J Gastrointest Surg (2014) 18:2009–2015 DOI 10.1007/s11605-014-2651-9
ORIGINAL ARTICLE
Assessment of a Complication Risk Score and Study of Complication Profile in Laparoscopic Distal Pancreatectomy Giuseppe Malleo & Roberto Salvia & Giuseppe Mascetta & Alessandro Esposito & Luca Landoni & Luca Casetti & Laura Maggino & Claudio Bassi & Giovanni Butturini
Received: 10 May 2014 / Accepted: 29 August 2014 / Published online: 20 September 2014 # 2014 The Society for Surgery of the Alimentary Tract
Abstract Objective This study assessed the patient-specific risk for major postoperative morbidity in a series of 100 laparoscopic distal pancreatectomies (LDP). Methods A previously established complication risk score (CRS), identifying body mass index (BMI), estimated blood loss (EBL), and pancreatic specimen length as determinants of postoperative morbidity were examined against the observed outcomes. In addition, multivariate analyses were performed to investigate risk factors specific to our study population. Results The postoperative morbidity rate was 49 %, major complication accounted for 12 %, and clinically relevant pancreatic fistulae (PF) were 13 %. The incidence of any complications, major complications, any PF, and clinically relevant PF did not vary appreciably when the CRS increased. The multivariate analysis indicated that male sex and an EBL ≥150 mL were independent predictors of major morbidity and clinically relevant PF. Conclusion In conclusion, the previously published CRS based on pre- and intraoperative factors was not able to predict the postoperative risk in our population. This is probably because risk scores may not be able to adjust for the case-mix (heterogeneity in baseline patient characteristics). According to our data, men and patients with EBL ≥150 mL are more likely to develop major postoperative complications after LDP. Keywords Laparoscopic distal pancreatectomy . Pancreatic fistula . Postoperative complications . Minimally invasive surgery
Introduction Direct outcome assessment has long been a staple in measuring the quality of surgical care. In recent years, the increasing application of minimally invasive techniques for the management of lesions located in the left pancreas has led to a number of studies assessing the complication profile of laparoscopic distal pancreatectomy (LDP), especially in comparison with the open procedure1–3. However, the number of centers performing a G. Malleo : R. Salvia : G. Mascetta : A. Esposito : L. Landoni : L. Casetti : L. Maggino : C. Bassi : G. Butturini (*) Unit of Surgery B, The Pancreas Institute, Department of Surgery and Oncology, G.B. Rossi Hospital, University of Verona Hospital Trust, P.Le L.A. Scuro 10, 37134 Verona, Italy e-mail: [email protected]
consistent number of laparoscopic pancreatic resections is small, such that only a few papers accrued 100 or more LDPs. As the international movements toward quality outcome reporting continues, it is essential that the effect of patient-specific risk assessment for postoperative morbidity and mortality is understood and evaluated. In this regard, records from patients undergoing LDP at nine American academic medical centers were recently analyzed to produce a model that identifies those at risk of postoperative adverse events, help in discussing patient-specific risk during the informed consent process, and allow comparison of perioperative outcomes between institutions. The resulting complication risk score (CRS) identified body mass index (BMI), estimated blood loss (EBL), and pancreatic specimen length as determinants of postoperative morbidity4 and was partially validated by a single institution5. The objective of this study was to investigate the complication profile and to examine this CRS against the observed outcomes in a series of LDP performed at a single, high-volume European center for pancreatic surgery. Furthermore, the risk factors for major morbidity specific to our study population were investigated.
2010
J Gastrointest Surg (2014) 18:2009–2015
Materials and Methods
Study of the Complication Profile
Study Design and Operative Details
Study of complication profile in the study population was carried out using univariate and multivariate analyses. For the univariate analysis, the dependent variables were the same as the paper by Weber et al.4. Continuous variables were dichotomized around the median value, and the chi-squared test (with Yates continuity correction in a 2×2 contingency table) was used to compare frequencies between groups. Fisher’s exact test was used when appropriate. The multivariate analysis was performed employing binary logistic regression. Two models were built using major complications and clinically relevant pancreatic fistula as dependent dichotic variables. Factors with a potential prognostic significance were entered in a hierarchical fashion (block 1=preoperative factors, block 2=intraoperative factors, univariate inclusion criteria of p8 cm, or estimated blood loss >150 mL4. We calculated the CRS (range 0–3) for each patient in our series and assessed whether the score correlated with the occurrence of any complications, major complications, any pancreatic fistula, and clinically relevant pancreatic fistula using chi-squared analysis. Furthermore, the observed incidence of the aforementioned endpoints was compared with the expected incidence, calculated from the original paper by Weber et al. The observed to expected (O/E) ratio was obtained dividing observed by expected cases. The 95 % confidence interval of the O/E ratio was calculated using the Wilson and Hilferty approximation of the exact Poisson distribution. The O/E ratio was considered significant when the 95 % confidence interval did not contain unity.
