The Diabetes-Pancreatic Cancer Risk Relationship over Time: A Systematic Review and Meta-Analysis

Abstract

Background/Objectives: The relationship between diabetes and pancreatic cancer (PCa) is controversial. In this meta-analysis and systematic review, we investigated diabetes and time since diagnosis as risk factors for PCa. Methods: Cohort and case-control studies were retrieved through a literature search. RevMan 5.4 software and a random effects model were used to estimate summary risks with their 95% confidence intervals (CIs), and the Newcastle–Ottawa Scale (NOS) was used to assess study quality. Results: Included were 23 studies representing 30,875,355 participants and 86,980 cases of PCa. The summary risk for the 14 case-control studies was 2.30 (95% CI: 2.03–2.62) and for the 9 cohort studies was 2.39 (95% CI: 2.09–2.73). The risk decreased with time after diabetes diagnosis: 3.27, 2.25, 1.55, and 1.12 for <2, 2–5, 5–10, and >10 years, respectively, in the case-control studies. The cohort studies also showed an increased risk of PCa in the first 2 years (4.29) and a decrease over time. Quality scores according to the NOS were 6–9 (good and fair quality), for an overall average of 7.82. Conclusions: Diabetes is a risk factor for PCa and this risk is much higher in the 2 years following diabetes diagnosis. In this period, the subgroup of patients who, through clinical follow-up and/or cancer screening, would have better clinical outcomes should be identified. Bearing in mind the poor survival rate for PCa, diabetes interventions focused on preventing onset and delaying progression via modifiable risk factors to reduce PCa incidence.

1. Introduction

Diabetes and cancer are both prevalent and on the rise. A number of recent studies and meta-analyses report a relationship between both entities [1]. Around 537 million adults worldwide have diabetes, most of whom have type 2 diabetes, and this number is expected to increase to 783 million by 2045. Worldwide, around 45% of people live with undiagnosed diabetes [2].

Cancer is the second cause of mortality in developed countries, and pancreatic cancer (PCa), which is growing in incidence, is among the most lethal cancers, with a 5-year survival rate of 10% [3]. PCa lethality is linked to the fact that the cancer is usually already very advanced by diagnosis as a result of mild and unclear symptom expression [4].

The risk of PCa increases with impaired glucose tolerance, including in persons who are overweight and with elevated serum glucose levels. The mechanisms underlying these associations possibly include the tumorigenic effect of hyperglycaemia, the mutagenic effect of hyperinsulinemia associated with obesity, and subclinical chronic inflammation caused by fatty infiltration of the pancreas [5]. However, as yet there is little direct evidence in support of these mechanisms.

Risk factors for developing PCa include a family history, obesity, alcohol consumption, smoking [6], and metabolic syndrome [7]. While some studies suggest that diabetes may be another risk factor, there is no consensus regarding this premise. The controversy lies in whether diabetes is more a consequence of the tumour rather than a risk factor [8]. For all these reasons, the role played by diabetes in PCa is unclear.

Previous meta-analyses indicate that the risk of diabetes is inversely related to the duration of diabetes [9,10,11]. A meta-analysis by Pang et al. [12] concludes that diabetes doubles the PCa risk and that this effect increases progressively with diabetes duration. However, in a 2015 meta-analysis of 44 studies by Song et al. [9], the average relative risk (RR) for diabetes as a risk factor for PCa was reported to be 1.64 (95% confidence interval (CI): 1.52–1.78), with risk decreasing with diabetes duration.

To advance our understanding of the relationship between diabetes and PCa, the literature needs to be combined and re-evaluated with a view to establishing specific prevention and management strategies.

Our aim, therefore, was to update the evidence regarding diabetes as a possible risk factor for PCa through a meta-analysis and systematic review. We also specifically studied the association between PCa risk and the time elapsed since diabetes diagnosis.

2. Materials and Methods

The work was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [13]. The study protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO, No. CRD42025646486). A summary of the methods and any deviations from the published protocol are described below.

Search strategy

A systematic search was carried out following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Using the keywords “pancreatic neoplasm” and “diabetes” combined using the AND operator, the PubMed, SCOPUS, and Web of Science databases were searched for observational epidemiological cohort and case-control studies published between 1 January 2015 and 30 April 2025.

Inclusion and exclusion criteria

The studies to be included in our meta-analysis were defined as case-control and cohort studies, published in peer-reviewed journals in English, French, or Spanish, and produced in Europe, North America, and Asia. The studies were required to have pancreatic cancer as the dependent variable and diabetes as a risk factor and to furnish estimated risk data, namely, RR, odds ratio (OR), or hazard ratio (HR) values with the corresponding CI. Studies were also included if they reported a point estimate of risk as adjusted RR, HR, or OR values with a 95% CI or reported data from which these values could be calculated. Excluded were studies that did not meet the previous criteria, descriptive studies, prevalence studies, previous meta-analyses, letters to the editor, and inaccessible articles.

Data extraction

Data were extracted independently by two researchers using a standardized protocol and data-recording form, and the results were subsequently reviewed by the other authors.

The data were stored in a database designed using Microsoft Excel. Following extraction and evaluation, the data were independently reviewed and evaluated. Titles and abstracts were reviewed, and subsequently, the entire texts of the finally selected articles were read.

