Abstract
Background
Chronic diseases may injure the pancreas and result in a reduction in pancreas volume (PV), which may adversely impact its critical role in endocrine and exocrine function. Less is known about whether individuals with smaller PV face higher risks of adverse clinical outcomes.
Methods
We estimated baseline age‐adjusted predicted PV (pPV) from abdominal magnetic resonance imaging in 36 592 UK Biobank participants. Since the initial magnetic resonance imaging occurred 9.0 (interquartile range, 7.5–10.1) years after baseline, we fit a linear regression equation that included age for the outcome of PV slope in 451 participants with repeat magnetic resonance imaging 2.8 (interquartile range, 2.2–4.8) years after the initial scan. Using these data, we calculated baseline age‐adjusted pPV slope by adding or subtracting the annual change to the initial PV measurement assuming linear change in PV over time. Multivariable‐adjusted proportional hazards models tested associations of pPV with incident chronic kidney disease, cardiovascular disease (CVD; myocardial infarction, heart failure, or stroke), and death.
Results
Participants (52.2% women) were 55±8 years old and had a pPV of 63.3±15.3 mL. During median follow‐up of 13.9 years, there were 757 incident chronic kidney disease, 1392 CVD, and 702 death events. In final adjusted models, participants in the first pPV quartile had a 1.67‐fold (95% CI, 1.34–2.07) and 1.28‐fold (95% CI, 1.10–1.49) increased risk of incident CVD and death compared with the fourth quartile, respectively. pPV had nonlinear associations with incident CVD and death (each P<0.001). pPV was not associated with incident chronic kidney disease.
Conclusions
Lower pPV is associated with increased risk of incident CVD and death but not incident chronic kidney disease. Future studies should investigate the pathophysiologic mechanisms driving these associations.
Keywords: cardiovascular disease, death, magnetic resonance imaging, pancreas volume
Subject Categories: Clinical Studies, Magnetic Resonance Imaging (MRI), Prognosis, Imaging, Cardiovascular Disease
Nonstandard Abbreviations and Acronyms
- pPV
predicted pancreas volume
- PV
pancreas volume
Research Perspective.
What Is New?
Lower pancreas volume, as assessed by magnetic resonance imaging, is associated with an increased risk of developing adverse cardiovascular events and death but not development of chronic kidney disease.
What Question Should Be Addressed Next?
Future studies should determine the mechanisms that drive these findings and whether interventional studies are able to prevent or stabilize decline in pancreas volume.
The pancreas plays a critical role in digestion and hormonal regulation of blood sugar levels. The pancreas primarily consists of exocrine tissue (>95%), which produces digestive enzymes to break down carbohydrates, proteins, and fats. 1 Despite constituting <5% of pancreatic mass, endocrine tissue produces enzymes vital for regulation of blood glucose levels and glandular secretion. Various diseases may cause damage to the pancreas resulting in atrophy. The injury occurs in the setting of acute or chronic diseases and may result in inflammation and subsequent development of fibrosis and atrophy of the pancreas. 2 , 3 , 4 In the most extreme cases, pancreatic insufficiency may develop after recurrent acute or chronic pancreatitis, long‐standing diabetes, autoimmune diseases, or pancreatectomy. 5 More subtle reductions in pancreas volume (PV) are observed in patients with type 1 and type 2 diabetes compared with patients without diabetes. 6 , 7 Since the pancreas is involved in the pathogenesis of several cardio–kidney–metabolic diseases, improved understanding of the risks associated with smaller PV may identify individuals at high risk for adverse clinical outcomes to enroll in trials for testing current or new therapies.
Cross‐sectional studies have documented differences in PV by diabetes status and body mass index. Fewer studies have been able to identify risk factors for smaller PV in cross‐section. 6 , 7 , 8 , 9 , 10 Similarly, a recent study demonstrated that lower PV may be causally linked with a higher risk of developing type 2 diabetes, 10 but there are limited data investigating the association of PV with adverse clinical outcomes. Moreover, inadequate sample sizes or follow‐up time limit the ability of other studies to study the association of PV with subsequent adverse clinical outcomes. To address this, we evaluated the associations of PV, as assessed by magnetic resonance imaging (MRI), with subsequent adverse clinical outcomes in participants from the UK Biobank.
METHODS
Source Population
The UK Biobank is a large, population‐based prospective cohort study, established to investigate the genetic and nongenetic determinants of disease in middle‐aged and older adults. The UK Biobank recruited >500 000 participants aged 40 to 69 years from England, Scotland, and Wales between 2006 and 2010. 11 , 12 The UK Biobank imaging substudy is an ongoing investigation to scan the brains, hearts, bones, and abdomens of 100 000 of the UK Biobank participants. 13 UK Biobank has approval from the North West Multi‐Centre Research Ethics Committee and obtained written informed consent from all participants before the study. All data and materials have been made publicly available by the UK Biobank and can be accessed at https://www.ukbiobank.ac.uk/use‐our‐data/apply‐for‐access. Data were extracted under access application 88 543.
Study Population
Of the UK Biobank participants, 37 601 participants had PV measurements available. These participants were further divided into separate analytic cohorts for each outcome of all‐cause death, incident cardiovascular disease (CVD), and incident chronic kidney disease (CKD) (Figure S1). Of these participants who had PV measurements, we excluded 237 participants due to the inability to calculate PV, 26 with operative procedures on their pancreas, and 746 outliers in the top and bottom 1% of PV, which resulted in 36 592 participants for the all‐cause death outcome (analytic cohort 1). For the outcome of incident CVD (analytic cohort 2), we excluded 722 participants with preexisting CVD, which resulted in 35 870 participants for this analysis. For the outcome of incident CKD (analytic cohort 3), we excluded 4538 participants who had prevalent CKD or were missing data for estimated glomerular filtration rate or urine albumin‐to‐creatinine ratio at baseline, which resulted in 32 054 participants for this analysis.
