Abstract
Background/Aim: The impact of perioperative fluid management on postoperative morbidity after pancreaticoduodenectomy (PD) remains uncertain. This study aimed to investigate the independent association between perioperative fluid balance (FB) and clinically relevant postoperative pancreatic fistula (POPF) in PD patients. Patients and Methods: A total of 243 consecutive open PD patients were included. Intra- and postoperative FB until postoperative day 3 were calculated, and their predictive performance for POPF was assessed using receiver operating characteristic (ROC) analysis. Propensity score (PS) was estimated as the probability of having higher FB, and factors associated with POPF were identified using crude and PS-adjusted logistic regression models. Results: POPF occurred in 60 patients (24.7%). ROC analysis showed the highest predictive value for total FB on postoperative days 1 and 2, with a cut-off value of 1,585 ml (area under the ROC curve=0.74). Patients with FB ≥1,585 ml had a significantly higher POPF rate (48.3%) compared to those with lower FB (11.0%, PS-adjusted p<0.001). Male sex, body mass index ≥25 kg/m2, non-pancreatic ductal adenocarcinoma, biliary drainage, main pancreatic duct diameter <3 mm, and higher FB showed significant associations with POPF in crude univariate analysis. Higher FB remained a significant factor in both crude multivariate and PS-adjusted analysis [crude multivariate: odds ratio (OR)=8.0; PS-adjusted univariate: OR=4.2; PS-adjusted multivariate: OR=6.1, all p<0.001]. Conclusion: Higher early postoperative FB, a potentially modifiable factor, may be independently associated with increased risk of POPF in PD patients.
Pancreaticoduodenectomy (PD) is a widely utilized procedure for treating pancreatic head and periampullary diseases, despite its inherent challenges and associated morbidity and mortality rates (1). Although high-volume institutions have witnessed a decline in in-hospital mortality rates over recent decades (1-3), overall complication rates remain high, ranging from 30% to 60% (1, 4-7). The significance of these complications lies in their potential to increase medical expenditure, reduce quality of life, impede adjuvant chemotherapy tolerance, and even diminish overall survival (8-10).
Among the complications arising from PD, postoperative pancreatic fistula (POPF) stands out as one of the most frequent and life-threatening (11). Despite advancements in surgical techniques and perioperative management, the incidence of POPF has not significantly decreased and continues to range from 3% to 45% (12). The potential benefits of a recently rapidly expanding minimally invasive approach (laparoscopic or robotic PD) in reducing POPF have not been clearly established (13). This persistence can be attributed, at least in part, to the uncorrectable nature of clinical risk factors associated with POPF, such as soft pancreatic texture, small pancreatic duct size, and patient comorbidities (11, 14-16). Consequently, efforts have primarily focused on refining surgical techniques, including anastomotic methods and the use of pancreatic ductal drainage (17, 18).
Perioperative fluid management represents a modifiable factor that can impact outcomes following major abdominal surgery. Prior studies have established a connection between excessive intravenous fluid administration and increased risks of cardiopulmonary complications, anastomotic leakage, and delayed recovery of gastrointestinal function (19-21). Conversely, inadequate fluid resuscitation can lead to tissue hypoperfusion, negatively affecting wound healing and organ function (22). Although several authors have investigated the influence of intraoperative fluid management on surgical outcomes, including POPF, in PD patients, the results have been inconsistent and inconclusive (23-27). While some studies suggest that excessive intraoperative fluid administration contributes to greater postoperative morbidity, others do not find a significant association. A meta-analysis involving seven retrospective trials and four randomized control trials failed to establish clear links between fluid regimens (“liberal” versus “restrictive”) during PD and early postoperative outcomes (26). Furthermore, fluid balance (FB), calculated by subtracting output from input, represents a crucial measure that could impact postoperative morbidity (28, 29). However, limited research has explored the potential influence of perioperative FB on the incidence of POPF (30-32).
In this retrospective study, we aimed to examine the effects of both intra- and early postoperative FB on postoperative morbidity in PD patients, with a particular focus on assessing the independent association between FB and POPF.
