J Gastrointest Surg DOI 10.1007/s11605-015-2786-3
ORIGINAL ARTICLE
Predictive Factors for Surgery Among Patients with Pancreatic Cysts in the Absence of High-Risk Features for Malignancy Susan Y. Quan & Brendan C. Visser & George A. Poultsides & Jeffrey A. Norton & Ann M. Chen & Subhas Banerjee & Shai Friedland & Walter G. Park
Received: 29 December 2014 / Accepted: 24 February 2015 # 2015 The Society for Surgery of the Alimentary Tract
Abstract Background Without a reliable biopsy technique for pancreatic cysts, consensus-based guidelines are used to guide surgical utilization. The primary objective of this study was to characterize the proportion of operations performed outside of these guidelines. Methods A 5-year retrospective review between July 1, 2007, and June 30, 2012, was performed of consecutive patients seen at a single tertiary medical center for a pancreatic cyst. Manual chart review for relevant clinical variables and cyst characteristics was performed. Results During this period, 148 patients underwent surgery, and of these, 23 (16 %) patients had no high-risk criteria by the 2006 Sendai criteria. None of these harbored high-grade dysplastic or cancerous lesions. A high cyst carcinoembryonic antigen (CEA) level (35 %), patient anxiety (26 %), and physician concern (22 %) were explicit reasons to proceed to surgery. An elevated cyst CEA level >192 ng/ml was the most significant predictor (OR 5.14 (95 % confidence interval (CI) 1.47–18.0) for surgery without high-risk criteria. Conclusion A high cyst CEA level was significantly associated with the decision to operate outside of consensus-based guidelines. The misuse of cyst CEA in the management of pancreatic cysts negatively impacts patient anxiety, increases physician uncertainty, and leads to surgery with minimal benefit. Keywords Pancreatic cyst . Biomarkers . Clinical practice pattern
Introduction The widespread use of high-resolution, cross-sectional imaging has made the management of pancreatic cystic neoplasms This paper was presented at the American Pancreas Association Meeting, November 2013, Miami, Florida. S. Y. Quan : A. M. Chen : S. Banerjee : S. Friedland : W. G. Park Department of Medicine, Stanford University School of Medicine, Stanford, CA 94305, USA B. C. Visser : G. A. Poultsides : J. A. Norton Department of Surgery, Stanford University School of Medicine, Stanford, CA 94305, USA W. G. Park (*) Pancreas Clinic, Stanford University School of Medicine, 300 Pasteur Drive, MC: 5187, Stanford, CA 94305, USA e-mail: [email protected]
(PCNs) a relatively common clinical conundrum. The vast majority of these cysts do not harbor cancer upon discovery but, up to one half, may harbor malignant potential.1 The goal of clinical management is to appropriately weigh the risks of cancer development with the risks of surgery. Consensusbased guidelines (a.k.a. Sendai criteria) were established in 2006 and, most recently, revised in 2012, to advise decisionmaking on appropriate surgical utilization of PCNs.2, 3 These guidelines, however, represent the best extrapolation of current evidence, and the accuracy of preoperative diagnoses for specific cyst type is suboptimal.4 Therefore, the appropriateness of surgery remains an individualized clinical judgment for each case. In this study, we sought to better understand the extent of surgical utilization for pancreatic cysts considered to have low risk for cancer as defined by the first consensus criteria. By characterizing this subpopulation of patients, we were able to identify possible factors that led to the decision to proceed with surgery. Better elucidation of these factors may provide insight into the daily challenges of managing these patients and further maximize the benefit from surgery.
