Decreasing Length of Stay After Pancreatoduodenectomy FREE

SINCE THE classic description of pancreatoduodenectomy (PD) in 1935,¹ this procedure has been associated with substantial morbidity and mortality. As recently as the 1970s, the average mortality was 20%.² Since that time, improved understanding of the pathophysiological features of the disease processes involved and improvements in surgical technique and perioperative care have contributed to an improved mortality rate. Several investigators^{3 - 11} have described the ability of experienced surgeons and high-volume hospitals to perform this procedure with minimal mortality, less than 4% in many centers. Unfortunately, reported complication rates have remained relatively constant, ranging from 35% to more than 50%.^{4 ,7 ,12 - 15}

Complications after surgery often prolong hospital stay and can be associated with a significantly increased cost in hospital care.¹⁶ After PD, length of stay (LOS) in uncomplicated cases is dictated by the time it takes for recovery of hemodynamic stability and bowel function and for the ability to resume adequate diet and activities of daily living. In patients who develop complications, these indicators of recovery may be delayed, and patients generally remain hospitalized until the complications are controlled. Therefore, the average LOS for an operative procedure in a cohort of patients can be affected by the number of patients with complications, the severity of the complications, and the recovery time for uncomplicated cases. In recent years, there has been pressure on surgeons from various sources to decrease LOS for all patients, with and without complications. We observed that LOS after PD was decreasing at our institution. This study determines if decreased LOS was due to a reduction in postoperative complications or earlier discharge for all patients regardless of complications.

The medical records of all patients undergoing PD at New York University Medical Center, New York, from January 9, 1986, to November 9, 1998, were retrospectively reviewed.

Demographic data recorded included age, sex, presence of diabetes mellitus, and dates of hospitalization, surgery, and discharge. Length of stay was measured as total in-hospital days, preoperative days, and postoperative days. Clinical variables evaluated included the operative procedure, management of the pancreatic remnant, the length of the procedure (in minutes), estimated blood loss (in milliliters), intraoperative transfusions (in units), and tumor histological features. Recovery indicators recorded were the times to nasogastric tube (NGT) removal, regular diet resumption, and pancreatic drain removal.

All procedures were performed by 1 of 10 surgeons (including S.G.M., E.N. T.H.G., H.L.P. and K.E.). The operative technique was described previously.³ Briefly, patients receive perioperative antibiotics, usually a second-generation cephalosporin. The gallbladder, distal common bile duct, duodenum, proximal jejunum, and head of the pancreas are resected in standard fashion. Management of the pancreatic remnant is determined by the surgeon at the time of the operation. The end-to-side pancreatojejunostomy is constructed by approximating jejunal and pancreatic ductal mucosa with interrupted, absorbable surgical sutures. An outer layer of interrupted, nonabsorbable seromuscular surgical sutures is used to approximate the jejunum to the pancreatic capsule. Invagination was performed using interrupted, nonabsorbable surgical sutures to approximate the seromuscular jejunum to the pancreatic capsule 2 cm from the transected end of the pancreas. An absorbable surgical suture was used to approximate the cut end of the pancreas to the jejunal mucosa. Pancreatojejunal internal silastic stents were used routinely. Oversewing of the pancreatic duct was performed with a nonabsorbable surgical suture, without pancreatoenteric anastomosis. Closed-suction Jackson-Pratt drains were placed adjacent to the pancreatic remnant at the time of the operation. Octreotide, erythromycin, tube gastrostomy, tube jejunostomy, and parenteral nutrition were not routinely used.

Major complications were defined as follows: delayed gastric emptying (DGE), inability to tolerate an oral diet after NGT removal, requirement for parenteral nutrition, reinsertion of an NGT for more than 5 days, or radiological confirmation of DGE; pancreatic fistula, drainage of more than 50 mL/d of amylase-rich fluid after postoperative day 10 or radiological or surgical confirmation of pancreatic anastomotic disruption; intra-abdominal abscess, presence of fluid in the abdominal cavity that was positive for microorganisms on a culture, necessitating percutaneous or operative drainage; sepsis, fever, hypotension, and end-organ damage requiring intensive care management; cerebrovascular accident, confirmed by examination and radiographs; pulmonary embolus, confirmed with a ventilation or perfusion scan, pulmonary angiogram, and/or spiral computed tomographic scan; gastrointestinal tract bleed, hemorrhage from the gastrointestinal tract requiring resuscitation by blood transfusion or endoscopic or operative intervention; acute myocardial infarction, associated with elevated enzyme levels, electocardiographic changes, or both; pneumonia, findings on a chest x-ray film associated with an elevated white blood cell count and positive sputum culture results necessitating antibiotic treatment; respiratory failure, requiring ventilator assistance for more than 48 hours after the operation or reintubation; renal failure, requiring renal dialysis; bowel obstruction, diagnosed radiographically or at surgery during the primary hospitalization or within 30 days; and mortality, in-hospital death or death within 30 days after the operation.