Abdominal MRI Protocol and Analysis
UK Biobank participants were recruited by email or letter from the National Health Service. Participants were scanned at 4 different imaging centers in Great Britain with a Siemens Aera 1.5T scanner (Syngo MRI D13; Siemens, Erlangen, Germany), which used a dual‐echo protocol that acquired overlapping images in 6 stations covering the body from the neck to knee within about 6 minutes. 13 , 14 PV was calculated from high‐resolution T1‐weighted 3‐dimensional acquisition sequence under a single expiration breath hold. Bias‐field correction was performed to reduce signal inhomogeneities in the normalized volume, which did not require additional preprocessing and has high reproducibility. The methods for calculation of PV have been previously described. All manual annotations were performed blind to any participant information or outcomes. 10 , 13 , 15
Exposure
The primary exposure was age‐adjusted predicted PV (pPV) at baseline. Since the initial abdominal MRI visit occurred 9.0 (interquartile range [IQR], 7.5–10.1) years after the baseline visit, we calculated the PV slope between abdominal MRI visits in 451 participants who returned for a repeat abdominal MRI. We back‐calculated baseline age‐adjusted pPV using a linear back‐extrapolation that accounted for a participant’s age using linear regression (Data S1).
Outcomes
In the prospective analyses, the primary outcomes were all‐cause death, incident CVD, and incident CKD. Incident CVD was defined as incident myocardial infarction, heart failure, or stroke. Secondary outcomes were the individual components of incident CVD. Incident outcomes were identified using International Classification of Diseases (ICD) codes in any primary care, hospital admission records, and death register records or office of Population Censuses and Surveys Classification of Interventions and Procedures version 4 codes in hospital inpatient data (Table S1). Outcomes were assessed from the baseline visit to the date of the outcome of interest.
Measurements and Assessment of Baseline Covariates
Data obtained at the baseline visit included sociodemographic characteristics, medical history, lifestyle behaviors, medications, standardized blood pressure measurements, anthropometric measures, and laboratory measurements. Ethnicity was self‐reported. Comorbid conditions were defined using self‐reported medical conditions or ICD, Tenth Revision (ICD‐10) codes in primary care and hospital inpatient data. Blood and urine samples were collected and analyzed at the central laboratory. Serum and urine creatinine were measured using an isotope dilution mass spectrometry–traceable method. Urine albumin was calculated using an immunoturbidometric method. Estimated glomerular filtration rate was calculated using the race‐free creatinine‐ and cystatin C–based Chronic Kidney Disease Epidemiology Collaboration 2021 equation. Urine albumin‐to‐creatinine ratio was calculated as spot urine albumin concentration divided by spot urine creatinine concentration from urine samples. If the urine albumin concentration was lower than the limit of the assay, the value was replaced by half of the minimum observed value and then divided by the urine creatinine concentration.
Statistical Analysis
To calculate baseline age‐adjusted pPV, we first calculated the rate of change (slope) in PV by subtracting the PV at MRI visit 1 from the PV at MRI visit 2 and dividing this result by the time elapsed between the 2 MRI visits. We fit a linear regression equation that included age as the predictor variable and PV slope as the outcome. To calculate baseline age‐adjusted pPV, we subtracted the predicted PV slope multiplied by the time elapsed between the baseline visit and MRI visit 1 from the PV at MRI visit 1 (Data S1). Continuous variables were expressed as mean±SD or median (IQR) and categorical variables were presented as counts with percentages. For skewed data distributions, we performed logarithmic transformation as appropriate. To assess differences in frequency distributions of categorical variables, normally distributed continuous variables, and nonnormally distributed continuous variables by quartiles of pPV, we used χ2, ANOVA, and Kruskal–Wallis tests, respectively. We evaluated clinical predictors of pPV using unadjusted and multivariable adjusted linear regression models.
We performed time‐to‐event analyses to examine the risks for the outcomes, evaluating pPV as a continuous variable (per doubling) and as quartiles. We used cause‐specific Cox proportional hazards regression to investigate the multivariable‐adjusted associations between pPV and adverse clinical outcomes. Subdistribution hazard competing risk models were not used because the goal was to understand the biological associations of pPV with each outcome. Our multivariable adjustment strategy was hierarchical and based on biological and clinical plausibility of baseline covariates as potential confounders of the association between pPV and outcomes. Model 1 was unadjusted; model 2 adjusted for age, sex, ethnicity, smoking status, alcohol use, multiple indices of deprivation, body mass index (BMI), waist‐to‐hip ratio, and body surface area. Model 3 included model 2 and further adjusted for diabetes status; history of cancer, heart failure, myocardial infarction, and stroke; use of antihypertensive medications; systolic blood pressure; total cholesterol; serum albumin; hemoglobin A1c; CRP (C‐reactive protein); estimated glomerular filtration rate; and urine albumin‐to‐creatinine ratio. We compared the Akaike information criterion in models that used age‐adjusted pPV as a linear, quadratic, and cubic term Akaike information criterion, which suggested that a linear term led to the best model fit (Table S2). We examined the possibly nonlinear relationship between pPV and outcomes with restricted cubic splines. Tests for nonlinearity used the likelihood ratio test, comparing the model with only the linear term to the model with the linear and cubic spline terms. 16 To evaluate potential prognostic thresholds of baseline age‐adjusted pPV with adverse clinical outcomes, we systematically evaluated the adjusted hazard ratios (HRs) across the full range of pPV from the multivariable adjusted Cox proportional hazards model using cubic splines. We defined the threshold as the lowest value at which the lower bound of the 95% CI for the HR remained above 1.0 for at least 5 consecutive spline points. This consecutive‐point criterion was implemented to reduce the risk of selecting unstable or spurious thresholds due to local fluctuations. This approach aligns with the principles described by Mazumdar et al, 17 which recommended identifying cut points within the multivariable model through systematic scanning of values to isolate prognostically distinct regions of elevated risk. We tested for statistical interaction between age (<55 versus ≥55 years), sex (male versus female), and diabetes status (yes versus no) in Cox models through multiplicative interaction terms. Since there were 28.5% missing covariate data, we used a multiple regression procedure with SAS PROC MI (SAS Institute, Cary, NC) to perform multiple imputation in the primary analysis. We generated 5 imputed data sets and imputed values for missing data on the basis of the observed data with the assumption that the data were missing at random. We combined the test results across the imputed data sets using the rules of Rubin with the SAS PROC MI ANALYZE procedure. 18 In sensitivity analyses, we performed the primary analysis using pPV data without multiple imputation of covariates. We also repeated the primary analyses by using the initial MRI visit PV data at baseline with multiple imputation of the missing covariate data. We confirmed no violations of the proportional hazards assumption using the Kolmogorov‐type supremum test and visual inspection by checking martingale residuals. Statistical analyses were performed using SAS software version 9.4 (SAS Institute). All statistical tests were 2‐sided, and P<0.05 was considered significant.
