Introduction

Hepatopancreatic (HP) surgery involves complex procedures that are associated with a high risk for perioperative complications.1,2 Even though mortality associated with HP surgery has decreased substantially over the past several decades, the risk of complications remains relatively high.1,3,4,5 The causes of complications are multifactorial with many studies reporting patient-, disease-, operative-, hospital-, and even social-specific factors that influence risk of morbidity.6,7,8,9,10 Previous research has largely focused on the impact of a single, specific postoperative complication on mortality risk. For example, Silber et al. reported that the development of any first postoperative complication was associated with, on average, a threefold increased odds of death.11 Many patients, however, experience multiple complications following major surgery. Data on the interaction effect of multiple concurrent postoperative complications relative to the risk of short-term mortality following HP surgery have not been reported. As such, the objective of the current study was to define the interaction effect of common and procedure-specific postoperative complications among patients undergoing HP surgery on 30-day mortality using a cohort of patients from the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP).

Materials and Methods

Data Sources and Study Population

Using the ACS-NSQIP Participant Use Data File (PUF) and the Procedure Targeted PUF, patients who underwent pancreatic and hepatic surgery between 2014 and 2016 were identified. Information in the final dataset included demographic characteristics, preoperative comorbidities, and perioperative clinical variables, as well as 30-day postoperative complications and mortality data. Eligible patients were selected based on Current Procedural Terminology codes (Supplementary Table 1). Patients who were outpatient (n = 208) or had the following preoperative comorbidities were excluded: acute renal failure/dialysis (n = 106), mechanical ventilation (n = 18), pneumonia (n = 0), preoperative sepsis (n = 341), wound infection (n = 159), and transfusions (n = 163). In addition, patients with missing data on weight (n = 63), height (n = 86), ASA status (n = 43), functional dependence (n = 62), operation time (n = 13), or the occurrence of pancreatic fistula (n = 162), delayed gastric emptying (n = 313), post-hepatectomy liver failure (PHLF) (n = 90), or bile leak (n = 0) were excluded. Finally, individuals who died on the day of surgery (n = 19) were excluded. The study did not require access to protected health information and therefore was exempt from review by The Ohio State University Institutional Review Board.

Outcomes

The primary outcome was the occurrence of 30-day mortality, which was defined as patient death within 30 days of the index operation. Six major common perioperative complications were evaluated: transfusion, organ space infection (OSI), sepsis/septic shock, pulmonary dysfunction, renal dysfunction, and cardiovascular complications. Certain outcomes were composite measures of more than one separate ACS-NSQIP outcome due to the low frequency of the individual metric. For example, pulmonary dysfunction was categorized as pneumonia, deep vein thrombosis (DVT) and pulmonary embolus (PE), unplanned intubation, and prolonged mechanical ventilation. Renal dysfunction included urinary tract infection (UTI), progressive renal insufficiency, and acute renal failure. Cardiovascular complications were defined as myocardial infarction, cardiac arrest, and cerebrovascular accident. For patients undergoing pancreatectomy, two procedure-specific complications were examined: pancreatic fistula and delayed gastric empting. Similarly, for patients undergoing hepatectomy, bile leak and PHLF were recorded. The definitions for these parameters in NSQIP are well established national/international criteria and can be found in the NSQIP code handbook.

Statistical Analysis

Patient characteristics and preoperative comorbidities were compared among patients who did not experience a postoperative complication versus patients who had at least one complication. The Pearson chi-square test was used to compare categorical variables and the Kruskal-Wallis rank test for continuous variables. The incidence of overall 30-day mortality among patients who experienced each of the major complications was analyzed. The Cox proportional hazard model was used to estimate the hazard ratios (HRs) with 95% confidence intervals (CI) for 30-day mortality.12 The two-way interaction effects of each combination of two complications on 30-day mortality were calculated using the relative excess risk due to interaction (RERI) in separate adjusted Cox models.13 As previously reported, the RERI method determines whether there is additive interaction in models that are inherently multiplicative, with additive interaction being the empirical method to determine biological interaction.14 All results were adjusted for age, sex, race, body mass index, emergency procedure, functional dependence, dyspnea, chronic obstructive pulmonary disease, current smoking, congestive heart failure, ascites, cancer, steroid use, bleeding disorder, preoperative hematocrit, procedure category, and log (operative time). Two-sided p values < 0.05 were used to evaluate statistical significance. All statistical analysis was performed with STATA 14.0 MP.

