Association of Pancreatic Cancer with Acute Pancreatitis: A Systematic Review and Meta-Analysis.

Despite advancements in diagnostics and treatment, pancreatic cancer (PC) remains one of the deadliest forms of cancer. Owing to its asymptomatic nature, it is often diagnosed at an advanced stage, resulting in a 5-year survival rate of 10% ( 1 ). Currently, PC ranks as the seventh leading cause of cancer-related mortality worldwide ( 1 – 5 ). Moreover, the global incidence of PC has more than doubled, rising from 195,000 cases in 1990 to 448,000 cases in 2017 ( 4 ). PC is now projected to become the second leading cause of cancer-related death. Surgical resection is the only treatment with curative potential; however, only 15%–20% are typically eligible for resection, stressing the need to identify high-risk populations for early detection and improved prognosis ( 1 – 5 ). Acute pancreatitis (AP), an inflammatory disorder of the pancreas, is a common gastrointestinal condition, affecting 34 per 100,000 people annually worldwide ( 6 ). While most AP cases resolve without complication, there were some reports suggesting a potential bidirectional relationship between AP and PC ( 7 ). It has been widely speculated that AP may serve as an early manifestation of PC, but emerging evidence suggest the association may not be limited to a short-term window. Large population-based studies observed that the association with PC, although attenuated, remained significantly elevated even after 5–10 years of follow-up ( 8 , 9 ) These long-term associations raise the possibility that AP may in some cases reflect not only an early symptom but also a marker of increased susceptibility to PC. Preclinical data has provided some biological rationale for this association. In the study by Carriere et al. ( 10 ) AP was induced in genetically engineered mouse models with oncogenic Kirsten rat sarcoma virus (KRAS) mutations, where it accelerated acinar-to-ductal metaplasia and progression to PC ( 10 ). While these findings support a role for AP in promoting carcinogenesis under certain genetic conditions, causal evidence in humans is lacking and uncertain. Therefore, AP is best considered both a potential indicator for earlier cancer detection and, in select contexts, a possible contributor to long-term risk ( 10 ). Two previous meta-analyses by Liu et al (2020) ( 11 ) and Zhang et al (2018) ( 12 ) reported markedly higher rates of PC after AP compared to individuals without AP, particularly within the first 2 years. They also found a time-dependent association, with the highest (20-fold) diagnosis rate occurring within the first 2 years after AP, followed by a persistent 2-fold rate of PC diagnosis beyond 2 years. Understanding the burden of PC in this population and quantifying the associated likelihood of PC diagnosis is essential for constructing necessary surveillance strategies and improving early detection. Furthermore, specific risk factors such as chronic pancreatitis (CP), diabetes mellitus (DM), and smoking may exacerbate this association. Accordingly, this systematic review and meta-analysis aimed to provide a comprehensive analysis and address knowledge gaps by evaluating the prevalence of PC among patients with AP, the time-dependent association of PC after AP, and the influence of key risk factors on the likelihood of PC being diagnosed in patients with AP.

