Increase diagnostic accuracy in differentiating focal type autoimmune pancreatitis from pancreatic cancer with combined serum IgG4 and CA19-9 levels

Pancreatology 14 (2014) 366e372

Contents lists available at ScienceDirect

Pancreatology journal homepage: www.elsevier.com/locate/pan

Original article

Increase diagnostic accuracy in differentiating focal type autoimmune pancreatitis from pancreatic cancer with combined serum IgG4 and CA19-9 levels Ming-Chu Chang a, Po-Chin Liang b, Shiow Jan c, Ching-Yao Yang d, Yu-Wen Tien d, Shu-Chen Wei a, Jau-Min Wong a, Yu-Ting Chang a, * a

Department of Internal Medicine, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan Department of Medical Imaging, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan Department of Laboratory Medicine, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan d Department of Surgery, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan b c

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 25 July 2014

Background/objectives: To distinguish autoimmune pancreatitis (AIP), especially focal type, from pancreatic cancer, is a greatest challenge for clinician. The aim of the study is to compare the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of combined serum IgG4 and CA19-9 levels to differentiate AIP from pancreatic cancer by HISORt, Asian and international consensus diagnostic criteria. Methods: We measured serum IgG4, CEA, and CA19-9 levels in 188 AIP patients, 86 non-AIP chronic pancreatitis patients, and 130 pancreatic cancer patients. The sensitivity, specificity, positive predictive value, negative predictive value and accuracy were compared with different diagnostic criteria. We also compared the diagnostic performance in patients with or without jaundice. Results: The serum level of IgG4 was significantly higher in AIP than those in non-AIP chronic pancreatitis and pancreatic cancer. The optimal cutoffs of IgG4 and CA19-9 to differentiate AIP from pancreatic cancer were 175 mg/dL and 85.0 U/ml based on ROC analysis. Combining IgG4 level over 280 mg/dL and CA19-9 below 85.0 U/ml could yield a best diagnostic accuracy (85.6%) to distinguishing AIP from pancreatic cancer in all of the HISORt, Asian and international consensus diagnostic criteria. With the combination of serological test, focal type AIP could be diagnosed with comparable accuracy as diffuse type AIP. Conclusions: Our study demonstrated that combined use of serum IgG4 (over 280 mg/dL) and CA19-9 9 (below 85.0 U/ml) together increases the diagnostic accuracy to distinguish AIP from pancreatic cancer non-invasively, especially in focal type autoimmune pancreatitis. Copyright © 2014, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.

Keywords: IgG4 CA19-9 Autoimmune pancreatitis Focal type Pancreatic cancer Diagnosis

Introduction Autoimmune pancreatitis (AIP) is a unique type of chronic pancreatitis characterized by elevated serum immunoglobulin G4 (IgG4), swelling of pancreas, irregular narrowing of main pancreatic duct, histological evidence of lymphoplasmacytic inflammation,

* Corresponding author. Department of Internal Medicine, National Taiwan University Hospital, College of Medicine, National Taiwan University, No. 7 Chung Shan South Road, Taipei, Taiwan. Tel.: þ886 2 23123456x63563; fax: þ886 2 23633658. E-mail address: [email protected] (Y.-T. Chang).

and a good response to steroid therapy [1]. Although some advance has been made in the diagnosis and treatment of AIP over the past years [2], the diagnosis of AIP is still a great clinical challenge, especially in the differential diagnosis from pancreatic cancer [3e5]. Hamano et al. firstly reported that elevation in serum IgG4 was a characteristic feature of AIP [6]. The elevated serum IgG4 level with different cut-offs has been used to constitute one of the items of different diagnostic criteria of AIP. In 2002, the Japan Pancreas Society proposed the diagnostic criteria for AIP based on imaging, serology, and histology. At that time, the serological criteria included elevated g-globulin, immunoglobulin G (IgG), and auto-

http://dx.doi.org/10.1016/j.pan.2014.07.010 1424-3903/Copyright © 2014, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.

M.-C. Chang et al. / Pancreatology 14 (2014) 366e372

antibodies [7]. In 2006, the revised Japanese criteria were modified and added IgG4 to the serological criteria [8]. In 2008, the Asian diagnostic criteria proposed to use IgG, and auto-antibodies or IgG4 as the serological criteria but g-globulin was deleted [9]. In the United States, the HISORt criteria allow only serum IgG4 as the serological criterion [10,11]. In 2011, the international consensus diagnostic criteria (ICDC) were proposed to use serum IgG4 alone as the serological criterion for type I AIP. In the ICDC criteria, the serum IgG4 level was further classified into two levels according to cutoffs set at 140 and 280 mg/dL to diagnose the type I AIP [12]. However, focal or tumor forming AIP, different from typical diffuse type AIP, is difficult to distinguish from pancreatic cancer [13]. The cutoff of serum IgG4 levels in differentiating pancreatic cancer from AIP with atypical pancreatic imaging (especially the focal type enlargement) remains to be elucidated. Sumimoto K. et al. recently compared the diagnostic ability of 5 criteria, including ICDC, Korean, Asian, JPS-2011, and HISORt criteria, in 61 patients with AIP and 56 patients with pancreatic cancer [14]. They found that by serology, the accuracy, sensitivity, and specificity of level 1 serology by ICDC criterion were 75.3%, 63.6%, and 96.7%, respectively, while those of level 2 were 91.8%, 90.9%, and 93.3%, respectively [14]. However, their results were based on a small population and needed to be verified in larger and different populations. CA19-9, considered as a biomarker in pancreatic cancer, could also elevate in other pancreatic diseases such as chronic pancreatitis or other pathological status [15]. Till now, it still lacks a simple serological test to diagnose and to differentiate AIP from pancreatic cancer. The first aim of this study was to evaluate whether the combined measurement of serum IgG4 and CA19-9 could increase the diagnostic accuracy for AIP from pancreatic cancer in different diagnostic settings including HISORT, Asian, and ICDC criteria in serology aspect. The second aim of this study was to determine the optimal way to diagnose focal type AIP and differentiate it from pancreatic cancer by serology.