Extracted data were as follows: authorship, publication year and region, and study type; mean study participant age, number of men and women, and sample size; number of patients with and without diabetes mellitus, number of patients diagnosed with PCa, presence of pre- and post-diagnosis diabetes (diagnosis of diabetes at least 6 months before pancreatic cancer in cases and diagnosis before study entry in controls), years with a diabetes diagnosis prior to the PCa diagnosis (<2 years, 2–5 years, 5–10 years, and >10 years), and OR/RR (and 95% CI) of diabetes for PCa; study quality; and other studied risk factors (excess weight, tobacco/alcohol use, and sedentarism).

Assessed as possible confounders were excess weight, tobacco/alcohol use, sedentarism, family history, and medication use (metformin or insulin) based on articles reviewed and previous meta-analyses.

Study quality

To evaluate quality, we used the Newcastle–Ottawa Scale (NOS), which rates the quality of non-randomized studies included in meta-analyses according to study group selection, study group comparability, and exposure/outcome of interest for case-control and cohort studies (4, 2, and 3 points, respectively, for a maximum of 9 points). According to the NOS, a good quality study should score 3 or 4 points for selection, 1 or 2 points for comparability, and 2 or 3 points for exposure/outcome. The scale scores were collected by two independent researchers for each study. Discrepancies between them were resolved with the intervention of a third researcher.

Statistical analysis

The different studies, grouped into case-control studies and cohort studies, were statistically described, and summary OR and RR values, respectively, were calculated along with their 95% CIs.

In view of the heterogeneity of the studies, for each study phase, a descriptive and narrative analysis of the results tables was carried out following the PRISMA guide structure [13]. Study heterogeneity was analysed using the Q test, considered significant for p < 0.1. The I² statistic was calculated to determine percentage variations between studies as a result of heterogeneity. If I² > 50%, heterogeneity was considered evident and a random effects model was used in the analyses; otherwise, a fixed effects model was used to pool the data.

Sensitivity analysis was performed to investigate the influence of specific studies on overall meta-analysis estimates by removing one study at each stage. To explore how association strengths varied, we performed stratified analyses of the data by sex, study design, study region, and years since diabetes diagnosis.

Publication bias was assessed using Begg’s test [14], with bias considered present when p < 0.05, and funnel plots were created to visually depict the relationship between study precision and effect size.

RevMan 5.4 software (Cochrane Collaboration, Oxford, UK) was used to analyse the data and to create the forest plots and JASP 0.19.3. software (University of Amsterdam, Amsterdam, The Netherlands) to produce the funnel plots.

3. Results

A literature search in PubMed, Scopus, and Web of Science located a total of 29,106 articles. The discarded articles were as follows: 20,267 non-observational studies; 4400 published before 1 January 2015; 2071 including participants aged <19 years; 30 in languages other than English, French, or Spanish; and 1458 for which the full text was unavailable. The titles and abstracts of the remaining 880 articles were reviewed and any article that did not meet the inclusion criteria was discarded (systematic reviews, duplicates, and articles whose title and/or abstract were not compatible with our study aim). Finally, the full text of each of the remaining 23 articles (9 cohort and 14 case-control studies) was reviewed (Figure 1).

The 23 articles included a total of 30,875,355 participants: 30,563,149 in cohort studies and 312,206 in case-control studies (Table 1 and Table 2, respectively). The sample included 86,980 PCa cases. The mean age of participants was 60.3 years (60.2 years in cohort studies and 60.5 years in case-control studies), for an age range of 19–93 years. The studies were conducted primarily in the USA (8/23), Europe (7/23), and East Asia (8/23), and all 23 articles were published in English (Table 1 and Table 2).

For the case-control studies, the summary OR was 2.30 (95% CI: 2.03–2.62) (Figure 2) [15,16,17,18,19,20,21,22,23,24,25,26,27,28]. All the studies reported OR estimates in a relatively narrow range between 1.56 [27] and 2.76 [18], except those by Antwi et al. [15] (OR 3.70) and Petrusel et al. [20] (OR 5.38). Overall, study heterogeneity was 75% (p < 0.001).

Table 1. Characteristics of case-control studies.

