Haigh PI, Bilimoria KY, DiFronzo LA. Early Postoperative Outcomes After Pancreaticoduodenectomy in the Elderly. Arch Surg. 2011;146(6):715-723. doi:10.1001/archsurg.2011.115
As our population ages, more elderly patients are undergoing major operations. Likewise, we should expect to encounter an increasing number of elderly patients who are potential candidates for pancreaticoduodenectomy because the incidence of pancreatic cancer is approximately 50-fold higher in octogenarians than in patients in their early 40s.1
The decision to perform a pancreaticoduodenectomy on an elderly patient is a difficult one because of the frequent comorbid conditions that may exist and the associated postoperative adverse events that are well documented to occur. The mortality after pancreaticoduodenectomy in the elderly is also debated, but many small case series have shown that the operation can be performed in patients older than 70 or 75 years with a low mortality rate that is no different than that of corresponding younger age groups.2- 11 The operation may therefore be justified in the elderly because of the apparent unimportance of chronological age. Unfortunately, the comparison of mortality by age in such studies is often unstable because of small sample sizes, particularly in the elderly group.
On the other hand, some case series have shown that the elderly have a higher mortality than younger patients.12,13 Studies that have used population databases have suggested that the mortality after pancreatectomy is uniformly higher in the elderly compared with that of younger patients.14- 19 Unfortunately, the ability to adjust for important clinical variables in such databases can be limited, and using administrative databases to assess complications may not be reliable.
This study was performed to determine the early postoperative outcomes after pancreaticoduodenectomy in the elderly using the American College of Surgeons–National Surgical Quality Improvement Program (ACS-NSQIP) database, a database that has validated and prospectively collected clinical data and that allows for clinical risk-adjustment.
The ACS-NSQIP participant-use files for 2005-2006 and 2007 were merged into a file that contained data on 363 897 patients that was prospectively collected from 183 participating hospitals.20 Any patient who underwent a pancreaticoduodenectomy was identified using Current Procedural Terminology codes 48150 (classic Whipple with pancreaticojejunostomy), 48152 (Whipple without pancreaticojejunostomy), 48153 (pylorus-preserving Whipple with pancreaticojejunostomy), or 48154 (pylorus-preserving Whipple without pancreaticojejunostomy). The synchronous use of gastrostomy or jejunostomy tubes was included. Patients who had synchronous resections of other organs, underwent emergency procedures, or had distant metastatic disease were excluded. In addition, patients who were American Society of Anesthesiologists class 4 or 5 or were receiving mechanical ventilation or dialysis were also excluded.
Patient demographic variables included age (>70 years [defined as elderly] vs ≤70 years), sex, and race. General preoperative variables included body mass index, whether the patient was a current cigarette smoker (within 1 year of operation), alcohol use of more than 2 drinks per day, and functional health status (independent, partially dependent, or totally dependent). Preoperative comorbid conditions included the presence of diabetes mellitus, dyspnea (with moderate exertion or at rest), history of chronic obstructive pulmonary disease, ascites, previous percutaneous cardiac intervention, previous cardiac surgery, hypertension requiring medication, history of revascularization procedure or amputation for peripheral vascular disease, history of rest pain or gangrene, and history of transient ischemic attack or stroke. Other preoperative variables included preexisting open wound, steroid use for chronic condition, weight loss of more than 10% in the preceding 6 months, bleeding disorders, prior operation within 30 days, and preoperative systemic sepsis within 48 hours before operation. Finally, the most recent preoperative laboratory values were collected within 90 days before the operation: sodium, creatinine, albumin, bilirubin, alkaline phosphatase, aspartate aminotransferase, white blood cells, hematocrit, platelets, and international normalized ratio. All laboratory results were categorized into values below, above, or within the normal range. Intraoperative factors included American Society of Anesthesiologists physical status class (1, 2, or 3), type of pancreaticoduodenectomy, histologic diagnosis (neoplasm of duodenum, pancreas, or bile duct vs other), the number of packed red blood cell units transfused intraoperatively (0, 1, 2, or ≥3), and the duration of operation (≤6 hours vs >6 hours).