This systematic review and meta-analysis was conducted according to the recommendation of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guideline ( 13 ) (see Supplementary Table S1, http://links.lww.com/CTG/B403 ) and the Cochrane Handbook ( 14 ). This review adhered to the protocol registered with PROSPERO ( 15 ) (CRD42023470350), with some deviations to enhance the scope and clarity of the analysis. The details of the deviations can be found in Supplementary Digital Content (see Document S1, http://links.lww.com/CTG/B403 ). The research project was conducted under the Systems Education model coordinated by the Centre of Translational Medicine, Semmelweis University, and the Hungarian Pancreatic Study Group ( 16 , 17 ). Studies were eligible for inclusion if they met the predefined criteria based on the Population, Exposure, Comparison, and Outcomes (PECO) framework ( 18 ). Eligible studies included adult populations (18 years or older) with a diagnosis of AP compared with populations without AP and reporting the diagnosis of PC in each group. We included studies that report the incidence or prevalence of PC in the AP population according to the Condition, Context, and Population (CoCoPop) framework. Furthermore, we conducted a subset analysis based on the Population, Factors, and Outcome (PFO) framework to assess the influence of certain clinical and demographic factors (e.g., sex, smoking, and DM) on the likelihood of PC diganosis in AP patients. Randomized controlled trials and observational studies were included, while non-human research, case reports, case series, conference abstracts, narrative reviews, systematic reviews, meta-analyses, and pediatric population studies were excluded. A systematic search was conducted in 3 electronic databases: MEDLINE (via PubMed), EMBASE, and Cochrane Central Register of Controlled Trials from database inception until July 14, 2025. All publications were downloaded without any restrictions or filters. In addition to the main search, references of included studies were incorporated into the selection process ( 19 ). The search key was constructed with 2 domains. The first domain contained different terms for AP, while the second domain contained different variations of PC. The full search key can be found in Supplementary Digital Content (see Document S2, http://links.lww.com/CTG/B403 ). Records retrieved from the systematic search were imported into EndNote 20, where duplicate articles were removed (J.L.). The selection process was conducted using the Rayyan Intelligent Systematic Review program ( 20 ). Four independent reviewers (J.L., I.C.-M., J.N., and E.B.G.) screened the records first by title and abstract, followed by full-text assessment according to the inclusion criteria. A fifth reviewer (S.B.) resolved any disagreements when necessary. Cohen's kappa coefficient was calculated after each screening phase to assess inter-reviewer agreement ( 21 ). Four reviewers (J.L., I.C.-M., J.N., and E.B.G.) extracted all relevant data independently into separate, standardized Microsoft Excel ( 22 ) spreadsheets. Discrepancies in data extraction were resolved through consensus or by consultation with a fifth reviewer (S.B.). The following data were extracted: study characteristics (first author, year of publication, study design, study period, country, and the number of involved centers); population demographics (total number of participants, age, and percentage of male individuals); characteristics of AP and non-AP individuals (age, percentage of male individuals, AP etiology, presence of smoking, DM, alcohol, history of gallstones, and CP); outcome measure values (prevalence, incidence, OR, and HR of PC in AP); and follow-up data, and for the evaluation of risk factors for PC in patients with AP, the number of PC in patients with AP who experienced the evaluated risk factor, those nonexposed to the factor, and the total number of exposed and nonexposed individuals with AP. The risk of bias in individual studies was independently assessed by 2 reviewers (J.L. and J.N.) using