Kaplan-Meier survival estimate for all patients with ductal adenocarcinoma of the pancreas undergoing pancreaticoduodenectomy between 1987 and 2005 (n = 182). The 5-year actuarial survival rate was 27.4%.
Actuarial survival rates for patients with adenocarcinoma of the pancreas undergoing a pancreaticoduodenectomy who had more than 400 mL vs 400 mL or less of intraoperative blood loss (P = .02).
Actuarial survival rates for patients with positive (n = 25) vs negative (n = 157) resection margins (P < .001).
Kaplan-Meier survival estimate for patients with well, moderately, and poorly differentiated adenocarcinoma of the pancreas (P = .001 for all 3 groups).
Actuarial survival rates for patients with adenocarcinoma of the pancreas who were node-negative vs node-positive (P < .001).
Actuarial survival rates for patients with adenocarcinoma of the pancreas without vs with perineural invasion (P < .001).
Actuarial survival rates for patients with adenocarcinoma of the pancreas who underwent pancreaticoduodenectomy between 1987 and 1995 vs between 1996 and 2005 (P = .03).
Kazanjian KK, Hines OJ, Duffy JP, Yoon DY, Cortina G, Reber HA. Improved Survival Following Pancreaticoduodenectomy to Treat Adenocarcinoma of the PancreasThe Influence of Operative Blood Loss. Arch Surg. 2008;143(12):1166-1171. doi:10.1001/archsurg.143.12.1166
Although the safety of pancreaticoduodenectomy has notably improved over the past several decades, the reported survival of patients with pancreatic cancer remains poor. We hypothesized that, in recent years, the survival of patients with pancreatic adenocarcinoma following pancreaticoduodenectomy has substantially improved.
Retrospective case series.
Major academic medical and pancreatic surgery center.
A total of 182 consecutive patients underwent pancreaticoduodenectomy for various diagnoses between 1987 and 2005. Patients from 1987-1995 were compared with patients from 1996-2005.
Pancreaticoduodenectomy for patients with a diagnosis of pancreatic adenocarcinoma.
Main Outcome Measures
Survival after pancreaticoduodenectomy and patient outcomes.
During the time period analyzed, 182 patients underwent pancreaticoduodenectomy to treat ductal adenocarcinoma. There were no operative deaths, and 86.3% of patients had an R0 resection. The 5-year survival rate for the entire group was 27.4%. However, survival improved from 15.8% to 35.5% during the study period. Both groups had equivalent demographic and pathological characteristics, and the only predictors of poor survival in multivariate analysis were operative blood loss of more than 400 mL (hazard ratio, 2.17), poorly differentiated tumors (3.03), lymph node metastases (1.92), perineural invasion (2.66), and undergoing an operation before 1996 (1.42).
The survival rate for patients undergoing pancreaticoduodenectomy to treat pancreatic cancer has substantially improved. This finding is partially owing to improved operative technique and limited operative blood loss.
Many reports about pancreatic ductal adenocarcinoma begin with a review of the bleak survival statistics associated with this neoplasm. The authors stress the advanced stage of the disease at presentation in 80% to 85% of patients and the fact that few of those afflicted are even candidates for surgical resection. Even among patients who undergo resection, the long-term survival rate is disappointingly low: 5-year postoperative survival rates approach 20% (Table 1).1- 9
In spite of these facts, it has been our impression that our own clinical experience has been better and that, especially in recent years, more patients are living longer following resection. For that reason, the specific aim of this study was to review the UCLA (University of California, Los Angeles) experience with resected pancreatic adenocarcinoma. We have focused on survival statistics and the factors that influenced them. This review consists of a consecutive and prospectively collected series of patients who underwent a pancreaticoduodenectomy (PD) or pylorus-preserving PD for ductal carcinoma of the head of the pancreas. Patients who underwent distal or total pancreatectomy were not analyzed. Our goal was to compare a modern series of patients who underwent resection with a previous series to determine whether survival was actually improving.
Between January 1, 1987, and December 31, 2005, 182 patients underwent PD for ductal adenocarcinoma of the pancreas at the UCLA Medical Center; all analyzed patients had at least 12 months of follow-up. To determine the primary pathological diagnosis and the extent of disease, all resected specimens were reviewed by a senior attending gastrointestinal pathologist (G.C.) with particular interest and expertise in pancreatic and periampullary cancers. Pathological characteristics analyzed included tumor size, grade, resection margin, and lymph node status as well as presence of perineural or lymphovascular invasion. Margins were considered positive if cancer was present at the pancreatic, bile duct, duodenal, or retroperitoneal soft-tissue resection line. The latter margin was identified and inked by the pathologist before cutting the specimen. If the initial (intraoperative frozen section) margin was positive but further resection yielded a negative margin, the margin was considered negative. Only primary pancreatic ductal adenocarcinomas were included in this analysis; adenocarcinomas of the bile duct, ampulla of Vater, duodenum, and those arising from intraductal papillary mucinous tumors of the pancreas or cystic neoplasms were excluded.