J Gastrointest Surg
Methods
Results
Study Population
Between July 1, 2007, and June 30, 2012, 486 patients with pancreatic cysts were identified. Of these patients, 148 (30 %) underwent surgery of which 23 (16 %) were considered low risk by consensus criteria. A descriptive summary of patientand cyst-related characteristics of (1) all patients, (2) those who underwent surgery, (3) those who underwent surgery with positive Sendai criteria, (4) those who underwent surgery with negative Sendai criteria, and (5) those observed is provided in Table 1. Table 1 shows that when compared to those observed (n= 338), those who underwent surgery (n=148) were younger (median age 61 vs. 64, p=0.007), had larger cysts (median cyst size 30 vs. 17 mm, p=0.0001), had a higher proportion of intracystic mural nodules (15 vs. 2 %, p=0.0001), had a higher proportion of cysts with a dilated duct (8 vs. 1 %, p=0.0001), were more symptomatic (54 vs. 37 %, p=0.0004), had higher cyst fluid CEA levels (median 631 vs. 73 ng/ml, p=0.0001), and had a higher proportion of atypical cytology (19 vs. 8 %) (p=0.0003). There were no differences observed as to the gender distribution, location of cysts, the proportion presenting with multiple cysts, the presence of cyst septations, the presence of multicystic morphology, or microcystic morphology. Among those who underwent surgery, 125 patients met Sendai criteria and 23 patients did not. Table 1 shows that the only significant differences between these two groups were characteristics that define the Sendai criteria including cyst size (median 33 vs. 18 mm, p=0.0001), proportion of intracystic nodules (18 vs. 0 %, p=0.03), and the presence of associated symptoms (64 vs. 0 %, p=0.00001). No other significant differences were observed. Patients who did not meet Sendai criteria would generally be observed, and these 23 patients were compared to the 338 patients who were observed. Table 1 shows that these patients were younger (median age 59 vs. 64, p=0.035), less symptomatic (0 vs. 37 %, p=0.0003), had fewer multifocal cysts (35 vs. 56 %, p=0.04), and had a higher CEA level (median 758 vs. 73 ng/ml, p=0.003). Table 2 summarizes the histological distribution of pancreatic cysts. Of the 148 patients who underwent surgery, 125 (85 %) met preoperative Sendai criteria for resection. Among these patients, 9 (7 %) were cystadenocarcinomas, 18 (14 %) were main or mixed duct intraductal papillary mucinous neoplasms (IPMNs), 40 (32 %) were branch duct IPMNs, 15 (12 %) were mucinous cystic neoplasms (MCNs), 16 (13 %) were serous cystic neoplasms (SCNs), and 17 (14 %) were pseudocysts. Twenty-three patients (16 %) did not meet preoperative Sendai criteria and included 14 (61 %) branch duct IPMNs, 3 (13 %) MCNs, 1 (4 %) SCN, 1 (4.5 %) pseudocyst, 1 (4.5 %) squamous cyst, 1 (4.5 %) area of focal pancreatitis, and 2 (9 %) neuroendocrine tumors (NET). There were no cystadenocarcinomas or main duct/mixed IPMNs in this group.
With approval from the Stanford University Hospital and Clinics Institutional Review Board, we performed a retrospective chart review covering a 5-year period (July 1, 2007, to June 30, 2012) of all adult patients (≥18 years) referred for endoscopic ultrasound (EUS) or surgery for a pancreatic cyst at Stanford University Hospital and Clinics. All medical records were manually reviewed to collect specific patient- and cyst-related characteristics. Patient characteristics of interest included age, gender, and the presence of symptoms potentially related to a pancreatic cyst. These symptoms included documentation of abdominal pain, weight loss, jaundice, and the presence or history of pancreatitis. Cyst characteristics of interest included cyst size, the presence of mural nodularity, main pancreatic duct (PD) diameter, carcinoembryonic