RESULTS
Study Population Characteristics
Of the 36 605 participants who had PV measurements, 451 participants underwent a repeat abdominal MRI with PV quantification. The median between the initial MRI scan and the repeat MRI scan was 2.8 (IQR, 2.2–4.8) years. We included age in the equation to predict PV slope (age β=0.07970, P=0.007). The mean PV slope was −0.42±0.60 mL/y. We multiplied the pPV slope by the time elapsed from baseline to MRI visit 1 and subtracted this value from the PV calculated at MRI visit 1 to calculate the baseline age‐adjusted pPV. Figure S2 shows histograms of PV at each of the MRI visits and the baseline age‐adjusted pPV. Table 1 shows baseline characteristics for the overall cohort and by quartiles of baseline age‐adjusted pPV. The mean pPV was 63.3±15.3 mL. The mean±SD age was 54.9±7.5 years; 52.2% of the participants were women; 2.6% of the participants had diabetes; mean BMI was 26.5±4.1 kg/m2; mean systolic blood pressure was 137±19 mm Hg; mean hemoglobin A1c was 5.4±0.5%; mean estimated glomerular filtration rate was 97.8± 12.8 mL/min per 1.73 m2; and median urine albumin‐to‐creatinine ratio was 5.7 (IQR, 3.5–9.3) mg/g. The baseline characteristics of the overall cohort compared with the subset who returned for a repeat MRI (n=451) were similar (Table S3).
Table 1.
Baseline Characteristics Across Quartiles of Age‐Adjusted pPV
| Baseline characteristics | All participants (n=36 592) | Quartile 1 (n=9148) (6.0–53.2 mL) | Quartile 2 (n=9148) (53.2–63.0) | Quartile 3 (n=9148) (63.0–3.5) | Quartile 4 (n=9148) (73.5–116.5) | P value |
|---|---|---|---|---|---|---|
| Age‐adjusted pPV, mL | 63.3±15.3 | 44.0±8.16 | 58.3±2.8 | 68.0±3.0 | 82.8±7.4 | <0.001 |
| Age, y | 54.9±7.5 | 53.7±7.6 | 54.75±7.4 | 55.3±7.4 | 56.0±7.3 | <0.001 |
| Female sex | 19 087 (52.2) | 6309 (69.0) | 5379 (58.8) | 4380 (47.9) | 3019 (33.0) | <0.001 |
| Race and ethnicity | ||||||
| White | 35 445 (96.9) | 8829 (96.5) | 8868 (96.9) | 8898 (97.3) | 8850 (96.7) | 0.02 |
| Black or Black British | 222 (0.6) | 75 (0.8) | 51 (0.6) | 40 (0.4) | 56 (0.6) | 0.009 |
| Asian or Asian British | 373 (1.0) | 95 (1.0) | 89 (1.0) | 94 (1.0) | 95 (1.0) | 0.97 |
| Chinese | 112 (0.3) | 30 (0.3) | 36 (0.4) | 26 (0.3) | 20 (0.2) | 0.18 |
| Mixed | 162 (0.4) | 53 (0.6) | 30 (0.3) | 33 (0.4) | 46 (0.5) | 0.03 |
| Other ethnic groups | 175 (0.5) | 46 (0.5) | 44 (0.5) | 36 (0.4) | 49 (0.5) | 0.54 |
| Missing | 103 (0.3) | 20 (0.2) | 30 (0.3) | 21 (0.2) | 32 (0.3) | |
| Smoking | ||||||
| Never | 22 138 (60.5) | 5843 (63.9) | 5621 (61.4) | 5509 (60.2) | 5165 (56.5) | <0.001 |
| Previous | 12 088 (33.0) | 2723 (29.8) | 2902 (31.7) | 3056 (33.4) | 3407 (37.2) | <0.001 |
| Current | 2279 (6.2) | 568 (6.2) | 598 (6.5) | 558 (6.1) | 555 (6.1) | 0.53 |
| Missing | 87 (0.2) | 14 (0.2) | 27 (0.3) | 25 (0.3) | 21 (0.2) | |
| Alcohol | ||||||
| Never | 932 (2.5) | 299 (3.3) | 273 (3.0) | 194 (2.1) | 166 (1.8) | <0.001 |
| Previous | 792 (2.2) | 230 (2.5) | 195 (2.1) | 178 (1.9) | 189 (2.1) | 0.05 |
| Current | 34 845 (95.2) | 8613 (94.2) | 8669 (94.8) | 8774 (95.9) | 8789 (96.1) | <0.001 |
| Missing | 23 (0.1) | 6 (0.1) | 11 (0.1) | 2 (0.0) | 4 (0.0) | |
| Diabetes | 944 (2.6) | 333 (3.6) | 204 (2.2) | 200 (2.2) | 207 (2.3) | <0.001 |
| Missing | 62 (0.2) | 17 (0.2) | 14 (0.2) | 17 (0.2) | 14 (0.2) | |
| Cancer | 2010 (5.5) | 550 (6.0) | 514 (5.6) | 495 (5.4) | 451 (4.9) | 0.01 |
| Missing | 89 (0.2) | 22 (0.2) | 37 (0.4) | 12 (0.1) | 18 (0.2) | |
| Prevalent CKD | 1327 (3.6) | 350 (3.8) | 321 (3.5) | 308 (3.4) | 348 (3.8) | 0.23 |
| Missing | 3211 (8.8) | 832 (9.1) | 757 (8.3) | 810 (8.9) | 812 (8.9) | |