the Joanna Briggs Institute (J.B.I.) proportions tool ( 23 ) for studies reporting incidence and prevalence data and the Quality in Prognostic Studies tool ( 24 ) for prognostic studies. The robvis (Risk-Of-Bias VISualization) tool was used to generate risk-of-bias plots for studies assessed with the JBI proportions tool ( 25 ). Details of each tool can be found in Supplementary Digital Content (see Document S3, http://links.lww.com/CTG/B403 ). As we assumed considerable between-study heterogeneity in all cases, a random-effects model was used to pool effect sizes in a frequentist framework. The different articles provided different outcome measures (which cannot be converted into each other without individual patient data); therefore, we provided separate meta-analyses on the measures of outcomes provided. OR was used as an effect size measure to assess the odds of PC among AP vs non-AP patients and to assess the odds of PC in AP patients in the presence vs absence of a given potential risk factor. To calculate the pooled OR, we extracted the total number of patients and those with the event of interest in each group separately from the studies if available. Otherwise, the OR provided among the eligible papers was used in a separate analysis. To assess the occurrence of PC among AP (vs no AP) patients, we pooled the hazards ratio (HR) and incidence rate (IR) ratios or (also from one-arm studies) the IR (in separate analysis). We performed subset analyses based on follow-up time categorized as short (0–24 months), medium (24–60 months), and long term (>60 months) (for IR, we performed calculations only for medium time values). In addition, we fitted a multivariate regression model to assess the HR (IR ratio) change in time, where several studies provided data for multiple time points. In addition, the prevalence of PC among patients with AP was also pooled because many articles only reported this measure. To assess the age of patients or the mean difference of age in the comparable groups and the mean time to PC from AP, we used mean or mean difference as effect size measure. We used the given mean and SD values from the articles, or we estimated these values from the quartiles (these studies symbolized with * on the forest plots). Results were considered statistically significant if the pooled 95% confidence interval (CI) did not contain the null value. We summarized the findings related to the meta-analysis in forest plots. Between-study heterogeneity was described by the between-study variance (τ 2 ) and the Higgins and Thompson's I 2 statistics ( 26 ). We reported the prediction interval if the study number was over 8. All statistical analyses were calculated by R software ( 27 ) using the meta ( 28 ) package for basic meta-analysis calculations and plots, dmetar ( 29 ) package for additional influential analysis calculations and plots, metafor ( 30 ) and clubSandwich ( 31 ) packages for the multivariate (correlated) calculations. For additional details on calculations, data synthesis, publication bias assessment, and influential analyses, please refer to Supplementary Digital Content (see Document S4, http://links.lww.com/CTG/B403 ).

The systematic search and reference chasing identified 36,461 and 859 records, respectively. We included a total of 61 studies in the systematic review, with 53 providing sufficient data for quantitative synthesis. The search and selection process is depicted in the PRISMA 2020 flow diagram (Figure 1 ). PRISMA flowchart of the article selection process. Of the 61 studies included in the systematic review, we had 43 retrospective cohort studies, 7 retrospective case-control studies, 9 prospective cohort studies, 1 cross-sectional study, and 1 international survey. The studies were published between 1985 and 2025 and collectively included around 40,000,000 participants, with the largest sample size coming from Khoudari et al (2020) ( 32 ) (32,970,850) followed by Munigala et al (2023) ( 33 ) (7,147,859). The included studies were conducted in diverse regions of the world, including 