antigen (CEA) level, and cytology. Imaging characteristics were collected from available computed tomography (CT) scans, magnetic resonance imaging (MRI) scans, and/or EUS reports. When more than one imaging modality was available, the largest reported cyst size and PD diameter were collected. Cyst size was recorded as the maximum dimension measured on cross-sectional imaging. The main PD was considered dilated when it measured ≥6 mm. Mural nodularity was considered present when noted on any available imaging modality report. Other cyst-related features of interest that were collected included the presence of septations and multiple cysts and the location of the cyst within the pancreas (proximal=head, distal=body and tail). Cyst fluid CEA levels and cytology results were collected from patients that underwent EUS with fine-needle aspiration (FNA). If a patient had cyst fluid obtained more than once, the highest cyst fluid CEA level recorded was used for analysis. For analysis, the 2006 Sendai criteria were used to define high- and low-risk pancreatic cysts for harboring cancer.2
Statistics Statistical analyses were performed using Stata 13.1 (Stata Corp., College Station, TX, USA). Descriptive statistics were performed using non-parametric assumptions. Chi-square test was used for univariate analysis of predictive factors of surgical utilization. Multivariate analysis with logistic regression was used to identify factors independently associated with surgical utilization in cases that did not meet Sendai criteria. Only univariate variables with p value 6 mm Presence of symptoms Proximal cyst location Multifocal cysts
302 (62 %) 64 (54–72) 20 (12–30) 149 (31 %) 28 (6 %) 15 (3 %) 204 (47 %) 265 (55 %) 127 (27 %)
85 (57 %) 61 (51–70) 30 (18–50) 79 (54 %) 21 (15 %) 12 (8 %) 80 (54 %) 76 (51 %) 30 (20 %)
70 (56 %) 62 (50–72) 33 (20–51) 79 (63 %) 21 (18 %) 12 (10 %) 80 (64 %) 64 (51 %) 28 (22 %)
15 (65 %) 59 (52–65) 18 (14–21) 0 (0 %) 0 (0 %) 0 (0 %) 0 (0 %) 8 (35 %) 2 (9 %)
217 (64 %) 64 (54–74) 17 (10–25) 70 (21 %) 7 (2 %) 3 (1 %) 124 (37 %) 189 (56 %) 97 (29 %)
Cyst septations Multicystic Microcystic Median CEA level (IQR, N) Atypical cytology
165 (35 %) 46 (9 %) 31 (6 %) 138 (7–933, N=325) 54 (11 %)
46 (31 %) 16 (11 %) 8 (5 %) 631 (61–2650, N=96) 28 (19 %)
37 (30 %) 15 (12 %) 7 (5 %) 555 (48–2831, N=82) 27 (22 %)
9 (39 %) 1 (4 %) 1 (4 %) 758 (229 –1719, N=14) 1 (4 %)a
119 (35 %) 30 (9 %) 23 (7 %) 73 (5–471, N=229)
a
26 (8 %)
Patient had cytology interpretation Batypical^ because it was suspicious for mucinous neoplasm (not frank high-grade dysplasia or cancer)
A manual review of the chart for those patients who underwent surgery without meeting Sendai criteria was performed to assess for case-specific reasons for proceeding with surgery. Among these 23 patients (Fig. 1), these included (1) a high cyst CEA (35 %), (2) patient anxiety and preference (26 %), (3) physician suspicion for malignancy (22 %), (4) interval increase in cyst size (9 %), (5) an interval rise in CEA (4 %), and (6) unknown (4 %). A univariate analysis of non-Sendai criteria was performed to identify potential predictors for surgery (Table 3). No
significant differences in age, cyst location, cyst morphology, CEA level, or cytology differentiated operations that met Sendai criteria compared to those that did not. When the group that underwent surgery without meeting any Sendai criteria was compared to those observed, they were younger with a higher proportion less than the 65 years old (p=0.03) and had a higher CEA level (p=0.01). By multivariate analysis, an elevated CEA level >192 ng/ml was the most significant factor (OR 5.14 (95 % confidence interval (CI) 1.47–18.0)) to predict resection among low-risk cases defined by Sendai criteria compared to those observed (Table 4).