| Blood pressure medications | 1292 (3.5) | 427 (4.7) | 377 (4.1) | 277 (3.0) | 211 (2.3) | <0.001 |
| Prevalent heart failure | 58 (0.2) | 16 (0.2) | 11 (0.1) | 13 (0.1) | 18 (0.2) | 0.57 |
| Prevalent myocardial infarction | 452 (1.2) | 91 (1.0) | 94 (1.0) | 123 (1.3) | 144 (1.6) | <0.001 |
| Prevalent stroke | 254 (0.7) | 67 (0.7) | 57 (0.6) | 62 (0.7) | 68 (0.7) | 0.75 |
| BMI, kg/m2 | 26.5±4.1 | 26.1±4.4 | 26.3±4.1 | 26.6±4.0 | 27.1±3.9 | <0.001 |
| Missing | 53 | 16 | 11 | 11 | 15 | |
| Body surface area, m2 | 1.9±0.2 | 1.8±0.2 | 1.8±0.2 | 1.9±0.2 | 2.0±0.2 | <0.001 |
| Missing | 53 | 16 | 11 | 11 | 15 | |
| Deprivation index | 11.3 (6.7–19.9) | 11.8 (6.9–20.9) | 11.5 (6.7–19.9) | 11.1 (6.6–19.5) | 11.0 (6.4–19.1) | <0.001 |
| Missing | 822 | 217 | 192 | 206 | 207 | |
| Waist‐to‐hip ratio | 0.9±0.4 | 0.8±0.1 | 0.9±0.1 | 0.9±0.1 | 0.9±0.1 | <0.001 |
| Missing | 33 | 10 | 8 | 5 | 10 | |
| Cholesterol, mg/dL | 221.4±41.9 | 220.1±42.0 | 223.6±41.9 | 221.9±42.0 | 220±41.8 | <0.001 |
| Missing | 2396 | 598 | 561 | 607 | 630 | |
| Systolic blood pressure, mm Hg | 136.8±18.8 | 134.5±18.9 | 135.9±18.7 | 137.7±18.8 | 139.3±18.5 | <0.001 |
| Missing | 2477 | 609 | 589 | 603 | 676 | |
| CRP, mg/L | 2.0±3.4 | 2.1±3.7 | 2.0±3.1 | 2.0±3.4 | 2.0±3.4 | 0.01 |
| Missing | 2462 | 608 | 580 | 626 | 648 | |
| Albumin, g/dL | 4.5±0.3 | 4.5±0.3 | 4.5±0.3 | 4.5±0.3 | 4.6±0.3 | <0.001 |
| Missing | 5372 | 1422 | 1289 | 1319 | 1342 | |
| Hemoglobin A1c, % | 5.4±0.5 | 5.4±0.6 | 5.4±0.4 | 5.3±0.4 | 5.4±0.4 | <0.001 |
| Missing | 2647 | 710 | 639 | 660 | 638 | |
| eGFR, mL/min per 1.73 m2 | 97.8±12.8 | 98.3±13.1 | 98.1±12.6 | 97.7±12.7 | 97.2±12.6 | <0.001 |
| Missing | 2438 | 608 | 571 | 617 | 642 | |
| UACR, mg/g | 5.7 (3.5–9.3) | 6.2 (3.9–10.1) | 5.8 (3.7–9.3) | 5.6 (3.5–9.2) | 5.1 (3.2–8.4) | <0.001 |
| Missing | 1187 | 309 | 283 | 299 | 296 | |
Data presented as mean±SD or median (interquartile range) for continuous variables and count (percentage) for categorical variables. P values represent differences across age‐adjusted pPV quartiles. BMI indicates body mass index; CRP, C‐reactive protein; eGFR, estimated glomerular filtration rate; pPV, predicted pancreas volume; and UACR, urine albumin‐to‐creatinine ratio.
Factors Associated With pPV
Factors associated with pPV in unadjusted and adjusted linear regression models are shown in Table 2. Of the demographics and socioeconomic factors, baseline age‐adjusted pPV values were 4.51 mL smaller in women versus men (95% CI, −5.07 to −3.96), 1.41 mL larger in previous smokers compared with never smokers (95% CI, 1.08–1.73), 1.28 mL larger in current alcohol drinkers versus never alcohol drinkers (95% CI, 0.31–2.25), and 0.56 mL smaller per SD increase in deprivation index (95% CI, −0.71 to −0.40). Of the comorbid conditions, baseline age‐adjusted pPV was 2.90 mL smaller in individuals with diabetes compared with individuals without diabetes (95% CI, −4.02 to −1.78) and 0.65 mL larger per SD increase in systolic blood pressure (95% CI, 0.470.83). Of the anthropometric measures, baseline age‐adjusted pPV was 2.35 mL larger per SD increase in body surface area (95% CI, 2.07–2.62), 0.46 mL larger per SD increase in waist‐to‐hip ratio (95% CI, 0.22–0.71), and 0.30 mL smaller per SD increase in BMI (95% CI, −0.54 to −0.07). The laboratory variables most strongly associated with baseline age‐adjusted pPV were higher hemoglobin A1c (0.67 mL smaller per SD increase [95% CI, −0.86 to −0.49]), higher total cholesterol (0.39 mL larger per SD increase [95% CI, 0.22–0.55]), and higher CRP (0.32 mL smaller per SD increase [95% CI, −0.48 to −0.16]).
Table 2.