Author and Year	Title	Country	Age	Sample	Men	Women	PCa Cases in Patients with Diabetes	Patients with Diabetes	Controls with Diabetes	Total Controls	OR	Confidence Interval	Others
Van Tran T, 2021 [23]	Risk factors of Pancreatic Cancer in Vietnam: A Matched Case-Control Hospital-Based Study	Vietnam	59	392	232	160	35	196	20	188	3.09	1.54–6.68	Cancer family history, tobacco use, alcohol, inflammatory diseases, hepatitis B virus infection
Li X, 2018 [26]	ABO Blood Group and Diabetes Mellitus Influence the Risk for Pancreatic Cancer in a Population from China	China	63.5	672	357	315	61	264	66	423	1.62	1.10–2.39	Family history of pancreatic cancer, tobacco, alcohol, chronic hepatitis B infection, chronic pancreatitis, ABO blood type
Olson SH, 2016 [17]	Weight Loss, Diabetes, Fatigue, and Depression Preceding Pancreatic Cancer	USA	61.9	973	514	459	63	463	23	463	3.55 (<3 years DM)	1.35–3.66	Body mass index, tobacco, family history of pancreatic cancer, history of prior cancer, weight loss, fatigue, depression, concentration
Valente R, 2017 [19]	Risk and protective factors for the occurrence of sporadic pancreatic endocrine neoplasms	Europa	59	840	237	603	35	281	58	603	2.09	1.27–3.45	Tobacco, alcohol, insulin, past medical history, family history of pancreatic cancer
Tan PS, 2023 [24]	Temporality of body mass index, blood tests, comorbidities and medication use as early markers for pancreatic ductal adenocarcinoma (PDAC): a nested case-control study	England	>18	288,356	143,912	145,444	6925	28,137	32,962	261,219	4.93	1.36–3.88	Body mass index, alcohol, tobacco, comorbidities, medications (metformin, insulin)
Ben Q, 2016 [16]	Risk factors for sporadic pancreatic neuroendocrine tumors: A case-control study	China	49.7	999	448	552	24	142	50	614	2.29	1.36–3.88	Tobacco, alcohol, family history of pancreatic cancer
Zheng Z, 2016 [18]	Risk Factors for Pancreatic Cancer in China: A Multicenter Case-Control Study	China	58.85	646	362	284	49	223	26	281	2.96	1.48–5.92	Obesity tobacco, family history of pancreatic cancer, tea and coffee consumption
Antwi SO, 2016 [15]	Pancreatic cancer: associations of inflammatory potential of diet, cigarette smoking and long-standing diabetes	USA	66.5	2573	1416	1157	211	817	151	756	3.27	2.58–4.17	Inflammatory diet, tobacco, duration of type II diabetes
Lu Y, 2015 [25]	New-onset type 2 diabetes, elevated HbA1c, anti-diabetic medications, and risk of pancreatic cancer	Sweden-England	20–75	5529	3166	2363	175	529	856	5000	2.36	1.94–2.89	Body mass index, alcohol, tobacco, fasting glucose, fasting insulin, allergies
Walker EJ, 2015 [27]	Metformin use among type 2 diabetics and risk of pancreatic cancer in a clinic-based case-control study	USA	61.5	1405	704	701	81	536	89	869	1.41	1.02–1.96	Body mass index, tobacco, alcohol, pancreatitis, family history of pancreatic cancer, diabetes duration, medication use (metformin, insulin…)
Petrusel L, 2020 [20]	Risk Factors in Pancreatic Adenocarcinoma: The Interrelation with Familial History and Predictive Role on Survival	Romania	63.4	624	376	248	115	279	36	312	4.1/new-onset: 11.78	2.58–31.42	Body mass index, alcohol, tobacco, coffee, chronic pancreatitis, family history of pancreatic cancer
Dayem Ullah AZM, 2021 [22]	Temporality of clinical factors associated with pancreatic cancer: a case-control study using linked electronic health records	England	55.1	8962	4569	4373	360	965	1045	4355	1.74/new-onset: 1.95	1.25–3.03	Comorbidities (gastrointestinal, cardiometabolic and respiratory), obesity, tobacco, alcohol
Farias AJ, 2020 [21]	Diabetes-related complications and pancreatic cancer incidence in the multiethnic cohort	USA	Tramo edad	2161	1025	1136	197	433	524	1728	1.22	1.09–1.36	Tobacco, alcohol, body mass index, comorbidities
Wang Y, 2016 [28]	Complex interplay between type 2 diabetes and pancreatic cancer: insights from observational and mendelian randomization analyses	China	67.34	1163	766	497	194	581	107	582	2.06	1.56–2.71	Duration of type II diabetes, obesity, tobacco, alcohol, medication (oral medication or insulin)

Table 1. Characteristics of case-control studies.

