1 Introduction

Acute pancreatitis is a potentially life-threatening inflammation of the pancreas. Its incidence is rising in many Western countries and is currently estimated at 20–40 cases per 100,000 person-years [1-3]. While short-term outcomes such as multiple organ failure, intensive care unit admissions, and in-hospital mortality have been extensively studied [4-10], less attention has been given to long-term morbidity following an episode of acute pancreatitis.

Acute pancreatitis induces an immense systemic inflammatory response, particularly in severe cases, resembling the response seen in sepsis [11-13]. Since sepsis is associated with an increased long-term mortality [14, 15], it is reasonable to hypothesize that acute pancreatitis may have a similar impact. Additionally, post-pancreatitis development of exocrine and endocrine dysfunction of the pancreas is well-described, conditions that are exacerbated by recurrent episodes of the disease [16, 17]. This dysfunction may also contribute to increased long-term mortality.

Some previous studies have reported on the long-term mortality after acute pancreatitis; however, the vast majority of these studies either lacked control groups, were of limited sample size, or were not designed to specifically study long-term mortality [3, 18-27]. However, recent work by Czapári et al. stands out as an exception [28]. They evaluated long-term mortality in a prospective multicentre cohort of 2613 individuals with acute pancreatitis. Their findings revealed an increased long-term mortality compared with the general population. However, the study had some limitations including the absence of a matched control group. Additionally, the methodology did not permit assessment of the isolated impact of a single episode of acute pancreatitis, excluding cases complicated by chronic pancreatitis, recurrent episodes, or coexisting malignant disease.

By using a large population-based cohort, the aim of this study was to provide a comprehensive assessment of the long-term mortality in adults with acute pancreatitis during the last 3 decades in Sweden.

2 Materials and Methods

3 Results

After exclusion of individuals who died during the index hospital stay (4422 individuals), the final cohort consisted of 89,465 individuals with a first-time episode of acute pancreatitis who were discharged from hospital alive and 890,837 matched controls (Table 1). Individuals were followed up for a mean of 10.0 years for a total of 10,155,039 person-years of follow-up. There were 33,764 deaths (37.7%) among individuals with acute pancreatitis and 265,403 deaths (29.8%) among controls. During follow-up, 17,522 individuals with acute pancreatitis (19.6%) experienced one or more episodes of recurrent acute pancreatitis and 6808 individuals (7.6%) developed chronic pancreatitis. The median time from hospital discharge to the first episode of recurrent acute pancreatitis was 249 days, while the median time to the development of chronic pancreatitis was 515 days. Notably, 3578 individuals (52.6%) who developed chronic pancreatitis had at least one prior episode of recurrent acute pancreatitis.

Individuals with acute pancreatitis had a higher prevalence of comorbidities, lower level of education and lower annual income compared with the control group. Among the acute pancreatitis cases, 45.2% were classified as gallstone-related, and 15.1% were classified as severe.

Kaplan–Meier estimates showed lower survival among individuals with acute pancreatitis compared with the pancreatitis-free control population (Supporting Information S1: eFigure 1). In the multivariable adjusted Cox model (Figure 1), the increased mortality was most pronounced during the first months of hospital discharge (HR 3.54 [95% CI 3.26–3.84] and 2.48 [95% CI 2.32–2.65] for the 0–1 and 1–3 monthtime strata, respectively). Thereafter, mortality declined but still remained significantly elevated throughout the study period (HR 1.27 [95% CI 1.24–1.30] in the long-term [> 10 years] time strata). The HRs were slightly attenuated after censoring for recurrent acute pancreatitis or chronic pancreatitis.

Subgroup analyses using time-stratified and restricted cubic spline models found increased HRs irrespective of cause and severity during the first 6 months after hospital discharge (Table 2, Figures 2 and 3). All causes and severities except non-severe gallstone-related acute pancreatitis were associated with increased mortality beyond 10 years of follow-up. HRs were generally higher in severe and non-gallstone-related episodes. After censoring for recurrent acute pancreatitis or chronic pancreatitis, the HRs attenuated towards 1 but remained increased in non-gallstone-related pancreatitis (HR 1.27 [95% CI 1.23–1.33] and HR 1.47 [95% CI 1.33–1.63] in the long-term [> 10 years] time strata for non-severe and severe episodes, respectively).

