.9%, a negative predictive value of
Table 3 Cut-off values for BV perfusional CT parameter.
Fig. 4. ROC curve for BV parameter. Area under the curve (AUC) is 0.798.
Criterion
Sensitivity
95% CI
Specificity
95% CI
< 1.3 <=8.5 <=9 <=10 <=10.4 <=12.3 <=14.5 <=14.8* <=15.7 <=18.5 <=19.2 <=26.5 <=28.3 <=37.3 <=47.2
0 35 35 40 45 70 70 80 80 85 85 90 90 100 100
0.0–16.8 15.4–59.2 15.4–59.2 19.1–63.9 23.1–68.5 45.7–88.1 45.7–88.1 56.3–94.3 56.3–94.3 62.1–96.8 62.1–96.8 68.3–98.8 68.3–98.8 83.2–100.0 83.2–100.0
100 100 91.67 91.67 83.33 83.33 75 75 58.33 58.33 41.67 41.67 25 25 0
73.5–100.0 73.5–100.0 61.5–99.8 61.5–99.8 51.6–97.9 51.6–97.9 42.8–94.5 42.8–94.5 27.7–84.8 27.7–84.8 15.2–72.3 15.2–72.3 5.5–57.2 5.5–57.2 0.0–26.5
Using ROC curves cut-off value for BV P-CT parameter with the best performance in terms of sensitivity and specificity in the characterization of high grade and low grade neoplasms.
232
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233
61.1% and a 75% accuracy. BV perfusion parameter with a cut-off of 14.8 ml/100 g can identify pancreatic high grade adenocarcinomas with a 80% sensitivity (95% CI = 0.625–0.975), a 75% specificity (95% CI = 0.505–0.995), a positive predictive value of 84.3%, a negative predictive value of 69.2% and a 78.1% accuracy. Considering both PEI and BV perfusion parameter with the related cut-off values, it is possible to characterize pancreatic ductal high grade adenocarcinomas with a 60% sensitivity (95% CI = 0.385–0.815), a 100% specificity, a positive predictive value of 100%, a negative predictive value of 60% and a 75% accuracy.
4. Discussion Perfusion CT (P-CT) is an imaging technique able to quantify the tissue enhancement after contrast material administration. Main P-CT fields of application are Neuroimaging (in diagnosing stroke and choosing its proper treatment) and Oncologic Imaging in terms of tumor assessment, characterization and staging, angiogenesis assessment, response to therapy prediction and response to therapy monitoring [20]. Many other applications are however emerging also in non oncologic fields such the assessment of changes in organ (liver, pancreas and spleen) perfusion in cirrhotic patients [21]. Regarding the CT perfusion technique applied in this study, although patient should not move at all during image acquisition, the movement provides less impact on image quality by using CT scanners based on slope method than other processing methods as reported in literature [17]. A shallow breathing during images acquisition is considered acceptable and patients must be instructed not to breathe deeply when feeling the hot flush due to the contrast agent. Pancreatic ductal adenocarcinoma is the most common pancreatic primary malignancy accounting for 80% of pancreatic malignancies. The main aims of imaging consist of tumor detection and characterization, playing an important role in disease management. CT study represents the modality of choice in diagnosis confirmation, providing accurate staging. In more than 95% of the patients pancreatic ductal adenocarcinoma is diagnosed at an advanced stage, in presence of locally advanced disease (vascular or perineural invasion) or distant metastases. Prognosis and therapeutic approaches depend on lesion resectability at the moment of clinical presentation. As reported by Barugola et al., up to 30% of resected patients die of disease within 1 year after surgery [4]. In this subgroup recurrence is very early, and survival is similar to that observed in patients with advanced disease undergoing antitumoral therapies alone. These so-called “early deaths” (ED) after resection for curative intent may be attributed either to inadequate pre- or intra-operative staging or to a particularly aggressive behavior of the disease. Thus, the ability to identify patients at high risk of ED before surgery is an important clinical goal to circumvent a demanding surgical intervention. According to some authors symptoms lasting >40 days, CA 19–9 >200 U/ml and G2 pathological grading are easy-to-obtain preoperative parameters in order to identify patients with a disease not suitable for front-up surgery, even if deemed resectable by high quality imaging [4,11]. According to Barugola et al. [4] those patients with level of CA 19–9 >200 U/ml and symptoms lasting >40 days should undergo fine needle biopsy: if pathology report shows a poorly differentiated or anaplastic tumor, a non-surgical therapeutical approach should be carried out. After neoadjuvant therapy (or, if necessary, palliative treatment), in case of disease stability or down-staging, patients could undergo surgical intervention. The present study aims to demonstrate P-CT ability in direct and preoperative characterization of high grade pancreatic ductal