Results

Preoperative Characteristics of Patients

Among 26,824 patients included in the study, 10,886 (40.5%) experienced at least one complication while 15,938 (59.5%) did not experience any postoperative complication (Table 1). The average age of patients who experienced at least one complication was 62 years; patients who experienced a complication were more likely to be female and non-white. The majority of patients who experienced a complication were ASA class 3 and were more likely to have preoperative risk factors such as diabetes, COPD, anemia, and functional dependence. Postoperative complications were more common among patients who underwent pancreatectomy than hepatectomy (69.2 vs 30.8%, p < 0.001).

Table 1 Characteristics of patients undergoing pancreatic and hepatic surgery
Full size table

Perioperative Complications and 30-Day Mortality

Overall, 30-day mortality was 1.3%; mortality was higher among patients who experienced at least one complication versus patients who did not experience a complication (3.0 vs 0.1%, p < 0.001) (Table 2). The most common complications were blood transfusion (16.9%, n = 4519), organ space infection (12.2%, n = 3273), and sepsis/septic shock (8.2%, n = 2205). Among patients who underwent pancreatic resection, pancreatic fistula was observed in 18.3% (n = 2967), while 12.5% (n = 2039) had delayed gastric emptying. Bile leak and PHLF were observed in 7.7% (n = 817) and 5.2% (n = 549) of patients who underwent liver resection, respectively. The incidence of postoperative mortality varied based on type of complication and was highest among patients who developed a cardiovascular event (34.4%, n = 165), PHLF (12.4%, n = 68), and pulmonary dysfunction (11.1%, n = 246) (Table 2). On univariable analysis, the occurrence of each type of complication was associated with an increased risk of 30-day mortality (Table 3). For example, organ space infection (HR 2.8, 95% CI 2.27–3.66) was associated with a roughly 3-fold increased risk of perioperative death. On multivariable analysis, after adjusting for risk factors, each specific complication remained associated with 30-day mortality, including organ space infection (HR 2.17, 95% CI 1.70–2.78) and cardiovascular complications (HR 38.26, 95% CI 30.27–48.37) (all p < 0.001).

Table 2 Perioperative complications and rate of mortality among patients undergoing pancreatic and hepatic surgery
Full size table
Table 3 Unadjusted and adjusted Cox proportional hazards models for 30-day mortality among patients undergoing hepatic and pancreatic surgery
Full size table

Impact of Postoperative Complications on Mortality

Among patients who developed a complication, 54.4% had only one complication, while 23.4, 12.0, 6.0, and 4.3% of patients had two, three, four, or five to eight complications, respectively (Fig. 1a). In addition, postoperative mortality increased exponentially with the concomitant increase in the number of complications (Fig. 1b). Specifically, patients without complications had a mortality of 0.1 versus 0.6% for patients with one complication, 2.7% for patients with two complications, 5.0% for those with three complications, 9.6% in patients with four complications, and 20.4% among patients who experienced 5–8 complications. To analyze the interactions among the various types of postoperative complications, the possible synergistic effects of eight different complication combinations on 30-day mortality was examined (Table 4). The combinations associated with a positive additive effect on mortality included: transfusion + renal dysfunction (RERI 12.3, 95% CI 5.2–19.4), pulmonary dysfunction + renal dysfunction (RERI 60.9, 95% CI 38.6–83.3), pulmonary dysfunction + cardiovascular complication (RERI 144.1, 95% CI 89.3–199.0), renal dysfunction + cardiovascular complication (RERI 48.5, 95% CI 10.6–86.3), sepsis/septic shock + pulmonary dysfunction (RERI 13.3, 95% CI 2.6–24.0), and sepsis/septic shock + renal dysfunction (RERI 11.5, 95% CI 4.4–18.7). The combination with the strongest synergistic effect on mortality was pulmonary dysfunction + cardiovascular complication, followed by the combination of pulmonary dysfunction + renal dysfunction. On the other hand, the most common combinations observed in the cohort were organ space infection + sepsis/septic shock (n = 1382, 5.1%), followed by pulmonary dysfunction + sepsis/septic shock (n = 800, 2.9%) and pulmonary dysfunction + renal dysfunction (1.3%, n = 349) (Table 5). The interactions among various types of postoperative complications among pancreatic (Supplementary Table 2) and hepatic (Supplementary Table 3) only patients were also examined. In particular, there was a synergistic effect of PHLF with pulmonary and renal dysfunction among patients undergoing hepatic resection.