Author and Year	Title	Country	Age	Sample	Men	Women	PCa Cases in Patients with Diabetes	Patients with Diabetes	Controls with Diabetes	Total Controls	OR	Confidence Interval	Others
Van Tran T, 2021 [23]	Risk factors of Pancreatic Cancer in Vietnam: A Matched Case-Control Hospital-Based Study	Vietnam	59	392	232	160	35	196	20	188	3.09	1.54–6.68	Cancer family history, tobacco use, alcohol, inflammatory diseases, hepatitis B virus infection
Li X, 2018 [26]	ABO Blood Group and Diabetes Mellitus Influence the Risk for Pancreatic Cancer in a Population from China	China	63.5	672	357	315	61	264	66	423	1.62	1.10–2.39	Family history of pancreatic cancer, tobacco, alcohol, chronic hepatitis B infection, chronic pancreatitis, ABO blood type
Olson SH, 2016 [17]	Weight Loss, Diabetes, Fatigue, and Depression Preceding Pancreatic Cancer	USA	61.9	973	514	459	63	463	23	463	3.55 (<3 years DM)	1.35–3.66	Body mass index, tobacco, family history of pancreatic cancer, history of prior cancer, weight loss, fatigue, depression, concentration
Valente R, 2017 [19]	Risk and protective factors for the occurrence of sporadic pancreatic endocrine neoplasms	Europa	59	840	237	603	35	281	58	603	2.09	1.27–3.45	Tobacco, alcohol, insulin, past medical history, family history of pancreatic cancer
Tan PS, 2023 [24]	Temporality of body mass index, blood tests, comorbidities and medication use as early markers for pancreatic ductal adenocarcinoma (PDAC): a nested case-control study	England	>18	288,356	143,912	145,444	6925	28,137	32,962	261,219	4.93	1.36–3.88	Body mass index, alcohol, tobacco, comorbidities, medications (metformin, insulin)
Ben Q, 2016 [16]	Risk factors for sporadic pancreatic neuroendocrine tumors: A case-control study	China	49.7	999	448	552	24	142	50	614	2.29	1.36–3.88	Tobacco, alcohol, family history of pancreatic cancer
Zheng Z, 2016 [18]	Risk Factors for Pancreatic Cancer in China: A Multicenter Case-Control Study	China	58.85	646	362	284	49	223	26	281	2.96	1.48–5.92	Obesity tobacco, family history of pancreatic cancer, tea and coffee consumption
Antwi SO, 2016 [15]	Pancreatic cancer: associations of inflammatory potential of diet, cigarette smoking and long-standing diabetes	USA	66.5	2573	1416	1157	211	817	151	756	3.27	2.58–4.17	Inflammatory diet, tobacco, duration of type II diabetes
Lu Y, 2015 [25]	New-onset type 2 diabetes, elevated HbA1c, anti-diabetic medications, and risk of pancreatic cancer	Sweden-England	20–75	5529	3166	2363	175	529	856	5000	2.36	1.94–2.89	Body mass index, alcohol, tobacco, fasting glucose, fasting insulin, allergies
Walker EJ, 2015 [27]	Metformin use among type 2 diabetics and risk of pancreatic cancer in a clinic-based case-control study	USA	61.5	1405	704	701	81	536	89	869	1.41	1.02–1.96	Body mass index, tobacco, alcohol, pancreatitis, family history of pancreatic cancer, diabetes duration, medication use (metformin, insulin…)
Petrusel L, 2020 [20]	Risk Factors in Pancreatic Adenocarcinoma: The Interrelation with Familial History and Predictive Role on Survival	Romania	63.4	624	376	248	115	279	36	312	4.1/new-onset: 11.78	2.58–31.42	Body mass index, alcohol, tobacco, coffee, chronic pancreatitis, family history of pancreatic cancer
Dayem Ullah AZM, 2021 [22]	Temporality of clinical factors associated with pancreatic cancer: a case-control study using linked electronic health records	England	55.1	8962	4569	4373	360	965	1045	4355	1.74/new-onset: 1.95	1.25–3.03	Comorbidities (gastrointestinal, cardiometabolic and respiratory), obesity, tobacco, alcohol
Farias AJ, 2020 [21]	Diabetes-related complications and pancreatic cancer incidence in the multiethnic cohort	USA	Tramo edad	2161	1025	1136	197	433	524	1728	1.22	1.09–1.36	Tobacco, alcohol, body mass index, comorbidities
Wang Y, 2016 [28]	Complex interplay between type 2 diabetes and pancreatic cancer: insights from observational and mendelian randomization analyses	China	67.34	1163	766	497	194	581	107	582	2.06	1.56–2.71	Duration of type II diabetes, obesity, tobacco, alcohol, medication (oral medication or insulin)

For the cohort studies, the summary HR was 2.39 (95% CI: 2.09–2.73) (Figure 3) [29,30,31,32,33,34,35,36,37]. With the sole exception of the study by Setiawan et al. [36] (RR 0.96, 95% CI: 0.78–1.18), all studies reported statistically significant estimates >2. Study heterogeneity overall was high, at 96% (p < 0.001).

Table 2. Characteristics of cohort studies.

Author and Year	Title	Country	Middle Ages	Sample	Men	Women	PCa Number	PCa Cases in Patients with Diabetes	Patients with Diabetes	PCa in Patients Without Diabetes	Patients’ Cohort Without Diabetes	HR	Confidence Interval	Other Parameters
Yuan C, 2020 [35]	Diabetes, weight change, and pancreatic cancer risk	USA	62.0	159,025	46,207	112,818	1116	135	196,382	735	3,970,434	2.97	2.31–3.82	Weight loss, body mass index, tobacco, alcohol, diabetes duration, race/ethnicity
Park BK, 2022 [30]	Lifestyle, body mass index, diabetes, and the risk of pancreatic cancer	South Korea	53.9	7,445,944	3,768,191	3,677,753	22,543	4434	802	18,109	6,644,308	1.48	1.43–1.53	Obesity, tobacco, alcohol, hyperlipidemia use, and physical activity
Lee DY, 2018 [34]	The influence of diabetes and antidiabetic medications on the risk of pancreatic cancer	Corea del Sur	>30	4,945,886	2,748,061	2,197,825	8589	2916	966,492	5673	3,979,394	2.2	2.12–2.32	Alcohol and comorbidities (chronic pancreatitis, acute pancreatitis, hepatitis B, hepatitis C), medication (oral or insulin)
Huang BZ, 2020 [38]	New-onset diabetes, longitudinal trends in metabolic markers, and risk of pancreatic cancer	USA	57.9	1,499,627	672,167	827,460	2002	759	1,675,673	937	506,034	3.17 (HR DM prevalent 1.85)	2.75–3.65 (1.67–2.05)	Body mass index, alcohol, family history of pancreatic cancer, pancreatitis
Er KC, 2016 [29]	Effect of glycemic control on the risk of pancreatic cancer	Taiwan	49.15	699,115	341,940	357,175	497	82	46,973	415	652,142	2.53	1.34–9.78	Alcohol, tobacco, obesity, chronic pancreatitis and other comorbidities
White MJ, 2023 [32]	The association of new-onset diabetes with subsequent pancreatic cancer	USA	51	2,561,684	1,275,720	1,285,964	1084	579	640,421	505	1,921,263	3.47	3.08–3.92	Tobacco, comorbidities
Setiawan VW, 2018 [36]	Pancreatic cancer following incident Diabetes in African Americans and Latinos	USA		48,995	21,483	27,512	408	128	15,833	280	33,162	2.39/new-onset Latino: 4.08/African American: 3.38	1.91–2.98/Latino: 2.76–6.03/African-American: 2.30–4.98)	Body mass index, tobacco, alcohol
Safadi H, 2024 [33]	Associations between diabetes and cancer: 10-year study	Hungary	61.4	3,681,774	1,678,889	1,719,388	17,328	892	93,852	16,436	3,796,764	2.294	2.099–2.507	Other cancers (breast, liver, colorectal, kidney…)
de Jong RGPJ, 2017 [37]	Impact of detection bias on the risk of gastrointestinal cancer in T2DM	Netherlands	63.9	34,038	17,343	16,695	34,038	53,804	34,038	34,038	34,038	4.1	1.09–1.36	Other cancers (breast, liver, colorectal, gastric)

Table 2. Characteristics of cohort studies.