adenocarcinomas in order to implement the identification of patients at high risk of ED, to provide each patient with the best treatment strategy. Thirty-two patients with locally advanced pancreatic ductal adenocarcinoma underwent P-CT imaging study and percutaneous US-guided fine needle biopsy. A statistically significant difference between high grade and low grade neoplasms was found in terms of median PEI and BV values resulting both from measurements performed on the large ROI (p = 0.012 for PEI and p = 0.004 for BV) and from the mean value of the measurements performed on the 6 small ROIs (p = 0.005 for PEI and p = 0.001 for BV). Using ROC curves the cut-off values for PEI and BV respectively with the best performance in terms of sensitivity and specificity in the characterization of high grade tumors have been chosen. PEI perfusional parameter with a cut-off of 17.8 HU can identify pancreatic high grade adenocarcinomas with a 75% accuracy. BV perfusional parameter with a cut-off of 14.8 ml/100 g can identify pancreatic high grade adenocarcinomas with a 78.1% accuracy. Considering both PEI and BV perfusional parameter with the related cut-off values, it is possible to characterize pancreatic ductal high grade adenocarcinomas with a 60% sensitivity, a 100% specificity, a positive predictive value of 100%, a negative predictive value of 60% and a 75% accuracy. According to our results, P-CT offers a significant contribution both to clinical and laboratory data in identifying high grade pancreatic ductal adenocarcinoma, and consequently patients at high risk of ED, more accurately than the suspicion given by duration of symptoms and CA 19–9 levels. In everyday clinical practice patients with pancreatic ductal adenocarcinoma classified as high grade at P-CT should undergo biopsy to achieve a grading confirmation even if resectable at imaging. Actually, standard of care for pancreatic lesions typical for ductal adenocarcinoma deemed resectable at imaging is surgical intervention, without pathological confirmation. Considering its accuracy values, P-CT could adequately select all patients with a high grade lesion and justify fine needle biopsy to confirm the grade itself, even in presence of a resectable disease at imaging. The limitations of this study are the relative small number of patients included and the analysis of a fine needle biopsy sample instead of the whole surgical specimen. Other prospective studies about resectable pancreatic lesions based on a larger population are needed. 5. Conclusion Perfusion CT can predict tumor grade of pancreatic adenocarcinoma. In particular, PEI and BV perfusion parameters proved their efficiency in identifying high grade pancreatic adenocarcinoma. References [1] Schima W, Ba-Ssalamah A, Kolblinger C, Kulinna-Cosentini C, Puespoek A, Gotzinger P. Pancreatic adenocarcinoma. European Radiology 2007;17(3):638–49. [2] D’Onofrio M, Gallotti A, Pozzi Mucelli R. Imaging techniques in pancreatic tumors. Expert Review of Medical Devices 2010;7(2):257–73. [3] Willett CG, Czito BG, Bendell JC, Ryan DP. Locally advanced pancreatic cancer. Journal of Clinical Oncology 2005;23(20):4538–44. [4] Barugola G, Partelli S, Marcucci S, et al. Resectable pancreatic cancer: who really benefits from resection? Annals of Surgical Oncology 2009;16(12):3316–22. [5] Barugola G, Partelli S, Crippa S, et al. Outcomes after resection of locally advanced or borderline resectable pancreatic cancer after neoadjuvant therapy. American Journal of Surgery 2012;203(2):132–9. [6] Zuckerman DS, Ryan DP. Adjuvant therapy for pancreatic cancer: a review. Cancer 2008;112(2):243–9. [7] Han SS, Jang JY, Kim SW, Kim WH, Lee KU, Park YH. Analysis of longterm survivors after surgical resection for pancreatic cancer. Pancreas 2006;32(3):271–5. [8] Czito BG, Willett CG, Clark JW, Fernandez Del Castillo C. Current perspectives on locally advanced pancreatic cancer. Oncology 2000;14(11):1535–45, discussion 46, 49–52.