Results Patient Characteristics The study population consisted of 100 patients. Table 1 outlines the demographic and operative details. Conversion to an open procedure was necessary in two cases because of an uncontrolled bleeding. The remaining 98 procedures were completed laparoscopically. Reoperation within the index admission was necessary in 10 patients. Five patients underwent drainage of infected collections (three procedures were performed laparoscopically and in one patient, concomitant splenectomy was necessary because the abscess diffusely involved the splenic hilum). Three patients underwent reexploration and hemostasis due to PPH (one procedure was performed laparoscopically). Two patients who underwent Warshaw’s procedure developed a splenic infarction and needed splenectomy, both on postoperative day 10. Postoperative mortality was nil. Assessment of the Complication Risk Score Forty-three patients had none of the complication risk factors, 27 patients had a CRS of 1, 29 had a CRS of 2, and only one had a CRS of 3. The only patient with a CRS of 3 had an uneventful postoperative course. Table 2 shows that the CRS did not correlate with any of the four outcome measures. In
J Gastrointest Surg (2014) 18:2009–2015
2011
Table 1 Clinical, surgical, and pathologic characteristics of the study population Study population (N=100) Sex Male Female Mean age (SD) Mean BMI (SD) Diagnosis Incidental Symptoms Mean serum Ca 19.9 (SD) Tumor location Body Tail Mean tumor size, mm (SD) Type of operation Distal pancreatectomy with splenectomy Spleen-preserving distal pancreatectomy Mean operative time, minutes (SD) Conversion to open procedures Yes No Pancreas transection Stapler Ultrasonic devices Sutures Stapler thickness (n=88) 35–38 mm 48 mm Use of staple line reinforcement (n=88) Yes No Intraoperative transfusion Yes No Estimated blood loss 48 years and a BMI >23.12 kg/m2 were associated with any complications, but there was no factor associated with major morbidity (Clavien-Dindo ≥3). Only male sex and an Table 3 Observed versus expected outcomes after laparoscopic distal pancreatectomy according to the complication risk score proposed by Weber et al.4 Complication risk score† Observed Expected O/E ratio 95 % CI Any complication 0 1 2 3 Major complication 0 1 2 3 Any pancreatic fistula 0 1 2 3 CR pancreatic fistula 0 1 2 3
20 14 15 0
6.02 9.18 12.47 0.56
3.32 1.53 1.20 0
2.03–5.13a 0.83–2.56 0.67–1.98 –
3 4 5 0
1.29 0 3.48 0.3
2.33 – 1.44 0
0.47–6.79 – 0.46–3.35 –
10 10 7 0
1.29 5.13 9.28 0.41
7.75 1.95 0.75 0
3.71–14.26a 0.93–3.59 0.30–1.55 –
3 6 4 0
0 0.54 3.77 0.33
– 11.1 1.06 0
– 4.06–24.18a 0.29–2.72 –
O/E observed/expected a
Statistically significant deviation from the expected frequency
%) %) %) %)
Clinically relevant fistula Yes (N=27) p
No (N=87)
10 (23.3 %) 0.537 40 (93.0 %) 10 (37.0 %) 21 (77.8 %) 7 (24.1 %) 25 (86.2 %) 0 1 (100.0 %)
Yes (N=13) p 3 (7.0 %) 6 (22.2 %) 4 (13.8 %) 0
0.311
estimated blood loss >150 mL were associated with clinically relevant pancreatic fistula. The analysis was repeated without dichotomizing continuous variables (comparison of means), and no factor resulted to be associated with any of the four endpoints. On multivariate analysis, male sex and an estimated blood loss >150 mL were associated with a greater probability of major complications (OR of 4.550 and 4.472). The second model showed the same two factors to be independent predictors of clinically relevant pancreatic fistulae (OR of 7.589 and 5.318). There were no outliers or collinearity of independent variables. Results of the multivariate analysis are showed in Table 5.
Discussion Despite the increasing application of minimally invasive techniques, LDP is not frequently performed at many institutions. As an example, data from the USA indicate that laparoscopy was utilized in 15 % of distal pancreatic resections from the Nationwide Inpatient Sample (NIS, 2003–2009) and in 27 % of the National Surgical Quality Improvement Project database (NSQIP, 2005–2010)12. Hence, the quality assessment and the risk stratification for patients undergoing LDP is based on a limited number of studies. The present paper indicates that LDP is a major surgical procedure, associated with an overall postoperative morbidity rate of 49 % and a pancreatic fistula rate of 27 %. However, nearly half of the pancreatic fistulae (13/27) were grade A (and thus clinically irrelevant), and major complications, defined according to the Clavien and Dindo classification, accounted for only 12 % of cases. Because almost all the procedures were performed for benign or low-grade neoplasms, it is likely that a soft pancreas texture has critically contributed to the overall fistula rate, although in minimally invasive surgery it is not even possible to assess manually the pancreatic parenchyma. The principal aim of our analysis was to assess a procedure-specific CRS established from a multiinstitutional data set of 219 LDP performed at nine American academic centers. This CRS, consisting of 1 point each for BMI >27 kg/m2, pancreatic specimen length >8 cm, or EBL >150 mL, was shown to be able to stratify patients into risk
0.428
1.000
1.000
0.111
0.518
0.520
0.509
0.678
0.357
23 (46.9 %) 26 (53.1 %) 22 (44.9 %) 27 (55.1 %) 27 (55.1 %) 22 (44.9 %) 21 (42.9 %) 28 (57.1 %) 31 (63.3 %) 18 (36.7 %) 42 (87.5 %) 6 (12.5 %) 24 (57.1 %) 18 (42.9 %) 40 (95.2 %) 2 (4.8 %) 46 (93.9 %) 3 (6.1 %) 27 (55.1 %) 22 (44.9 %)
Incidental 29 (56.9 %) Symptoms 22 (43.1 %) Tumor location Body 23 (45.1 %) Tail 28 (54.9 %) Tumor size ≤35 mm 28 (54.9 %) >35 mm 23 (45.1 %) Operative time ≤200 min 31 (60.8 %) >200 min 20 (39.2 %) Operation type DP-S 28 (54.9 %) SPDP 23 (45.1 %) Pancreas transection (n=98) Stapler 46 (92.0 %) Ultrasonic 4 (8.0 %) Stapler thickness (n=88) 35–38 mm 22 (47.8 %) 48 mm 24 (52.2 %) Staple line reinforcement (n=88) No 42 (91.3 %) Yes 4 (8.7 %) Intraoperative transfusion No 50 (98.0 %) Yes 1 (2.0 %) Estimated blood loss 32 (62.7 %) 48 BMI ≤23.12 kg/m2 >23.12 kg/m2 Diagnosis