367

women) with cytological or/and pathologically confirmed adenocarcinoma of pancreas were enrolled, including 3 patients with stage 1, 10 patients with stage 2, 23 patients with stage 3, and 84 patients with stage 4 disease. There were 86 patients with nonautoimmune chronic pancreatitis (CP) (idiopathic CP; 53 men and 33 women) during the same period were enrolled. The diagnosis of non-autoimmune chronic pancreatitis was based on following criteria: typical histological findings; calcification of the pancreas in imaging or histological examination; recurrent pancreatitis with abnormal pancreatogram, or steatorrhea [17]. Serum IgG, IgG4, CA19-9, and CEA levels were measured in the three groups. Serum IgG4 was measured by particle-enhanced immunonephelometry (Nephelometer, Dade Behring, Germany). The institutional ethics committee approved this study. Informed consent was obtained from patients to analyze their data anonymously. All the patients' medical charts were reviewed and the patients' demographic data, including age, sex, serological studies, image studies, and clinical manifestations etc. were collected. Statistical analysis We calculated the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of serum IgG4 and CA19-9 levels for the diagnosis of AIP with different diagnostic criteria. The statistical calculations were carried out using SPSS 17 statistical software (SAS Institute, Cary, NC). The relationship between the serum IgG4 and CA19-9 level and various clinicopathologic characteristics was evaluated by using ANOVA or chi-square test for binary variables. All reported P value was 2-sided. Differences with a P value less than 0.05 were considered to be statistically significant. The optimal cut-off points for discriminating between AIP and pancreatic cancer were sought by constructing ROC curves, which were generated by calculating the sensitivities and specificities of IgG4 and CA 19-9 data at several predetermined cut-off points.

Methods Results Study participants Between Jan 1996 and Dec 2012, we consecutively collected 188 patients with AIP (95 men and 93 women) at National Taiwan University Hospital, a tertiary referred center for management of pancreatic diseases in Taiwan [16]. All the patients with AIP fulfilled the HISORt criteria (158/188, 84.0%), or Asian diagnostic criteria (162/188, 86.2%), or the ICDC criteria (168/188, 89.4%) for AIP (Table 1). A total of consecutive 130 patients (65 men and 65

Table 1 The distribution of diagnostic criteria for autoimmune pancreatitis (n ¼ 188). Diagnostic criteria

N

% in each criterion

Fit HISORt criterion Group A:based on diagnostic pancreatic histology Group B: based on typical imaging þ serology Group C: based on response to steroid Fit Asian criterion Based on Imaging þ serology Based on Imaging þ histopathology LPSP Based on Imaging þ response to steroid Fit ICDC criterion Definite Probable Not otherwise specified

158 30 52 119 162 143 30 133 168 116 35 17

100% 19.0% 32.9% 75.3% 100% 88.3% 18.5% 82.1% 100% 69.0% 20.8% 10.1%

ICDC: international consensus diagnostic criteria. LPSP: lymphoplasmacytic sclerosing pancreatitis not otherwise specified (NOS) of ICDC criteria.

Serum IgG4, CA19-9, and CEA levels in AIP, chronic pancreatitis, and pancreatic cancer patients The enrolled patients consisted of 188 patients with AIP, 86 patients with non-AIP chronic pancreatitis, and 130 patients with pancreatic cancer. The patients' mean age was 51.4 years (range, 33e78 years), 49.3 years (range, 18e82 years), and 60.9 years (range, 32e78 years) in AIP, non-AIP chronic pancreatitis, and pancreatic cancer (Table 2). The age was not statistically different in AIP fit in different diagnostic criteria (Table 2). The IgG4 level was measured in 162 patients among 188 patients with AIP, 121 patients with pancreatic cancer, and all patients with non-AIP chronic pancreatitis. The mean levels of IgG4 and CA19-9 levels were shown in Table 2. The mean concentrations (mean ± SEM) of IgG4 in patients with AIP were 346.6 ± 56.2 mg/dL, statistically significantly higher than those in patients with pancreatic cancer, 119.2 ± 23.9 mg/dL, and those in patient with non-AIP chronic pancreatitis, 69.5 ± 5.0 mg/dL respectively (P < 0.0001 for each comparison, Table 2). The mean level of serum IgG4 level in stage 1, 2, 3, and 4 of pancreatic cancer were 45.10, 136.3, 85.47, and 127.2 mg/dL respectively. The mean level (mean ± SEM) of CA19-9 in patients with AIP were 49.8 ± 8.8 U/mL, statistically significantly lower than those in patients with pancreatic cancer, 5593.8 ± 987.1 mg/dL (P < 0.0001, Table 2). The mean level of serum CA19-9 level in stage 1, 2, 3, and 4 of pancreatic cancer were 58.5, 380.7, 1667.9, and 720.5 U/mL respectively .The mean CEA level in patients with pancreatic cancer was 5.9 ± 2.1 U/mL, statistically