Author and Year	Title	Country	Middle Ages	Sample	Men	Women	PCa Number	PCa Cases in Patients with Diabetes	Patients with Diabetes	PCa in Patients Without Diabetes	Patients’ Cohort Without Diabetes	HR	Confidence Interval	Other Parameters
Yuan C, 2020 [35]	Diabetes, weight change, and pancreatic cancer risk	USA	62.0	159,025	46,207	112,818	1116	135	196,382	735	3,970,434	2.97	2.31–3.82	Weight loss, body mass index, tobacco, alcohol, diabetes duration, race/ethnicity
Park BK, 2022 [30]	Lifestyle, body mass index, diabetes, and the risk of pancreatic cancer	South Korea	53.9	7,445,944	3,768,191	3,677,753	22,543	4434	802	18,109	6,644,308	1.48	1.43–1.53	Obesity, tobacco, alcohol, hyperlipidemia use, and physical activity
Lee DY, 2018 [34]	The influence of diabetes and antidiabetic medications on the risk of pancreatic cancer	Corea del Sur	>30	4,945,886	2,748,061	2,197,825	8589	2916	966,492	5673	3,979,394	2.2	2.12–2.32	Alcohol and comorbidities (chronic pancreatitis, acute pancreatitis, hepatitis B, hepatitis C), medication (oral or insulin)
Huang BZ, 2020 [38]	New-onset diabetes, longitudinal trends in metabolic markers, and risk of pancreatic cancer	USA	57.9	1,499,627	672,167	827,460	2002	759	1,675,673	937	506,034	3.17 (HR DM prevalent 1.85)	2.75–3.65 (1.67–2.05)	Body mass index, alcohol, family history of pancreatic cancer, pancreatitis
Er KC, 2016 [29]	Effect of glycemic control on the risk of pancreatic cancer	Taiwan	49.15	699,115	341,940	357,175	497	82	46,973	415	652,142	2.53	1.34–9.78	Alcohol, tobacco, obesity, chronic pancreatitis and other comorbidities
White MJ, 2023 [32]	The association of new-onset diabetes with subsequent pancreatic cancer	USA	51	2,561,684	1,275,720	1,285,964	1084	579	640,421	505	1,921,263	3.47	3.08–3.92	Tobacco, comorbidities
Setiawan VW, 2018 [36]	Pancreatic cancer following incident Diabetes in African Americans and Latinos	USA		48,995	21,483	27,512	408	128	15,833	280	33,162	2.39/new-onset Latino: 4.08/African American: 3.38	1.91–2.98/Latino: 2.76–6.03/African-American: 2.30–4.98)	Body mass index, tobacco, alcohol
Safadi H, 2024 [33]	Associations between diabetes and cancer: 10-year study	Hungary	61.4	3,681,774	1,678,889	1,719,388	17,328	892	93,852	16,436	3,796,764	2.294	2.099–2.507	Other cancers (breast, liver, colorectal, kidney…)
de Jong RGPJ, 2017 [37]	Impact of detection bias on the risk of gastrointestinal cancer in T2DM	Netherlands	63.9	34,038	17,343	16,695	34,038	53,804	34,038	34,038	34,038	4.1	1.09–1.36	Other cancers (breast, liver, colorectal, gastric)

Analysis by geographic region (Asia, USA, and Europe) yields very similar estimates across case-control and cohort studies (Supplementary Figures S1 and S2), and sensitivity analysis, removing one study at each stage, also reveals no significant changes in the ORs or HRs (Supplementary Figures S3 and S4).

Case-control studies looking at recent-onset diabetes (<2 years) reported a summary risk of 3.27 (95% CI 2.09–2.73) (Figure 4). This risk decreased with increasing time since previous diabetes diagnosis. For case-control studies with a previous diabetes diagnosis of 2–5 years, the summary risk was 2.25 (95% CI 1.64–3.08), for the 5–10 year period it was 1.55 (95% CI 1.50–1.61), and for the period of more than 10 years it was 1.12 (95% CI 0.77–1.63) (Figure 4).

Case and control studies. Cohort studies also showed an increased PCa risk in the first 2 years after diabetes diagnosis (RR 4.29; 95% CI: 2.17–8.47). The risk appeared to decrease over time, although note that the corresponding number of studies was relatively small (Figure 5).