Factors Associated With Baseline Age‐Adjusted pPV
| Univariable analysis | Multivariable analysis | |||
|---|---|---|---|---|
| β (95% CI) | P value | β (95% CI) | P value | |
| Female sex | −8.33 (−8.63 to −8.03) | <0.001 | −4.51 (−5.07 to −3.96) | <0.001 |
| Asian or Asian British | −0.10 (−1.66 to 1.47) | 0.90 | 1.43 (−0.09 to 2.95) | 0.06 |
| Black or Black British | −2.87 (−4.89 to −0.85) | 0.005 | −1.50 (−3.44 to 0.44) | 0.13 |
| Chinese | −2.22 (−5.06 to 0.63) | 0.13 | 0.76 (−1.96 to 3.48) | 0.58 |
| Mixed | −0.23 (−2.60 to 2.13) | 0.85 | 1.58 (−0.67 to 3.82) | 0.17 |
| Other ethnic groups | −0.51 (−2.83 to 1.81) | 0.67 | 0.91 (−1.26 to 3.09) | 0.41 |
| Previous smokers | 2.12 (1.78 to 2.46) | <0.001 | 1.41 (1.08 to 1.73) | <0.001 |
| Current smokers | 0.56 (−0.10 to 1.22) | 0.09 | −0.11 (−0.75 to 0.52) | 0.72 |
| Previous alcohol use | 1.59 (0.14 to 3.04) | 0.03 | 0.02 (−1.38 to 1.41) | 0.98 |
| Current alcohol use | 3.44 (2.44 to 4.43) | <0.001 | 1.28 (0.31 to 2.25) | 0.01 |
| BMI, per 1 SD kg/m2 | 1.44 (1.28 to 1.59) | <0.001 | −0.30 (−0.54 to −0.07) | 0.01 |
| Deprivation index, per 1 SD unit | −0.61 (−0.77 to −0.45) | <0.001 | −0.56 (−0.71 to −0.40) | <0.001 |
| Waist‐to‐hip ratio, per 1 SD | 3.34 (3.19 to 3.49) | <0.001 | 0.46 (0.22 to 0.71) | <0.001 |
| Hemoglobin A1c, per 1 SD, % | −0.45 (−0.61 to −0.29) | <0.001 | −0.67 (−0.86 to −0.49) | <0.001 |
| CRP, per 1 SD, mg/L | −0.21 (−0.37 to −0.05) | 0.01 | −0.32 (−0.48 to −0.16) | <0.001 |
| Albumin, per 1 SD, g/dL | 0.52 (0.35 to 0.69) | <0.001 | −0.02 (−0.18 to 0.15) | 0.84 |
| Cholesterol, per 1 SD, mg/dL | 0.01 (−0.15 to 0.17) | 0.92 | 0.39 (0.22 to 0.55) | <0.001 |
| eGFR, per 1 SD, mL/min per 1.73m2 | −0.50 (−0.66 to −0.34) | <0.001 | 0.02 (−0.14 to 0.18) | 0.77 |
| Natural log UACR, per 1 SD, mg/g | −1.05 (−1.21 to −0.89) | <0.001 | 0.02 (−0.14 to 0.18) | 0.83 |
| Systolic blood pressure, per 1 SD, mm Hg | 1.53 (1.36 to 1.69) | <0.001 | 0.65 (0.47 to 0.83) | <0.001 |
| Blood pressure medications | −3.97 (−4.82 to −3.12) | <0.001 | −0.66 (−1.50 to 0.18) | 0.12 |
| Body surface area, per 1 SD, m2 | 3.89 (3.74 to 4.04) | <0.001 | 2.35 (2.07 to 2.62) | <0.001 |
| Diabetes | −2.95 (−3.94 to −1.96) | <0.001 | −2.90 (−4.02 to −1.78) | <0.001 |
| Heart failure | 1.32 (−2.63 to 5.26) | 0.51 | −0.78 (−4.62 to 3.05) | 0.69 |
| Myocardial infarction | 2.88 (1.46 to 4.30) | <0.001 | 0.68 (−0.72 to 2.08) | 0.34 |
| Stroke | 0.08 (−1.81 to 1.97) | 0.93 | −0.33 (−2.14 to 1.48) | 0.72 |
| Cancer | −1.17 (−1.86 to −0.48) | <0.001 | −0.42 (−1.08 to 0.24) | 0.21 |
White is the reference group for race; never smoker is the reference group for smoking status; never alcohol use is the reference group for alcohol drinking status. BMI indicates body mass index; CRP, C‐reactive protein; eGFR, estimated glomerular filtration rate; pPV, predicted pancreas volume; and UACR, urine albumin‐to‐creatinine ratio.
Association of pPV With Death
During a median follow‐up time of 13.9 (IQR, 13.1–14.5) years, 702 participants died. Table 3 shows the unadjusted and multivariable‐adjusted HRs with 95% CIs according to baseline age‐adjusted pPV as a continuous variable and by quartiles. Compared with participants in the highest quartile, participants were at 1.67‐, 1.28‐, and 1.25‐fold increased risks of death in quartiles 1, 2, and 3, respectively, in multivariable models adjusted for demographics, socioeconomic factors, anthropometric measures, comorbid conditions, and laboratory variables. Restricted cubic spline analysis revealed a nonlinear association between pPV and death (P<0.001; Figure 1A), and the spline had a statistically significant association with death in the final adjusted model (P<0.001). We identified a threshold effect at a baseline age‐adjusted pPV level of 37.6 mL. Participants who had baseline age‐adjusted pPV values <37.6 mL were at 1.67‐fold increased risk of death compared with participants with values ≥37.6 mL (Table S4). We found no evidence of statistical interaction by age (P=0.94), sex (P=0.53), or diabetes status (P=0.63) (Table S5).
Table 3.
Association of pPV With Death, Incident CVD, and Incident CKD
| Continuous,* HR (95% CI) | Quartile 1, HR (95% CI) (6.0–53.2 mL) | Quartile 2, HR (95% CI) (53.2–63.0 mL) | Quartile 3, HR (95% CI) (63.0–73.5 mL) | Quartile 4, HR (95% CI) (73.5–116.5 mL) | |
|---|---|---|---|---|---|
| Death | |||||
| N | 36 592 | 9148 | 9148 | 9148 | 9148 |
| Events | 702 | 190 | 164 | 179 | 169 |
| Model 1 | 1.05 (0.98–1.13) | 1.14 (0.92–1.40) | 0.98 (0.79–1.22) | 1.07 (0.87–1.32) | Reference |
| Model 2 | 1.25 (1.16–1.35) | 1.78 (1.43–2.20) | 1.31 (1.06–1.63) | 1.27 (1.03–1.57) | Reference |
| Model 3 | 1.22 (1.16–1.35) | 1.67 (1.34–2.07) | 1.28 (1.03–1.60) | 1.25 (1.01–1.54) | Reference |
| Incident CVD | |||||
| N | 35 870 | 8967 | 8968 | 8968 | 8967 |
| Events | 1392 | 330 | 302 | 355 | 405 |
| Model 1 | 0.92 (0.87–0.97) | 0.81 (0.70–0.94) | 0.74 (0.64–0.86) | 0.88 (0.76–1.01) | Reference |
| Model 2 | 1.12 (1.06–1.18) | 1.35 (1.16–1.57) | 1.04 (0.90–1.21) | 1.07 (0.93–1.23) | Reference |
| Model 3 | 1.10 (1.04–1.16) | 1.28 (1.10–1.49) | 1.02 (0.88–1.19) | 1.06 (0.92–1.22) | Reference |
| Incident CKD | |||||
| N | 32 054 | 8013 | 8014 | 8014 | 8013 |
| Events | 757 | 207 | 180 | 187 | 183 |
| Model 1 | 1.05 (0.98–1.13) | 1.14 (0.93–1.39) | 0.99 (0.80–1.21) | 1.03 (0.84–1.26) | Reference |
| Model 2 | 1.17 (1.08–1.26) | 1.50 (1.22–1.85) | 1.20 (0.97–1.47) | 1.14 (0.93–1.40) | Reference |
| Model 3 | 1.08 (0.99–1.16) | 1.21 (0.98–1.49) | 1.11 (0.90–1.37) | 1.06 (0.86–1.30) | Reference |
Model 1: Unadjusted. Model 2: adjusted for age, sex, ethnicity, smoking status, alcohol use, multiple indices of deprivation, body mass index, waist‐to‐hip ratio, and body surface area. Model 3: model 2 and further adjusted for diabetes status; history of cancer, heart failure, myocardial infarction, and stroke; use of antihypertensive medications; systolic blood pressure; total cholesterol; hemoglobin A1c; serum albumin; C‐reactive protein; estimated glomerular filtration rate; and natural log transformed urine albumin‐to‐creatinine ratio. CKD indicates chronic kidney disease; CVD, cardiovascular disease; HR, hazard ratio; and pPV, predicted pancreas volume.