Minor complications were defined as follows: wound infection, involving skin and subcutaneous tissue, purulent drainage, and/or deliberate opening of the wound by the surgeon for a positive culture result; urinary tract infection, positive urine culture result with no more than 2 species of microorganisms, necessitating antibiotic therapy; bile leak, radiological or surgical demonstration of a biliary anastomotic disruption; cholangitis, bile that was positive for microorganisms on a culture without a bile leak and a clinical course consistent with cholangitis treated with antibiotics; arrhythmia, abnormal cardiac rhythm of new onset necessitating pharmacological intervention for greater than 24 hours after the operation; Clostridium difficile colitis, postoperative diarrhea associated with a positive stool C difficile toxin assay result; urinary retention, inability to void spontaneously, requiring prolonged bladder catheter placement, pharmacological intervention, or surgery; deep venous thrombosis, immediate onset of postoperative leg swelling or pain with iliac or femoral thrombus diagnosed by duplex ultrasonography or computed tomographic scan; and prolonged fever, fevers persisting several days to weeks in the late postoperative period without an identifiable source.

The data were divided into early and late groups based on operative date: group 1 (January 9, 1986, through December 21, 1992) and group 2 (February 16, 1993, through November 9, 1998). All values are reported as the median (range) unless otherwise stated. The Mann-Whitney test was used to compare median values. Data on LOS for patients who died in the hospital were considered censored and not included in the analysis of LOS. To compare proportions, the Pearson χ² test or the Fisher exact test was used where appropriate. Multivariate analysis was performed using linear regression to delineate variables associated with prolonged LOS; only variables representing more than 5% of the total cohort were included in the analysis. P values are reported for a 2-tailed test, with P<.05 accepted as significant.

Between January 9, 1986, and November 9, 1998, 215 PDs were performed at New York University Medical Center. The median age of the patients was 67 years (range, 18-91 years), with 96 female patients (45%) and 119 male patients (55%). Most patients were white (n=188), with 14 African American, 8 Asian, and 5 Hispanic patients. Forty-seven (22%) of the patients had evidence of diabetes mellitus preoperatively.

When divided into 2 groups based on operative date before or after January 1, 1993, the early group (group 1) contained 104 patients and the late group (group 2) had 111 patients. A comparison of the groups by demographic characteristics is shown in Table 1. Group 1 was significantly younger and contained a greater percentage of male patients than group 2.

Table 2 depicts the distribution of clinical variables by group. Sixty-four percent of the patients underwent pylorus-preserving PD. One hundred sixty-four patients (76%) underwent PD for a malignant neoplasm. Most (70%) of the pancreatojejunostomies were anastamosed end to side. Pancreatic invagination was used in 17%, and total pancreatectomy was performed in 3%. In the early group, 20 patients were treated by oversewing the pancreatic duct without anastomosis. These patients have been previously described.³ Group 1 contained patients with significantly fewer pancreatic cancers and a greater proportion of ampullary cancers. The median operative time was slightly longer in group 2.

The median total hospital LOS for the entire cohort was 20 days (range, 5-117 days), the median preoperative stay was 2 days (range, 0-36 days), and the median postoperative stay was 15 days (range, 4-95 days). The median preoperative and postoperative LOSs, arranged by year of operation, are depicted in Figure 1. Figure 2 shows the comparison of total, preoperative, and postoperative LOSs for the 2 groups. The median total LOS for group 2 is significantly lower than that for group 1 (15 days [range, 5-61 days] vs 26 days [range, 13-117 days]; P<.001). The median preoperative LOS was significantly shorter in group 2 vs group 1 (2 days [range, 0-36 days) vs 4 days [range, 0-28 days]; P<.001), and postoperative stay was also shorter for group 2 vs group 1 (12 days [range, 4-58 days] vs 19 days [range, 11-95 days]; P<.001).

Significant associations with total hospital LOS are displayed in Table 3. Operative year was inversely proportional to LOS. Patient age was directly proportional to LOS. Men had a significantly longer LOS. Diabetes and race were not predictive of increased LOS. Length of stay after surgery for benign and malignant disease was identical. Patients who underwent percutaneous transhepatic biliary stenting had a significantly longer total LOS, while patients who underwent preoperative endoscopic biliary stenting had a significantly shorter total LOS. Percutaneous drainage was more likely to be performed in group 1, and endoscopic drainage was more likely to be performed in group 2. The preoperative LOS directly correlated with the total LOS.