Sex Female Male Age
Variable
0.066
0.405
0.591
1.000
0.355
1.000
0.649
0.578
0.951
0.649
1.000
0.356
0.122
p
70 (95.9 %) 3 (4.1 %) 45 (61.4 %) 28 (38.4 %)
58 (90.6 %) 6 (9.4 %)
32 (50.0 %) 32 (50.0 %)
64 (88.9 %) 8 (11.1 %)
42 (57.5 %) 31 (42.5 %)
39 (53.4 %) 34 (46.6 %)
39 (53.4 %) 34 (46.6 %)
31 (42.5 %) 42 (57.5 %)
39 (53.4 %) 34 (46.6 %)
40 (54.8 %) 33 (45.2 %)
39 (53.4 %) 34 (46.6 %)
60 (82.2 %) 13 (17.8 %)
No (N=73)
Any fistula
Table 4 Univariate analysis of factors associated with postoperative complications after laparoscopic distal pancreatectomy
26 (96.3 %) 1 (3.7 %) 14 (51.9 %) 13 (48.1 %)
24 (100.0 %) 0
14 (58.3 %) 10 (41.7 %)
24 (92.3 %) 2 (7.7 %)
17 (63.0 %) 10 (37.0 %)
13 (48.1 %) 14 (51.9 %)
16 (59.3 %) 11 (40.7 %)
14 (51.9 %) 13 (48.1 %)
13 (48.1 %) 14 (51.9 %)
11 (40.7 %) 16 (59.3 %)
11 (40.7 %) 16 (59.3 %)
19 (70.4 %) 8 (29.6 %)
Yes (N=27)
0.513
1.000
0.183
0.647
0.726
0.794
0.808
0.769
0.541
0.808
0.306
0.368
0.312
p
84 (96.6 %) 3 (3.4 %) 55 (63.2 %) 32 (36.8 %)
70 (92.1 %) 6 (7.9 %)
41 (53.9 %) 35 (46.1 %)
76 (88.4 %) 10 (11.6 %)
50 (57.5 %) 37 (42.5 %)
47 (54.0 %) 40 (46.0 %)
46 (52.9 %) 41 (47.1 %)
38 (43.7 %) 49 (56.3 %)
48 (55.2 %) 39 (44.8 %)
46 (52.9 %) 41 (47.1 %)
46 (52.9 %) 41 (47.1 %)
72 (82.8 %) 15 (17.2 %)
No (N=87)
12 (92.3 %) 1 (7.7 %) 4 (30.8 %) 9 (69.2 %)
12 (100.0 %) 0
5 (41.7 %) 7 (58.3 %)
12 (100.0 %) 0
9 (69.2 %) 4 (30.8 %)
5 (38.5 %) 8 (61.5 %)
9 (69.2 %) 4 (30.8 %)
7 (53.8 %) 6 (42.6 %)
4 (30.8 %) 9 (69.2 %)
5 (38.5 %) 8 (61.5 %)
4 (30.8 %) 9 (69.2 %)
7 (53.8 %) 6 (42.6 %)
Yes (N=13)
Clinically relevant fistula
0.050
0.432
0.591
0.631
0.355
0.616
0.453
0.420
0.698
0.179
0.501
0.234
0.027
p
J Gastrointest Surg (2014) 18:2009–2015 2013
0.364 5 (38.5 %) 8 (61.5 %) 49 (56.3 %) 38 (43.7 %) 1.000 26 (53.1 %) 23 (46.9 %)
49 (55.7 %) 39 (44.3 %)
5 (41.7 %) 7 (58.3 %)
0.545
40 (54.8 %) 33 (45.2 %)
14 (51.9 %) 13 (48.1 %)
0.971
7 (53.8 %) 4 (30.8 %) 2 (15.4 %) 0 7 (53.8 %) 6 (46.2 %) 64 (73.6 %) 20 (23.0 %) 2 (2.3 %) 1 (1.1 %) 65 (74.7 %) 22 (25.3 %) 0.519 0.330 17 (63.0 %) 8 (29.6 %) 2 (7.4 %) 0 17 (63.0 %) 10 (37.0 %) 54 (74.0 %) 16 (21.9 %) 2 (2.7 %) 1 (1.4 %) 55 (75.3) 18 (24.7 %) 0.750 0.735 8 (66.7 %) 4 (33.3 %) 0 0 8 (66.7 %) 4 (33.3 %) 0.479 0.428 33 (67.3 %) 13 (26.5 %) 3 (6.1 %) 0 33 (67.3 %) 16 (32.7 %)
63 (71.6 %) 20 (22.7 %) 4 (4.5 %) 1 (1.1 %) 64 (72.7 %) 24 (27.3 %)
Yes (N=13) No (N=87) p Yes n=49
No (N=88)
Yes (N=12)
p
No (N=73)
Yes (N=27)
p
classes for different outcome metrics, including major complications and major pancreatic fistula4. Major complications varied from 3 to 30 % when the complication risk score increased from 0 to 3, respectively, and the risk of clinically relevant pancreatic fistulae increased from 0 to 30 %4. This CRS supported the concept that that a combination of patient and tumor factors is the main determinant of procedurespecific complications and that patients with obese body habitus and larger or more proximal tumors requiring extended resections have the greatest likelihood of developing pancreatic fistulae and other associated complications. The CRS was recently applied at a single institution, and not only was shown to predict overall complications and fistulae as originally published but also resulted to be a reliable tool for assessing other adverse perioperative outcomes. However, several cases of this validation set had been used for the development of the original CRS, such that the patients’ overlap may have prevented a true validation of the score. Remarkably, a subset analysis excluding those patients used in the original formulation of the CRS only partially validated the findings5. The authors concluded that study of the CRS in a non-overlapping patient population was necessary for validation of this clinical tool. In terms of major complications and clinically relevant fistulae, our results are comparable with the American multiinstitutional experience, which reported a major complication rate of 11 %, and a clinically relevant pancreatic fistula rate of 10 %. The greater incidence of any complications observed in our series (49 versus 39 %) is attributable to pulmonary complications (pleural effusion and pneumonia), which accounted for 22 %. When applying the CRS to our study population, we found surprisingly that an increasing score did not correlate with the observed outcomes. In particular, 20/49 patients with any complications and 10/27 patients with a pancreatic fistula had a score of zero. The rate of complications and fistulae did not vary appreciably when the CRS increased. In addition, the observed morbidity and fistula rate Table 5 Multivariate analysis assessing independent predictors of major complications and clinically relevant pancreatic fistula after laparoscopic distal pancreatectomy
28 (54.9 %) 23 (45.1 %)
38 (74.5 %) 11 (21.6 %) 1 (2.0 %) 1 (2.0 %) 39 (76.5 %) 12 (23.5 %)
No n=51
Factor
≥150 mL Pathologic diagnosis PCN NEN PDAC Other Cystic Solid Specimen length ≤70 mm >70 mm
Clinically relevant fistula Any fistula Major complications Any complications Variable
Table 4 (continued)
0.117 0.182
J Gastrointest Surg (2014) 18:2009–2015
p
2014
p value
OR (Exp[B])
95 % CI
Model 1. Dependent dichotic variable = major complications Hosmer-Lemeshow test p=0.955, prediction accuracy of 89.7 % Sex (M) 0.032 4.550 1.139–18.169 EBL (≥150 mL) 0.032 4.472 1.133–17.650 Model 2. Dependent dichotic variable = clinically relevant pancreatic fistula Hosmer-Lemeshow test p=0.912, prediction accuracy of 88.6 % Sex (M) 0.005 7.589 1.252–12.590 EBL (≥150 mL) 0.024 5.318 1.819–11.657
J Gastrointest Surg (2014) 18:2009–2015