Hazard ratios are per 1 SD decrease in exposure of interest.
Figure . Restricted cubic spline analyses for the association of age‐adjusted pPV with death, incident CVD, and incident CKD.
Restricted cubic spline model reflects fully adjusted model for covariates described in model 3 of Table 3. Baseline age‐adjusted pPV had a nonlinear association with (A) death (P<0.001), (B) incident CVD (P<0.001), and (C) incident CKD (P<0.001). Thresholds of prognostic significance are noted in red for death and incident CVD. Mean baseline age‐adjusted pPV (63 mL) is the reference. Knots are placed at the 25th, 50th, and 75th percentiles of age‐adjusted pPV. CKD indicates chronic kidney disease; CVD, cardiovascular disease; and pPV, predicted pancreas volume.
Association of pPV With Incident CVD
During a median follow‐up time of 13.8 (IQR, 13.1–14.5) years, 1392 participants experienced an incident CVD event. In the final adjusted model, participants in the lowest quartile had a significantly increased risk of incident CVD (HR, 1.28 [95% CI, 1.10–1.49]; Table 3). Restricted cubic spline analysis demonstrated a nonlinear association between baseline age‐adjusted pPV and incident CVD (P<0.001; Figure 1B), and the spline had a statistically significant associated with incident CVD (P<0.001). We identified a threshold effect at a baseline age‐adjusted pPV value of 34.8 mL. Participants who had baseline age‐adjusted pPV values <34.9 mL were at 1.59‐fold increased risk of death compared with participants with values ≥34.9 mL (Table S4). We found no evidence of statistical interaction by age (P=0.05), sex (P=0.08), or diabetes status (P=0.12) for the association between baseline age‐adjusted pPV and incident CVD (Table S6). Table S7 shows the association of baseline age‐adjusted pPV with each component of the composite CVD outcome. In the final adjusted model, participants in the lowest quartile of pPV had a 1.73‐fold increased risk of incident heart failure compared with participants in the highest quartile. Participants with smaller pPV were at nominally higher risks of incident myocardial infarction, but this did not reach statistical significance after multivariable adjustment. There was no statistically significant association between pPV and incident stroke.
Association of pPV With Incident CKD
During a median follow‐up time of 13.9 (IQR, 13.1–14.5) years, 757 participants developed CKD. After multivariable adjustment, participants in the lower quartiles of pPV were at nominally higher risks of incident CKD, but these associations did not reach statistical significance (Table 3). Baseline age‐adjusted pPV had a nonlinear association with incident CKD (P<0.001; Figure 1C), but the spline did not have a statistically significant association with incident CKD (P=0.41). We found no evidence of statistical interaction by age (P=0.48), sex (P=0.09), or diabetes status (P=0.67) for the association between pPV and incident CKD (Table S8).
Sensitivity Analyses
We repeated the primary analysis as a complete case analysis since we used multiple imputation for missing covariate data. The associations of pPV with all‐cause death, incident CVD, and incident CKD were qualitatively unchanged (Table S9). Since we estimated PV at baseline age‐adjusted pPV assuming a linear change over time, we repeated the primary analysis by placing the initial MRI visit PV measurement at baseline, which did not qualitatively change the results (Table S10).
DISCUSSION
In this prospective cohort study of >36 000 middle‐aged and older adults, we found that smaller abdominal MRI‐derived pPV was independently associated with higher risks of all‐cause death and experiencing a CVD event. We also identified several demographic, socioeconomic, anthropometric, comorbid, and laboratory variables that were significantly associated with pPV in cross‐section. Despite the importance of these variables in explaining an individual’s pPV, our prospective findings remained robust after multivariable adjustments. Taken together, our findings suggest that PV may play a critical role in the development of adverse clinical outcomes, and future studies should determine the mechanisms that drive these findings and whether interventional studies are able to prevent or stabilize the decline in PV.