Fig. 1
figure 1

a Number of complications among patients experiencing at least one complication following liver and pancreatic surgery. b Rate of 30-day mortality according to the number of complications

Full size image
Table 4 Relative excess risk due to interaction for 30-day postoperative mortality based on adjusted Cox proportional hazards models of the occurrence of two perioperative complications among patients undergoing hepatic and pancreatic procedures
Full size table
Table 5 Frequency of specific complication combinations, with significant synergistic interactions, among patients undergoing hepatic and pancreatic surgery
Full size table

Discussion

Postoperative complications have a significant and deleterious impact on patient-reported experiences, hospital costs, cancer recurrence, and perioperative mortality.15,16,2,17,1,18 Therefore, an enhanced understanding of the relationship between postoperative complications and risk of mortality is important to improve outcomes among patients undergoing major HP surgery. In this population-based study, the interaction effects among postoperative complications and the derivative impact on postoperative mortality were defined. Among the over 10,000 patients who experienced at least one complication, almost half of all patients experienced multiple complications. In turn, while overall mortality associated with HP surgery was low (< 2%), risk of death was strongly associated with postoperative complications. In fact, the risk of mortality incrementally increased as the number of postoperative complications increased. Data from the current study demonstrated that the increase in mortality associated with multiple complications was exponential, rather than additive in nature. The relative excess risk due to interaction (RERI) was determined among patients who developed multiple complications and the final mortality risk was noted to be greater than the simple sum of the individual mortality risks.13 Specifically, while mortality was only 0.6% among patients who experienced one complication, the incidence of mortality grew to almost 10% among patients who had four complications. The type of complications also affected mortality, as several specific combinations of complications had a synergistic effect that led to an observed mortality risk greater than would have been expected (Table 4). Collectively, the data demonstrated that complications were not independent events; rather, multiple complications—and specific combinations of complications—significantly altered mortality risk associated with HP surgery. Both the number of complications and the specific type of complication(s) can influence risk of mortality following a surgical procedure.17,19 Furthermore, complications are not independent events; rather, index complications can markedly alter a patient’s subsequent risk of developing additional complications.20,21 For example, Kim et al. reported that multiple complications increased the risk of other perioperative complications.22 Specifically, the development of acute kidney injury (AKI), acute respiratory failure, or sepsis/septic shock alone increased the risk of developing the other two complications.22 In another study, Tevis et al. reported on the interactions between specific complications among patients who underwent a range of general surgical procedures.23 In this study, the authors reported an over fourfold increase in mortality among patients who had multiple complications compared with patients who had only 0–1 complication.23 Certain complications such as coma, septic shock, and failure to wean off the ventilator were strongly correlated with subsequent additional complications. In a separate study, Varley et al. reported that each additional complication that a patient experienced after pancreaticoduodenectomy was associated with a 48% increased odds of death (OR 1.48, 95% CI, 1.31–1.58).24 In the current study, both the number (Fig. 1b) and type of complication (Table 3) strongly correlated with risk of mortality among patients undergoing HP surgery. Indeed, the adjusted risk of death was strongly associated with cardiovascular complications, pulmonary dysfunction, and PHLF. Furthermore, 30-day mortality exponentially increased from 0.6% among patients with one complication to 20.4% among patients who had five or more complications. Collectively, the data strongly suggest that complications have a synergistic effect on mortality in the perioperative period. In turn, identification and prevention of early complications may be important in preventing a cascading risk of subsequent complications and increased mortality.