in patients with a CRS=0 were significantly greater than expected according to the report by Weber et al. Similarly, clinically relevant pancreatic fistulae were more frequent than expected in patients with a CRS=1 (O/E ratio of 11.1), and the incidence did not increase when the CRS was greater. Of note, the CRS (0=low risk versus 1–2=intermediate risk) was not predictive of other adverse outcomes, such as longer operative time, percutaneous drain insertion, and length of hospital stay. The poor performance of the CRS may be due to difference in baseline parameters among institutions, such as the proportion of patients with a BMI >27kg/m2 (34.8 % in the paper by Kneuertz et al. and 16 % in this series). This may have caused an inhomogeneous distribution of patients into risk classes, as suggested by the fact that only one individual in our series had a CRS of 3 (high risk of complications), whereas in the paper by Kneuertz et al., the proportion of patients at high risk was substantially greater than expected, indicating a greater operative complexity5. We then sought to evaluate in our study population the risk factors for the same outcome measures; in the univariate analysis, continuous variables were dichotomized around the median value. Only age >48 years and a BMI greater than 23.1 kg/m2 resulted to be associated with any postoperative complications, whereas no factor was found to be associated with major complications. When limiting the analysis to pancreatic fistula, only male sex and an EBL >150 mL were associated with a clinically relevant event. Different subanalyses were performed using the cutoffs proposed by Weber et al. for continuous variables, or even without dichotomizing the variables around the median value; but no significant associations were found. Although the univariate analysis showed only a small number of significant associations, we constructed a multivariate model with univariate inclusion criteria of p150 mL were independent predictors of major morbidity. We did not try to customize the CRS or to construct an alternative complication score, because a sample size of 100 cases would lead to an underpowered model. In conclusion, although the rate of major complications and clinically relevant pancreatic fistulae was similar to that observed by other case series, the previously published CRS based on pre- and intraoperative factors was not able to predict the postoperative risk in our study population. This is probably because risk scores may not be able to adjust for the casemix (heterogeneity in the baseline characteristics of the index population) and may not be optimally calibrated across groups. In our series, males were more likely to be older and with more comorbidities, thus being at greater risk for
2015
postoperative morbidity. The other prognostic factor was EBL, which serves as a surrogate marker of technical operative complexity and hampers optimal visualization of the surgical field. As the number of procedures performed across institutions becomes greater, robust modeling with large development and validation samples is needed to better identify the determinants of complications after LDP, which—despite being a major procedure—remains in expert hands a safe operation with zero mortality. Financial Support Giuseppe Malleo was supported by the Fondazione Italiana per la ricerca sulle malattie del pancreas (FIMP).
References 1. Venkat R, Edil BH, Schulick RD, Lidor AO, Makary MA, Wolfgang CL. Laparoscopic distal pancreatectomy is associated with significantly less overall morbidity compared to the open technique: a systematic review and meta-analysis. Ann Surg 2012;255:1048– 1059. 2. Pericleous S, Middleton N, McKay SC, Bowers KA, Hutchins RR. Systematic Review and Meta-Analysis of Case-Matched Studies Comparing Open and Laparoscopic Distal Pancreatectomy. Pancreas 2012;41:993–1000. 3. Abu Hilal M, Takhar AS. Laparoscopic left pancreatectomy: current concepts. Pancreatology 2013;13:443–448. 4. Weber SM, Cho CS, Merchant N, Pinchot S, Rettammel R, Nakeeb A, Bentrem D, Parikh A, Mazo AE, Martin RCG, Scoggins CR, Ahmad SA, Kim HJ, Hamilton N, Hawkins W, Schmidt CM, Kooby DA. Laparoscopic left pancreatectomy: complication risk score correlates with morbidity and risk for pancreatic fistula. Ann Surg Oncol 2009; 6:2825–2833. 5. Kneuertz PJ, Patel SH, Chu CK, Fisher SB, Maithel SK, Sarmiento JM, Weber SM, Staley CA, Kooby DA. Laparoscopic Distal Pancreatectomy: Trends and Lessons Learned Through an 11-Year Experience. J Am Coll Surg 2012;215:167–176. 6. Melotti G, Butturini G, Piccoli M, Casetti L, Bassi C, Mullineris B, Lazzaretti MG, Pederzoli P. Laparoscopic Distal Pancreatectomy. Ann Surg 2007;246:77–82. 7. Kimura W, Inoue T, Futakawa N, Shinkai H, Han I, Muto T. Spleenpreserving distal pancreatectomy with conservation of the splenic artery and vein. Surgery 1996;120:885–890. 8. Ferrone CR, Konstantinidis IT, Sahani DV, Wargo JA, Fernandez-del Castillo C, Warshaw AL. Twenty-Three Years of the Warshaw Operation for Distal Pancreatectomy With Preservation of the Spleen. Ann Surg 2011;253:1136–1139. 9. DeOliveira ML, Winter JM, Schäfer M, Cunningham SC, Cameron JL, Yeo CJ, Clavien PA. Assessment of Complications After Pancreatic Surgery. Ann Surg 2006;244:931–937. 10. Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J, Neoptolemos J, Sarr M, Traverso W, Buchler M. Postoperative pancreatic fistula: An international study group (ISGPF) definition. Surgery 2005;138:8–13. 11. Wente MN, Veit JA, Bassi C, Dervenis C, Fingerhut A, Gouma DJ, Izbicki JR, Neoptolemos JP, Padbury RT, Sarr MG; Yeo CJ, Buchler MW. Postpancreatectomy hemorrhage (PPH)–An International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery 2007;142: 20–25. 12. Rosales-Velderrain A. National trends in resection of the distal pancreas. World J Gastroenterol 2012;18:4342.