Smaller PV may occur due to alterations in organ development or as a by‐product of parenchymal loss in the endocrine and exocrine‐related areas of the pancreas. While the mechanisms that lead to smaller PV before chronic disease onset are incompletely understood, diseases that injure cells responsible for the endocrine and exocrine functions likely lead to inflammation and cellular apoptosis with subsequent loss of function over time. 2 , 3 , 4 Consistent with this hypothesis, histological analyses demonstrated that patients with type 2 diabetes may have reductions in β‐cell mass by 30% to 60% compared with patients without diabetes. 19 , 20 These changes may occur due to excessive β‐cell workload that results in damage through oxidative stress and inflammation that results in apoptosis and subsequent loss of β‐cell mass over time. While obesity may increase PV, 21 β cells may still experience greater workload due to the need of insulin secretion to maintain normoglycemia even in the absence of hyperglycemia, 22 and this increased workload may eventually lead to β‐cell damage over time. A number of factors may influence exocrine function in the pancreas, but there may also be an interplay between the endocrine and exocrine pancreas. 23 , 24 Both obesity and diabetes may contribute to acinar cell injury by inducing inflammation and subsequent fibrosis in the exocrine pancreas. 4 Similarly, individuals who develop chronic pancreatitis or cancer may have atrophy and fibrosis of the exocrine pancreas, and these individuals may have reductions in their β cells with some patients eventually developing hyperglycemia with or without overt diabetes. 23 , 25 Since the exocrine pancreas is critical for its role in regulation of digestion, clinical or subclinical impairments in digestion due to loss of exocrine tissue could potentially lead to adverse clinical outcomes. Prior studies have demonstrated an intricate link between the gut microbiome, exocrine function, and pancreas size. 26 Population‐based studies showed that measures of pancreatic exocrine function may alter intestinal microbiota composition and its affiliated plasma metabolites. 27 Additional observational data found metabolites produced by gut microbiota were associated with inflammation and adverse CVD events, 28 , 29 , 30 which suggests that clinical or subclinical impairments in digestion could lead to adverse clinical outcomes. Although not all individuals with smaller PV have overt phenotypes of endocrine or exocrine insufficiency, we speculate that individuals may still have functional deficiencies that impair homeostatic mechanisms and drive systemic inflammation that ultimately contribute to adverse clinical outcomes. Additional research is required to identify mechanisms that drive smaller PV in individuals with and without chronic disease that leads to adverse clinical outcomes.
PV increases until young adulthood, remains stable throughout adulthood, and begins to decline in older adults. 7 , 31 While a prior study across the life span demonstrated that men have larger PV than women, 31 these data conflicted with a prior meta‐analysis that did not find differences by sex. 7 Prior cross‐sectional studies demonstrated that patients with diabetes had smaller PV compared with patients without diabetes. 6 , 7 , 9 Conversely, higher BMI was associated with larger PV. 31 , 32 Prior literature also found inconsistent results for the associations of smoking status and alcohol use with PV. 7 , 33 Consistent with prior studies, female sex and history of diabetes were predictors of smaller baseline age‐adjusted pPV in this study. Similar to prior work, our unadjusted analysis showed that individuals with higher BMI had larger pPV, but our multivariable‐adjusted analysis showed that higher BMI was associated with smaller pPV. In these analyses, higher body surface area remained significantly associated with larger pPV after multivariable adjustment, which included other anthropometric measures such as waist‐to‐hip ratio. Our results also found that prior smoking status and current alcohol use were each associated with larger pPV after multivariable adjustment. These results conflict with prior studies that found inconsistent or weak associations of smoking status and alcohol use with PV. Since both smoking and alcohol use may be associated with inflammation and development of cardiometabolic disease, additional investigation is needed to elucidate the mechanisms surrounding the intersection of smoking, alcohol use, metabolic syndrome, and PV. Given the potential hypothesis that inflammation may serve a role in smaller PV, we found that individuals with higher levels of CRP had smaller pPV. We further add that a higher systolic blood pressure, worse socioeconomic status, and higher hemoglobin A1c remained associated with smaller pPV after multivariable adjustment. Collectively, our results add to the prior literature and provide additional predictors of abdominal MRI‐derived PV in a large cohort of individuals.
Prior literature that evaluated the prognostic value of PV is mostly limited to patients with pancreatitis or pancreatic cancer, 34 , 35 with fewer studies evaluating the future risk of cardio–kidney–metabolic diseases. A prior analysis in >30 000 UK Biobank participants found that smaller PV may have a causal role in type 2 diabetes, which supports the hypothesis that smaller PV may precede adverse clinical outcomes. Our study leveraged repeated measurements of abdominal MRI‐derived PV in the subset of participants who returned for abdominal MRI to demonstrate that PV declined minimally over time. These results are similar to an analysis over the life span that suggested PV remains relatively stable during adulthood, consistent with the age range of UK Biobank participants included in this study. 31 Our prospective results found that smaller baseline age‐adjusted pPV are associated with higher risks of death and incident CVD. While the risk for incident CKD was nominally higher in individuals with a smaller pPV, these results did not reach statistical significance. Although there were many incident CKD events, the relatively smaller sample size of the analytic cohort for our CKD analyses compared with the analytic cohorts for all‐cause death and CVD may have accounted for this nonsignificant finding for the risk of incident CKD. A recent small study of patients with type 1 diabetes that underwent repeated abdominal MRI‐derived PV identified that PV declined over time in patients with type 1 diabetes, and smaller PV may identify individuals at risk of progression to later stages of type 1 diabetes and pancreatic diseases. While this study found that individuals who progress had >10% decline in PV, it is not clear what threshold may be clinically important. 36 Our threshold analysis demonstrated that baseline age‐adjusted pPV values <37.6 mL and <34.8 mL may be inflection points for death and incident CVD, respectively. While these values are ≈2 SD lower than the mean value in the cohort, this may be a conservative estimate since this is where the lower 95% CI bound was consistently >1.0. It is possible that risk begins at higher values, which requires further investigation in future studies. Importantly, our finding of an association of smaller pPV with higher risks of incident CVD and death were independent of multiple relevant clinical risk factors, including comorbid conditions and anthropometric measures, which are known factors that affect PV. Interestingly, a number of UK Biobank participants did not have cardiovascular, kidney, or metabolic diseases at baseline, and smaller PV still remains a risk factor for future development of adverse clinical outcomes, which requires further investigation into the mechanisms that lead to smaller PV in the absence of known chronic diseases or drive smaller PV over time.
Strengths of this study include leveraging the UK Biobank, which includes a large, well‐characterized cohort of individuals who underwent abdominal MRI‐derived PV with detailed covariate data to permit comprehensive multivariable adjustment. Our study has several limitations that warrant consideration. Despite the large cohort of individuals included in this study, there is limited racial and ethnic diversity, as >95% of the study cohort were White individuals, and our findings require replication. Our study cohort had adequate follow‐up time to record sufficient events for adverse clinical outcomes, but event ascertainment relied on administrative codes for incident CVD and CKD. We do not have histopathologic evidence of endocrine or exocrine pancreas volume to draw conclusions on the relevant areas of the pancreas, as assessed by MRI. We do not have other functional measures of exocrine or endocrine function that may help assess subclinical dysfunction in our multivariable models, which should be considered in follow‐up cohort studies. Although we adjusted for smoking status in our multivariable adjusted models, we did not capture other measures of smoking that may provide information about pancreas atrophy. 37 We estimated PV at baseline by using PV measured at 2 time points assuming a linear change in PV over time, but our analyses remained robust in our sensitivity analysis that used the raw PV value. Although we used multiple imputation to account for missingness of covariate data, our results remained similar in a complete case sensitivity analysis. Due to the observational design, there may be residual confounding from unmeasured covariates.