Research on the underlying mechanisms associated with the interactions among complications is scarce. In particular, data on which complications interact synergistically to elevate the risk of mortality remain poorly defined. Kim et al. reported synergistic interactions among AKI, acute respiratory failure (ARF), sepsis/septic shock, and stroke among patients undergoing general surgery procedures.25 The authors theorized that multifactorial mechanisms including biological, clinical, and social factors were responsible. In addition, the management of one complication may lead to or exacerbate a subsequent complication. For example, aggressive resuscitation of bleeding (via transfusion) or AKI (via intravenous fluids) may lead to pulmonary complications. Complex physiopathological process may also contribute to inter-organ “crosstalk”—i.e., the adverse effect of one malfunctioning organ on the function of another.26 Crosstalk among organ systems can particularly occur related to kidney, lung, heart, brain, and liver dysfunction.27,26,28,29,30 In turn, perturbation of an individual process can increase the risk of developing a subsequent complication in a new distant organ system. To this point, we investigated the synergistic effects of specific combinations of complications on death by measuring the relative excess risk due to interaction (RERI). Indeed, several combinations of complications elevated the risk of death greater than expected such as pulmonary dysfunction + renal dysfunction, pulmonary dysfunction + cardiovascular complication, sepsis/septic shock + pulmonary dysfunction, and sepsis/septic shock + renal dysfunction. Furthermore, among patients undergoing hepatectomy, PHLF interacted synergistically with pulmonary and renal dysfunction to markedly increase mortality. Further research is needed to elucidate the mechanisms that underlie the interactions among postoperative complications and identify ways to intervene in the deadly cascade associated with multiple complications.

The consistent and accurate classification of postoperative outcomes has important implications on hospital accreditation, quality initiatives, and research endeavors. Currently, the Clavien-Dindo grading system is the most widely adopted complication scoring system. The Clavien-Dindo system assigns a score of I-V based on the severity of the worst postoperative complication experienced by the patient.31 The Clavien-Dindo system is, however, an ordinal scale and fails to take into account that potential important differences in severity between the complication grades. For example, grade I versus II complications may not equal the differences between grade II and III complications; in addition, all grade III complications may not be equal in severity. In addition, the system is predicated on just one complication (i.e., the “most severe” complication). These limitations have led to the development of other complication schemes such as the comprehensive complication index (CCI).32 Data from the current study highlight the importance of the number, severity, and specific types of complications that occur in the postoperative period and suggest that the CCI, or grading systems like it, may be more appropriate for research and quality initiatives. The ability to rescue patients after major postoperative complications has also been increasingly identified as an important metric of high quality hospitals. Rescue from complications is especially important following HP surgery as morbidity can be high—the incidence of complications was 40.5% in the current study.33,34,35 In particular, the occurrence of multiple complications can be a strong risk factor for failure to rescue following HP surgery.24 The ability to rescue patients from multiple complications, especially complications that have a synergistic and deleterious effect, should be the focus of quality and safety initiatives.

The current study had several limitations. Similar to other studies that utilized administrative data, there are inherent limitations to the ACS-NSQIP dataset.36 Specifically, data on complications were pre-determined according to the ACS-NSQIP catalog and the definitions established by ACS-NSQIP. However, standard variable definition and rigorous clinical abstraction available through the ACS-NSIQP provided reliable and valid surgical outcomes assessment measures.37 Complication data collected through the ACS-NSQIP are also generally more accurate than data obtained from administrative, claims data sources.38,39 Due to the very low frequency of certain complications, some outcomes were analyzed as composite measures rather than as separate entities. While these complications were grouped according to the organ system involved or common underlying physiopathology, future studies will need to evaluate the impact of individual complications such as myocardial infarction, stroke, or pulmonary embolism. The exact etiology of each complication and its specific contribution to death could also not be directly assessed. Therefore whether complications and mortality were a result of disease stage, overall health status, underlying liver disease, or a combination of all these factors could not be fully elucidated.

In conclusion, both the number and specific type of complication impacted the incidence of postoperative mortality among patients undergoing HP surgery. Furthermore, certain complications interacted in a synergistic manner, leading to a greater than expected increase in the risk of short-term mortality. Further studies are needed to better understand the biological, clinical, and psychosocial factors that underlie the synergistic interactions between postoperative complications. A high level of vigilance in the postoperative period may facilitate early identification of patients at risk for multiple complications, thereby rescuing these patients early on and reducing postoperative mortality following major liver and pancreatic surgery.