There was no significant difference in postoperative LOS by the type of procedure performed (classic vs pylorus preserving). Operative year was inversely proportional to postoperative LOS (Table 4). Management of the pancreatic remnant by end-to-side anastamosis resulted in a shorter LOS than the other methods. Oversewing the pancreatic duct and total pancreatectomy resulted in a significantly longer hospitalization. Operative time was directly proportional to postoperative LOS, while estimated blood loss and the number of intraoperative transfusions were not related to LOS. Patients who underwent percutaneous transhepatic biliary stenting did not have a significantly longer postoperative LOS, while patients who underwent preoperative endoscopic biliary stenting had a significantly shorter postoperative LOS. Preoperative LOS directly correlated with postoperative LOS. The development of any complication, any major complication, or any minor complication was associated with increased LOS.

Overall, 53% of the patients had at least 1 complication, with a median postoperative LOS of 21 days (range, 8-95 days), vs 12 days (range, 4-47 days) for those without any complications (P<.001). Fifty-nine percent of the patients in group 1 had a complication, while 42% of the patients in group 2 developed complications (P=.02).

Thirty-seven percent of all patients developed major complications, with a postoperative LOS of 24 days (range, 8-95 days), vs 13 days (range, 4-55 days) for patients without major complications (P<.001). Lengths of stay for patients in groups 1 and 2 with and without major complications are depicted in Figure 3. Forty-nine percent of the patients in group 1 had major complications, while 25% of the patients in group 2 had major complications (P<.001). As expected, patients with complications had a longer postoperative LOS: in group 1, the median was 25 days (range, 15-95 days) for those with complications vs 15 days (range, 11-47 days) for those without major complications (P<.001), and in group 2, the median was 20 days (range, 8-58 days) in patients with complications vs 11 days (range, 4-55 days) for those without complications (P<.001) (Figure 3). The median postoperative LOS was significantly shorter for group 2 compared with group 1 for patients with (20 vs 25 days; P=.05) and without (11 vs 15 days; P<.001) major complications.

A comparison of complication rates between groups 1 and 2 is shown in Table 5. There is a significant decrease in the rate of DGE and sepsis in group 2. There was a trend toward a decreased rate of pancreatic fistulae in group 2 as well (P=.08). As shown in Table 4, DGE and pancreatic fistulae were associated with a significantly increased postoperative LOS. In group 2, there was a significant increase in minor complications, including wound infection and urinary tract infection. The 30-day mortality decreased from 3% in group 1 to 0% in group 2. The 30-day mortality for the entire cohort was 1.4%.

The median number of days to removal of NGTs, resumption of a regular diet, and removal of pancreatic drains (recovery indicators) for patients with and without major complications in groups 1 and 2 are depicted in Figure 4. All recovery indicators were shorter in group 2 compared with group 1 for patients with major complications (NGT removal, 5.5 days [range, 0-33 days] vs 9 days [range, 6-36 days] [P<.005]; regular diet resumption, 11 days [range, 5-43 days] vs 14 days [range, 8-75 days] [P=.03]; and drain removal, 13.5 days [range, 4-47 days] vs 20 days [range, 8-120 days] [P<.005]) (Figure 4, top) and for patients without complications (NGT removal, 5 days [range, 1-15 days] vs 7 days [range, 4-10 days]; regular diet resumption, 8 days [range, 3-15 days] vs 9 days [range, 7-15 days]; and drain removal, 10 days [range, 5-38 days] vs 12 days [range, 9-19 days] [P<.001 for all]) (Figure 4, bottom). As expected, the median number of days for most recovery indicators was shorter in patients without complications compared with those with complications in group 1 (NGT removal, regular diet resumption, and drain removal, P<.001) and group 2 (NGT removal, P=.21; regular diet resumption, P<.001; and drain removal, P=.02).

All preoperative and postoperative factors were entered into a multivariate linear regression model. The significant predictors of total LOS are displayed in Table 6, along with their regression coefficient and confidence interval. Patient age, use of a percutaneous biliary stent, and complications, including DGE, pancreatic fistula, biliary complications, and infectious complications, all were independently associated with increased total LOS. Independent predictors of postoperative LOS included age and the development of complications. Operative year was an independent predictor of decreased postoperative and decreased total LOS.

Length of stay for PD has progressively decreased at New York University Medical Center during the past decade. This is consistent with reports^{7 - 9 ,16 - 17} from other institutions. We have seen that the rates of major complications have declined in the later years. The percentage of patients without any complications has increased from 36% to 52%. Our data indicate that management of patients without complications has become streamlined, leading to earlier discharge. In patients who develop complications, principles of care have been adjusted to streamline inpatient management and convert patients to outpatient management when medically appropriate.