Patients and Methods
Patient enrollment, data collection, and ethics statement. A total of 252 consecutive patients who underwent pancreaticoduodenectomy (PD) at our institute between January 2010 and December 2022 were retrospectively identified. The inclusion criteria were as follows: 1) patients who underwent elective open PD, including pylorus-preserving PD (PPPD), subtotal stomach-preserving PD (SSPPD), and conventional PD, with lymphadenectomy for any indication; 2) those who required PD combined with right hemicolectomy, minor liver resection, or vascular resection and reconstruction. The exclusion criteria were as follows: 1) patients who underwent total pancreatectomy and duodenectomy; 2) those who required major liver resection with PD; 3) those who lacked important data, including fluid records; 4) those with maintenance hemodialysis. A total of 243 patients were finally enrolled and analyzed. Data, including patient age, sex, American Society of Anesthesiologists physical status (ASA-PS), body mass index (BMI), comorbidities, indication for surgery, main pancreatic duct diameter (MPD), presence or absence of preoperative biliary drainage, preoperative laboratory values, types of PD, operative time, blood loss, perioperative fluid input/output, occurrence of POPF, and several other outcome measures, were extracted from electronic anesthesia and medical records. The preoperative estimated glomerular filtration rate (eGFR) was calculated using the Japanese GFR equation (33). The MPD was measured at the resection line of the pancreas using preoperative computed tomography images. Pancreatic gland texture, which has been discussed as a factor associated with POPF, was not included in the analysis because it strongly depends on the individual surgeon’s experience and lacks objective definition and criteria. This study was approved by the Institutional Ethics Committee of Meiwa Hospital (approval no.: 2023-05) as a retrospective analysis of collected data in accordance with the ethical standards of the World Medical Association’s Declaration of Helsinki. Individual written consent for this retrospective analysis was waived.
Surgery and postoperative management. Both anesthesia and surgery were performed by an experienced anesthetic and surgical team. However, the anesthetic approaches, surgical procedures, and postoperative care processes varied without strict standardization at our institution. Intraoperative fluid infusion was controlled by the anesthesiologist to achieve sufficient diuresis and maintain mean arterial pressure. The type of PD initially transitioned from conventional PD to PPPD and eventually to almost all SSPPD. Following PD, pancreaticojejunostomy with an external pancreatic duct stent was routinely performed in a duct-to-mucosa fashion using the modified Blumgart method (17), except for a few patients in the early study period who underwent the modified Kakita method (34). Choledochojejunostomy and antecolic gastrojejunostomy or duodenojejunostomy with Brown anastomosis were then performed. Two intra-abdominal closed-suction drains were placed near the pancreaticojejunostomy and choledochojejunostomy sites. All patients received an intraoperatively placed jejunostomy catheter for enteral feeding. Nasogastric tubes were rarely used. In combination with parenteral nutrition, enteral feeding was generally initiated on postoperative day (POD) 2 at 300 ml/day and gradually increased until sufficient solid food intake met the requirements. Patients were allowed a limited oral water intake but no food for the first 3 days postoperatively. The postoperative maintenance fluid volume aimed for 35-40 ml/kg/day, with additional intravenous fluid administered for resuscitation or replacement based on patient conditions such as low blood pressure, inadequate urine output (<200 ml/8 h), fever with sweating, and vomiting. No prophylactic octreotide was administered to prevent POPF. Drain amylase levels were routinely measured on POD 3, and the timing of drain removal, replacement, or irrigation thereafter was at the discretion of the attending surgeons.
Calculation of perioperative FB. Intraoperative FB in milliliters per operative hour (ml/h) was calculated as the total volume of intravenous fluids administered (including crystalloids, colloids, and blood products) minus estimated blood loss and urine output. Postoperative FB in milliliters (ml) was calculated as the difference between the total fluid administered (intravenous, enteral, oral) and the total output during the first three postoperative days (POD 1, 2, and 3). Blood product administration during this period was minimal. The postoperative output included urine volume, discharge from intra-abdominal drains, external pancreatic duct stent, and nasogastric tube (if inserted). Additionally, FB per preoperative body weight (FB/BW) was calculated and included in the analysis to account for individual variations.
Outcomes. The predefined primary outcome measure was POPF, classified as clinically relevant grade B and C fistulas according to the guidelines established by the International Study Group in Pancreatic Surgery (12). Secondary outcome measures included drain amylase levels on POD 3, post-pancreatectomy hemorrhage (PPH), overall morbidities graded using the Clavien-Dindo (C-D) classification system (35), length of hospital stay, and in-hospital mortality.