6 from England, 13 from the United States, 20 from Europe, and 22 from Asia, providing a broad representation of populations. A detailed summary of the baseline characteristics of the included studies is provided in Table 1 . Main characteristics of included studies AP, acute pancreatitis; CP, chronic pancreatitis; DM, diabetes mellitus; HR, hazards ratio; IR, incidence rate; MD, mean difference; NA, not available/applicable; OR, odds ratio; PC, pancreatic cancer; SM, single mean; USA, United States of America. A total of 19 ( 32 , 34 , 36 , 37 , 40 – 43 , 46 , 47 , 49 , 55 , 56 , 61 , 63 , 66 , 69 , 71 , 80 ) studies comprising 360,517 patients with AP were included in the analysis. Across the included studies, the pooled prevalence of PC in patients with AP was 2% (CI: 2–4%; I 2 = 99.5%, CI: 99%–99.5%) (Figure 2 ). Prevalence of pancreatic cancer in acute pancreatitis. AP, acute pancreatitis; CI, confidence interval; PC, pancreatic cancer. Data from 8 studies ( 8 , 9 , 33 , 38 , 47 , 54 , 73 , 76 ) were pooled to evaluate the frequency of PC over time after an initial episode of AP (Figure 3 ). The likelihood of PC being diagnosed was highest near the episode of AP and gradually declined over time; however, the association with PC diagnosis remained significantly high even in the long term. In addition, we pooled HR from the same 8 studies ( 8 , 9 , 33 , 38 , 47 , 54 , 73 , 76 ) that adjusted for key confounders, including age and sex; some also adjusted for smoking, alcohol consumption, comorbidities (e.g., DM, gallstone disease, Charlson comorbidity index), and socioeconomic factors (Table 2 ). The HRs were pooled together in a multivariate model and stratified by follow-up period after AP: short-term (60 months; HR: 1.71; CI: 1.22–2.40; I 2 = 76.6%, CI: 50.9%–88.8%). For individual figures, please refer to Supplementary Digital Content (see Supplementary Figures S2-4, http://links.lww.com/CTG/B403 ). Risk rate of pancreatic cancer according to follow-up time after acute pancreatitis. PC, pancreatic cancer. Risk rate of pancreatic cancer following acute pancreatitis stratified according to follow-up periods AP, acute pancreatitis; CI, confidence interval; HR, hazards ratio, No., Number; PC, pancreatic cancer. The association between several clinical and demographic factors and the likelihood of PC diagnosis among patients with AP was explored. Subsequent CP ( 44 , 52 , 54 , 61 , 76 , 80 ) diagnosed after an episode of AP demonstrated a strong association with PC (OR: 3.71, CI: 2.00–6.90; I 2 = 85%, CI: 70%–93%). Patients with AP older than 50 years were also strongly associated with PC (OR: 4.04, CI: 2.73–5.97; I 2 = 12%, CI: 0%–87%) (see Supplementary Figure S16, http://links.lww.com/CTG/B403 ) ( 46 , 47 , 54 , 61 ). Among metabolic factors, new-onset diabetes (NOD) ( 46 , 56 , 61 , 80 ) carried a significant association with subsequent PC diagnosis (OR: 2.22, CI: 1.02–4.84; I 2 = 21%, CI: 0%–92%) (see Supplementary Figure S8, http://links.lww.com/CTG/B403 ). Smoking ( 33 , 46 , 61 , 81 ) showed a nonsignificant trend toward higher PC diagnosis (OR: 1.49, CI: 0.61–3.63; I 2 = 74%, CI: 27%–91%) (see Supplementary Figure S10, http://links.lww.com/CTG/B403 ). Regarding AP etiologies, idiopathic AP ( 8 , 52 , 78 , 80 , 81 ) was associated with nearly a 3-fold higher likelihood of PC diagnosis compared with nonidiopathic AP (OR: 2.97, CI: 1.44–6.13, I 2 = 63%, CI: 3%–86%) (see Supplementary Figure S18, http://links.lww.com/CTG/B403 ). By contrast, the effects of gallstone-related AP (see Supplementary Figure S9, http://links.lww.com/CTG/B403 ) ( 8 , 33 , 52 , 54 , 61 , 78 , 80 , 81 ) and alcoholic AP (see Supplementary Figure S5, http://links.lww.com/CTG/B403 ) ( 8 , 47 , 52 , 58 , 78 , 80 , 81 ) were evaluated, revealing no significant associations with PC (Table 3 ). Influence of clinical and demographic factors on the likelihood of pancreatic cancer diagnosis in acute pancreatitis AP, acute pancreatitis; CI, confidence interval; CP, chronic pancreatitis; DM, diabetes mellitus, No, number; OR, odds ratio; PC, pancreatic