ORIGINAL ARTICLE
Assessment of a Complication Risk Score and Study of Complication Profile in Laparoscopic Distal Pancreatectomy Giuseppe Malleo & Roberto Salvia & Giuseppe Mascetta & Alessandro Esposito & Luca Landoni & Luca Casetti & Laura Maggino & Claudio Bassi & Giovanni Butturini
Received: 10 May 2014 / Accepted: 29 August 2014 / Published online: 20 September 2014 # 2014 The Society for Surgery of the Alimentary Tract
Abstract Objective This study assessed the patient-specific risk for major postoperative morbidity in a series of 100 laparoscopic distal pancreatectomies (LDP). Methods A previously established complication risk score (CRS), identifying body mass index (BMI), estimated blood loss (EBL), and pancreatic specimen length as determinants of postoperative morbidity were examined against the observed outcomes. In addition, multivariate analyses were performed to investigate risk factors specific to our study population. Results The postoperative morbidity rate was 49 %, major complication accounted for 12 %, and clinically relevant pancreatic fistulae (PF) were 13 %. The incidence of any complications, major complications, any PF, and clinically relevant PF did not vary appreciably when the CRS increased. The multivariate analysis indicated that male sex and an EBL ≥150 mL were independent predictors of major morbidity and clinically relevant PF. Conclusion In conclusion, the previously published CRS based on pre- and intraoperative factors was not able to predict the postoperative risk in our population. This is probably because risk scores may not be able to adjust for the case-mix (heterogeneity in baseline patient characteristics). According to our data, men and patients with EBL ≥150 mL are more likely to develop major postoperative complications after LDP. Keywords Laparoscopic distal pancreatectomy . Pancreatic fistula . Postoperative complications . Minimally invasive surgery
Introduction Direct outcome assessment has long been a staple in measuring the quality of surgical care. In recent years, the increasing application of minimally invasive techniques for the management of lesions located in the left pancreas has led to a number of studies assessing the complication profile of laparoscopic distal pancreatectomy (LDP), especially in comparison with the open procedure1–3. However, the number of centers performing a G. Malleo : R. Salvia : G. Mascetta : A. Esposito : L. Landoni : L. Casetti : L. Maggino : C. Bassi : G. Butturini (*) Unit of Surgery B, The Pancreas Institute, Department of Surgery and Oncology, G.B. Rossi Hospital, University of Verona Hospital Trust, P.Le L.A. Scuro 10, 37134 Verona, Italy e-mail: [email protected]
consistent number of laparoscopic pancreatic resections is small, such that only a few papers accrued 100 or more LDPs. As the international movements toward quality outcome reporting continues, it is essential that the effect of patient-specific risk assessment for postoperative morbidity and mortality is understood and evaluated. In this regard, records from patients undergoing LDP at nine American academic medical centers were recently analyzed to produce a model that identifies those at risk of postoperative adverse events, help in discussing patient-specific risk during the informed consent process, and allow comparison of perioperative outcomes between institutions. The resulting complication risk score (CRS) identified body mass index (BMI), estimated blood loss (EBL), and pancreatic specimen length as determinants of postoperative morbidity4 and was partially validated by a single institution5. The objective of this study was to investigate the complication profile and to examine this CRS against the observed outcomes in a series of LDP performed at a single, high-volume European center for pancreatic surgery. Furthermore, the risk factors for major morbidity specific to our study population were investigated.
2010
J Gastrointest Surg (2014) 18:2009–2015
Materials and Methods
Study of the Complication Profile
Study Design and Operative Details
Study of complication profile in the study population was carried out using univariate and multivariate analyses. For the univariate analysis, the dependent variables were the same as the paper by Weber et al.4. Continuous variables were dichotomized around the median value, and the chi-squared test (with Yates continuity correction in a 2×2 contingency table) was used to compare frequencies between groups. Fisher’s exact test was used when appropriate. The multivariate analysis was performed employing binary logistic regression. Two models were built using major complications and clinically relevant pancreatic fistula as dependent dichotic variables. Factors with a potential prognostic significance were entered in a hierarchical fashion (block 1=preoperative factors, block 2=intraoperative factors, univariate inclusion criteria of p8 cm, or estimated blood loss >150 mL4. We calculated the CRS (range 0–3) for each patient in our series and assessed whether the score correlated with the occurrence of any complications, major complications, any pancreatic fistula, and clinically relevant pancreatic fistula using chi-squared analysis. Furthermore, the observed incidence of the aforementioned endpoints was compared with the expected incidence, calculated from the original paper by Weber et al. The observed to expected (O/E) ratio was obtained dividing observed by expected cases. The 95 % confidence interval of the O/E ratio was calculated using the Wilson and Hilferty approximation of the exact Poisson distribution. The O/E ratio was considered significant when the 95 % confidence interval did not contain unity.