368

M.-C. Chang et al. / Pancreatology 14 (2014) 366e372

Table 2 Comparison of demographic data in patients with autoimmune pancreatitis (AIP), non-AIP chronic pancreatitis (CP), and pancreatic cancer (PC). Characteristics Age, years Fit any one Fit HISORt Fit Asian Fit ICDC Focal type Gender Fit any one Fit HISORt Fit Asian Fit ICDC Focal type IgG4, mg/dL Fit any one Fit HISORt Fit Asian Fit ICDC Focal type CA19-9, U/mL Fit any one Fit HISORt Fit Asian Fit ICDC Focal type CEA, U/mL Fit any one Fit HISORt Fit Asian Fit ICDC Focal type

AIP (n ¼ 188)

Non-AIP CP (n ¼ 86)

PC (n ¼ 130)

51.7 ± 14.5 51.5 ± 16.2 52.7 ± 16.9 52.4 ± 16.5 49.9 ± 61.5 M/F (m%) 94/94(50%) 88/65(52.4%) 84/65(51.9%) 82/65(51.9%) 43/65(46.2%)

49.1 ± 11.2

52.3 ± 16.2

0.321

54(63.0%)

65(50%)

0.726

70.3 ± 47.0

119.2 ± 205.9

<0.0001

25.6 ± 23.3

5593.8 ± 10210.7

<0.0001

1.3 ± 1.1

5.9 ± 20.9

<0.0001

346.6 370.6 374.5 372.6 228.9 59.8 47.1 51.1 51.2 56.6

± ± ± ± ±

± ± ± ± ±

714.5 772.0 768.2 753.4 304.4

119.7 108.4 128.3 28.3 154.7

2.17 ± 4.58 2.3 ± 4.9 2.3 ± 4.9 2.3 ± 4.9 1.8 ± 4.2

P

P: ANOVA for AIP, non-AIP chronic pancreatitis and pancreatic cancer. Value presented as mean ± SD; SD: standard deviation. Focal type means focal type autoimmune pancreatitis in ICDC criteria (criteria P level 2).

significantly higher than those in patients with AIP, 2.2 ± 0.3 U/mL (P < 0.0001, Table 2). Autoimmune pancreatitis patients A total of 188 autoimmune pancreatitis patients with complete record and follow up were enrolled for the study analysis. All of the patients fit at least one of the diagnostic criteria including HISORt, Asian, or ICDC. In 168 (89.4%) AIP patients fit ICDC criteria, 116 patients were definite AIP, 35 patients were probable AIP, and 17 patients were classified as not otherwise specified (NOS). There were 30 (19.0%) patients with histological diagnosis of AIP (Table 1). All 30 patients with histology proved AIP were LPSP. The mean follow up period of all AIP patients were 5.2 years (range, 0.5e12years). There are 116 (61.74%) patients with extrapancreatic involvement, including 18 (9.6%) with liver involvement, 53 (28.2%) with biliary involvement, 9 (4.8%) with kidney involvement, 58 (27.6%) with salivary involvement, 5 (2.7%) retroperitoneal involvement, 4 (2.1%) with thyroid involvement, 1 (0.5%) with pituitary involvement, and 6 (3.2%) with ulcerative colitis. Among all the patients, there were 119 (63.3%) patients received steroid treatment. The response rate of steroid is nearly 98% (116/119). There were 90 (48.4%) patients having auto-antibodies. Among them, antinuclear antibody (ANA) and Rheumatic factor (RF) were the most common ones in AIP patients. For diagnosis of AIP by serology criteria, the HISORt criterion has the highest sensitivity (84.2%) but lowest specificity (60%) compared with Asian and ICDC criteria. The ICDC criterion has the highest accuracy (74.3%) to diagnose AIP in our population.