Data collection and analysis for this study were approved by the UCLA Institutional Review Board. The review of the prospective database included patient demographic characteristics (age, race, and sex), surgical data (type of resection, operative blood loss, and duration of surgical procedure), postoperative morbidity and mortality, length of hospital stay, and long-term survival rate. A pancreatic fistula was defined as drainage of more than 30 mL of amylase-rich fluid, which is at least 3 times the upper limit of normal for serum amylase concentrations, in a 24-hour period after the fifth postoperative day. Postoperative mortality was defined as death occurring in the first 30 postoperative days or before discharge from the hospital. Delayed gastric emptying was defined as intolerance to oral intake and the need for nasogastric decompression after the seventh postoperative day.
Follow-up data on long-term survival were obtained through medical record review, direct patient contact, and review of US Social Security Administration records. Survival and the factors influencing it were determined for the entire cohort. A separate analysis was also done for 2 time periods, and these periods were compared: 1987-1995 (n = 57) and 1996-2005 (n = 125).
Overall, 75% of the resections in this series were performed by one surgeon (H.A.R.), so variations in technique were limited. The operative approach for PD has been described previously.10- 12 We performed a standard lymph node dissection in all cases, which included clearance of the soft tissues and nodes along the right side of the superior mesenteric artery and anterior to the aorta. The pancreaticojejunostomy was performed as a 2-layer anastomosis, in which the inner layer incorporated the duct; pancreatic duct stents were not used. Details of this technique have been published previously.12 The hepaticojejunostomy was performed as a single-layer anastomosis. For standard PD, an antrectomy was performed. For pylorus-preserving PD, the duodenum was transected 2 to 4 cm distal to the pylorus, and the right gastric artery was divided. Gastro-(duodeno)jejunostomy was performed in a retrocolic position in most cases. Two closed-suction drains (10-mm flat Jackson-Pratt) were placed, one near the biliary anastomosis and the other near the pancreatic anastomosis, and a T tube was used to stent the hepaticojejunostomy if the bile duct was less than 1 cm in diameter. Patients left the operating room with a nasogastric tube in place, which was removed the next morning.
Data are presented as mean (standard error of the mean) and median unless otherwise indicated. Differences in discrete demographic, operative, and pathological variables were analyzed with a Fischer exact test. Survival estimates were generated using the Kaplan-Meier method, and survival curves were compared using the log-rank test. A multivariate Cox proportional hazards model was established to determine which factors independently affected postoperative survival. Statistical significance was achieved at P ≤ .05. Statistical analyses were performed using SAS statistical software (SAS Institute Inc, Cary, North Carolina).
For the entire cohort of 182 patients, the median age was 65 years, and 56.1% were male. Based on the preference of the operating surgeon, 109 patients (59.9%) underwent pylorus-preserving PD, and 73 (40.1%) underwent standard PD. Vein resection was performed in only 3 patients (1.6%). The mean operative time was 7.2 (0.2) hours (median, 6.7 hours). The mean operative blood loss for the entire series was 504 (41) mL (median, 400 mL), with 43 patients (23.6%) receiving transfused blood during their hospital course. Resected ductal adenocarcinoma characteristics are included in Table 1. Of patients in the series, 86.3% had an R0 resection. The median tumor size was 2.5 cm, and 36.8% of the tumors were poorly differentiated; lymph node metastases were found in 51.1% and perineural invasion in 61.5%. There were no operative deaths in the entire series of patients, but postoperative morbidity occurred in 42.9%. The most common complications were delayed gastric emptying (14.3%) and pancreatic fistula (7.1%). Short-term outcomes are summarized in Table 2.
The overall 5-year Kaplan-Meier actuarial survival rate for the entire cohort of 182 patients with ductal adenocarcinoma was 27.4% (Figure 1). The median survival was 28.5 months, and median duration of follow-up was 26 months (mean, 35  months). The actual 5-year survival rate was 24%. Univariate analysis showed that neither patient age nor the presence of lymphovascular invasion had a significant influence on postoperative survival (Table 1). On the other hand, survival was significantly shorter among patients who experienced higher operative blood loss, transfusions, positive resection margins, poorer tumor differentiation, larger tumors, lymph node metastases, and perineural invasion (Table 1) (Figures 2, 3, 4, 5, and 6). We performed a subgroup analysis of 57 patients who underwent resection from 1987-1995 compared with 125 patients who underwent the procedure from 1996-2005. This demonstrated a significant improvement in survival for the latter group. The 5-year survival rate was 15.8% for the earlier cohort compared with 35.5% for the later cohort (P = .03) (Figure 7).