Statistics. No statistical sample size calculations were performed. Categorical variables are described as counts and percentages, and continuous variables are expressed as medians with interquartile ranges (IQRs) of 25% to 75%. Logistic regression was used to assess the performance of perioperative FB in predicting POPF by analyzing the receiver operating characteristics (ROC) curve and area under the ROC curve (AUC) statistics. The cut-off value was obtained with Youden’s index and evaluated for sensitivity and specificity. Based on the cut-off value for FB that yielded the highest AUC, the patients were divided into higher and lower FB groups. Between-group differences in baseline and surgical characteristics were analyzed using Fisher’s exact test for categorical variables and the Mann–Whitney U-test for continuous variables (p-values in Table I). To account for potential confounding and selection biases, propensity score (PS), representing the probability of having higher FB for each patient, was calculated using multivariate logistic regression based on the relevant 29 variables (listed in Table I). Model discrimination was assessed using AUC. The regression analysis was then repeated to compare the higher and lower FB groups for each covariate, with the comparison adjusted for the PS (adjusted p-values in Table I). To investigate the association of FB with POPF and other outcomes, in addition to crude analysis, PS-adjusted analysis was performed where PS was used as an adjusting covariate in logistic or linear regression models. For model selection in multivariate analysis, variables with a p-value less than 0.1 in crude univariate analysis were identified. A subgroup analysis using the PS-adjusted method was performed to further explore the association between FB and POPF according to several patients’ characteristics. All statistical analyses were conducted using R 4.2.2 software (Foundation for Statistical Computing, Vienna, Austria), and p<0.05 was considered significant.
Comparison of baseline and surgical characteristics of patients with higher and lower fluid balance (FB).
Results
Demographic and clinical features of the study patients. The study population consisted of 125 men and 118 women with a median age of 71 years (range=64-78 years). The majority of patients were ASA-PS 3 (n=180; 74.1%). The most common diseases were pancreatic ductal adenocarcinoma (PDAC) (n=153; 63.0%), followed by cholangiocarcinoma (n=50, 20.6%), and ampullary carcinoma (n=18; 7.4%). One hundred and forty-five patients (59.7%) with obstructive jaundice underwent biliary drainage as a bridge to surgery. Two hundred and twenty-three patients (91.8%) underwent SSPPD, 13 patients (5.4%) underwent conventional PD, and seven patients underwent PPPD. The median operative time and estimated blood loss were 453 (395-538) minutes and 540 (275-928) ml, respectively (Table I).
Intra- and postoperative FB and their associations with POPF. Sixty (24.7%) of the 243 patients developed POPF. Intra- and postoperative fluid input, output, and balance are summarized in Table II. Univariate logistic regression analysis showed that FB on POD 1, POD 2, total FB of POD 1 and 2, and total FB from POD 1 to 3 were significantly associated with the occurrence of POPF. Intraoperative FB and FB on POD 3 did not show statistically significant differences. Similar trends were observed regarding FB/BW. According to ROC analysis, the total FB of POD 1 and 2 had the largest discriminative ability for the occurrence of POPF [AUC=0.74, 95% confidence interval (CI)=0.66-0.82], with an optimal cut-off value of 1,585 ml, sensitivity of 72%, and specificity of 75% (Figure 1). Using this cut-off, patients were divided into higher (≥1,585 ml; n=89) and lower (<1,585 ml; n=154) FB groups. The POPF rate was 48.3% (95%CI=37.6-59.2%) and 11.0% (95%CI=6.6-17.1%) in patients in the higher and lower FB groups, respectively (Table III). However, when patient characteristics were compared, significant differences were observed between the higher and lower FB groups in terms of diagnosis, serum albumin levels, frequency of MPD <3 mm, and biliary drainage (Table I). After adjusting for the PS, however, none of the covariates included in creating the score showed significant differences between the higher and lower FB groups, as shown by the adjusted p-values in Table I. The AUC for the PS derivation model was 0.81 (95%CI=0.75-0.86), indicating good model discrimination.
Perioperative fluid input, output, fluid balance (FB), and predictive performance of FB for postoperative pancreatic fistula (POPF).