Table 2 Distribution of pancreatic cysts evaluated by endoscopic ultrasound and surgery over a 5-year period Total number of pancreatic cysts Underwent surgery Positive preoperative Sendai criteria Cyst type distribution Main/mixed IPMN Branch duct IPMN Mucinous cystic neoplasm Serous cystic neoplasm Pseudocyst Cancera Otherb
Discussion
486 148 (30 %) 125 (85 %) Sendai (+) (N=125) 18 (14 %) 40 (32 %) 15 (12 %) 16 (13 %) 17 (14 %) 9 (7 %) 10 (8 %)
With increasing recognition of the prevalence of pancreatic cysts,5, 6 appropriate decision-making between resection and Sendai (−) (N=23) 0 (0 %) 14 (61 %) 3 (13 %) 1 (4.5 %) 1 (4.5 %) 0 (0 %) 4 (17 %)
IPMN intraductal papillary mucinous neoplasm a
IPMN cancer/cystic adenocarcinoma (neuroendocrine tumors not included)
b The four other cases not meeting Sendai criteria were focal pancreatitis, benign squamous cyst, and two neuroendocrine tumors
Fig. 1 Distribution of reasons for surgery when Sendai criteria were not met
J Gastrointest Surg Table 3 Univariate analysis of non-Sendai criteria factors to predict surgical cases that met Sendai criteria compared to those that did not and to compare those who underwent surgery without meeting Sendai criteria versus those observed
Independent variables
Sendai (+) vs. Sendai (−) surgery cases
Sendai (−) surgery cases vs. observed
Age >65 Proximal cyst location Multifocal cysts Cyst septations Multicystic Microcystic CEA level >192 Atypical cytology
2.73 (0.96–7.84), p=0.06 0.45 (0.18–1.13), p=0.09 3.08 (0.68–14.0), p=0.15 0.65 (0.26–1.66), p=0.37 3.03 (0.38–24.2), p=0.30 1.31 (0.15–11.22), p=0.80 0.42 (0.11–1.50), p=0.18 4.96 (0.64–38.7), p=0.13
0.33 (0.12–0.90), p=0.03 2.38 (0.98–5.76), p=0.06 0.24 (0.06–1.07), p=0.06 1.27 (0.53–3.05), p=0.60 0.49 (0.06–3.75), p=0.50 0.65 (0.08–5.05), p=0.68 4.71 (1.37–16.2), p=0.01 0.76 (0.10–5.97), p=0.80
surveillance remains an increasingly important clinical dilemma. Without a safe and reliable biopsy technique, the 2006 Sendai consensus guidelines were written to delineate clinical and radiographic features associated with a high risk of malignancy.2 These features included (1) cyst size greater than 3 cm, (2) presence of an intracystic mural nodule, (3) presence of positive cytology, (4) presence of an associated dilated main duct, and (5) associated symptoms attributed to the cyst. In the absence of these features, surveillance was recommended with the caveat that, given the overall low quality of evidence, individualized physician judgment and patient preference to do otherwise were acceptable. Consequently, we sought to retrospectively evaluate the frequency of patients undergoing surgery for pancreatic cysts in the absence of high-risk features from 2007 to 2012. This period was specifically chosen because it allowed a 1-year period for guideline dissemination and ended just when the guidelines were revised.3 We observed that of the 148 patients who underwent surgical resection for a pancreatic cyst, 23 (16 %) patients did not have any of the five high-risk features as defined by the Sendai criteria. While none of these cysts harbored adenocarcinoma, there were five cysts (22 %) that proved to be appropriate for surgical resection by current guidelines that had they been accurately diagnosed preoperatively (three MCNs and two NETs). An in-depth medical chart review was performed of these 23 cases to determine explicit reasons for proceeding to surgery. An elevated cyst fluid CEA (35 %), patient anxiety and preference (26 %), and physician suspicion for malignancy (22 %) were the most commonly stated reasons for proceeding to surgery despite acknowledging the absence of high-risk features as defined by Sendai criteria. When compared to those patients with pancreatic cysts that underwent observation (n = 338), these patients were younger and had significantly elevated cyst CEA levels. Cyst CEA level in this group that underwent surgery despite an absence of high-risk features was highest (mean 64,961 ng/ml (95 % CI 0–195,948)) when compared to those observed (mean 2904 ng/ml (95 % CI 165–5,645)) (p65 Proximal cyst location Multifocal cysts CEA level >192