cancer; RAP, recurrent acute pancreatitis; RF, risk factor. Other factors, including AP severity (see Supplementary Figures S20-22, http://links.lww.com/CTG/B403 ) ( 42 , 46 , 61 , 80 , 81 ), recurrent AP (see Supplementary Figure S19, http://links.lww.com/CTG/B403 ) ( 33 , 42 , 54 , 76 , 80 ), and male sex (see Supplementary Figure S27, http://links.lww.com/CTG/B403 ) ( 8 , 46 , 49 , 54 , 76 , 81 ), did not exhibit significant associations with PC diagnosis. Patients who developed PC after an episode of AP tended to be younger at diagnosis compared with PC patients without a history of AP (mean difference: −4.8 years, CI: −9.0 to −0.7; 7 studies, n = 7,184 vs 113,738) (see Supplementary Figure S14, http://links.lww.com/CTG/B403 ) ( 8 , 45 , 53 , 62 , 64 , 72 , 75 ). The interval from AP to PC diagnosis was typically short, with a pooled mean time of 1.2 years (CI: 0.4–1.9; 7 studies, n = 620) (see Supplementary Figure S15, http://links.lww.com/CTG/B403 ) ( 8 , 49 , 51 , 52 , 77 , 78 , 80 ). Nevertheless, as shown in large population-based cohorts, the association between AP and subsequent PC diagnosis persisted even at longer follow-up, though at lower magnitude. Importantly, PC diagnosed after AP was more likely to be detected at an earlier stage (OR: 1.45, CI: 1.26–1.67; 3 studies, n = 6,302 vs 98,298) (see Supplementary Figure S17, http://links.lww.com/CTG/B403 ), which may reflect comprehensive diagnostic evaluation triggered by the AP episode. With respect to tumor location, most cancers occurred in the pancreatic head (61%, CI: 46%–75%; 9 studies, n = 6,855), followed by the body (12%, CI: 6%–22%; 6 studies, n = 4,429) and tail (6%, CI: 3%–10%; 6 studies, n = 4,429) (see Supplementary Figures S24-26, http://links.lww.com/CTG/B403 ). Methodological quality of 42 studies was assessed with the JBI proportions checklist where 29 studies had an overall low risk of bias, 13 studies had moderate, and 1 study had a high risk. With the Quality in Prognostic Studies tool, we assessed 33 studies, where 21 studies had low risk, 11 studies had moderate risk, and 1 study had high risk of bias. The main reasons for the potential bias in the studies were the lack of consideration for the confounders and the lack of details in how the prognostic factor and outcomes were measured. The risk of bias assessments for each outcome can be found in Supplementary Digital Content (see Supplementary Tables S2-3, http://links.lww.com/CTG/B403 ). Moderate-to-high heterogeneity was observed in all evaluated outcomes, which may arise from several factors. First, potential confounders such as smoking, alcohol consumption, and DM—known risk factors for PC—could contribute to variations in the associations across studies. Second, differences in study design and demographics may play a role; while some studies were conducted in single-center settings, others used national databases with broader, more diverse cohorts. In addition, the inclusion of studies from multiple regions and countries may introduce further heterogeneity, as the IR of PC differs based on the region. Among the analyses, publication bias could be assessed only for the prevalence of PC which did not suggest any small-study effects ( P = 0.5376), indicating no statistical evidence of publication bias. Visualization with funnel plots and the Egger's test result (with Peters' modification) can be found in Supplementary Digital Content (see Supplementary Figures S1, http://links.lww.com/CTG/B403 ).

Patients with AP have a higher likelihood of PC diagnosis, especially within the first 2 years. Although the risk decreases with time, it remains higher than in the general population long term. Newly diagnosed CP, NOD, idiopathic AP may further elevate the likelihood of PC diagnosis. PC diagnosed after AP tends to occur at a younger age, more often at an earlier stage, typically in the pancreatic head. These findings warrant careful investigation in the selected subpopulation, while future research is required to clarify causality and generate adequate surveillance protocols.

