Introduction

Surgical resection provides the only hope of a cure for patients with pancreatic ductal adenocarcinoma (PDAC), and the only chance for long-term survival [1, 2]. After resection, patients with localized PDAC can expect a 5-year survival rate of about 20% [3]. When the cancer is located in the pancreatic head, neck, or uncinate process, the standard operation for curative resection is a pancreaticoduodenectomy (PD). In the past two decades, several advancements have contributed to improving postoperative outcome after pancreatic surgery, including subspecialization, gained experience, and advances in intensive care treatment [4]. Currently, elective pancreatic resections are associated with mortality rates below 5% in specialized high-volume medical centers [5–8]. However, postoperative morbidity after pancreatic resection remains high, ranging from 30% to 60%. Postoperative morbidity leads to prolonged hospital stay, and occasionally, to the need of intensive care [5–7, 9, 10]. The interdisciplinary and experienced approach to patients with complications is important to reduce the ‘failure-to-rescue’ rate in those patients, and to achieve the low mortality rates of specialized centers [11]. Intensive care unit (ICU) services play a pivotal role in the management of postoperative complications in pancreatic centers. However, the overall survival of patients requiring ICU admission appears to be worse after pancreaticoduodenectomy for both benign and malignant pathologies [12]. Thus, patients with resectable or borderline resectable PDAC must be carefully selected in order to achieve an uncomplicated postoperative outcome and an improved quality of life after surgery.

The aims of this study were to identify and characterize patients with resectable PDAC that required either immediate and/or delayed postoperative ICU treatment, and assess the factors that influenced outcome.

Patients and methods

Patients

Between October 2001 and June 2008, all patients that underwent PD for PDAC of the pancreatic head at the Department of General, Visceral, and Transplant Surgery, University of Heidelberg, Germany were identified from a prospectively recorded electronic database. Informed consent was obtained from the patients in accordance with the Helsinki Declaration and serum collection and analysis was approved by the local ethical committee (decision number 301/2001). Patients with ampullary adenocarcinomas, intraductal papillary mucinous neoplasms, distal bile-duct carcinomas, and other malignant and benign pancreatic pathologies were excluded from the study.

Selected patients that presented with advanced local tumor invasion with encasement of the hepatic or mesenteric arteries underwent neoadjuvant chemoradiation.

Surgical technique

The pancreatic resection of choice was a standard pylorus-preserving PD (PPPD) as previously described [13]. When the separation of the tumor from the superior mesenteric vein or the portal vein was not achievable, venous resection and reconstruction were performed to accomplish an R0 resection, where all the margins were histologically free of the tumor. The standard lymphadenectomy in the case of a resectable PDAC comprised complete dissection of the hepatoduodenal ligament, the hepatic artery, the retropancreatic tissue, and right side of the superior mesenteric artery. An extended lymphadenectomy included dissection of the interaorto-caval tissue and the left side of the superior mesenteric artery. Extended radical resection was defined as PD with multivisceral resection (e.g., colon or partial liver resection) or vascular resection (other than the superior mesenteric or portal vein) and reconstruction.

Postoperative care

All patients were routinely extubated in the operating room and transferred to the recovery room that is equivalent to an intensive care unit setting. After a monitoring period (generally less than 6–18 h), patients were routinely transferred to the surgical ward. Patients with minor hemodynamic instability that required low-dose vasopressors were placed in an intermediate care unit (IMC, no ventilator support) for a period of 1–2 days, and then transferred to the surgical ward.

Patients that required extended ICU admission for more than 24 consecutive hours were defined as “ICU patients”, since 99% of the patients were transferred to lower care wards during the first 24 h. Among these, two subgroups were defined according to the time of ICU admission. Patients admitted to the ICU immediately after the operation and subsequent recovery period were defined as “immediate ICU patients”, and patients transferred to the ICU after 24 h of postoperative stay in a lower care ward were defined as “delayed ICU patients”.

Analyzed factors and definitions for complications and tumor stages

The ICU admission data were analyzed regarding the prevalence of the following preoperative and perioperative factors: age, sex, history of smoking, diabetes mellitus (DM), coronary artery disease (CAD), chronic obstructive pulmonary disease (COPD), chronic renal insufficiency (CRI), body mass index (BMI), American Society of Anesthesiology (ASA) score, New York Heart Association (NYHA) score, preoperative blood values (creatinine, albumin, bilirubin, hemoglobin, C-reactive protein, prothrombin time, and international normalized ratio), neoadjuvant therapy, intraoperative estimated blood loss (EBL), blood cell transfusions within 24 h of the operation start time, length of operation, operative characteristics (PPPD/classic PD, extended radical resection, extended lymphadenectomy, portal-/mesenteric vein resection, and intraoperative radiation therapy), and tumor pathology (TNM stage, lymph node ratio, and tumor grade).