Receiver operating characteristic curve of total fluid balance on postoperative day 1 and 2 as a predictor of postoperative pancreatic fistula. The area under the receiver operating characteristic curve was 0.74 (p<0.001; 95% confidence interval=0.66-0.82). The optimal cut-off value was 1,585 ml, with a sensitivity of 72% and specificity of 75%.
Outcomes.
Risk factor analysis for POPF. To investigate the association between higher FB and POPF, crude and PS-adjusted regression analyses were performed. In crude univariate analysis, male sex, obesity with BMI ≥25 kg/m2, non-PDAC, preoperative biliary drainage, MPD <3mm, and higher FB were associated with the occurrence of POPF. Crude multivariate [odds ratio (OR)=8.0, 95%CI=3.6-17.5; p<0.001], PS-adjusted univariate (OR=4.2, 95%CI=2.0-8.6; p<0.001), and PS-adjusted multivariate (OR=6.1, 95%CI=2.6-14.3; p<0.001) regression analyses confirmed higher FB as an independent factor associated with POPF (Table IV). Furthermore, obesity, non-PDAC, and MPD < 3mm were also significantly associated with POPF in crude multivariate and PS-adjusted univariate analyses. The PS-adjusted subgroup analysis stratified by age, sex, ASA-PS, and some risk factors for POPF showed that higher FB tended to be consistently associated with POPF, while only the elderly group aged 75 years or older did not show a significant difference (Figure 2).
Crude and propensity score (PS)-adjusted logistic regression analysis of the effect of fluid balance (FB) on postoperative pancreatic fistula (POPF).
Forest plot of subgroup analysis for the association between fluid balance and postoperative pancreatic fistula. FB: Fluid balance; PS: propensity score; OR; odds ratio; CI: confidence interval; ASA-PS: American Society of Anesthesiologists physical status; BMI: body mass index; PDAC: pancreatic ductal adenocarcinoma; MPD: main pancreatic duct diameter.
Other outcomes. The overall postoperative morbidity rate in the entire cohort was 51.4%, with C-D grades II and IIIa being the most common morbidities (Table III). In PS-adjusted regression, patients with higher FB had a higher rate of overall morbidity (p=0.005), particularly C-D grade IIIa complications, and had a longer hospital stay (p=0.03) than those with lower FB. Most of the C-D grade IIIa complications were due to drain exchange and fistulography in the treatment of POPF. In a crude analysis, drain amylase levels on POD 3 were significantly higher in patients with higher FB than those with lower FB. However, this difference was no longer significant after adjusting for PS (p=0.06). Of the five patients who developed PPH, four were treated with interventional radiology (IVR) procedures. However, one patient in the lower FB group underwent reoperation. Of the three in-hospital deaths, one patient in the higher FB group died from multiple organ failure following IVR for PPH. The other two patients died from myocardial infarction and nosocomial SARS-CoV-2 infection, respectively. No significant differences were observed in PPH and in-hospital mortality between patients with higher and lower FB (Table III).
Discussion
The present study examined the role of perioperative FB in patients undergoing PD, with a focus on its impact on POPF. Patients with higher early postoperative FB had a higher incidence of POPF, as well as increased overall morbidity and length of hospital stay, compared with those with lower FB. Additionally, this study identified other predictive factors for POPF, including obesity, diseases other than PDAC, and small MPD, which are crucial to consider when investigating the relationship between FB and POPF. Our main findings are consistent with a few previous studies in the field. In 2018, Winer et al. analyzed 104 patients who underwent PD and found that high net FB over the first 72 h was a significant predictor of POPF (30). Similarly, two independent studies by Lapisatepun et al. and Jeon et al. in 2022 found a positive association of POPF with cumulative FB/BW at POD 3 and FB until POD 2, respectively (31, 32). These associations were determined through multivariate analysis. In our study, we contribute to the existing literature by utilizing PS adjustment to evaluate the independent association between FB and POPF. This method allowed us to account for multiple risk factors for POPF and potential confounders, providing a more robust analysis. To the best of our knowledge, our study is the first to explore this association using PS adjustment in the context of PD.