2.58 (0.89–7.43), p=0.08 0.47 (0.19–1.22), p=0.12
0.36 (0.13–1.02), p=0.05 1.92 (0.76–4.84), p=0.17
–
0.27 (0.06–1.20, p=0.09
–
5.14 (1.47–18.0), p=0.01
Only variables with a p value
ORIGINAL ARTICLE
Predictive Factors for Surgery Among Patients with Pancreatic Cysts in the Absence of High-Risk Features for Malignancy Susan Y. Quan & Brendan C. Visser & George A. Poultsides & Jeffrey A. Norton & Ann M. Chen & Subhas Banerjee & Shai Friedland & Walter G. Park
Received: 29 December 2014 / Accepted: 24 February 2015 # 2015 The Society for Surgery of the Alimentary Tract
Abstract Background Without a reliable biopsy technique for pancreatic cysts, consensus-based guidelines are used to guide surgical utilization. The primary objective of this study was to characterize the proportion of operations performed outside of these guidelines. Methods A 5-year retrospective review between July 1, 2007, and June 30, 2012, was performed of consecutive patients seen at a single tertiary medical center for a pancreatic cyst. Manual chart review for relevant clinical variables and cyst characteristics was performed. Results During this period, 148 patients underwent surgery, and of these, 23 (16 %) patients had no high-risk criteria by the 2006 Sendai criteria. None of these harbored high-grade dysplastic or cancerous lesions. A high cyst carcinoembryonic antigen (CEA) level (35 %), patient anxiety (26 %), and physician concern (22 %) were explicit reasons to proceed to surgery. An elevated cyst CEA level >192 ng/ml was the most significant predictor (OR 5.14 (95 % confidence interval (CI) 1.47–18.0) for surgery without high-risk criteria. Conclusion A high cyst CEA level was significantly associated with the decision to operate outside of consensus-based guidelines. The misuse of cyst CEA in the management of pancreatic cysts negatively impacts patient anxiety, increases physician uncertainty, and leads to surgery with minimal benefit. Keywords Pancreatic cyst . Biomarkers . Clinical practice pattern
Introduction The widespread use of high-resolution, cross-sectional imaging has made the management of pancreatic cystic neoplasms This paper was presented at the American Pancreas Association Meeting, November 2013, Miami, Florida. S. Y. Quan : A. M. Chen : S. Banerjee : S. Friedland : W. G. Park Department of Medicine, Stanford University School of Medicine, Stanford, CA 94305, USA B. C. Visser : G. A. Poultsides : J. A. Norton Department of Surgery, Stanford University School of Medicine, Stanford, CA 94305, USA W. G. Park (*) Pancreas Clinic, Stanford University School of Medicine, 300 Pasteur Drive, MC: 5187, Stanford, CA 94305, USA e-mail: [email protected]
(PCNs) a relatively common clinical conundrum. The vast majority of these cysts do not harbor cancer upon discovery but, up to one half, may harbor malignant potential.1 The goal of clinical management is to appropriately weigh the risks of cancer development with the risks of surgery. Consensusbased guidelines (a.k.a. Sendai criteria) were established in 2006 and, most recently, revised in 2012, to advise decisionmaking on appropriate surgical utilization of PCNs.2, 3 These guidelines, however, represent the best extrapolation of current evidence, and the accuracy of preoperative diagnoses for specific cyst type is suboptimal.4 Therefore, the appropriateness of surgery remains an individualized clinical judgment for each case. In this study, we sought to better understand the extent of surgical utilization for pancreatic cysts considered to have low risk for cancer as defined by the first consensus criteria. By characterizing this subpopulation of patients, we were able to identify possible factors that led to the decision to proceed with surgery. Better elucidation of these factors may provide insight into the daily challenges of managing these patients and further maximize the benefit from surgery.