Serum IgG4 for the differentiation of AIP from pancreatic cancer The median serum IgG4 level was significantly higher in AIP patients than in the other 2 groups of patients (Table 2, P < 0.0001). There were 105 (64.8%) of 162 AIP patients with IgG4 level over 140 mg/dL. There were 49 (30.2%) of 162 AIP patients with IgG4 level over 280 mg/dL. There were 4 (4.6%) of 86 non-AIP chronic pancreatitis patients with IgG4 level over 140 mg/dL and no any patients of non-AIP chronic pancreatitis with IgG4 level over 280 mg/dL. There were 18 (20.0%) and 5 (5.6%) of 90 pancreatic cancer patients with available IgG4 level over 140 mg/dL and 280 mg/dL respectively. The sensitivity and specificity of serum IgG4 for the diagnosis of AIP at the IgG4 level of 140 mg/dL from non-AIP chronic pancreatitis was 88% and 92%. The sensitivity and specificity of serum IgG4 for the diagnosis of all 188 AIP from pancreatic cancer at the IgG4 level of 140 mg/dL was 64.8% and 80.4% (Table 3). The sensitivity and specificity of serum IgG4 for the diagnosis of “focal type” AIP (mimicking pancreatic cancer in imaging) from pancreatic cancer at the IgG4 level of 140 mg/dL was only 58.9% and 80.0% (Table 3). If we set the cutoff of IgG4 at 280 mg/dL to differentiation AIP from pancreatic cancer, the specificity reached to 94.4%, both in all AIP and focal type AIP (Table 3). The ROC curve and area under the curve of IgG4 at 280 mg/dL, CA19-9 at 85 U/ml, and combined both were shown in Fig. 1. Serum CA19-9 in AIP, non-AIP CP and pancreatic cancer The CA19-9 level was measured in 186 patients with AIP, 130 patients with pancreatic cancer and 86 patients with non-AIP chronic pancreatitis. The mean CA19-9 level was 49.8 U/ml (range, below 2e1200 U/ml), 25.9 U/ml (range, below 1e1094 U/ ml), and 5593.8 U/ml (range, 3e58898 U/ml) in patients with AIP, non-AIP chronic pancreatitis, and pancreatic cancer, respectively (Table 2). The median serum CA19-9 level was significantly higher in pancreatic cancer patients than those in the other 2 groups of patients (P < 0.0001). The CA19-9 level was increased in 51 of 186 (27.4%) AIP patients if we set 37 U/ml as the upper limit of normal range. Seventeen of 86 cases (19.8%) of non-AIP chronic pancreatitis and 108 of 130 cases (83.1%) of pancreatic cancer showed elevated serum CA19-9 levels more than 37 U/ml. The sensitivity and specificity of serum CA19-9 using 37 U/ml as cut-off for the diagnosis of AIP from pancreatic cancer was 83.7% and 72.6% (Table 3). By operating the ROC curve, the optimal cutoff of CA19-9 in differentiating the AIP form pancreatic cancer was 85.0 U/ml with specificity up to 91.0% (Table 3). Combined use of serum IgG4 and CA19-9 in differentiation of AIP from pancreatic cancer To differentiate AIP from pancreatic cancer, the most optimal cutoff value of IgG4 in this study was 175 mg/dL, at which the sensitivity and specificity was 67.5% and 90.4% respectively (AUC, 0.724, CI 0.66e0.79). The most optimal cutoff value of CA19-9 was 85.0 U/mL, at which the sensitivity and specificity was 76.6% and 91.0%, respectively (AUC, 0.875; CI: 0.83e0.93). When combined the IgG4 over 140 mg/dL and CA19-9 9 below 37 U/ml, the sensitivity was 64.4% and specificity was 91.9% with a diagnostic accuracy about 82.1% in differentiating AIP from pancreatic cancers in our population (Table 3). If we combined the IgG4 at 280 mg/dL and CA19-9 at 85.0 U/ml, the specificity was 96.9% with a diagnostic accuracy 86.9% in differentiating AIP from pancreatic cancers (Table 3).

M.-C. Chang et al. / Pancreatology 14 (2014) 366e372 Table 3 Diagnostic performance of IgG4 (mg/dL) and CA19-9 9 (U/mL) with different cutoffs of IgG4 and C19-9 in all 188 autoimmune pancreatitis (AIP), 168 AIP fir ICDC criteria, 158 AIP fit HISORt criterai,162 fit Asian criteria and 108 focal type AIP compared to 130 pancreatic cancer (PC). Cut-offs IgG4 over 140 Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP IgG4 over 280 Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP CA19-9 above 37 Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP CA19-9 above 85 Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP IgG4 below 140 and Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP IgG4 below 140 and Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP IgG4 below 280 and Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP IgG4 below 280 and Fit any one Fit HISORt Fit Asian Fit ICDC Focal type AIP

SN

SP

PPV

NPV

Accuracy

64.8 84.2 65.9 70.0 58.9

80.4 60.0 80.0 80.0 80.0

85.4 84.2 82.9 82.4 70.5

55.8 60.0 61.5 66.7 70.6

70.2 71.8 71.6 74.3 70.6

30.2 34.0 33.3 34.1 24.7

94.4 94.4 94.4 94.4 94.4

90.7 90.7 89.9 89.1 78.3

42.9 47.2 49.1 48.2 60.7

53.2 57.3 58.1 60.0 63.9

83.7 68.8 83.7 83.7 83.7

72.6 84.9 70.3 70.9 76.3

67.9 68.8 83.1 72.4 73.1

86.5 84.9 70.8 82.6 77.2

77.1 76.4 76.4 77.0 78.4

73.8 73.8 73.8 73.8 73.8 CA19-9 64.4 64.4 64.4 64.4 64.4 CA19-9 54.4 54.4 54.4 54.4 54.4 CA19-9 78.9 78.9 78.9 78.9 78.9 CA19-9 68.9 68.9 68.9 68.9 68.9

91.4 91.0 90.9 90.7 90.3 above 37 91.9 92.4 93.2 91.7 93.2 above 85 97.5 97.9 97.7 97.5 97.3 above 37 86.4 86.1 86.4 85.8 89.0 above 85 96.9 97.2 96.9 96.7 95.9

85.7 86.5 87.3 88.1 91.4

83.3 81.7 90.6 79.0 71.2

84.2 83.5 83.2 82.7 80.7

80.3 84.1 86.6 85.9 92.1

82.3 82.1 79.4 77.5 68.0

82.1 81.6 81.5 82.0 77.3

92.5 94.2 94.2 84.2 96.1

79.4 77.5 75.9 74.1 63.4

82.1 81.2 80.2 79.0 73.6

82.8 78.0 79.8 80.7 89.9

88.1 86.7 85.7 84.4 77.4

83.7 83.3 83.3 82.9 83.9

92.5 93.9 93.9 93.9 95.4

84.9 83.3 82.1 80.6 84.4

86.9 86.3 85.6 84.8 85.6

All AIP: patients fit any one of the HISORt, Asian or ICDC diagnostic criteria; ICDC: patients fit ICDC diagnostic criteria; focal type AIP means focal type AIP in AIP patients fit ICDC (criteria P level 2). SN: sensitivity, SP: specificity, PPV: positive predictive value, NPV: negative predictive value.