To support the appropriateness of such a comparison, we compared the clinicopathological profiles of patients in the 2 cohorts (Table 3). There were no significant differences in demographic and pathological variables for the 2 operative eras, except for a greater frequency of lymphovascular invasion in the 1996-2005 cohort (P = .006). The only operative factor that differed was the median operative blood loss for the 1996-2005 cohort, which was significantly less than that for the 1987-1995 cohort (350 mL vs 475 mL; P = .004). A multivariate Cox proportional hazards model controlling for all factors was established (Table 4). Independent prognostic factors for poor survival included operative blood loss of more than 400 mL, poorly differentiated tumors, lymph node metastases, female sex, and perineural invasion. Undergoing an operation after 1996 was an independent prognostic factor for improved survival. Finally, when we compared survival duration among patients with blood loss less than 400 mL from 1987-1995 and 1996-2005, there was no difference (29 months vs 33 months; P = .92).
During the past 2 decades, there has been a significant improvement in the operative mortality rate for patients undergoing PD but only minimal improvement in long-term survival rate when the operation is performed to treat pancreatic ductal adenocarcinoma.1- 9 Recently, several reports from Europe and the United States demonstrated 5-year survival rates of 25% for pancreatic ductal adenocarcinoma after R0 resection.3,5 Our results are comparable in that we found a 30.8% 5-year survival rate after R0 resection. However, in the present study, we report an overall 5-year actuarial survival rate of 27.4% after PD for ductal adenocarcinoma of the head of the pancreas for combined R0 and R1 resections. More important, there has been a significant improvement in survival at our institution within the relatively modern period this series represents (1987-2005). The 5-year survival rate for patients who underwent PD during the first half of the study period (1987-1995) was only 15.8%; this improved dramatically to 35.5% for the latter half (1996-2005; P = .03). Indeed, multivariate analysis determined that PD performed to treat pancreatic cancer during 1996-2005 was an independent favorable prognostic factor for long-term survival. To ensure that this significantly improved survival for the more recent period was not owing to the shorter follow-up time, the analysis was also performed limiting the recent era to include patients from only 1996-2002 (n = 102). The subgroup analysis and comparison to the 1987-1995 cohort was consistent, demonstrating an improved 5-year survival rate of 32.2% (P = .043).
What factors could account for this improvement? First, the recent time frame of our series suggests that the patients might have benefited from improved diagnostic techniques (eg, helical computed tomographic scanning and endoscopic ultrasonography) so that the disease was diagnosed at an earlier stage. Nevertheless, the mean diameter of resected tumors in the 2 eras reviewed was similar (2.79 vs 2.78 cm; P = .91). In addition, all patients were symptomatic at the time of their diagnosis, which suggests that the tumors were advanced enough to produce symptoms, usually obstructive jaundice and pain. These facts argue against the possibility that the diagnosis was made earlier in the more recent time period. Second, compared with the earlier cohort, a larger proportion of the more recent patient group may have received adjuvant therapy, which is increasingly being recognized as the standard of care following resection. It is also true that more effective treatments may now be available.13,14 Although we did not have adequate data for such an analysis (many patients continued care with oncologists at distant institutions), recently published data suggest that this is not likely to be the explanation for improved survival. For example, in the well-controlled ESPAC-1 (European Study Group for Pancreatic Cancer 1) study published in 2004,13 the 5-year survival in the group that received adjuvant chemotherapy was 21.1%, a good bit less than the best 5-year survival rates we observed. Moreover, many patients in our series also had radiation therapy, which in the ESPAC-1 trial was associated with an even less favorable outcome. Because most of these ESPAC-1 patients were likely treated during the same time period as our patients, it seems reasonable to make such a comparison. Nevertheless, we admit that the influence of adjuvant therapy in our patients remains unknown.
We also found no evidence that the pathological characteristics of the tumors were more favorable in the recent era. For example, there were no differences in a number of prognostic factors, including the degree of tumor differentiation, the incidence of R1 resections (positive resection margins), or the frequency of lymph node metastases or perineural invasion in the 2 groups. The one factor that did differ actually tended to favor a poorer survival rate in the more recent group of patients: 33.6% of the latter group showed lymphovascular invasion compared with 14.0% in the earlier group (P = .006).