Results Patient Characteristics The study population consisted of 100 patients. Table 1 outlines the demographic and operative details. Conversion to an open procedure was necessary in two cases because of an uncontrolled bleeding. The remaining 98 procedures were completed laparoscopically. Reoperation within the index admission was necessary in 10 patients. Five patients underwent drainage of infected collections (three procedures were performed laparoscopically and in one patient, concomitant splenectomy was necessary because the abscess diffusely involved the splenic hilum). Three patients underwent reexploration and hemostasis due to PPH (one procedure was performed laparoscopically). Two patients who underwent Warshaw’s procedure developed a splenic infarction and needed splenectomy, both on postoperative day 10. Postoperative mortality was nil. Assessment of the Complication Risk Score Forty-three patients had none of the complication risk factors, 27 patients had a CRS of 1, 29 had a CRS of 2, and only one had a CRS of 3. The only patient with a CRS of 3 had an uneventful postoperative course. Table 2 shows that the CRS did not correlate with any of the four outcome measures. In
J Gastrointest Surg (2014) 18:2009–2015
2011
Table 1 Clinical, surgical, and pathologic characteristics of the study population Study population (N=100) Sex Male Female Mean age (SD) Mean BMI (SD) Diagnosis Incidental Symptoms Mean serum Ca 19.9 (SD) Tumor location Body Tail Mean tumor size, mm (SD) Type of operation Distal pancreatectomy with splenectomy Spleen-preserving distal pancreatectomy Mean operative time, minutes (SD) Conversion to open procedures Yes No Pancreas transection Stapler Ultrasonic devices Sutures Stapler thickness (n=88) 35–38 mm 48 mm Use of staple line reinforcement (n=88) Yes No Intraoperative transfusion Yes No Estimated blood loss 48 years and a BMI >23.12 kg/m2 were associated with any complications, but there was no factor associated with major morbidity (Clavien-Dindo ≥3). Only male sex and an Table 3 Observed versus expected outcomes after laparoscopic distal pancreatectomy according to the complication risk score proposed by Weber et al.4 Complication risk score† Observed Expected O/E ratio 95 % CI Any complication 0 1 2 3 Major complication 0 1 2 3 Any pancreatic fistula 0 1 2 3 CR pancreatic fistula 0 1 2 3
20 14 15 0
6.02 9.18 12.47 0.56
3.32 1.53 1.20 0
2.03–5.13a 0.83–2.56 0.67–1.98 –
3 4 5 0
1.29 0 3.48 0.3
2.33 – 1.44 0
0.47–6.79 – 0.46–3.35 –
10 10 7 0
1.29 5.13 9.28 0.41
7.75 1.95 0.75 0
3.71–14.26a 0.93–3.59 0.30–1.55 –
3 6 4 0
0 0.54 3.77 0.33
– 11.1 1.06 0
– 4.06–24.18a 0.29–2.72 –
O/E observed/expected a
Statistically significant deviation from the expected frequency
%) %) %) %)
Clinically relevant fistula Yes (N=27) p
No (N=87)
10 (23.3 %) 0.537 40 (93.0 %) 10 (37.0 %) 21 (77.8 %) 7 (24.1 %) 25 (86.2 %) 0 1 (100.0 %)
Yes (N=13) p 3 (7.0 %) 6 (22.2 %) 4 (13.8 %) 0
0.311
estimated blood loss >150 mL were associated with clinically relevant pancreatic fistula. The analysis was repeated without dichotomizing continuous variables (comparison of means), and no factor resulted to be associated with any of the four endpoints. On multivariate analysis, male sex and an estimated blood loss >150 mL were associated with a greater probability of major complications (OR of 4.550 and 4.472). The second model showed the same two factors to be independent predictors of clinically relevant pancreatic fistulae (OR of 7.589 and 5.318). There were no outliers or collinearity of independent variables. Results of the multivariate analysis are showed in Table 5.
Discussion Despite the increasing application of minimally invasive techniques, LDP is not frequently performed at many institutions. As an example, data from the USA indicate that laparoscopy was utilized in 15 % of distal pancreatic resections from the Nationwide Inpatient Sample (NIS, 2003–2009) and in 27 % of the National Surgical Quality Improvement Project database (NSQIP, 2005–2010)12. Hence, the quality assessment and the risk stratification for patients undergoing LDP is based on a limited number of studies. The present paper indicates that LDP is a major surgical procedure, associated with an overall postoperative morbidity rate of 49 % and a pancreatic fistula rate of 27 %. However, nearly half of the pancreatic fistulae (13/27) were grade A (and thus clinically irrelevant), and major complications, defined according to the Clavien and Dindo classification, accounted for only 12 % of cases. Because almost all the procedures were performed for benign or low-grade neoplasms, it is likely that a soft pancreas texture has critically contributed to the overall fistula rate, although in minimally invasive surgery it is not even possible to assess manually the pancreatic parenchyma. The principal aim of our analysis was to assess a procedure-specific CRS established from a multiinstitutional data set of 219 LDP performed at nine American academic centers. This CRS, consisting of 1 point each for BMI >27 kg/m2, pancreatic specimen length >8 cm, or EBL >150 mL, was shown to be able to stratify patients into risk
0.428
1.000
1.000
0.111
0.518
0.520
0.509
0.678
0.357
23 (46.9 %) 26 (53.1 %) 22 (44.9 %) 27 (55.1 %) 27 (55.1 %) 22 (44.9 %) 21 (42.9 %) 28 (57.1 %) 31 (63.3 %) 18 (36.7 %) 42 (87.5 %) 6 (12.5 %) 24 (57.1 %) 18 (42.9 %) 40 (95.2 %) 2 (4.8 %) 46 (93.9 %) 3 (6.1 %) 27 (55.1 %) 22 (44.9 %)
Incidental 29 (56.9 %) Symptoms 22 (43.1 %) Tumor location Body 23 (45.1 %) Tail 28 (54.9 %) Tumor size ≤35 mm 28 (54.9 %) >35 mm 23 (45.1 %) Operative time ≤200 min 31 (60.8 %) >200 min 20 (39.2 %) Operation type DP-S 28 (54.9 %) SPDP 23 (45.1 %) Pancreas transection (n=98) Stapler 46 (92.0 %) Ultrasonic 4 (8.0 %) Stapler thickness (n=88) 35–38 mm 22 (47.8 %) 48 mm 24 (52.2 %) Staple line reinforcement (n=88) No 42 (91.3 %) Yes 4 (8.7 %) Intraoperative transfusion No 50 (98.0 %) Yes 1 (2.0 %) Estimated blood loss 32 (62.7 %) 48 BMI ≤23.12 kg/m2 >23.12 kg/m2 Diagnosis