Focal type AIP and pancreatic cancer There were 108 (57.4%) of 188 patients with focal type AIP. The mean serum level of IgG4 was 228.9 ± 35.6 mg/dL in focal type AIP, compared to 442.8 ± 91.5 mg/dL in those in diffuse type AIP (p ¼ 0.058). The age, gender and serum levels of CEA and CA19-9 in patients with focal type AIP were not different from those in diffuse type of AIP in our study (Table 2). The sensitivity of serum IgG4 at 140 mg/dL to diagnose focal AIP was only 58.9% (Table 3). Furthermore, the accuracy of IgG4 alone or CA19-9 alone at different cut-offs in different criteria to differentiate focal type AIP from pancreatic cancer was lower compared to combined measurement of IgG4 and CA19-9 (Table 3). If we combined the IgG4 at 280 m/dL and C19-9 at 85.0 U/ml as cut-offs, the accuracy of

369

diagnose focal type AIP from pancreatic cancer could reached to 85.6% (Fig. 2). Diagnostic performance of IgG4 and CA19-9 in patient with and without jaundice Table 4 showed the diagnostic performance of IgG4 and CA19-9 in patients without and without jaundice. The accuracy of diagnosis of AIP from pancreatic cancer was not influenced by jaundice if we combined the IgG4 at 280 m/dL and C19-9 at 85.0 U/ml (Fig. 3, Tables 4 and 6). Table 5 showed the diagnostic performance of IgG4 and CA19-9 in focal type AIP patients and non-metastatic pancreatic cancer stratified by jaundice. The accuracy of diagnosis of focal type AIP patients with jaundice from non-metastatic pancreatic cancer was 80.0% if we combined the IgG4 at 280 m/ dL and C19-9 at 85.0 U/ml, compared to 86.5% in non-jaundiced group (Tables 5 and 6). Discussion Hamano et al. reported that serum IgG4 level was a good marker to differentiating sclerosing pancreatitis from other biliarypancreatic diseases with very high sensitivity (95%) and high specificity (97%) in the initial report with relatively small case number of AIP patients (20 patients with AIP, 20 normal controls) [18]. In a larger cohort of patients with a wide variety of pancreatic diseases, Ghazale et al. showed that elevated serum IgG4 levels are characteristic but not diagnostic of AIP [19]. In their study, they found lower sensitivity and specificity (76% and 93% respectively) for IgG4 over 140 mg/dL by analyzing 45 AIP patients and 135 pancreatic cancer patients. Choi et al. has reported similar sensitivity and specificity (73.3% and 95.1% respectively) for IgG4 over 141 mg/dL which determined as the most optimal cutoff value in 35 AIP patients and 76 pancreatic cancer patients in Korean [20]. We observed lower sensitivity of using previous defined IgG4 levels to diagnose AIP and to differentiate it from pancreatic cancer in our study. The larger sample size we enrolled might contribute to the result. Our current study demonstrated the IgG4 level at 175 mg/dL was determined as the most optimal cutoff value with sensitivity and specificity of 87.5% and 93.4% in Chinese in Taiwan, similar to our previous reports [16,21,22]. The optimal cutoff level of IgG4 to diagnose AIP in our study was the highest one than those been reported before. To further investigate the lower sensitivity of serum IgG4 to diagnose AIP in our population, we compared the profiles of seropositive and seronegative AIP patients. The demographic profiles of AIP patients with normal IgG4 levels were not different from those of seropositive AIP patients in age, gender, extrapancreatic involvements, and treatment response to steroid. Our observation was different from the previous report by Ghazale et al. who found that the seronegative AIP patients were more common in female and had less extrapancreatic organ involvement [19]. The seronegative rate of our AIP was 37.0%, which was higher compared with their series (24.0%). The entity of seronegative AIP was worthy further study with larger cases series and in different populations. In our study, the patients with AIP were classified as type 1 AIP. It is important to note that about 37% of AIP patients had normal serum IgG4 levels even in type 1 AIP. This highlights the fact that seronegative AIP can occur, and is not uncommon using current diagnostic criteria. Furthermore, the utility of IgG4 in type 2 AIP which were more common in United States and Europe should be verified. The reported specificity of serum IgG4 for diagnosis of AIP was similar in our study compared with previously reported ones from Japan, United States and Korea [18e20]. We found 20% pancreatic cancer patients had elevated IgG4 levels in our population. The

370

M.-C. Chang et al. / Pancreatology 14 (2014) 366e372

Fig. 1. Receiver operating characteristic (ROC) curve in the differential diagnosis of autoimmune pancreatitis and pancreatic cancer. a. IgG4: 280 mg/d (area under the curve: 95% CI: 0.31e0.44), b. CA19-9: 85 U/ml (area under the curve: 95% CI: 0.76e0.88), c. IgG4 280 mg/dl and CA19-9 85 U/ml (area under the curve: 95% CI: 0.77e0.89).