Some studies reported noteworthy results, e.g., the OR for poorly controlled diabetes in the study by Er et al. [29] was 3.61 (95% CI: 1.34–9.78). Most studies not only focused on diabetes as a risk factor, but also included other variables that could influence cancer development, primarily tobacco use (OR 1.96, 95% CI: 1.24–3.09), [15,16,17,18,19,20,21,25,30,32,38] alcohol use (OR 1.94, 95% CI: 1.11–3.06) [16,19,20,23,25,38], obesity/body mass index [18], family history of PCa (OR 1.66, 95% CI: 1.11–2.47) [16,18,19,20], and unintentional weight loss prior to cancer diagnosis (HR 6.75, 95% CI: 4.55–10.00) as reported by Yuan et al. [35]. However, a significantly smaller number adjusted for important variables such as family history of pancreatic cancer, medication, and glycaemic control. Few studies reported the association between changes in HbA1c in diabetes and PCa. The studies by Lu et al. [25] and Tan et al. [24] showed that increasing HbA1c levels, which potentially indicate increased severity of diabetes, was associated with an increased risk of PCa compared with diabetic patients with small changes in HbA1c levels after diabetes diagnosis. This result may further corroborate the pivotal role of new-onset diabetes with typical hyperglycaemia in the development of PCa. However, the study by Walker et al. [27] does not support these results. Specific anti-diabetic medications carry different risks of PCa, with a particularly increased risk among insulin users, and gradually less increased risks among users of a combination of oral anti-diabetic medications, sulphonylurea, and metformin, although these associations warrant further research [25].

The NOS quality review resulted in scores ranging between 6 and 9 (fair and good quality), for an overall average of 7.82 (

Table 3. Quality Newcastle–Ottawa scale: case-control studies.

Author	Title	Year	Quality	Results
Van Tran T [23]	Risk factors of Pancreatic Cancer in Vietnam: A Matched Case-Control Hospital-Based Study	2021	Selection: 4 Comparability: 2 Exposure: 3	9 Good
Dayem Ullah AZM [22]	Temporality of clinical factors associated with pancreatic cancer: a case-control study using linked electronic health records	2021	Selection: 3 Comparability: 1 Exposure: 2	7 Good
Li X [26]	ABO Blood Group and Diabetes Mellitus Influence the Risk for Pancreatic Cancer in a Population from China	2018	Selection: 3 Comparability: 2 Exposure: 3	8 Good
Olson SH [17]	Weight Loss, Diabetes, Fatigue, and Depression Preceding Pancreatic Cancer	2016	Selection: 4 Comparability: 2 Exposure: 3	9 Good
Valente R [19]	Risk and protective factors for the occurrence of sporadic pancreatic endocrine neoplasms	2017	Selection: 3 Comparability: 1 Exposure: 3	7 Good
Tan PS [24]	Temporality of body mass index, blood tests, comorbidities and medication use as early markers for pancreatic ductal adenocarcinoma (PDAC): a nested case-control study	2023	Selection: 3 Comparability: 2 Exposure: 3	8 Good
Ben Q [16]	Risk Factors for Sporadic Pancreatic Neuroendocrine Tumors: A Case-Control Study	2016	Selection: 3 Comparability: 2 Exposure: 3	8 Good
Zheng Z [18]	Risk Factors for Pancreatic Cancer in China: A Multicenter Case-Control Study	2016	Selection: 4 Comparability: 2 Exposure: 3	9 Good
Antwi SO [15]	Pancreatic cancer: associations of inflammatory potential of diet, cigarette smoking and long-standing diabetes	2016	Selection: 3 Comparability: 1 Exposure: 2	6 Fair
Lu Y [25]	New-onset type 2 diabetes, elevated HbA1c, anti-diabetic medications, and risk of pancreatic cancer	2015	Selection: 4 Comparability: 2 Exposure: 3	9 Good
Walker EJ [19]	Metformin use among type 2 diabetics and risk of pancreatic cancer in a clinic-based case-control study	2015	Selection: 4 Comparability: 2 Exposure: 3	9 Good
Petrusel L [20]	Risk Factors in Pancreatic Adenocarcinoma: the Interrelation with Familial History and Predictive Role on Survival	2020	Selection: 3 Comparability: 1 Exposure: 2	6 Fair
Farias AJ [21]	Diabetes-related complications and pancreatic cancer incidence in the multiethenic cohort	2020	Selection: 4 Comparability: 2 Exposure: 3	9 Good
Wang Y [28]	Complex interplay between type 2 diabetes mellitus and pancreatic cancer: insights from observational and mendelian randomization analyses	2025	Selection: 3 Comparability: 2 Exposure: 3	8 Good

and

Table 4. Quality Newcastle-Ottawa scale: cohort studies.

Author	Title	Year	Quality	Results
Yuan C [35]	Diabetes, Weight Change, and Pancreatic Cancer Risk	2020	Selection: 2 Comparability: 2 Outcomes: 3	7 Good
Park BK [30]	Lifestyle, body mass index, diabetes, and the risk of pancreatic cancer in a nationwide population-based cohort study with 7.4 million Korean subjects	2022	Selection: 4 Comparability: 2 Outcomes: 3	9 Good
Huang BZ [38]	New-Onset Diabetes, Longitudinal Trends in Metabolic Markers, and Risk of Pancreatic Cancer in a Heterogeneous Population	2020	Selection: 3 Comparability: 1 Outcomes: 3	7 Good
Er KC [29]	Effect of glycemic control on the risk of pancreatic cancer: A nationwide cohort study	2016	Selection: 3 Comparability: 1 Outcomes: 3	7 Good
White MJ [32]	The association of new-onset diabetes with subsequent diagnosis of pancreatic cancer-novel use of a large administrative database	2023	Selection: 4 Comparability: 2 Outcomes: 3	9 Good
Lee DY [34]	The influence of diabetes and antidiabetic medications on the risk of pancreatic cancer: a nationwide population-based study in Korea	2018	Selection: 3 Comparability: 1 Outcomes: 3	7 Good
Setiawan VW [36]	Pancreatic cancer following Incident Diabetes in African Americans and Latinos: The multiethnic cohort	2018	Selection: 4 Comparability: 1 Outcomes: 3	8 Good
Safadi H [33]	Associations between diabetes and cancer: A 10-year national population-based retrospective cohort study	2024	Selection: 4 Comparability: 1 Outcomes: 2	7 Good
de Jong RGPJ [37]	Impact of detection bias on the risk of gastrointestinal cancer and its subsites in type 2 diabetes mellitus	2017	Selection: 4 Comparability: 1 Outcomes: 3	8 Good

).