J Gastrointest Surg
Methods
Results
Study Population
Between July 1, 2007, and June 30, 2012, 486 patients with pancreatic cysts were identified. Of these patients, 148 (30 %) underwent surgery of which 23 (16 %) were considered low risk by consensus criteria. A descriptive summary of patientand cyst-related characteristics of (1) all patients, (2) those who underwent surgery, (3) those who underwent surgery with positive Sendai criteria, (4) those who underwent surgery with negative Sendai criteria, and (5) those observed is provided in Table 1. Table 1 shows that when compared to those observed (n= 338), those who underwent surgery (n=148) were younger (median age 61 vs. 64, p=0.007), had larger cysts (median cyst size 30 vs. 17 mm, p=0.0001), had a higher proportion of intracystic mural nodules (15 vs. 2 %, p=0.0001), had a higher proportion of cysts with a dilated duct (8 vs. 1 %, p=0.0001), were more symptomatic (54 vs. 37 %, p=0.0004), had higher cyst fluid CEA levels (median 631 vs. 73 ng/ml, p=0.0001), and had a higher proportion of atypical cytology (19 vs. 8 %) (p=0.0003). There were no differences observed as to the gender distribution, location of cysts, the proportion presenting with multiple cysts, the presence of cyst septations, the presence of multicystic morphology, or microcystic morphology. Among those who underwent surgery, 125 patients met Sendai criteria and 23 patients did not. Table 1 shows that the only significant differences between these two groups were characteristics that define the Sendai criteria including cyst size (median 33 vs. 18 mm, p=0.0001), proportion of intracystic nodules (18 vs. 0 %, p=0.03), and the presence of associated symptoms (64 vs. 0 %, p=0.00001). No other significant differences were observed. Patients who did not meet Sendai criteria would generally be observed, and these 23 patients were compared to the 338 patients who were observed. Table 1 shows that these patients were younger (median age 59 vs. 64, p=0.035), less symptomatic (0 vs. 37 %, p=0.0003), had fewer multifocal cysts (35 vs. 56 %, p=0.04), and had a higher CEA level (median 758 vs. 73 ng/ml, p=0.003). Table 2 summarizes the histological distribution of pancreatic cysts. Of the 148 patients who underwent surgery, 125 (85 %) met preoperative Sendai criteria for resection. Among these patients, 9 (7 %) were cystadenocarcinomas, 18 (14 %) were main or mixed duct intraductal papillary mucinous neoplasms (IPMNs), 40 (32 %) were branch duct IPMNs, 15 (12 %) were mucinous cystic neoplasms (MCNs), 16 (13 %) were serous cystic neoplasms (SCNs), and 17 (14 %) were pseudocysts. Twenty-three patients (16 %) did not meet preoperative Sendai criteria and included 14 (61 %) branch duct IPMNs, 3 (13 %) MCNs, 1 (4 %) SCN, 1 (4.5 %) pseudocyst, 1 (4.5 %) squamous cyst, 1 (4.5 %) area of focal pancreatitis, and 2 (9 %) neuroendocrine tumors (NET). There were no cystadenocarcinomas or main duct/mixed IPMNs in this group.
With approval from the Stanford University Hospital and Clinics Institutional Review Board, we performed a retrospective chart review covering a 5-year period (July 1, 2007, to June 30, 2012) of all adult patients (≥18 years) referred for endoscopic ultrasound (EUS) or surgery for a pancreatic cyst at Stanford University Hospital and Clinics. All medical records were manually reviewed to collect specific patient- and cyst-related characteristics. Patient characteristics of interest included age, gender, and the presence of symptoms potentially related to a pancreatic cyst. These symptoms included documentation of abdominal pain, weight loss, jaundice, and the presence or history of pancreatitis. Cyst characteristics of interest included cyst size, the presence of mural nodularity, main pancreatic duct (PD) diameter, carcinoembryonic