Fig. 2. Accuracy of IgG4 and CA19-9 at different cutoffs for diagnosis of autoimmune pancreatitis. Any means fit any of the ICDC, HISORt or Asian criteria.

clinical profiles of pancreatic cancer patients with and without elevated IgG4 levels were similar. Besides, there were 4.7% of our non-AIP chronic pancreatitis patients had elevated serum IgG4 in our study. Of the 4 patients with non-AIP chronic pancreatitis who had elevated IgG4 (ranged from 147 to 278 mg/dL), three of them were female and three were younger than 30 year-old. Their pancreatic morphology were pancreatic atrophy and/or dilated main pancreatic duct which were different from the current imaging diagnostic criteria for AIP. Of these non-AIP chronic pancreatitis patients with elevated IgG4 levels, only one patient had available tissue proof without any histological evidence of LPSP. Besides, the clinical presentations and their pancreatic imaging studies were not different from those in non-AIP chronic pancreatitis with normal IgG4 level. In our study, the sensitivity and specificity of serum IgG4 to differentiate AIP from non-AIP chronic pancreatitis were 88.9% and 92.3% at the level of 135 mg/dL (AUC:0.906) based on ROC analysis. These findings suggest that the elevations of IgG4 in patients with other pancreatic diseases do not

M.-C. Chang et al. / Pancreatology 14 (2014) 366e372 Table 4 Diagnostic performance of IgG4 (mg/dL) and CA19-9 9 (U/mL) in patients without and with jaundice. Cut-offs IgG4 over 140 All Without jaundice With jaundice IgG4 over 280 All Without jaundice With jaundice CA19-9 above 37 All Without jaundice With jaundice CA19-9 above 85 All Without jaundice With jaundice IgG4 below 140 and All Without jaundice With jaundice IgG4 below 140 and All Without jaundice With jaundice Focal type AIP IgG4 below 280 and All Without jaundice With jaundice IgG4 below 280 and All Without jaundice With jaundice

SN

SP

PPV

NPV

Accuracy

Table 5 Diagnostic performance of IgG4 (mg/dL) and CA19-9 9 (U/mL) in patient with focal autoimmune pancreatitis (AIP) and non-metastatic pancreatic cancer (PC) stratified by jaundice. Cut-offs

64.8 32.0 15.4

80.4 40.4 27.9

85.4 68.0 84.6

55.8 59.6 72.1

70.2 38.7 21.8

30.2 16.0 15.4

94.4 80.9 54.4

90.7 84.0 84.6

42.9 19.1 45.6

53.2 67.2 35.3

83.7 81.6 84.6

72.6 85.9 58.5

67.9 18.4 15.4

86.5 14.1 42.5

77.1 84.7 71.3

73.8 73.7 73.9 CA19-9 above 64.4 44.0 72.3 CA19-9 above 54.4 36.0 61.5

91.4 96.9 85.1 37 91.9 95.7 86.8 85 97.5 98.9 95.6

85.7 26.3 26.1

83.3 3.0 14.9

84.2 90.1 79.3

80.3 56.0 7.7

82.3 4.3 13.2

82.1 84.9 79.7

92.5 64.0 38.5

79.4 1.1 4.41

82.1 85.7 78.9

CA19-9 above 78.9 60.0 86.2 CA19-9 above 68.9 52.0 75.4

37 86.4 90.4 80.9 85 96.9 2.1 4.4

82.8 40.0 13.8

88.1 9.6 19.1

83.7 84.0 83.5

92.5 0.5 24.6

84.9 97.9 95.6

86.9 88.2 86.7

indicate undiagnosed concomitant AIP. Elevated IgG4 levels have been associated with some diseases, including pemphigus [23], atopic dermatitis [24] and filariasis [25] et al. In our cases, we did not identify any of these diseases in our non-AIP patients with elevated IgG4 levels. The reason for IgG4 elevation in non-AIP CP patients needs further study. One great challenge for clinician is to distinguish AIP from pancreatic cancer, especially those presented as a focal pancreatic mass (focal type AIP). From the imaging studies, the both could be almost the same if no other collateral evidences of AIP could be obtained. It requires clinical suspicion and careful judgment based on a good understanding of both diseases. Presence of cachexia and severe pain are highly suggestive of but not specific to pancreatic

371

IgG4 over 140 All Without jaundice With jaundice IgG4 over 280 All Without jaundice With jaundice CA19-9 above 37 All Without jaundice With jaundice CA19-9 above 85 All Without jaundice With jaundice IgG4 below 140 and All Without jaundice With jaundice IgG4 below 140 and All Without jaundice With jaundice IgG4 below 280 and All Without jaundice With jaundice IgG4 below 280 and All Without jaundice With jaundice

SN

SP

PPV

NPV

Accuracy

30.0 50.0 9.0

41.1 48.0 26.1

80.0 50.0 90.9

58.9 52.0 73.9

34.9 48.3 17.8

6.7 25.0 4.3

75.3 82.0 63.6

93.3 75.0 95.6

24.6 18.0 36.4

55.3 74.1 33.3

75.0 60.0 80.8

76.3 87.3 55.3

25.0 40.0 19.2

23.7 12.7 44.7

75.9 83.0 65.6

90.3 94.5 84.2 37 93.2 96.0 86.9 85 97.3 98.0 95.1 37 89.0 94.0 78.3 85 95.9 96.0 95.7