On visual inspection, the meta-analysis funnel plots reflecting the association between diabetes and PCa are largely symmetrical, indicating a low likelihood of publication bias (Supplementary Figures S5 and S6, for case-control and cohort studies, respectively).

4. Discussion

Our meta-analysis of 23 studies conducted in different parts of the world and with different populations points to a significantly increased PCa risk in patients with diabetes: compared to non-diabetics, the overall risk in patients with diabetes was 2.30 for case-control studies and 2.39 for cohort studies.

Meta-analyses conducted a few years earlier reported similar but slightly lower PCa risk values than those observed in our study: Huxley et al. [11], 1.82; Batabyal et al. [10], 1.97; Song et al. [9], 1.64; and more recently, Pang et al. [12], 1.87. In contrast, later meta-analyses reported higher risk values: Zhang et al. [39], 3.69; Mellenthin et al. [40], 3.35; and Yang et al. [41], 2.69 (

Table 5. Comparison with previous meta-analyses of pancreatic cancer risk in diabetes.

Author [Reference]	Year	Cohort Studies	Case-Control Studies	Total Studies	RR/OR * Summary	95% CI *
Huxley [11]	2005	19	17	36	1.82	1.66–1.89
Batabyal [10]	2013	50	39	89	1.97	1.78–2.18
Song [9]	2015	21	23	44	1.64	1.52–1.78
Pang [12]	2017	22	0	22	1.87	1.48–2.37
Zhang [39]	2018	0	26	26	3.69	3.12–4.37
Mellenting [40]	2022	13	8	21	3.35	2.75–4.09
Yang [41]	2023	0	8	8	2.69	2.52–2.88

* RR, Risk ratio; OR odds ratio; CI, confidence interval.

). The evidence accumulated across all these meta-analyses would suggest that diabetes is a strong and consistent risk factor for PCa. The slow but steady increase in the incidence of PCa in the population suggests recent changes in the risk profile of patients [4]. An aging population, more comorbidities, and a massive increase in overweight may explain the increase in recent estimates. Also, the considerable heterogeneity among studies, for example, study design, setting, or comparator drugs, might explain the different results. Furthermore, reverse causality cannot be ruled out in almost all of these original studies.

A key issue emerging in some of the included studies, e.g., that by Lu et al. [25], is the relevance of the years elapsing from diabetes diagnosis to PCa diagnosis. Overall, PCa risk was much higher when diabetes was diagnosed <2 years prior to the PCa diagnosis, suggesting the possibility of a paraneoplastic manifestation of the tumour. Very high risk was observed in the meta-analyses by Batabyal et al. [10] (RR 6.69 for 1 year prior to PCa) and Zhang et al. [39] (RR 3.69 for <2 years prior to PCa). In both meta-analyses, this risk was reduced in later periods, reflecting an inverse relationship with years elapsed since diabetes diagnosis. Song et al. [9], excluding studies with diabetes diagnosis <2 years, also reported an inverse relationship with years since diabetes diagnosis, estimating a risk of 1.64 for ≥2 years, 1.58 for ≥5 years, and 1.50 for ≥10 years. Shen et al. [42] also underlined the relevance of the years since diabetes diagnosis for PCa development, and also of prediabetes, as a risk factor for PCa in patients with high basal glycaemia. Also, in the case-control study by Wang et al., [28] using Mendelian randomization, the causal relationship between diabetes and Pca was explored. This study supports the association between newly diagnosed type 2 diabetes mellitus (less than 2 years) and an increased risk of PCa, with insulin therapy, fibroblast growth factor (FGF4), and sulfhydryloxidase 2 mediating this pathway.

Our subgroup analysis also shows that, in both case-control and cohort studies, the PCa risk is likewise greater in the first 2 years and subsequently decreases in the other analysed periods (2–5 years, 5–10 years, and >10 years) (Figure 6).

These results would support the hypothesis that diabetes in the 2 years prior to PCa diagnosis could be a paraneoplastic manifestation and a consequence of PCa rather than a risk factor for PCa. The underlying mechanisms are unclear, but it is hypothesized that tumour-induced alterations in pancreatic islet cells may contribute to impaired glucose metabolism. By directly destroying insulin-producing β cells, PCa can disrupt normal pancreatic function [43]. Furthermore, systemic inflammation associated with PCa and tumour-derived factors, such as islet amyloid polypeptide [44], may induce insulin resistance, a further contributing to hyperglycaemia. This observation highlights the potential role of hyperglycaemia as an early warning sign of PCa. Following surgical resection of PCa, the dynamics of glycaemic control can vary significantly among patients. Some individuals may experience an improvement in glycaemic status due to the removal of tumour burden [45,46], suggesting a relevant role of tumour in diabetes. Nevertheless, the study by Wang et al., [28] using Mendelian randomization would support the hypothesis of diabetes as a risk factor of PCa in the early years of the diagnosis, so this is an open issue for future research.