In this systematic review and meta-analysis, we found a clinically relevant prevalence of PC in AP, an increased likelihood of PC diagnosis after AP, and the potential influence of specific clinical and demographic factors. These findings strengthen the association between AP and PC, which warrants increased clinical awareness. The prevalence of PC among patients with AP in our meta-analysis was 2%, suggesting that at a given time point, approximately 2 out of every 100 patients with AP have a concurrent or previously undiagnosed PC. Although this may seem low, it is clinically relevant given the poor prognosis of PC and the potential for it being missed when AP occurs. This highlights the need for careful investigation for underlying malignancy in patients with AP. The underlying mechanisms of PC-AP association remain unclear and are likely multifactorial ( 2 ). In the short term, AP often represents the first clinical manifestation of an occult cancer, where reverse causation, careful diagnostic workup after AP, potential surveillance bias, or a delay in the diagnosis of pre-existing PC likely explain the increase in PC diagnoses within the first 2 years. Our meta-analysis revealed a gradual time-dependent decline in the frequency of PC diagnosis after AP but still remained significant. These estimates were adjusted for various confounding variables, suggesting that the observed trends were independent of these variables. Many studies ( 8 , 9 , 38 , 39 , 47 , 49 , 53 , 54 , 56 , 75 , 80 , 83 , 87 , 88 , 90 ) have reported similar trends. Pang et al (2018) ( 88 ) reported a HR of 9.99 (CI: 3.20–31.16) in a cohort of 1,079 AP patients with a 2-year lag period. Beyond 2 years after AP, the likelihood of PC diagnosis gradually decreases; however, it persists for at least 5 years, with Kirkegard et al (2018) ( 8 ) and Sadr-Azodi et al (2018) ( 54 ) reporting a 2-fold increased hazards even after 10 years. This suggests that while some early cases may be due to pre-existing PC, AP itself could still contribute to long-term pancreatic carcinogenesis, potentially through chronic inflammation, fibrosis, or metabolic alterations. Large registry studies have also shown that patients with AP-related PC are diagnosed at younger ages and more often at earlier stages (see Supplementary Figure S17, http://links.lww.com/CTG/B403 ) ( 60 , 61 , 75 ), consistent with earlier detection triggered by the pancreatitis episode. Our analysis shows distinct clinical characteristics of PC diagnosed in AP. Compared with patients without prior AP, those with AP were diagnosed with PC at a younger age (see Supplementary Figure S14, http://links.lww.com/CTG/B403 ) ( 8 , 45 , 53 , 62 , 64 , 72 , 75 ), and the interval from AP to cancer diagnosis was typically short (see Supplementary Figure S15, http://links.lww.com/CTG/B403 ) ( 8 , 49 , 51 , 52 , 77 , 78 , 80 ). This supports that AP can serve as an initial manifestation that prompts earlier clinical evaluation, thereby resulting in the earlier detection of PC. Indeed, the higher likelihood of early-stage diagnosis among AP patients suggests that AP-triggered investigations may uncover PC at a resectable stage, which has important implications for prognosis. Furthermore, the predominance of tumors in the pancreatic head is consistent with the pathophysiology of AP because head lesions are more likely to obstruct the pancreatic duct and precipitate an attack, with Singh et al ( 75 ) observing that the head of pancreas cancer occurred twice as commonly in patients with prior AP compared with those without AP. While these findings strengthen the rationale for targeted surveillance after AP—particularly in older patients or those with idiopathic etiology—they should be interpreted with caution. The younger age at diagnosis may reflect detection bias, and the short interval again raises the possibility of reverse causation, where AP represents the first clinical manifestation of an occult tumor rather than an independent risk factor. Nonetheless, the combination of earlier stages at diagnosis and distinct anatomical distribution emphasizes the need for clinicians to maintain a high level of suspicion for PC in patients presenting with AP. We performed subset analyses to gain a deeper insight into how different risk factors influence the odds of developing PC in AP patients. Among the factors analyzed, subsequently diagnosed CP after AP was significantly associated with an increased odds of PC. This aligns with evidence suggesting that the transition from AP to CP reflects underlying pancreatic injury and fibrosis, which is known to contribute to carcinogenesis ( 91 ). However, a study conducted by Jeon et al (2020) ( 89 ) reported negligible risk of PC in CP patients with prior AP (HR: 0.92; 95% CI: 0.49–1.73) likely because of differences in study population, as their focus was exclusively on patients with CP. Nonetheless, our findings suggest the importance of close monitoring of patients with AP who progress to CP. In addition to CP, recurrent AP also appears to be frequently associated with PC. Two studies ( 33 , 54 ) reported a higher association of PC with increasing number of recurrent episodes of AP, with one study ( 54 ) finding the highest rate of PC diagnosis when multiple episodes occurred near the initial episode of AP (<2 years). Although our analysis could not reach statistical significance (see Supplementary Figure S19, http://links.lww.com/CTG/B403 ), there is a trend in all except one study ( 42 ) that suggests increased odds for PC. By contrast, alcoholic or biliary etiology did not significantly increase the likelihood of PC diagnosis, though both analyses exhibited substantial heterogeneity. In addition, both are self-limiting, acute, inflammatory conditions with a specific cause that does not necessarily lead to progressive pancreatic damage. However, alcoholic pancreatitis can result in chronic inflammation and fibrosis if alcohol abuse continues, but this was not clear within the included studies. While alcoholic and biliary etiology were insignificant, idiopathic AP was found to have a significant association with PC, with Hussein et al ( 80 ) reporting that 3% of patients with AP who developed PC had idiopathic AP compared with 1% in the nonidiopathic group (see Supplementary Figure S18, http://links.lww.com/CTG/B403 ). Smoking and prevalent DM in AP were not significantly associated with PC. New-onset DM in AP, however, was observed to have a significant association with PC diagnosis (see Supplementary Figure S8, http://links.lww.com/CTG/B403 ) ( 56 , 61 , 80 ). Choi et al ( 67 ) found a 6-fold increased risk in patients with DM with AP compared with those without AP. In addition, Jeon et al ( 68 ) found a higher prevalence of PC in early-stage patients with DM with AP compared with late-stage patients with DM (5% vs 2%), suggesting that new-onset DM may reflect early pancreatic carcinogenesis rather than being an independent risk factor. The lack of statistical significance in our subset analyses of the pancreatitis etiologies, smoking, and prevalent DM may be due to sample size limitations, heterogeneity, or insufficient long-term follow-up to notice delayed carcinogenic effects. The lack of association between PC and the other assessed factors indicates that the progression to CP or the presence of NOD, rather than AP itself or traditional risk factors, plays a key role in carcinogenesis. Our comprehensive analysis incorporates a robust methodology and a large, diverse sample, enhancing the generalizability of our findings. In addition, despite considerable heterogeneity in the time-dependent analysis, new-onset CP, and OR of PC in AP, the consistency of the effect estimates across diverse studies reinforces the validity and generalizability of our findings. Furthermore, the time-stratified analyses allowed us to distinguish the short-term clustering of PC diagnoses, likely reflecting reverse causation, from the persistently elevated long-term association. In addition, we evaluated clinically relevant modifiers such as age, idiopathic AP, and NODs, and we summarized tumor characteristics, including stage and location, which have not been addressed in prior meta-analyses. However, some limitations must be acknowledged. First, the retrospective design of most of the included studies limits causal inference and introduces potential selection bias. In addition, there was considerable heterogeneity across multiple analyses, suggesting variations in diagnostic criteria, follow-up durations, and study designs which may have influenced the pooled estimates. Several confounding lifestyle factors, including smoking or alcohol, could not be fully accounted for in each analysis. Furthermore, three studies ( 9 , 73 , 76 ) included in the time-dependent analysis ( Figures S2-4, http://links.lww.com/CTG/B403 ) used partially overlapping populations from the Korean National Health Insurance Service (NHIS) and the National Health Screening Program (NHSP) databases. To provide a comprehensive overview, we included these studies in the analysis while acknowledging the potential overlap. Leave-one-out sensitivity analyses excluding the overlapping studies showed that the pooled hazard ratios remained materially unchanged, supporting the robustness of our findings, with only S. Park et al.( 76 ) exerting notable influence ( Figures S28-30, http://links.lww.com/CTG/B403 ) Finally, although the association between AP and PC diagnosis remained significant after 5 years, causality cannot be inferred. Translating scientific results into real-world benefits has high importance ( 92 , 93 ). Our findings suggest that AP can serve as an early sign for PC, especially in patients older than 50 years, those with idiopathic AP, those who develop NOD, and those who progress to CP. Given this, clinicians should maintain a high level of awareness in these subgroups with careful consideration of follow-up or detailed diagnostic evaluation. Future prospective studies should be conducted, implementing risk stratification models, integrating biomarkers, and conducting biology-focused research in humans to clarify the causal relationship between AP and PC, if any.