41.6 60.0 34.6

9.7 54.5 15.8

81.4 86.2 76.6

46.7 87.0 31.8

6.8 4.0 13.0

81.6 84.5 77.8

60.0 87.5 45.5

2.7 2.0 4.3

80.6 86.2 75.6

33.3 62.5 2.7

11.0 6.0 21.7

82.5 86.2 77.8

46.7 75.0 36.4

4.1 4.0 4.3

83.5 86.2 80.0

58.3 40.0 65.4 CA19-9 above 53.3 12.5 68.2 CA19-9 above 40.0 12.5 54.5 CA19-9 above 66.7 37.5 77.3 CA19-9 above 53.3 25.0 63.6

cancer. Besides, patients with Lewis antigen negative blood group [Le(a-/b-)] would have normal CA 19-9 level. In our study, the prevalence rate of Lewis antigen negative blood group, about 10e15%, was similar to the reported rate in our population. Furthermore, factors such as jaundice, infection, and renal insufficient et al. could variously influence the level of CA19-9. The optimal cut-off of CA19-9 in our study was 85.0 U/ml with a 76.6% of sensitivity and 91.0% of specificity respectively based on ROC analysis (AUC, 0.875; CI: 0.83e0.93) to differentiate AIP from pancreatic cancer. In this study, we also stratified the patients by presence of jaundice for analysis. The diagnostic accuracy (86.7%) was not influenced if we combined both IgG4 and CA19-9 even in Table 6 Accuracy of IgG4 (mg/dL) and CA19-9 9 (U/mL) in patient autoimmune pancreatitis (AIP) and pancreatic cancer (PC) stratified by jaundice. Cut-offs

Accuracy AIP and PC

Fig. 3. Accuracy of IgG4 and CA19-9 at different cutoffs for diagnosis of 188 autoimmune pancreatitis patients from 130 pancreatic cancer stratified by jaundice.

Focal AIP and nonmetastatic PC

Jaundice

All

Without

With

All

Without

With

IgG4 over 140 IgG4 over 280 CA19-9 above 37 CA19-9 above 85 IgG4 below 140 and CA19-9 above 37 IgG4 below 140 and CA19-9 above 85 IgG4 below 280 and CA19-9 above 37 IgG4 below 280 and CA19-9 above 85

70.2 53.2 77.1 84.2 82.1

38.7 67.2 84.7 90.1 84.9

21.8 35.3 71.3 79.3 79.7

34.9 55.3 75.9 81.4 81.6

48.3 74.1 83.0 86.2 84.5

17.8 33.3 65.6 76.6 77.8

82.1

85.7

78.9

80.6

86.2

75.6

83.7

84.0

83.5

82.5

86.2

77.8

86.9

88.2

86.7

83.5

86.2

80.0

372

M.-C. Chang et al. / Pancreatology 14 (2014) 366e372

jaundiced patients (Table 4, Table 6, and Fig. 2). In contrast, if we only use CA19-9 or IgG4 alone to differentiate AIP from PC, the accuracy were lower in patients with jaundice, compared to those without jaundice. Therefore, the combined use of serum IgG4 and CA19-9 to distinguish AIP, especially focal type, from pancreatic cancer with cutoffs defined 280 mg/dL for IgG4 and 85.0 U/ml for CA19-9 had sensitivity, specificity and accuracy 68.9%, 95.9% and 85.6%. The combined measurement of IgG4 and CA19-9 works much better than using IgG4 or CA19-9 alone (Fig. 1) and not influenced by presence of jaundice. The most challenging cases for clinician were differentiation focal type AIP from non-metastatic pancreatic cancer. In this study, we also compared the diagnostic accuracy of IgG4 and CA19-9 in focal type AIP and non-metastatic pancreatic cancer. The best diagnostic accuracy were combined the use of IgG4 (over 280 mg/dL) and CA19-9 9 (below 85.0 U/ml) (Tables 5 and 6). In addition, we divided our cases into two sets: the test set and the validation set. The IgG4 and CA19-9 cutoffs obtained in the test set to differentiate focal type AIP from nonmetastatic pancreatic cancer could well differentiate the focal type AIP from non-metastatic pancreatic cancer in the validation set. This might help clinically in differentiating focal type AIP from pancreatic cancers before invasive procedure or empirical steroid therapy. In summary, we have showed that elevated serum IgG4 levels are a characteristic but not a diagnostic feature of AIP. Our study demonstrated that combined use of serum IgG4 (over 280 mg/dL) and CA19-9 9 (below 85.0 U/ml) increases the diagnostic accuracy to distinguish AIP from pancreatic cancer non-invasively, especially in focal type autoimmune pancreatitis. Acknowledgments This work was supported by grants from National Science Council, Taiwan (NSC 94-2314-B-002-272) and NTUH (National Taiwan University Hospital) NTUH-95-M-22, NTUH-97-M-1001; NSC 102-2321-B-002-083; MOHW103-TD-B-111-04; Liver Disease Prevention & Treatment Research Foundation and New Century Health Care Promotion Foundation. The authors are thankful for the help of department of Laboratory Medicine in National Taiwan University Hospital for technical support and analysis of serum IgG4. The authors express their deep sense of gratitude to all of the individuals for agreeing to participate in the study. References [1] Yoshida K, Toki F, Takeuchi T, Watanabe S, Shiratori K, Hayashi N. Chronic pancreatitis caused by an autoimmune abnormality. Proposal of the concept of autoimmune pancreatitis. Dig Dis Sci 1995;40:1561e8. [2] Kamisawa T, Chari ST, Lerch MM, Kim MH, Gress TM, Shimosegawa T. Recent advances in autoimmune pancreatitis: type 1 and type 2. Gut 2013;62: 1373e80. [3] Kamisawa T, Imai M, Yui Chen P, Tu Y, Egawa N, Tsuruta K, et al. Strategy for differentiating autoimmune pancreatitis from pancreatic cancer. Pancreas 2008;37:e62e7.