Postoperative death and morbidities within 30 days of pancreaticoduodenectomy were examined and compared between the elderly and the younger group. In ACS-NSQIP, morbidity events are collected irrespective of whether the patient was an inpatient or outpatient or was readmitted to the same or to a different hospital. Morbidities included wound occurrences (surgical site infection and wound disruption), respiratory occurrences (outpatient pneumonia, unplanned intubation, pulmonary embolism, and need for ventilator support for >48 hours), urinary tract occurrences (acute renal failure, progressive renal insufficiency, or urinary tract infection), neurologic occurrences (stroke or coma), cardiac occurrences (cardiac arrest or myocardial infarction), and other occurrences (bleeding requiring >4 units of packed red blood cells within 72 hours postoperatively, deep venous thrombosis, sepsis or septic shock, and return to the operating room). Having at least 1 morbidity was also assessed, and the mortality was determined separately in patients with complications. Finally, the mean length of hospital stay was compared between the 2 age groups.
Categorical predictor variables were compared between the 2 age groups using the χ2 test, and the mean length of hospital stay was compared between the 2 age groups using the t test. For univariate analyses, contingency table analysis was applied to test for an association between the predictor variable and morbidity or mortality. Likelihood of morbidity (individual and at least 1) or mortality was estimated by applying multivariate logistic regression models adjusting for preoperative variables, preoperative laboratory values, and intraoperative variables that had P values less than .10 on univariate analysis. Age was also modeled as a continuous variable, but in ACS-NSQIP, any patient 90 years or older is classified as 90 years or older. Therefore, for 6 patients, they were artificially coded numerically as “90” for the analysis that used age as a continuous variable. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. All P values were 2-tailed, and .05 was the threshold for statistical significance on final multiple regression models and on comparisons of the predictor variables in each age group. All analyses were performed using SPSS, version 15.1 (SPSS, Inc, Chicago, Illinois).
There were 2610 patients who underwent pancreaticoduodenectomy. The mean (SD) age was approximately 64 (13) years, with a range from 16 to more than 90 years. There were 977 patients (37.4%) older than 70 years of age: of these, 755 (77.3%) were in their 70s, 216 (22.1%) in their 80s, and 6 (0.6%) who were at least 90 years of age. Characteristics of the patients in the elderly group compared with those of the younger group are shown in Table 1. The elderly group had a smaller percentage of black patients than did the younger group. The body mass index was higher in the elderly. The elderly group tended to have more preoperative comorbidities—particularly diabetes, pulmonary problems, cardiac disease, and neurologic disease. The younger group had more smokers and alcohol consumers.
Regarding intraoperative variables, the elderly group had more patients with higher American Society of Anesthesiologists class and more patients with malignant neoplasms (Table 1). In addition, the elderly were more likely to receive blood transfusions and a higher number of transfused units. The elderly cohort had shorter operations. The distribution of the 4 types of operations performed was similar between the 2 groups. It is notable that the few patients who apparently had operations without having any kind of pancreatic anastomosis were most likely coding errors, and inasmuch as such operations are described, it is uncommon practice in the United States to occlude the pancreatic duct in the remnant pancreas.
Laboratory values were more frequently outside the normal range in the elderly group, except for aspartate transaminase, alkaline phosphatase, and leukocyte count (Table 2). Preoperative hypoalbuminemia, anemia, thrombocytopenia, renal insufficiency, and prolonged international normalized ratio were more common in the elderly.
At least 1 morbidity occurred in 40.7% of the elderly compared with 34.0% of the younger group (P = .001) (Table 3). After adjusting for other variables on multivariate analysis, elderly patients had a higher likelihood of having at least 1 morbidity (OR, 1.27; 95% CI, 1.06-1.51; P = .01). Respiratory occurrences were more common in the elderly on univariate analysis, specifically with unplanned intubation and ventilator use for more than 48 hours. However, after adjusting for other comorbidities, elderly age was not an independent predictor for these adverse events. Acute renal failure and urinary tract infection were more common in the elderly on univariate analysis, with elderly age remaining as a significant predictor for urinary tract infection after adjusting for other comorbidities (OR, 1.44; 95% CI, 1.04-1.99; P = .03). Cardiac arrest and stroke were more common in the elderly on univariate analysis but were not significant on multivariate analysis. Wound occurrences, including surgical site infections, were no different between the 2 age groups.