Sex Female Male Age
Variable
0.066
0.405
0.591
1.000
0.355
1.000
0.649
0.578
0.951
0.649
1.000
0.356
0.122
p
70 (95.9 %) 3 (4.1 %) 45 (61.4 %) 28 (38.4 %)
58 (90.6 %) 6 (9.4 %)
32 (50.0 %) 32 (50.0 %)
64 (88.9 %) 8 (11.1 %)
42 (57.5 %) 31 (42.5 %)
39 (53.4 %) 34 (46.6 %)
39 (53.4 %) 34 (46.6 %)
31 (42.5 %) 42 (57.5 %)
39 (53.4 %) 34 (46.6 %)
40 (54.8 %) 33 (45.2 %)
39 (53.4 %) 34 (46.6 %)
60 (82.2 %) 13 (17.8 %)
No (N=73)
Any fistula
Table 4 Univariate analysis of factors associated with postoperative complications after laparoscopic distal pancreatectomy
26 (96.3 %) 1 (3.7 %) 14 (51.9 %) 13 (48.1 %)
24 (100.0 %) 0
14 (58.3 %) 10 (41.7 %)
24 (92.3 %) 2 (7.7 %)
17 (63.0 %) 10 (37.0 %)
13 (48.1 %) 14 (51.9 %)
16 (59.3 %) 11 (40.7 %)
14 (51.9 %) 13 (48.1 %)
13 (48.1 %) 14 (51.9 %)
11 (40.7 %) 16 (59.3 %)
11 (40.7 %) 16 (59.3 %)
19 (70.4 %) 8 (29.6 %)
Yes (N=27)
0.513
1.000
0.183
0.647
0.726
0.794
0.808
0.769
0.541
0.808
0.306
0.368
0.312
p
84 (96.6 %) 3 (3.4 %) 55 (63.2 %) 32 (36.8 %)
70 (92.1 %) 6 (7.9 %)
41 (53.9 %) 35 (46.1 %)
76 (88.4 %) 10 (11.6 %)
50 (57.5 %) 37 (42.5 %)
47 (54.0 %) 40 (46.0 %)
46 (52.9 %) 41 (47.1 %)
38 (43.7 %) 49 (56.3 %)
48 (55.2 %) 39 (44.8 %)
46 (52.9 %) 41 (47.1 %)
46 (52.9 %) 41 (47.1 %)
72 (82.8 %) 15 (17.2 %)
No (N=87)
12 (92.3 %) 1 (7.7 %) 4 (30.8 %) 9 (69.2 %)
12 (100.0 %) 0
5 (41.7 %) 7 (58.3 %)
12 (100.0 %) 0
9 (69.2 %) 4 (30.8 %)
5 (38.5 %) 8 (61.5 %)
9 (69.2 %) 4 (30.8 %)
7 (53.8 %) 6 (42.6 %)
4 (30.8 %) 9 (69.2 %)
5 (38.5 %) 8 (61.5 %)
4 (30.8 %) 9 (69.2 %)
7 (53.8 %) 6 (42.6 %)
Yes (N=13)
Clinically relevant fistula
0.050
0.432
0.591
0.631
0.355
0.616
0.453
0.420
0.698
0.179
0.501
0.234
0.027
p
J Gastrointest Surg (2014) 18:2009–2015 2013
0.364 5 (38.5 %) 8 (61.5 %) 49 (56.3 %) 38 (43.7 %) 1.000 26 (53.1 %) 23 (46.9 %)
49 (55.7 %) 39 (44.3 %)
5 (41.7 %) 7 (58.3 %)
0.545
40 (54.8 %) 33 (45.2 %)
14 (51.9 %) 13 (48.1 %)
0.971
7 (53.8 %) 4 (30.8 %) 2 (15.4 %) 0 7 (53.8 %) 6 (46.2 %) 64 (73.6 %) 20 (23.0 %) 2 (2.3 %) 1 (1.1 %) 65 (74.7 %) 22 (25.3 %) 0.519 0.330 17 (63.0 %) 8 (29.6 %) 2 (7.4 %) 0 17 (63.0 %) 10 (37.0 %) 54 (74.0 %) 16 (21.9 %) 2 (2.7 %) 1 (1.4 %) 55 (75.3) 18 (24.7 %) 0.750 0.735 8 (66.7 %) 4 (33.3 %) 0 0 8 (66.7 %) 4 (33.3 %) 0.479 0.428 33 (67.3 %) 13 (26.5 %) 3 (6.1 %) 0 33 (67.3 %) 16 (32.7 %)
63 (71.6 %) 20 (22.7 %) 4 (4.5 %) 1 (1.1 %) 64 (72.7 %) 24 (27.3 %)
Yes (N=13) No (N=87) p Yes n=49
No (N=88)
Yes (N=12)
p
No (N=73)
Yes (N=27)
p
classes for different outcome metrics, including major complications and major pancreatic fistula4. Major complications varied from 3 to 30 % when the complication risk score increased from 0 to 3, respectively, and the risk of clinically relevant pancreatic fistulae increased from 0 to 30 %4. This CRS supported the concept that that a combination of patient and tumor factors is the main determinant of procedurespecific complications and that patients with obese body habitus and larger or more proximal tumors requiring extended resections have the greatest likelihood of developing pancreatic fistulae and other associated complications. The CRS was recently applied at a single institution, and not only was shown to predict overall complications and fistulae as originally published but also resulted to be a reliable tool for assessing other adverse perioperative outcomes. However, several cases of this validation set had been used for the development of the original CRS, such that the patients’ overlap may have prevented a true validation of the score. Remarkably, a subset analysis excluding those patients used in the original formulation of the CRS only partially validated the findings5. The authors concluded that study of the CRS in a non-overlapping patient population was necessary for validation of this clinical tool. In terms of major complications and clinically relevant fistulae, our results are comparable with the American multiinstitutional experience, which reported a major complication rate of 11 %, and a clinically relevant pancreatic fistula rate of 10 %. The greater incidence of any complications observed in our series (49 versus 39 %) is attributable to pulmonary complications (pleural effusion and pneumonia), which accounted for 22 %. When applying the CRS to our study population, we found surprisingly that an increasing score did not correlate with the observed outcomes. In particular, 20/49 patients with any complications and 10/27 patients with a pancreatic fistula had a score of zero. The rate of complications and fistulae did not vary appreciably when the CRS increased. In addition, the observed morbidity and fistula rate Table 5 Multivariate analysis assessing independent predictors of major complications and clinically relevant pancreatic fistula after laparoscopic distal pancreatectomy
28 (54.9 %) 23 (45.1 %)
38 (74.5 %) 11 (21.6 %) 1 (2.0 %) 1 (2.0 %) 39 (76.5 %) 12 (23.5 %)
No n=51
Factor
≥150 mL Pathologic diagnosis PCN NEN PDAC Other Cystic Solid Specimen length ≤70 mm >70 mm
Clinically relevant fistula Any fistula Major complications Any complications Variable
Table 4 (continued)
0.117 0.182
J Gastrointest Surg (2014) 18:2009–2015
p
2014
p value
OR (Exp[B])
95 % CI
Model 1. Dependent dichotic variable = major complications Hosmer-Lemeshow test p=0.955, prediction accuracy of 89.7 % Sex (M) 0.032 4.550 1.139–18.169 EBL (≥150 mL) 0.032 4.472 1.133–17.650 Model 2. Dependent dichotic variable = clinically relevant pancreatic fistula Hosmer-Lemeshow test p=0.912, prediction accuracy of 88.6 % Sex (M) 0.005 7.589 1.252–12.590 EBL (≥150 mL) 0.024 5.318 1.819–11.657
J Gastrointest Surg (2014) 18:2009–2015