antigen (CEA) level, and cytology. Imaging characteristics were collected from available computed tomography (CT) scans, magnetic resonance imaging (MRI) scans, and/or EUS reports. When more than one imaging modality was available, the largest reported cyst size and PD diameter were collected. Cyst size was recorded as the maximum dimension measured on cross-sectional imaging. The main PD was considered dilated when it measured ≥6 mm. Mural nodularity was considered present when noted on any available imaging modality report. Other cyst-related features of interest that were collected included the presence of septations and multiple cysts and the location of the cyst within the pancreas (proximal=head, distal=body and tail). Cyst fluid CEA levels and cytology results were collected from patients that underwent EUS with fine-needle aspiration (FNA). If a patient had cyst fluid obtained more than once, the highest cyst fluid CEA level recorded was used for analysis. For analysis, the 2006 Sendai criteria were used to define high- and low-risk pancreatic cysts for harboring cancer.2
Statistics Statistical analyses were performed using Stata 13.1 (Stata Corp., College Station, TX, USA). Descriptive statistics were performed using non-parametric assumptions. Chi-square test was used for univariate analysis of predictive factors of surgical utilization. Multivariate analysis with logistic regression was used to identify factors independently associated with surgical utilization in cases that did not meet Sendai criteria. Only univariate variables with p value 6 mm Presence of symptoms Proximal cyst location Multifocal cysts
302 (62 %) 64 (54–72) 20 (12–30) 149 (31 %) 28 (6 %) 15 (3 %) 204 (47 %) 265 (55 %) 127 (27 %)
85 (57 %) 61 (51–70) 30 (18–50) 79 (54 %) 21 (15 %) 12 (8 %) 80 (54 %) 76 (51 %) 30 (20 %)
70 (56 %) 62 (50–72) 33 (20–51) 79 (63 %) 21 (18 %) 12 (10 %) 80 (64 %) 64 (51 %) 28 (22 %)
15 (65 %) 59 (52–65) 18 (14–21) 0 (0 %) 0 (0 %) 0 (0 %) 0 (0 %) 8 (35 %) 2 (9 %)
217 (64 %) 64 (54–74) 17 (10–25) 70 (21 %) 7 (2 %) 3 (1 %) 124 (37 %) 189 (56 %) 97 (29 %)
Cyst septations Multicystic Microcystic Median CEA level (IQR, N) Atypical cytology
165 (35 %) 46 (9 %) 31 (6 %) 138 (7–933, N=325) 54 (11 %)
46 (31 %) 16 (11 %) 8 (5 %) 631 (61–2650, N=96) 28 (19 %)
37 (30 %) 15 (12 %) 7 (5 %) 555 (48–2831, N=82) 27 (22 %)
9 (39 %) 1 (4 %) 1 (4 %) 758 (229 –1719, N=14) 1 (4 %)a
119 (35 %) 30 (9 %) 23 (7 %) 73 (5–471, N=229)
a
26 (8 %)
Patient had cytology interpretation Batypical^ because it was suspicious for mucinous neoplasm (not frank high-grade dysplasia or cancer)
A manual review of the chart for those patients who underwent surgery without meeting Sendai criteria was performed to assess for case-specific reasons for proceeding with surgery. Among these 23 patients (Fig. 1), these included (1) a high cyst CEA (35 %), (2) patient anxiety and preference (26 %), (3) physician suspicion for malignancy (22 %), (4) interval increase in cyst size (9 %), (5) an interval rise in CEA (4 %), and (6) unknown (4 %). A univariate analysis of non-Sendai criteria was performed to identify potential predictors for surgery (Table 3). No
significant differences in age, cyst location, cyst morphology, CEA level, or cytology differentiated operations that met Sendai criteria compared to those that did not. When the group that underwent surgery without meeting any Sendai criteria was compared to those observed, they were younger with a higher proportion less than the 65 years old (p=0.03) and had a higher CEA level (p=0.01). By multivariate analysis, an elevated CEA level >192 ng/ml was the most significant factor (OR 5.14 (95 % confidence interval (CI) 1.47–18.0)) to predict resection among low-risk cases defined by Sendai criteria compared to those observed (Table 4).