[4] Ghazale A, Chari ST, Smyrk TC, Levy MJ, Topazian MD, Takahashi N, et al. Value of serum IgG4 in the diagnosis of autoimmune pancreatitis and in distinguishing it from pancreatic cancer. Am J Gastroenterol 2007;102: 1646e53. [5] Morselli-Labate AM, Pezzilli R. Usefulness of serum IgG4 in the diagnosis and follow up of autoimmune pancreatitis: a systematic literature review and meta-analysis. J Gastroenterol Hepatol 2009;24:15e36. [6] Kamisawa T. IgG4-positive plasma cells specifically infiltrate various organs in autoimmune pancreatitis. Pancreas 2004;29:167e8. [7] Kamisawa T, Okazaki K, Kawa S. Diagnostic criteria for autoimmune pancreatitis in Japan. World J Gastroenterol 2008;14:4992e4. [8] Okazaki K, Kawa S, Kamisawa T, Naruse S, Tanaka S, Nishimori I, et al. Clinical diagnostic criteria of autoimmune pancreatitis: revised proposal. J Gastroenterol 2006;41:626e31. [9] Otsuki M, Chung JB, Okazaki K, Kim MH, Kamisawa T, Kawa S, et al. Asian diagnostic criteria for autoimmune pancreatitis: consensus of the Japan-Korea symposium on autoimmune pancreatitis. J Gastroenterol 2008;43:403e8. [10] Chari ST, Smyrk TC, Levy MJ, Topazian MD, Takahashi N, Zhang L, et al. Diagnosis of autoimmune pancreatitis: the Mayo clinic experience. Clin Gastroenterol Hepatol 2006;4:1010e6. quiz 934. [11] Chari ST, Takahashi N, Levy MJ, Smyrk TC, Clain JE, Pearson RK, et al. A diagnostic strategy to distinguish autoimmune pancreatitis from pancreatic cancer. Clin Gastroenterol Hepatol 2009;7:1097e103. [12] Shimosegawa T, Chari ST, Frulloni L, Kamisawa T, Kawa S, Mino-Kenudson M, et al. International consensus diagnostic criteria for autoimmune pancreatitis: guidelines of the international association of pancreatology. Pancreas 2011;40:352e8. [13] Matsumoto I, Shinzeki M, Toyama H, Asari S, Goto T, Yamada I, et al. A focal mass-forming autoimmune pancreatitis mimicking pancreatic cancer with obstruction of the main pancreatic duct. J Gastrointest Surg 2011;15:2296e8. [14] Sumimoto K, Uchida K, Mitsuyama T, Fukui Y, Kusuda T, Miyoshi H, et al. A proposal of a diagnostic algorithm with validation of international consensus diagnostic criteria for autoimmune pancreatitis in a Japanese cohort. Pancreatology 2013;13:230e7. [15] Chang MC, Chang YT, Su TC, Yang WS, Chen CL, Tien YW, et al. Adiponectin as a potential differential marker to distinguish pancreatic cancer and chronic pancreatitis. Pancreas 2007;35:16e21. [16] Chang MC, Jan IS, Liang PC, Jeng YM, Yang CY, Tien YW, et al. PRSS1 but not SPINK1 variants increase the risk of type 1 autoimmune pancreatitis. J Gastroenterol Hepatol 2014 Jun 9. http://dx.doi.org/10.1111/jgh.12649. [Epub ahead of print]. [17] Chang MC, Chang YT, Tien YW, Liang PC, Wei SC, Wong JM. Association of tumour necrosis factor alpha promoter haplotype with chronic pancreatitis. Gut 2006;55:1674e6. [18] Hamano H, Kawa S, Horiuchi A, Unno H, Furuya N, Akamatsu T, et al. High serum IgG4 concentrations in patients with sclerosing pancreatitis. N Engl J Med 2001;344:732e8. [19] Ghazale A, Chari ST, Zhang L, Smyrk TC, Takahashi N, Levy MJ, et al. Immunoglobulin G4-associated cholangitis: clinical profile and response to therapy. Gastroenterology 2008;134:706e15. [20] Choi EK, Kim MH, Lee TY, Kwon S, Oh HC, Hwang CY, et al. The sensitivity and specificity of serum immunoglobulin G and immunoglobulin G4 levels in the diagnosis of autoimmune chronic pancreatitis: Korean experience. Pancreas 2007;35:156e61. [21] Chang MC, Chang YT, Tien YW, Liang PC, Jan IS, Wei SC, et al. T-cell regulatory gene CTLA-4 polymorphism/haplotype association with autoimmune pancreatitis. Clin Chem 2007;53:1700e5. [22] Chang MC, Chang YT, Wei SC, Kuo CH, Liang PC, Wong JM. Autoimmune pancreatitis associated with high prevalence of gastric ulcer independent of Helicobacter pylori infection status. Pancreas 2009;38:442e6. [23] Shirakata Y, Shiraishi S, Sayama K, Miki Y. Subclass characteristics of IgG autoantibodies in bullous pemphigoid and pemphigus. J Dermatol 1990;17: 661e6. [24] Aalberse RC, Van Milligen F, Tan KY, Stapel SO. Allergen-specific IgG4 in atopic disease. Allergy 1993;48:559e69. [25] Hussain R, Poindexter RW, Ottesen EA. Control of allergic reactivity in human filariasis. Predominant localization of blocking antibody to the IgG4 subclass. J Immunol 1992;148:2731e7.