Pancreatoduodenectomy has evolved from a procedure with significant morbidity and an often unacceptable mortality rate to a safer procedure, with acceptable morbidity and mortality,¹⁸ used for the treatment of various malignant and benign indications.¹⁹ In the present study, 20% of the procedures were performed for benign conditions. Until recently, major complication rates have remained relatively constant despite many modifications to the operative technique.¹² Some researchers^{10 - 11} have recently reported lower complication rates in centers with a higher volume. This could be due to increased experience and comfort with the procedure by the operating surgeons and an experienced team of nursing and support staff responsible for the perioperative care of the patient.

The importance of the management of the pancreatic remnant for complications has been previously described.³ In group 2, we performed more end-to-side pancreatojejunostomies and did not oversew the pancreatic duct in any patient. These data on the pancreatic anastamosis technique were incorporated into the multivariate analysis and were not found to be a significant predictor of increased LOS. Therefore, the change in our operative technique was not an independent factor leading to decreased LOS. Operative time was an independent factor predictive of increased total LOS but not postoperative LOS; as such, it may be a marker of a prolonged hospital course. There was no difference in LOS when the classic Whipple operation was compared with the pylorus-preserving technique.

One area that has changed during the period covered by this study is the preoperative examination of these patients. Improved imaging modalities²⁰ and a decreased need for invasive testing²¹ have led to a decrease in the requirement for a preoperative hospital stay. In the past year, preoperative hospital stay has decreased to a median of 0 days, indicating that outpatient preoperative diagnostic workup and same-day admission for surgery is standard. In addition, the increasing use of endoscopically placed stents before surgical referral has allowed outpatient evaluation and management of the patient with obstructive jaundice. The decreasing use of percutaneous drains may also shorten hospital stays by decreasing preoperative stays and lowering complication rates. In the present study, patients with endoscopically placed biliary stents had a shorter hospital stay than those receiving percutaneous stents or no stent at all. However, the presence of a biliary stent was an independent predictor of total, but not postoperative, LOS on multivariate analysis. The value, if any, of preoperative biliary drainage in patients undergoing PD is unclear,^{22 - 23} and a prospective, randomized study is warranted. At present, we recommend endoscopic biliary stenting only in jaundiced patients with significant comorbidity preventing prompt surgical intervention.²²

Several recent studies^{7 - 9 ,16 ,24} of patients undergoing PD have reported median LOSs of 10 to 16 days. The period encompassed by the later group (group 2) in the present study is coincident with the arrival of managed care contracts at our medical center. This period also saw the introduction of clinical pathways for several other gastrointestinal tract procedures at our institution. The principles of management from those pathways may have overlapped into the management of patient hospitalizations after PD, as seen in the shortened recovery indicators for all patients, with and without complications, in the later group. With the introduction of managed care, improved ambulatory and home care have allowed earlier discharge for all patients undergoing surgery, including PD. Recent studies demonstrated no benefit to prolonged use of NGTs²⁵ and pancreatic drains²⁶ and showed no benefit to the use of total parenteral nutrition¹⁷ or enteral tube feeding²⁷ in the patient without complications. Since operative year was an independent factor in the multivariate analysis, our data indicate that this trend toward the streamlining of perioperative care is a major determinant of LOS for PD.

Although our rate of minor complications increased in the later period, this was not a significant factor influencing hospital stay. This increase may be due to prospective data collection since 1995. However, the multivariate analysis identified major complications traditionally associated with PD,^{5 ,8} such as DGE and pancreatic fistula, as being independent predictors of increased LOS. It is possible to discharge patients without complications who have pancreatic fistulae in a timely fashion. Our data confirm those of Yeo et al,¹⁵ who cited DGE as a highly significant factor prolonging hospitalization in patients undergoing PD.

In summary, we have shown a decrease in overall, preoperative, and postoperative LOS after PD during the past 12 years at our medical center. This significant reduction in LOS is due to a reduction in the incidence of major complications and to streamlining perioperative management for all patients.

Presented in part at the second annual meeting of the Americas Hepato-Pancreato-Biliary Association, Ft Lauderdale, Fla, February 20, 1999.

We dedicate this article to the memory of John H. C. Ranson, BM, BCh, MA, the S. A. Localio professor of surgery, New York University School of Medicine, New York.

Reprints: Stuart G. Marcus, MD, Department of Surgery, New York University Medical Center, 530 First Ave, Suite 6B, New York, NY 10016 (e-mail: smarcus@pop.nychhc.org).