In clinical practice, the accumulation of fluid status can exhibit variability even with the administration of the same amount of fluid, indicating that FB may hold more significance than the sheer volume of fluid infused (32). One potential mechanism through which higher FB might heighten the risk of POPF is by inducing edema in the pancreatic remnant and small bowel (24, 25, 27, 30-32). Fluid overload can result in increased interstitial pressure, thereby diminishing blood flow and impeding the healing process of anastomosis (21). Furthermore, excessive fluid has been linked to impaired lung function and an augmented likelihood of pulmonary complications following various surgical procedures (20, 21, 27, 32). These consequences can lead to tissue hypoxia, an elevated susceptibility to infection, and compromised anastomotic healing. Nonetheless, the exact causal relationship and directionality between higher FB and POPF remain uncertain. If pancreatic juice leaks immediately after surgery, it can trigger localized peritonitis characterized by heightened capillary permeability, culminating in fluid retention, tissue edema, and subsequently, an elevation in FB (29). The absence of a significant association between intraoperative FB and POPF observed in this study and some prior studies may be partly explained by a mechanism in which POPF precedes the occurrence of higher FB. However, it is noteworthy that in this study, the median value of drain amylase level on POD 3 in the higher FB group was unexpectedly low, being less than three times the upper limit of the normal serum value. A certain proportion of patients indeed developed POPF on POD 4 or later. Thus, given the potential variation in the timing of POPF occurrence among patients, it is plausible that both of these mechanisms coexist.
Optimizing FB in patients undergoing PD holds significant clinical importance. However, achieving optimal FB can be challenging, given the complexity of fluid output control, including urine volume and drain volume management. Furthermore, the definition of higher FB as a risk for POPF still varies across limited studies (30-32), making it difficult to generalize. Similar to Lapisatepun et al.’s study (31), we calculated FB/BW as a means of evaluating FB relative to physique, but the discriminative ability of FB/BW in predicting POPF was slightly inferior to that of FB alone. We used preoperative body weight to adjust FB because we lacked daily body weight data up to 3 days after surgery. Consequently, the accuracy of the FB/BW values might have been compromised. It is necessary to accumulate further research to determine the optimal FB for individual patients and develop strategies to achieve it.
Although our study observed a relatively high incidence of POPF and prolonged hospital stays compared to previous reports, the incidence of grade C POPF, PPH, and mortality remained relatively low (23-25, 29-32). This can be attributed to conservative drain management, which may increase the incidence of grade B POPF but does not significantly raise the risk of severe complications (36). Notably, we did not find a significant association between higher FB and grade C POPF, PPH, or mortality, likely due to the low incidence of these outcomes in our cohort.
Several limitations of our study should be acknowledged. First, being a single-center retrospective study with a relatively small sample size, there may be inherent limitations related to selection bias and statistical measurements, which could limit the external validity of our findings. Second, the study did not account for the effects of specific fluid types and the use of diuretics, which can impact circulating plasma volume and FB. Third, the actual FB may be underestimated considering factors such as insensible perspiration and unrecorded additional fluid loss. Last, the analysis did not incorporate certain parameters that could be relevant to POPF, such as drain management method, anastomosis technique, and surgeon expertise, either due to difficulties in categorization or potential bias in case distribution.
In summary, our study highlights the independent association between higher early postoperative FB and increased risk of POPF in patients undergoing PD. These findings underscore the potential for postoperative FB as a modifiable factor to mitigate the risk of POPF. Nonetheless, further research is necessary to validate our results, establish standardized criteria for defining higher FB, and develop tailored strategies to optimize FB in PD patients.
Acknowledgements
The Authors would like to thank all the patients and medical staff at the institution who contributed to this study.
Footnotes
Authors’ Contributions
Shinichi Ikuta conceived and designed the study, collected and analyzed the data, and wrote the manuscript. Takayoshi Nakajima, Meidai Kasai and Masataka Fujikawa assisted in data collection, analysis, and manuscript preparation. Tsukasa Aihara contributed to the data interpretation, statistical analysis, and manuscript editing. Naoki Yamanaka provided guidance throughout the study, critically reviewed the manuscript, and supervised the project. All Authors have read and approved the final version of the manuscript.
Conflicts of Interest
The Authors declare that they have no conflicts of interest in relation to this study.
- Received June 1, 2023.
- Revision received June 28, 2023.
- Accepted June 29, 2023.
- Copyright © 2023 International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.