For periods >2 years, diabetes remains a risk for PCa, although with moderate values that decrease with the passing of time from the diabetes diagnosis. The underlying mechanisms linking diabetes to PCa risk are not entirely clear but may involve several pathways. Hyperinsulinism and insulin resistance, both characteristic of diabetes, have been proposed as mediators in this association [47]. The fact that insulin, which has mitogenic effects, can promote cell proliferation and inhibit apoptosis may contribute to PCa development and progression [48]. Chronic inflammation, another feature of diabetes, has also been implicated in PCa development [49,50]. Likewise, alterations in insulin-like growth factor signalling pathways may also play a role in PCa development [51,52]. In particular, studies have shown that IGF receptors are frequently overexpressed in cancer, with corresponding changes in the circulating levels of IGF peptides. The IGF system and insulin signalling pathway simultaneously play important roles in hyperinsulinemia, insulin resistance and tumour pathogenesis, which are more likely to be the potential mechanisms in a variety of human cancers [51,52].

HbA1c reflects average blood sugar levels over a period of weeks/months. Some studies have demonstrated that improving HbA1c for people with types 1 or 2 diabetes cuts the risk of microvascular complications. Therefore, deterioration of HbA1c may increase the risk of diabetes complications, including cancer. Although a limited number of existing studies examined the role of elevated HbA1c on pancreatic cancer, the increased HbA1c levels indicate poorly controlled diabetes, which, reasonably, contributes to the metabolic derangement that may lead to pancreatic cancer development [25].

The role of anti-diabetic medication in PCa etiology has been proposed to either increase the risk of cancer (insulin) or ameliorate the risk (metformin), but results are inconsistent. Also, the meta-analyses on metformin and PCa showed inconsistent results [25] and the considerable heterogeneity between studies might explain the different results.

Our study has some important implications. The fact that the prognosis for PCa is poor [3] and that little progress has been made in the last 40 years in relation to improving survival would point to two key issues. First, reducing risk to reduce incidence is crucial, given the survival rate once PCa is diagnosed. Risk can be reduced by lifestyle modifications—primarily dietary changes, increased physical activity, and weight loss—aimed at preventing or delaying diabetes progression [53]. Second, patients who could benefit from clinical monitoring need to be identified, e.g., patients with newly diagnosed diabetes who have other risk factors should be candidates for close monitoring for early tumour detection in the 2 years following their diagnosis—as proposed for patients with new-onset diabetes [10,54], and also for patients with obesity [55], most especially if the obesity is central, as determined in the meta-analysis by Moore et al. [56]. We do not recommend routine screening, as no biomarker or imaging test has proven cost-effective to date. However, a risk scoring system to identify diabetic patients at higher risk would be interesting. Factors such as abdominal obesity, poor glycaemic control, a family history of pancreatic cancer, or unintentional weight loss indicate an increased risk.

The strengths of our meta-analysis include the large number of patients in the included studies, which provided sufficient statistical power to detect an association between diabetes and PCa risk; the fact that the validity of the results is confirmed by subgroup analysis of times/periods of PCa diagnosis in patients with diabetes; and finally, the random effects modelling that took heterogeneity between studies into account.

The main limitations are summarized as follows: firstly, patient selection bias was possible since the studies were observational; secondly, distinguishing between short-term and long-term diabetes was not possible, as there is no consensus regarding unified criteria to define the time elapsing from diabetes diagnosis to PCa diagnosis; third, the fact that some studies do not distinguish diabetes type I and type II could imply an underestimation of risk, although the potential bias generated by diabetes type 1 is considered minimal given that the age of the patients, the use of oral anti-diabetic agents, and their risk profile suggest that the majority of patients have type 2 diabetes. Fourth, case-control studies use the date PCa diagnosis to select cases and their respective controls. In cases with a history of diabetes diagnosis of less than two years, there is a possibility that diabetes would be a consequence of PCa, rather than a risk factor. This is more likely if there is a delay in the diagnosis of PCa and if the patient is diagnosed at a more advanced stage. This possibility could generate reverse causality bias, which could explain part of the elevated risk of PCa in the first two years after diabetes diagnosis. However, the study by Wang et al. [28] using Mendelian randomization supports the association between newly diagnosed type 2 diabetes mellitus (less than 2 years) and an increased risk of Pca. Fifth, analysis stratified by time since diabetes diagnosis (<2, 2–5, 5–10, and >10 years) reveals this to be the main source of heterogeneity, although the variability of the adjustment factors used across studies does not rule out these factors also being part of the observed heterogeneity. Sixth, some studies may exhibit survival bias, which could underestimate the risk of diabetes in patients with a longer history of diagnosis. However, the different studies took into account the age of the participants in the logistic regression models, which could partially offset this effect. Finally, as the causality of pancreatic cancer is largely unclear, residual confounding by unknown risk factors may result in chance errors.

5. Conclusions

We find that diabetes is a risk factor for PCa, and that the PCa risk is much higher in the 2 years immediately following diabetes diagnosis. Identifying the subgroup of patients corresponding to this period of 2 years for clinical monitoring and/or cancer screening could improve outcomes. However, until definitive evidence is available on the usefulness of screening, and bearing in mind the poor survival rate for PCa, interventions focused on preventing diabetes onset and delaying diabetes progression via modifiable risk factors could reduce PCa incidence.