Overall mortality in the 2610 patients was 2.7%. Elderly patients had a higher likelihood of mortality compared with that of younger patients (4.3% vs 1.7%; OR, 2.58; 95% CI, 1.59-4.18; P < .001) (Table 4). After accounting for all variables, advanced age independently predicted higher mortality (adjusted OR, 2.01; 95% CI, 1.18-3.43; P = .01). In a separate model, using age as a continuous variable, for every 1-year increment in age, mortality increased (OR, 1.03; 95% CI, 1.01-1.06; P = .02; other parameter estimates are not shown, but there were insignificant absolute changes, and no relative changes to that presented in Table 4). Variables that independently predicted for mortality were age, hypernatremia, classic Whipple without pancreaticojejunostomy, and number of red blood cell units transfused during the operation (Table 4).
In patients who had at least 1 morbidity, mortality was 10.1% in the elderly compared with 4.1% in the younger patients (P = .002).
The mean (SD) length of hospital stay was 12.9 (10.9) days in the younger group compared with 14.2 (11.1) days in the elderly group (P = .002).
This population-based study of ACS-NSQIP hospitals provides 30-day risk-adjusted outcomes on patients who underwent pancreaticoduodenectomy. The elderly, defined as those older than 70, had a higher mortality and were more likely to have at least 1 adverse event after a pancreaticoduodenectomy compared with the mortality and likelihood of younger patients.
It has been argued that pancreaticoduodenectomy can be performed safely in the elderly, and that advanced age alone should not be used as a reason to not perform the operation. Many of these studies, however, are from single institutions and may not accurately represent patient outcomes in the hospitals throughout the United States where a majority of pancreaticoduodenectomies are actually performed.2- 11 In addition, in a number of these studies, mortality was actually higher in the older age cohort, but the study was underpowered to demonstrate a statistical difference.11,21,22 Our study corroborates the conclusions from other population-based studies in which advanced age has been demonstrated to be an independent predictor of mortality in elderly patients undergoing pancreaticoduodenectomy.15- 17
In our study, evaluating mortality in only those patients with complications, there was an even greater difference in mortality between older and younger patients. The failure-to-rescue rate provides a measure of the degree to which providers responded to adverse occurrences that developed on their watch.23,24 A study by Ghaferi et al25 using the ACS-NSQIP database found that in overall “very high mortality” hospitals in patients with major complications after general or vascular surgery, the mortality was 21.4% compared with 12.5% in “very low mortality” hospitals, even though the rate of complications in each hospital group was similar, suggesting that hospitals with higher rates of mortality may not recognize and may not treat such complications as expeditiously or as expertly as those with lower rates of mortality. The higher failure-to-rescue rate from a complication of surgery in the elderly in our study also may indicate that the elderly are not as easily salvaged after a complication has occurred. However, it also could be true that if a patient of advanced age with pancreatic cancer had a significant complication and the outlook was grim, the patient or family may not have chosen to be “rescued,” thereby artificially increasing the failure-to-rescue rate.
Notably, overall mortality was low at 2.7% for the entire study population; it was also lower in the elderly subgroup (4.3%) compared with mortality for elderly patients in previous population-based studies, which ranged from 4.4% to 15.5%.15,16,19,26 This difference may reflect improvements in patient selection, operative technique, and perioperative care in the most recent decade because prior studies have examined patient populations over a longer time frame, typically beginning in the 1990s, whereas our study population represents a sample from a shorter and more recent time frame of 2005 to 2007. Supporting this concept are the results of a recent population-based study based on data from Texas hospitals, in which Riall and colleagues15 showed that mortality after pancreatectomy decreased in each successive year of the study and that the year of operation was actually an independent predictor of mortality.
Other factors that may contribute to the observed differences in mortality are the study population itself and the study hospitals where care was rendered. Patients undergoing pancreaticoduodenectomy in hospitals enrolled in ACS-NSQIP may represent a relatively unique population, and more important, the hospital perioperative care in institutions participating in ACS-NSQIP may be different and perhaps even superior, as reflected in a presumed interest in studying and improving surgical outcomes. Mortality differences between young and old in hospitals that have an interest in quality improvement may be dampened.
It is interesting that many of the clinical factors that are often associated with operative risk were not found to influence mortality independently. For example, diabetes, body mass index, coexistent chronic obstructive pulmonary disease, dyspnea, hypertension, cardiac history, stroke, and all laboratory values except for sodium level had little effect on mortality.