M. D’Onofrio et al. / European Journal of Radiology 82 (2013) 227–233 [9] Picozzi VJ, Pisters PW, Vickers SM, Strasberg SM. Strength of the evidence: adjuvant therapy for resected pancreatic cancer. Journal of Gastrointestinal Surgery 2008;12(4):657–61. [10] Barugola G, Falconi M, Bettini R, et al. The determinant factors of recurrence following resection for ductal pancreatic cancer. Journal of the Pancreas 2007;8(Suppl. 1):132–40. [11] Yeo CJ, Cameron JL. Pancreatic cancer. Current Problems in Surgery 1999;36(2):59–152. [12] Park MS, Klotz E, Kim MJ, et al. Perfusion CT: noninvasive surrogate marker for stratification of pancreatic cancer response to concurrent chemo- and radiation therapy. Radiology 2009;250(1):110–7. [13] Kuwahara K, Sasaki T, Kuwada Y, Murakami M, Yamasaki S, Chayama K. Expressions of angiogenic factors in pancreatic ductal carcinoma: a correlative study with clinicopathologic parameters and patient survival. Pancreas 2003;26(4):344–9. [14] Hattori Y, Gabata T, Zen Y, Mochizuki K, Kitagawa H, Matsui O. Poorly enhanced areas of pancreatic adenocarcinomas on late-phase dynamic computed tomography: comparison with pathological findings. Pancreas 2010;39(8):1263–70. [15] Wang ZQ, Li JS, Lu GM, Zhang XH, Chen ZQ, Meng K. Correlation of CT enhancement, tumor angiogenesis and pathologic grading
[16]
[17] [18]
[19] [20] [21]
233
of pancreatic carcinoma. World Journal of Gastroenterology 2003;9(9): 2100–4. Bellomi M, Viotti S, Preda L, D’Andrea G, Bonello L, Petralia G. Perfusion CT in solid body-tumours, Part II: clinical applications and future development. Radiologia Medica 2010;115(6):858–74. Kambadakone AR, Sahani DV. Body perfusion CT: technique, clinical applications, and advances. Radiologic Clinics of North America 2009;47(1):161–78. Kambadakone AR, Sharma A, Catalano OA, Hahn PF, Sahani DV. Protocol modifications for CT perfusion (CTp) examinations of abdomen-pelvic tumors: impact on radiation dose and data processing time. European Radiology 2011;21(6):1293–300. Miles KA. Perfusion CT for the assessment of tumour vascularity: which protocol? British Journal of Radiology 2003;76(Spec no. 1):S36–42. Miles KA. Perfusion imaging with computed tomography: brain and beyond. European Radiology 2006;16(Suppl. 7):M37–43. Motosugi U, Ichikawa T, Sou H, Morisaka H, Sano K, Araki T. Multi-organ perfusion CT in the abdomen using a 320-detector row CT scanner: Preliminary results of perfusion changes in the liver, spleen, and pancreas of cirrhotic patients. European Journal of Radiology 2012;81(10):2533–7.