Complications were graded according to the severity score introduced by Pierre-Alain Clavien [14]. Briefly, grade I was considered a minor deviation from uncomplicated postoperative status that did not require pharmacologic treatment or surgical, endoscopic, or radiological interventions; grade II required additional intravenous treatment; grade III required interventional treatment; grade IV was the presence of life-threatening complications that required intermediate care or ICU management; and grade V was death. Perioperative mortality (grade V) was defined as death during postoperative hospitalization.

Postoperative pancreatic fistulas (POPFs) included Grade B and C fistulas according to the ISGPF definition [15]. Histological tumor findings were obtained from patient pathology reports and classified according to the World Health Organization grading system.

Adjuvant treatment and follow-up

Following surgical resection, patients were treated according to, and/or included in the European Study Group for Pancreatic Cancer (ESPAC)-3 or ESPAC-1 trials, or patients were enrolled in the CapRI-Trial [16, 17]. Most patients were followed regularly in the outpatient clinic. The patients that were followed by their primary physician were contacted until April 2009 in order to determine the follow-up and overall survival rate.

Statistical analysis

SAS software (Release 9.1, SAS Institute, Inc., Cary, NC, USA) was used for statistical analysis. Quantitative variables are expressed as the median and range. Comparisons between subgroups of patients with respect to quantitative variables were performed using the Mann–Whitney U test or the Kruskall–Wallis test. Categorical variables were analyzed with Fisher’s exact test. Logistic regression analyses were computed to identify factors determining ICU admission. Overall survival from the date of pancreatic resection was calculated by the Kaplan–Meier estimate. Patients alive at the last follow-up were censored. Five hundred and twenty-six patients completed the follow-up, and 14 (2.6%) patients had incomplete follow-ups. Survival curves were examined with the log rank test. Categorical variables were analyzed with Fisher’s exact test. Two-sided P values were always computed and a difference was considered statistically significant at P ≤ 0.05.

Results

Analyzed population

During the study period, 836 consecutive PDs were performed for different tumors of the pancreatic head. Five hundred and forty patients had a final histopathologic diagnosis of PDAC. Resection of the pancreatic head tumor was achieved through a PPPD in 449 (83.1%), and through a classic PD in 91 (16.9%) of the cases. Portal vein resection was performed in 124 cases (23.0%) and extended radical resection in 52 cases (9.6%). Intraoperative radiotherapy (IORT) was applied to 28 patients (5.2%). A standard PD (either PPPD or classic PD) was defined as a PD without IORT, portal vein and extended radical resection, and extended lymphadenectomy, and was performed in 335 cases (62.0%). These data and the pathologic stages of the resected tumors are summarized in Table 1. The median follow-up of the 177 patients that survived through April 2009 was 21.6 months (interquartile range, 13.1–35.4 months).

Table 1 Operative technique and tumor stage of 540 pancreatic head resections for pancreatic ductal adenocarcinoma
Full size table

The demographic and clinical data for the entire cohort of all 540 patients are given in Table 2. Two hundred and nineteen patients (40.5%) had at least one comorbidity, including COPD, DM, CRI, and CAD. In addition, 61 patients (11.4%) had a pathological NYHA score, and 44 patients (8.1%) had a BMI greater than 30 kg/m2.

Table 2 Demographic and clinical data (n = 540 patients)
Full size table

Postoperative course and complications

After the operation, the patients stayed in the recovery room for a median of 15 h (interquartile range, 6–18 h) and then were transferred to the intermediate care unit or to the surgical ward. Packed red blood cell (PRBC) transfusions were administered to 186 patients (34.4%) within the first 24 h from the start of the operation, with a median of two PRBC transfusion units (interquartile range, 2–3) per patient. During the later hospital stay (>24 h after the operation), 109 patients (20.2%) required transfusions with a median of three PRBC units per patient (interquartile range, 2–9).