Guarantor of the article: Specific author contributions: J.L.: conceptualization, data curation, investigation, project administration, writing–original draft. I.C.-M.: conceptualization, data curation, investigation, writing–review and editing. J.N.: conceptualization, data curation, investigation, writing–review and editing. E.B.G.: conceptualization, data curation, investigation, writing–review and editing. D.S.V.: conceptualization, formal analysis, investigation, validation, visualization, writing–review and editing. E.Á.Sz.: conceptualization, methodology, supervision, writing–review and editing. M.O.: conceptualization, supervision, writing–review and editing. R.P.: conceptualization, supervision, writing–review and editing. S.B.: conceptualization, methodology, supervision, writing–review and editing. P.H.: conceptualization, methodology, supervision, writing–review and editing. All authors certify that they have participated sufficiently in the work to take public responsibility for the content, including participation in the concept, design, analysis, writing, or revision of the manuscript. Financial support: Funding was provided by the Centre for Translational Medicine, Semmelweis University, Hungarian Ministry of Innovation and Technology, National Research, Development and Innovation Fund (TKP2021-EGA-23 to P.H.), NKFIH project grants K131996 and K147265 (to P.H.). This study was supported by Semmelweis University Research and Innovation Fund. Potential competing interests: None to report. Ethical approval: No ethical approval was required for this systematic review with meta-analysis, as all data were already published in peer-reviewed journals. No patients were involved in the design, conduct or interpretation of our study. The datasets used in this study can be found in the full-text articles included in the systematic review and meta-analysis. Data availability: All the data analyzed in this study are available in the full text of the studies and supplementary material included. Study Highlights WHAT IS KNOWN ✓ Pancreatic cancer is sometimes diagnosed shortly after an episode of acute pancreatitis. ✓ The long-term association of pancreatic cancer in patients with acute pancreatitis remains uncertain. WHAT IS NEW HERE ✓ Patients with acute pancreatitis (AP) have a higher likelihood of pancreatic cancer (PC) diagnosis, especially within the first 2 years. Although the association decreases with time, it remains significant long term. ✓ Newly diagnosed chronic pancreatitis (CP), new-onset diabetes, idiopathic AP may further elevate the likelihood of PC diagnosis. PC diagnosed after AP tends to occur at a younger age, more often at an earlier stage, typically in the pancreatic head. ✓ Pancreatic cancer is sometimes diagnosed shortly after an episode of acute pancreatitis. ✓ The long-term association of pancreatic cancer in patients with acute pancreatitis remains uncertain. ✓ Pancreatic cancer is sometimes diagnosed shortly after an episode of acute pancreatitis. ✓ The long-term association of pancreatic cancer in patients with acute pancreatitis remains uncertain. ✓ Patients with acute pancreatitis (AP) have a higher likelihood of pancreatic cancer (PC) diagnosis, especially within the first 2 years. Although the association decreases with time, it remains significant long term. ✓ Newly diagnosed chronic pancreatitis (CP), new-onset diabetes, idiopathic AP may further elevate the likelihood of PC diagnosis. PC diagnosed after AP tends to occur at a younger age, more often at an earlier stage, typically in the pancreatic head. ✓ Patients with acute pancreatitis (AP) have a higher likelihood of pancreatic cancer (PC) diagnosis, especially within the first 2 years. Although the association decreases with time, it remains significant long term. ✓ Newly diagnosed chronic pancreatitis (CP), new-onset diabetes, idiopathic AP may further elevate the likelihood of PC diagnosis. PC diagnosed after AP tends to occur at a younger age, more often at an earlier stage, typically in the pancreatic head.