The importance of analyzing the outcomes of major cancer operations, such as pancreaticoduodenectomy, in the elderly cannot be understated, given the issues of a growing elderly population, the higher incidence of malignancy in older individuals, and the spiraling costs of health care. It is well established that people older than 65 years are the most rapidly growing subset of the United States population, and it is estimated that by the year 2030, individuals older than 65 will account for 20% of the total population. By 2050, 5% of the population will be composed of people older than 85 years.27 Furthermore, pancreatic cancer is a disease of older age; the annual incidence of pancreatic cancer is only 2 cases per 100 000 in patients aged 40 to 44, but it is 100 cases per 100 000 in patients aged 80 to 84 years.1 Thus, the number of elderly patients with pancreatic cancer is expected to grow significantly in the coming decades.
A potential limitation of the study is the inability to analyze the effect of hospital volume on short-term outcomes. A number of previous studies have reported improved survival in patients undergoing pancreaticoduodenectomy in high-volume centers, generally defined as those performing more than 10 operations per year, according to 2004 Leapfrog criteria.15,28- 31 The subjects of this study represented those having operations in one of the 183 hospitals participating in the ACS-NSQIP program. Although the presence of high-volume hospitals can be determined with respect to overall participation in ACS-NSQIP, the hospital at which the operation was performed is censored. Thus, it is uncertain what proportion of pancreaticoduodenectomies were performed at high-volume centers and unclear what effect this has on short-term outcomes in these patients. These data would certainly be of value because prior studies have shown improved outcomes specifically in elderly patients when pancreaticoduodenectomy is performed at a high-volume center.15,17
Other limitations of the study are the absence of data on recognized complications of pancreatectomy that can cause additional morbidity and even mortality; specifically, it would be valuable to know the effect of age on complications such as pancreatic leak or fistula and on delayed gastric emptying. This information will be included in the next generation of ACS-NSQIP. Also, information is lacking on whether patients had preoperative biliary stenting, another factor that can influence complications.32
Another highly relevant issue as it relates to clinical decision making is weighing the increased risks of complex surgery in elderly patients and balancing it with the potential for survival benefit in this population. Our study does not provide long-term survival data. However, several studies have demonstrated that 5-year survival in elderly patients undergoing pancreaticoduodenectomy for pancreatic cancer may be worse than that of younger patients. In a single institution study, Fong et al4 showed that, although short-term outcomes for patients older than 70 were equivalent to those of younger subjects undergoing pancreaticoduodenectomy, the 5-year survival rate of 21% in the elderly group was significantly worse than that of younger patients (29%). In a population-based study that examined both short-term outcomes using the National Inpatient Sample and long-term survival outcomes using the Surveillance Epidemiology and End Results database, Finlayson and colleagues16 demonstrated that both short- and long-term results were affected by advanced age in patients undergoing resection for lung, esophageal, and pancreatic malignant neoplasms. In this study, the 5-year survival for patients aged 60 to 65 years undergoing pancreatectomy was 16% compared with a 5-year survival of 11% in patients aged 80 and older. The difference was not statistically significant, however, because of the relatively small representative samples of patients having pancreatic resection in each respective age group. Therefore, the knowledge of increased operative risk has to be critically evaluated when considering the use of pancreatic resection in an elderly patient because the longer-term survival will likely be shorter.
In summary, in this risk-adjusted study from participating hospitals in the ACS-NSQIP, chronological age independently predicted for morbidity and mortality after pancreaticoduodenectomy. Advanced age also may have a role in failure to rescue because the mortality in those who have complications is higher in the elderly than in younger patients.
Correspondence: Philip I. Haigh, MD, MSc, FRCSC, Department of Surgery, Kaiser Permanente Los Angeles Medical Center, 4760 Sunset Blvd, Los Angeles, CA 90027.
Accepted for Publication: May 17, 2010.
Author Contributions: Dr Haigh had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Haigh. Acquisition of data: Haigh and Bilimoria. Analysis and interpretation of data: Haigh, Bilimoria, and DiFronzo. Drafting of the manuscript: Haigh and DiFronzo. Critical revision of the manuscript for important intellectual content: Haigh, Bilimoria, and DiFronzo. Statistical analysis: Haigh and Bilimoria. Administrative, technical, and material support: DiFronzo. Study supervision: Haigh.
Financial Disclosure: None reported.
Role of the Sponsor: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating in the ACS-NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
Previous Presentations: This study was presented at the 81st Annual Meeting of the Pacific Coast Surgical Association; February16, 2010; Maui, Hawaii.