The overall hospital morbidity rate was 54.2% (Table 3). Surgical complications occurred in 41.1% of the patients. DGE (19.2%) and postpancreatectomy hemorrhage (7.0%) were the most common surgical complications. The severity classifications showed 61.2% minor complications (grades I and II). Grade V complications (deaths) were mostly caused by septic multi-organ failure (n = 14, 56% of all grade V complications). The postoperative overall in-house mortality rate was 4.6%, and the 30-day mortality was 2.6%. After a standard PD, the in-house and 30-day mortality were 3.6% and 2.1%, respectively (Table 4). Interestingly, patients that required ICU treatment had a 19.8% (n = 19) mortality rate, and 18 of these 19 patients were “delayed ICU patients”. In other words, the mortality in patients that had a delayed ICU admission was 39.1% (18 of 46 patients). Relaparotomy for complications was indicated in 51 patients (9.4%).

Table 3 Postoperative complications (n = 540 patients)
Full size table
Table 4 Postoperative mortality after pancreatic head resection for pancreatic ductal adenocarcinoma
Full size table

Characterization of ICU patients

Overall, 17.8% of the patients (96/540) were admitted postoperatively to the ICU for more than 24 h. Fifty patients (9.3%) remained in the ICU after the operation (“immediate ICU patients”), and 41 patients (7.6%) were transferred to the ICU from the surgical ward or from the intermediate care unit at a later time point (“delayed ICU patients”). Five patients (0.9%) required both immediate and delayed ICU admission. The median postoperative day of transfer to the ICU for “delayed ICU patients” was day 7 (interquartile range, 3–16). For these patients, the median length of the ICU stay was 7 days (interquartile range, 4–17). In contrast, the “immediate ICU patients” required significantly shorter stays in the ICU (median 2 days, interquartile range, 2-3; P < 0.0001). The indications for ICU admission are listed in Table 5.

Table 5 Indications for ICU admission in 96 patients
Full size table

Ten patients required delayed ICU admission for a diagnosis of POPF (21.7% of “delayed ICU patients”). In three patients, the admission was only for the diagnosis of POPF. In the remaining seven cases, POPF was associated with other complications, including sepsis (four cases), bleeding (two cases), and respiratory dysfunction (one case). In seven of the 18 “delayed ICU patients” that died (38.9%), a POPF was present.

The median postoperative hospital stay for patients admitted to the ICU was significantly longer compared to those not admitted (17 versus 12 days, respectively; P < 0.0001). In fact, the hospital stay was similar for “immediate ICU patients” (median 13 days; interquartile range, 11–18) and patients not admitted. However, “delayed ICU patients” required significantly longer hospital stays (median 37 days, interquartile range, 18–56).

Next, we compared operative parameters and tumor characteristics in patients admitted and those not admitted to the ICU. The ICU group had longer mean operating times (391 versus 361 min, respectively; P = 0.0023) and higher mean EBLs (1,580 versus 1,117 ml, respectively; P < 0.0001). The ICU group also required significantly more RPBC transfusions within the first 24 h after the start of the operation (54.2% versus 30.2% of patients, respectively P < 0.0001). The ICU group also had a higher prevalence of extended lymphadenectomies (8.3% versus 3.2%, respectively; P = 0.0395) and extended radical resections (16.7% versus 8.1%, respectively, P = 0.0202), but there was no significant difference between the two groups in the prevalence of superior mesenteric or portal vein resection (ICU, 26.0%; no ICU, 22.3%; P = 0.4246). There was a trend towards more frequent use of intraoperative radiotherapy in the ICU group (ICU, 9.4%; no ICU, 4.3%; P = 0.0700). Tumor size, lymph node status and ratio, metastatic disease (TNM stage), and cases with residual macroscopic tumor (R2 resections) were not significantly different in patients admitted and those not admitted to the ICU.

Predictors of ICU admission

On univariate analysis, several clinical preoperative factors were associated with ICU admission, including advanced age (>60 years), history of CAD, CRI, and DM (orally treated DM had the highest odds ratio of 2.89), ASA III score, and a pathological NYHA score. Among preoperative blood values, high creatinine (>0.9 mg/dl), high total bilirubin (≥2 mg/dl), high C-reactive protein (≥12 mg/l), low albumin (≤35 g/l), and low hemoglobin (<12 g/l) were significantly associated with ICU admission. Of the intraoperative factors, high EBL (≥1,500 ml), long operating time (>7 h), intraoperative PRBC transfusion, and extended radical resection were positive predictors for postoperative ICU admission.