Table 2 Distribution of pancreatic cysts evaluated by endoscopic ultrasound and surgery over a 5-year period Total number of pancreatic cysts Underwent surgery Positive preoperative Sendai criteria Cyst type distribution Main/mixed IPMN Branch duct IPMN Mucinous cystic neoplasm Serous cystic neoplasm Pseudocyst Cancera Otherb
Discussion
486 148 (30 %) 125 (85 %) Sendai (+) (N=125) 18 (14 %) 40 (32 %) 15 (12 %) 16 (13 %) 17 (14 %) 9 (7 %) 10 (8 %)
With increasing recognition of the prevalence of pancreatic cysts,5, 6 appropriate decision-making between resection and Sendai (−) (N=23) 0 (0 %) 14 (61 %) 3 (13 %) 1 (4.5 %) 1 (4.5 %) 0 (0 %) 4 (17 %)
IPMN intraductal papillary mucinous neoplasm a
IPMN cancer/cystic adenocarcinoma (neuroendocrine tumors not included)
b The four other cases not meeting Sendai criteria were focal pancreatitis, benign squamous cyst, and two neuroendocrine tumors
Fig. 1 Distribution of reasons for surgery when Sendai criteria were not met
J Gastrointest Surg Table 3 Univariate analysis of non-Sendai criteria factors to predict surgical cases that met Sendai criteria compared to those that did not and to compare those who underwent surgery without meeting Sendai criteria versus those observed
Independent variables
Sendai (+) vs. Sendai (−) surgery cases
Sendai (−) surgery cases vs. observed
Age >65 Proximal cyst location Multifocal cysts Cyst septations Multicystic Microcystic CEA level >192 Atypical cytology
2.73 (0.96–7.84), p=0.06 0.45 (0.18–1.13), p=0.09 3.08 (0.68–14.0), p=0.15 0.65 (0.26–1.66), p=0.37 3.03 (0.38–24.2), p=0.30 1.31 (0.15–11.22), p=0.80 0.42 (0.11–1.50), p=0.18 4.96 (0.64–38.7), p=0.13
0.33 (0.12–0.90), p=0.03 2.38 (0.98–5.76), p=0.06 0.24 (0.06–1.07), p=0.06 1.27 (0.53–3.05), p=0.60 0.49 (0.06–3.75), p=0.50 0.65 (0.08–5.05), p=0.68 4.71 (1.37–16.2), p=0.01 0.76 (0.10–5.97), p=0.80
surveillance remains an increasingly important clinical dilemma. Without a safe and reliable biopsy technique, the 2006 Sendai consensus guidelines were written to delineate clinical and radiographic features associated with a high risk of malignancy.2 These features included (1) cyst size greater than 3 cm, (2) presence of an intracystic mural nodule, (3) presence of positive cytology, (4) presence of an associated dilated main duct, and (5) associated symptoms attributed to the cyst. In the absence of these features, surveillance was recommended with the caveat that, given the overall low quality of evidence, individualized physician judgment and patient preference to do otherwise were acceptable. Consequently, we sought to retrospectively evaluate the frequency of patients undergoing surgery for pancreatic cysts in the absence of high-risk features from 2007 to 2012. This period was specifically chosen because it allowed a 1-year period for guideline dissemination and ended just when the guidelines were revised.3 We observed that of the 148 patients who underwent surgical resection for a pancreatic cyst, 23 (16 %) patients did not have any of the five high-risk features as defined by the Sendai criteria. While none of these cysts harbored adenocarcinoma, there were five cysts (22 %) that proved to be appropriate for surgical resection by current guidelines that had they been accurately diagnosed preoperatively (three MCNs and two NETs). An in-depth medical chart review was performed of these 23 cases to determine explicit reasons for proceeding to surgery. An elevated cyst fluid CEA (35 %), patient anxiety and preference (26 %), and physician suspicion for malignancy (22 %) were the most commonly stated reasons for proceeding to surgery despite acknowledging the absence of high-risk features as defined by Sendai criteria. When compared to those patients with pancreatic cysts that underwent observation (n = 338), these patients were younger and had significantly elevated cyst CEA levels. Cyst CEA level in this group that underwent surgery despite an absence of high-risk features was highest (mean 64,961 ng/ml (95 % CI 0–195,948)) when compared to those observed (mean 2904 ng/ml (95 % CI 165–5,645)) (p65 Proximal cyst location Multifocal cysts CEA level >192
2.58 (0.89–7.43), p=0.08 0.47 (0.19–1.22), p=0.12
0.36 (0.13–1.02), p=0.05 1.92 (0.76–4.84), p=0.17
–
0.27 (0.06–1.20, p=0.09
–
5.14 (1.47–18.0), p=0.01
Only variables with a p value