In sensitivity analyses, the HRs were comparable to those in the main model when (i) restricting the study period from 2000 to 2019 (Supporting Information S1: eTable 4) and (ii) restricting the analyses to individuals without comorbidities or low educational level (Supporting Information S1: eTable 5). In individuals younger than 50 years, there were 3326 (12.8%) deaths after hospital discharge among individuals with acute pancreatitis and 9346 (3.6%) deaths among the control population. Kaplan–Meier estimates showed lower survival among individuals younger than 50 years with acute pancreatitis, particularly for severe nongallstone-related episodes, compared to the pancreatitis-free population (Supporting Information S1: eFigure 2). The trends in HRs in time-stratified Cox models among individuals younger than 50 years were comparable to those in the whole population (Supporting Information S1: eTable 6). After censoring for recurrent acute pancreatitis or chronic pancreatitis, the HRs were attenuated towards 1 for gallstone-related acute pancreatitis (HR 1.09 [95% CI 0.92–1.30] and HR 1.49 [95% CI 0.95–2.34] in the long-term [> 10 years] time strata for non-severe and severe acute pancreatitis, respectively). The HRs for non-gallstone-related acute pancreatitis remained increased when censoring for recurrent acute pancreatitis or chronic pancreatitis (HR 1.95 [95% CI 1.76–2.16] and HR 2.51 [95% CI 1.93–3.25] in the long-term [> 10 years] time strata for non-severe and severe episodes, respectively).

When comparing subgroups of non-gallstone-related acute pancreatitis, there was a trend of higher mortality in non-severe alcohol-related episodes compared with nonsevere nongallstone–nonalcohol-related episodes. However, after censoring the analysis for recurrent acute pancreatitis or chronic pancreatitis, the HRs for non-severe alcohol-related acute pancreatitis attenuated to the same levels as the HRs for non-severe non-gallstone-non-alcohol-related episodes (Supporting Information S1: eTable 7). An increased long-term mortality was observed for both nonalcohol–nongallstone-gallstone- and alcohol-related acute pancreatitis. In a long-term setting (> 10 years), after censoring for recurrent acute pancreatitis or chronic pancreatitis, there was no statistically significant association between alcohol-related acute pancreatitis and mortality compared with the control population (HR 1.29 [95% CI 0.88–1.89]). The corresponding HR for nonalcohol–nongallstone-related episodes was statistically significantly increased (HR 1.46 [95% CI 1.32–1.62]).

4 Discussion

In this population-based cohort study, we found that acute pancreatitis was associated with an increased mortality after hospital discharge that remained increased even after more than 10 years of follow-up. Mortality was higher in non-gallstone-related and severe episodes and seemed to be mediated in part via recurrent acute pancreatitis and development of chronic pancreatitis as well as via higher prevalence of comorbidities and lower socioeconomic status.

The current study had a number of strengths, including its nation-wide coverage and use of validated registers, the large number of acute pancreatitis cases and the use of a matched pancreatitis-free control population. This provided a good internal validity and sufficient statistical power to conduct detailed subgroup and sensitivity analyses. The results are likely to have a high generalizability to other populations in Western countries that have comparable demographics and access to health care.

One of the study weaknesses was the lack of detailed data on certain confounders, such as smoking, alcohol consumption, physical activity and body mass index [31]. We attempted to reduce the impact of residual confounding from these lifestyle factors by adjusting the analyses for educational level and annual income, which are highly correlated to the missing confounders [32-34]. Furthermore, the estimates remained virtually unchanged in sensitivity analyses that (i) were restricted to healthier individuals with a lower risk of mortality (by excluding individuals with comorbidities and those with a low level of education) and (ii) were stratified nongallstone-related acute pancreatitis by alcohol abuse. Finally, the analyses were censored for recurrent acute pancreatitis or chronic pancreatitis, two factors highly associated with smoking [35] and alcohol abuse [36]. An additional limitation is that the diagnosis but not the cause of acute pancreatitis has been validated in the Swedish Patient Register [37]. In particular, alcohol-induced acute pancreatitis was likely to be underreported in our study. The diagnostic code for alcohol-related acute pancreatitis (K85.2) was not introduced until 2006 and seems to have been underutilized since then. One study found that 59% of all acute pancreatitis cases were classified as “unspecified” (K85.9) and that only 3.5% were classified as alcohol-induced in the Swedish Patient Register [37]. To address this, we included diagnostic codes for alcohol abuse in the definition of alcohol-induced acute pancreatitis. However, the Swedish Patient Register does not capture diagnoses from primary care and many individuals with high alcohol consumption do not seek care or receive a diagnosis of alcohol abuse. Consequently, it is likely that our definition of alcohol-induced acute pancreatitis underestimated the true proportion of alcohol-induced acute pancreatitis in the study population. On the other hand, the positive predictive value of the used definition was most likely very high, and the large sample size ensured sufficient power to assess mortality in this subgroup.