However, on multivariate analysis, only the pathological NYHA score, diagnosis of DM, operating time longer than 420 min, and intraoperative PRBC transfusions were confirmed as independent risk factors for ICU admission (Table 6). Tumor size (T-stage) was not a significant uni- or multivariate factor; however, it is important to note that only 16 of 540 tumors were classified as T1 or T2 (96.3% were T3 tumors, 0.7% were T4 tumors; see also Table 1).

Table 6 Multivariate logistic regression analyses of parameters associated with immediate or delayed ICU stay in 517 pancreaticoduodenectomies for ductal adenocarcinomas of the pancreatic head (likelihood ratio test: p < 0.0001; 23 patients with missing values were excluded)
Full size table

Impact of ICU admission on long-term survival

The data demonstrated that patients that required a postoperative ICU stay after R0/1 pancreaticoduodenectomy had lower long-term survival rates. Indeed, ICU admission (immediate or delayed ICU admission) was a significant negative predictor for survival (P = 0.0155, Fig. 1). Interestingly, postoperative morbidity (including grade III and IV complications) did not negatively influence survival (P = 0.5438, Fig. 2). Considering the main indications for immediate ICU admission (monitoring and fluid management) and for delayed ICU admission (see also Table 3), it was unexpected that “immediate ICU patients” had a worse survival similar to the group of “delayed ICU patients”, when hospital mortality was excluded from the statistical analysis (P = 0.4470, Fig. 3).

Fig. 1
figure 1

Overall survival after R0/1 pancreaticoduodenectomy in 454 patients with pancreatic ductal adenocarcinoma, grouped according to the requirement of extended ICU admission (P = 0.0155; hospital mortalities were excluded). R0 microscopically tumor-free resection margins, R1 microscopically detected tumor cells within 1 mm of the resection margin

Full size image
Fig. 2
figure 2

Overall survival after R0/1 pancreaticoduodenectomy in 454 patients with pancreatic ductal adenocarcinoma, grouped according to postoperative morbidity (P = 0.5438; hospital mortalities were excluded). R0 microscopically tumor-free resection margins, R1 microscopically detected tumor cells within 1 mm of the resection margin

Full size image
Fig. 3
figure 3

Overall survival after R0/1 pancreaticoduodenectomy in 67 patients with pancreatic ductal adenocarcinoma and ICU admission, grouped according to indication for ICU admission (P = 0.4470; hospital mortalities were excluded). R0 microscopically tumor-free resection margins, R1 microscopically detected tumor cells within 1 mm of the resection margin

Full size image

Discussion

In the last three decades, advances in operative techniques and in perioperative patient care have improved the postoperative outcome for patients that are amenable to radical pancreatic resection. In selected patients with advanced tumor growth of a PDAC, surgeons can aim at a complete tumor resection through an extended radical resection (histologically tumor-free resections margins = R0) in order to achieve the best outcome for the patients [18]. The present study included a high percentage of those extended radical resections and cases with IORT that however had an increased mortality compared to standard pancreatic head resections.

Even high-volume pancreatic centers report an approximate 20% rate of major complications following PD that result in prolonged hospital stays, the need for invasive interventional procedures, and ICU utilization [7, 19]. Intensive care enables physicians to improve the immediate postoperative outcome of patients with major complications after PD. Several groups have studied ICU utilization after pancreatic surgery and its short-term effect on patient outcome [12, 20, 21]. To the authors’ knowledge, the present study is the largest analysis of postoperative ICU treatment for patients that underwent a PD for PDAC. Recently, Bentrem et al. reported a low survival rate of patients with ICU admission after PD for different benign and malignant pancreatic head tumors [12]. That study included 295 patients with PDAC; however, the survival analysis was based on a heterogeneous subset of 431 periampullary adenocarcinomas, and there were no details on the radicality of the surgery. A rational explanation for the lower patient survival rate after ICU admission may be that those patients underwent extended radical resection (e.g., multivisceral resection), or had larger tumors than patients not admitted to the ICU. It therefore remained unclear whether ICU admission independently reduced long-term survival.