in patients with a CRS=0 were significantly greater than expected according to the report by Weber et al. Similarly, clinically relevant pancreatic fistulae were more frequent than expected in patients with a CRS=1 (O/E ratio of 11.1), and the incidence did not increase when the CRS was greater. Of note, the CRS (0=low risk versus 1–2=intermediate risk) was not predictive of other adverse outcomes, such as longer operative time, percutaneous drain insertion, and length of hospital stay. The poor performance of the CRS may be due to difference in baseline parameters among institutions, such as the proportion of patients with a BMI >27kg/m2 (34.8 % in the paper by Kneuertz et al. and 16 % in this series). This may have caused an inhomogeneous distribution of patients into risk classes, as suggested by the fact that only one individual in our series had a CRS of 3 (high risk of complications), whereas in the paper by Kneuertz et al., the proportion of patients at high risk was substantially greater than expected, indicating a greater operative complexity5. We then sought to evaluate in our study population the risk factors for the same outcome measures; in the univariate analysis, continuous variables were dichotomized around the median value. Only age >48 years and a BMI greater than 23.1 kg/m2 resulted to be associated with any postoperative complications, whereas no factor was found to be associated with major complications. When limiting the analysis to pancreatic fistula, only male sex and an EBL >150 mL were associated with a clinically relevant event. Different subanalyses were performed using the cutoffs proposed by Weber et al. for continuous variables, or even without dichotomizing the variables around the median value; but no significant associations were found. Although the univariate analysis showed only a small number of significant associations, we constructed a multivariate model with univariate inclusion criteria of p150 mL were independent predictors of major morbidity. We did not try to customize the CRS or to construct an alternative complication score, because a sample size of 100 cases would lead to an underpowered model. In conclusion, although the rate of major complications and clinically relevant pancreatic fistulae was similar to that observed by other case series, the previously published CRS based on pre- and intraoperative factors was not able to predict the postoperative risk in our study population. This is probably because risk scores may not be able to adjust for the casemix (heterogeneity in the baseline characteristics of the index population) and may not be optimally calibrated across groups. In our series, males were more likely to be older and with more comorbidities, thus being at greater risk for
2015
postoperative morbidity. The other prognostic factor was EBL, which serves as a surrogate marker of technical operative complexity and hampers optimal visualization of the surgical field. As the number of procedures performed across institutions becomes greater, robust modeling with large development and validation samples is needed to better identify the determinants of complications after LDP, which—despite being a major procedure—remains in expert hands a safe operation with zero mortality. Financial Support Giuseppe Malleo was supported by the Fondazione Italiana per la ricerca sulle malattie del pancreas (FIMP).
References 1. Venkat R, Edil BH, Schulick RD, Lidor AO, Makary MA, Wolfgang CL. Laparoscopic distal pancreatectomy is associated with significantly less overall morbidity compared to the open technique: a systematic review and meta-analysis. Ann Surg 2012;255:1048– 1059. 2. Pericleous S, Middleton N, McKay SC, Bowers KA, Hutchins RR. Systematic Review and Meta-Analysis of Case-Matched Studies Comparing Open and Laparoscopic Distal Pancreatectomy. Pancreas 2012;41:993–1000. 3. Abu Hilal M, Takhar AS. Laparoscopic left pancreatectomy: current concepts. Pancreatology 2013;13:443–448. 4. Weber SM, Cho CS, Merchant N, Pinchot S, Rettammel R, Nakeeb A, Bentrem D, Parikh A, Mazo AE, Martin RCG, Scoggins CR, Ahmad SA, Kim HJ, Hamilton N, Hawkins W, Schmidt CM, Kooby DA. Laparoscopic left pancreatectomy: complication risk score correlates with morbidity and risk for pancreatic fistula. Ann Surg Oncol 2009; 6:2825–2833. 5. Kneuertz PJ, Patel SH, Chu CK, Fisher SB, Maithel SK, Sarmiento JM, Weber SM, Staley CA, Kooby DA. Laparoscopic Distal Pancreatectomy: Trends and Lessons Learned Through an 11-Year Experience. J Am Coll Surg 2012;215:167–176. 6. Melotti G, Butturini G, Piccoli M, Casetti L, Bassi C, Mullineris B, Lazzaretti MG, Pederzoli P. Laparoscopic Distal Pancreatectomy. Ann Surg 2007;246:77–82. 7. Kimura W, Inoue T, Futakawa N, Shinkai H, Han I, Muto T. Spleenpreserving distal pancreatectomy with conservation of the splenic artery and vein. Surgery 1996;120:885–890. 8. Ferrone CR, Konstantinidis IT, Sahani DV, Wargo JA, Fernandez-del Castillo C, Warshaw AL. Twenty-Three Years of the Warshaw Operation for Distal Pancreatectomy With Preservation of the Spleen. Ann Surg 2011;253:1136–1139. 9. DeOliveira ML, Winter JM, Schäfer M, Cunningham SC, Cameron JL, Yeo CJ, Clavien PA. Assessment of Complications After Pancreatic Surgery. Ann Surg 2006;244:931–937. 10. Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J, Neoptolemos J, Sarr M, Traverso W, Buchler M. Postoperative pancreatic fistula: An international study group (ISGPF) definition. Surgery 2005;138:8–13. 11. Wente MN, Veit JA, Bassi C, Dervenis C, Fingerhut A, Gouma DJ, Izbicki JR, Neoptolemos JP, Padbury RT, Sarr MG; Yeo CJ, Buchler MW. Postpancreatectomy hemorrhage (PPH)–An International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery 2007;142: 20–25. 12. Rosales-Velderrain A. National trends in resection of the distal pancreas. World J Gastroenterol 2012;18:4342.