We are left with the conclusion that the principal factor influencing long-term survival was operative blood loss. This is not a new observation for pancreatic cancer procedures; it was one of the prognostic factors identified in the benchmark series of 526 patients who underwent PD at The Johns Hopkins Hospital in 2000.7 Those authors observed a similar relationship between operative blood loss and survival rate, but the threshold used in that analysis was blood loss of 750 mL, and the 5-year survival rate was 17%. The median operative blood loss for this series was 400 mL, which is substantially less than any previously reported series of which we are aware. But even at this lower threshold, blood loss of more than 400 mL was a strong predictor of poorer survival in both univariate and multivariate analyses. There is an obvious relationship between operative blood loss and transfusion requirement, and it has been hypothesized that in other types of cancer (eg, colon cancer) the immunosuppressive effects of transfusions explain the poorer survival rate.15 Others place less importance on the transfusion variable and propose that the association may be owing to other factors that are still unexplained.16 In the present study, the transfusion requirement itself was not an independent predictor of survival in multivariate analysis. Therefore, while the reasons for this effect on survival by operative blood loss are not understood completely, the correlation is a strong one, and blood loss should be minimized as much as technically feasible. In a way, this speaks to the importance of the experience of the operating surgeon. Thus, it seems intuitive that increasing expertise with the technical performance of the operation is likely to result in lower operative blood loss for any complex surgical procedure. The fact that the senior author (H.A.R.) performed 50% of the resections in the earlier time period, and 90% in the later one, is consistent with this idea. Another observation that supports the idea that operative blood loss is a critical determinant of prognosis comes from a subgroup analysis of the 2 eras including only patients with estimated blood loss of less than 400 mL (n = 109). Of the patients who had blood loss of less than 400 mL, the median duration of survival was similar between the 2 periods (29 vs 33 months; P = .92). This suggests the greater importance of operative blood loss compared with the time period when the operation was performed. We suspect that the significantly better survival experienced by the more recent group of patients was owing to the fact that a greater proportion of them had blood loss of less than 400 mL.
Consistent with recent prospective trials, we found no significant difference in survival depending on whether pylorus-preserving or standard PD was performed.17,18 Several histopathological predictors of survival were found in resected specimens, including resection margin status, tumor differentiation, tumor size, lymph node status, and perineural invasion.
Our results are in agreement with previous reports on the importance of tumor differentiation, tumor size, resection margins, and lymph node involvement.1- 9 Poor tumor differentiation was the strongest predictor of postoperative survival in multivariate analysis (hazard ratio, 3.03; 95% confidence interval, 2.0-4.5; P < .001). We also demonstrated the importance of perineural invasion, a less-recognized predictor of survival in patients with pancreatic adenocarcinoma.19,20 In 61.5% of resected specimens, perineural invasion was seen on routine pathological sectioning. The median survival for patients without perineural invasion was 43.1 months compared with only 19.5 months for those with perineural invasion (5-year survival rate: 40.3% vs 18.7%; P < .001). In fact, perineural invasion posed a survival disadvantage to a greater degree than lymph node metastases in both univariate and multivariate analyses.
In conclusion, we found significant improvement in survival rate, to a degree previously unreported, among patients with pancreatic ductal adenocarcinoma who underwent PD. The reasons for this improvement are likely to be multifactorial. However, our data suggest the importance of technical factors in the performance of the operation and the desirability of minimizing operative blood loss. In addition to reaffirming several established clinicopathological predictors of survival, our analysis also revealed the importance of perineural invasion in predicting outcome in this disease.
Correspondence: Howard A. Reber, MD, Section of Gastrointestinal Surgery, University of California, Los Angeles, School of Medicine, Building 72-215 CHS, 10833 Le Conte Ave, Los Angeles, CA 90095-6904 (firstname.lastname@example.org).
Accepted for Publication: September 2, 2007.
Author Contributions: Drs Kazanjian, Hines, and Reber had full access to all the data in the study and take full responsibility for the integrity and accuracy of the data analysis. Study concept and design: Kazanjian, Hines, and Reber. Acquisition of data: Kazanjian, Duffy, Yoon, Cortina, and Reber. Analysis and interpretation of data: Kazanjian, Hines, Yoon, and Reber. Drafting of the manuscript: Kazanjian, Hines, and Reber. Critical revision of the manuscript for important intellectual content: Kazanjian, Hines, Duffy, Yoon, Cortina, and Reber. Statistical analysis: Reber. Administrative, technical, or material support: Hines, Yoon, Cortina, and Reber. Study supervision: Hines and Reber.
Financial Disclosure: None reported.