The present study showed that patients that had postoperative ICU treatment after PDAC resection in fact had longer and more extended radical operations (in detail: higher EBLs, a higher percentage of extended lymphadenectomy, and more extended radical resections) than patients that did not require ICU treatment. Nevertheless, we found that a long operative duration and the need for blood transfusions were the only two independent risk factors for ICU utilization among those operative characteristics. Moreover, the survival analysis confirmed that patients with PDAC had a lower chance of long-term survival when they underwent postoperative immediate or delayed ICU treatment. Although delayed ICU treatment was associated with an increased mortality rate and hospital stay compared to immediate ICU treatment, both groups had similarly low long-term survival rates.

In general, the long-term outcome after PD can be improved by improving the criteria for patient selection before surgical resection, optimizing the preoperative condition of patients according to these criteria in a limited period of time, and, of course, avoiding long operations and large amounts of blood loss. There is no question that surgeons strive to limit operative durations and the amount of blood loss, but patients also benefit from an R0 resection, which in some cases necessitates an extended radical resection. However, the present study identified two novel independent, significant risk factors for postoperative ICU admission: preoperative DM and heart failure (advanced NYHA scores).

Previously, DM was known to contribute to complications after PD; high levels of blood glucose aggravated the outcome in surgically and critically ill patients [22, 23]. According to the analysis by Bentrem et al., DM was not identified as a significant predictor for ICU admission, despite a trend in higher admission rates for diabetic patients [12]. In our cohort, the greater prevalence of diabetic patients (28.9% versus 15.5% in [12]) and the higher total number of patients might have contributed to the difference in statistical significance.

Patients treated with oral anti-diabetic drugs had a greater risk for ICU admission than patients treated with insulin or untreated patients. This might be explained by differences in long-term glycemic control; however, we did not collect data on glycemic control in the weeks before surgery, and thus, we did not investigate this hypothesis.

A pathological NYHA score has yet not been described as an independent factor for ICU admission, but its identification was not surprising. Only a few studies have analyzed the role of the NYHA score as a risk factor for complications after gastrointestinal surgery, but none have considered it for pancreatic resections [24]. The NYHA classification reflects the functional capacity of patients with heart failure and weighs the limitation of physical activity. Among patients that required intensive care, one principal indication for immediate and delayed ICU admissions was a compromised hemodynamic status. Presumably, patients with a pathological NYHA score are more likely to have an impaired hemodynamic response to PD. Furthermore, we observed a linear risk correlation with increasing NYHA score: patients with NYHA II and III were more likely to be admitted to the ICU than patients with NYHA I (odds ratio, 3.75 versus 2.16, respectively).

Historically, preoperative patient-related factors that were associated with a greater incidence of surgical complications after PD included severe jaundice, renal insufficiency, heart disease, and obesity [19, 25]. In the present study, most of these factors were not confirmed on a multivariate analysis for ICU admission. In a recent analysis, obesity was the only independent factor associated with ICU admission when immediate transfers to the ICU from the recovery room were excluded [12]. The percentage of patients with a BMI greater than 30 kg/m2 (the definition of obesity) were comparable between the present and the above-cited study (8.1% versus 8.6% in [12]). Nonetheless, our multivariate analysis did not confirm increased BMI as a univariate or multivariate risk factor for ICU admission. Likewise, other investigators have found that obese patients did not have a significant predisposal for serious complications after PD [26]. Therefore, it remains unclear whether obesity has a significant impact on the incidence of complications and the need for intensive care.

Conclusions

Considering the high percentage of advanced tumor stages (80% of the patients had lymph node or distant metastases at the time of the surgery) and vascular or extended radical resections, the analyzed population was mainly composed of high-risk and medically demanding patients. About 18% of the patients that underwent PD for PDAC required extended ICU admission (including immediate and delayed extended ICU admissions). The hospital mortality in “delayed ICU patients” was high (39.1%). POPF was the indication for delayed ICU admission in 21.7% of the cases, and POPF was present in 38.9% of the patients that died after delayed ICU admission.

A short operative duration and low EBL (or no need for intraoperative RPBC transfusion) appeared to be the most important determinants for avoiding ICU admission and for promoting an optimal, procedure-related long-term outcome.

We identified two new factors that predisposed patients to extended postoperative ICU admission. Based on these factors, we recommend that surgeons should critically assess the potential for long and complicated resections for PDAC in patients with DM or advanced heart failure. If possible, in a reasonable time span before surgery, optimization of these preoperative pathological conditions should be considered. Future studies should evaluate whether better preoperative glycemic control or improvement of heart failure can reduce postoperative ICU admission following PD.