Further, the current dataset lacked the components necessary to apply the Revised Atlanta classification for severity. Therefore, the variables used to define disease severity were selected to approximate the Atlanta classification as closely as possible. Notably, it was not possible to differentiate between moderately severe and severe episodes without substantial risk of misclassification bias; as a result, these categories were combined and collectively referred to as “severe”. Also, the register-based data collection constitutes a risk of misclassification bias with respect to covariate status. The Swedish Patient Register does not contain diagnoses from the primary health care, which may contribute to under diagnosis of comorbidity status. However, such misclassification would most likely be non-differential with respect to the future occurrence of acute pancreatitis. Also, though not a limitation per se, it should be noted that the exclusion of individuals who died during the index hospital stay affected many baseline characteristics of the cohort, including a lower proportion of severe disease in the current study compared to the entire SwePan population (13.7% vs. 18.3%). Therefore, the findings should not be used to infer such characteristics of the overall acute pancreatitis population.

To our knowledge, the work by Czapári et al. is the only previous population-based assessment of long-term mortality in acute pancreatitis compared with a control population. They found an increased long-term mortality in acute pancreatitis and, importantly, noted that the first year post-discharge mortality was higher than the in-hospital mortality (5.5% vs. 3.5%). The leading cause of death during the first year post-discharge was end-stage cancer, which suggests an overrepresentation of malignant disease in individuals with acute pancreatitis. The second leading cause of death in their study was pancreatitis-related sepsis, indicating recurrent acute pancreatitis or development of chronic pancreatitis. In our study, interestingly, the long-term mortality remained increased when adjusting for malignant disease and when censoring for development of recurrent acute pancreatitis or chronic pancreatitis. These findings suggest that the association between acute pancreatitis and long-term mortality is mediated through multiple pathways.

Previous research on long-term mortality in acute pancreatitis is less comparable to the current study. For example, a single-center retrospective study from Finland by Karjula et al. found a 4 times higher mortality rate among in acute pancreatitis compared to a matched control population with a median follow-up time of 9.5 years [27]. However, the study was limited by the sample size (1644 cases) and the single-center study design. The external validity was further limited by the high proportion of cases with alcohol-related acute pancreatitis (71.4%). A recent large Danish population-based study found a trend of decreasing long-term (31–365 days) mortality in acute pancreatitis during 1988–2018 (adjusted mortality rate ratio 0.64 [95% CI 0.56–0.74] comparing 2013–2017 to 1988–1992) [3]. The study design did not include a pancreatitis-free control population, and thus, the results are not directly comparable to our study.

Chronic pancreatitis is associated with an increase in long-term mortality [38]. In the current study, the development of chronic pancreatitis was found to contribute to the increased long-term mortality after a first-time episode of acute pancreatitis. However, the exposure-outcome association remained after censoring for chronic pancreatitis, indicating that acute pancreatitis is associated with long-term mortality independent of future occurrence of chronic pancreatitis, particularly if the episodes are severe or non-gallstone-related. Importantly, the associations observed in our study were stronger when the analyses were limited to individuals under 50 years of age. This does not necessarily indicate an age-specific effect on mortality. Instead, the higher HRs estimates most likely reflect the lower number of expected deaths in the younger control group.

There are several potential mechanisms through which acute pancreatitis may lead to increased long-term mortality. In addition to the cancer-related mortality and the directly pancreatitis-related mortality noted by Czapári et al. development of post-pancreatitis diabetes mellitus is likely to contribute to the long-term increase in mortality [17]. Further, the systemic inflammatory stress caused by acute pancreatitis may in some individuals result in persistent organ dysfunction [39], non-resolved inflammatory responses [40], and acceleration of pre-existing conditions [41]. Our study highlights the need for further research to better understand the causes of death and the development of chronic diseases across different time periods following acute pancreatitis. Improved understanding of these factors is likely to support the development of tailored follow-up of patients with acute pancreatitis to treat or prevent long-term consequences of the disease.

In conclusion, in this population-based cohort study, we found increased long-term mortality after an episode of acute pancreatitis. The association was in part, but not fully, explained by comorbidities, recurrent acute pancreatitis, and development of chronic pancreatitis. The results stress the need for patient-tailored follow-up to prevent or treat sequelae following acute pancreatitis. Future studies should assess the cause-specific death and development of chronic diseases in patients with acute pancreatitis.

Author Contributions

Concept and design: All authors; Data analysis: D.S., O.S.-A., V.O.; Manuscript preparation: D.S., O.S.-A.; Manuscript revision: All authors; Final approval: All authors.

Ethics Statement

The study was approved by the Regional Ethical Review Board in Stockholm, Sweden (Ref. no. 2010/920-31/4 and 2015/0090-32).

Conflicts of Interest

The authors declare no conflicts of interest.