In summary, we found that lower baseline age‐adjusted pPV is associated with increased risk of death and incident CVD events. These findings were independent of a number of clinical variables associated with pPV, some of which have not been previously reported in the literature. These findings build upon existing yet limited literature describing the prognostic value of smaller PV, and additional investigations should determine whether extraction of features from PV images through agnostic data characterization algorithms, such as radiomics, 38 provide new insights into an individual’s risk of adverse clinical outcomes. Future studies should also identify the mechanisms that lead to smaller PV and whether existing or emerging therapies may reduce the risk of declining PV over time.
Sources of Funding
This work was supported by institutional funds at the University of Illinois Chicago (A.S.) and Boston Medical Center (S.S.W.). A.S. is supported by National Institutes of Health grants R01DK139321, R01HL180499, K23DK120811, and U01AI163081; Kidney Precision Medicine Project Opportunity Pool grant under award U2CDK114886; and the American Society of Nephrology Carl W. Gottschalk Research Scholar Award. I.M.S. is supported by National Institutes of Health grant K01DK136973, the American Society of Nephrology Carl W. Gottschalk Research Scholar Award, and the Boston University Department Career Investment Award. P.B. is supported by National Institutes of Health grants R01DK129211, R01DK132399, R01HL165433, R01DK137844, R01DK138915, and U01DK142249; Breakthrough T1D (3‐SRA‐2022‐1097‐M‐B, 3‐SRA‐2022‐1243‐M‐B, and 3‐SRA‐2022‐1230‐M‐B), University of Washington Medicine Diabetes Institute, and Seattle Children’s Research Institute.
Disclosures
A.S. reports personal fees from Amgen/Horizon Therapeutics PLC, CVS Caremark, Novo Nordisk, AstraZeneca, Bayer AG, and FNIH. P.B. reports serving or having served as a consultant for AstraZeneca, Bayer, Bristol‐Myers Squibb, Boehringer Ingelheim, Eli Lilly, LG Chemistry, Sanofi, Novo Nordisk, and Amgen/Horizon Therapeutics PLC. P.B. also serves or has served on the advisory boards or steering committees of AstraZeneca, Bayer, Boehringer Ingelheim, Novo Nordisk, and XORTX. P.B. receives cooperative grant funding from Novo Nordisk and the Renal Pre‐Competitive Consortium. The remaining authors have no disclosures to report.
Supporting information
Data S1
Tables S1–S10
Figures S1–S2
Acknowledgments
The research has been conducted using the UK Biobank Resource under application 88543. P.T. and A.S. were responsible for the concept and design of the study. P.T., C.M., and A.S. were responsible for statistical analyses. All authors interpreted the data. Each author contributed important intellectual content during manuscript drafting or revision and agrees to be personally accountable for the individual’s own contributions and to ensure that questions pertaining to the accuracy or integrity of any portion of the work, even one in which the author was not directly involved, are appropriately investigated and resolved, including documentation in the literature if appropriate.
This work was presented at the American Diabetes Association Scientific Sessions, June 20–23, 2025, in Chicago, IL.
This manuscript was sent to Yen‐Hung Lin, MD, PhD, Associate Editor, for review by expert referees, editorial decision, and final disposition.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.125.042348
For Sources of Funding and Disclosures, see page 10.
References
- 1. Karpinska M, Czauderna M. Pancreas‐its functions, disorders, and physiological impact on the mammals’ organism. Front Physiol. 2022;13:807632. doi: 10.3389/fphys.2022.807632 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Homo‐Delarche F, Calderari S, Irminger J‐C, Gangnerau M‐N, Coulaud J, Rickenbach K, Dolz M, Halban P, Portha B, Serradas P. Islet inflammation and fibrosis in a spontaneous model of type 2 diabetes, the GK rat. Diabetes. 2006;55:1625–1633. doi: 10.2337/db05-1526 [DOI] [PubMed] [Google Scholar]
- 3. Eguchi K, Manabe I. Macrophages and islet inflammation in type 2 diabetes. Diabetes Obes Metab. 2013;15(Suppl 3):152–158. doi: 10.1111/dom.12168 [DOI] [PubMed] [Google Scholar]
- 4. Matsuda A, Makino N, Tozawa T, Shirahata N, Honda T, Ikeda Y, Sato H, Ito M, Kakizaki Y, Akamatsu M, et al. Pancreatic fat accumulation, fibrosis, and acinar cell injury in the Zucker diabetic fatty rat fed a chronic high‐fat diet. Pancreas. 2014;43:735–743. doi: 10.1097/mpa.0000000000000129 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Ewald N, Hardt PD. Diagnosis and treatment of diabetes mellitus in chronic pancreatitis. World J Gastroenterol. 2013;19:7276–7281. doi: 10.3748/wjg.v19.i42.7276 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Garcia TS, Rech TH, Leitao CB. Pancreatic size and fat content in diabetes: a systematic review and meta‐analysis of imaging studies. PLoS One. 2017;12:e0180911. doi: 10.1371/journal.pone.0180911 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. DeSouza SV, Singh RG, Yoon HD, Murphy R, Plank LD, Petrov MS. Pancreas volume in health and disease: a systematic review and meta‐analysis. Expert Rev Gastroenterol Hepatol. 2018;12:757–766. doi: 10.1080/17474124.2018.1496015 [DOI] [PubMed] [Google Scholar]
- 8. Löhr M, Klöppel G. Residual insulin positivity and pancreatic atrophy in relation to duration of chronic type 1 (insulin‐dependent) diabetes mellitus and microangiopathy. Diabetologia. 1987;30:757–762. doi: 10.1007/bf00275740 [DOI] [PubMed] [Google Scholar]
- 9. Macauley M, Percival K, Thelwall PE, Hollingsworth KG, Taylor R. Altered volume, morphology and composition of the pancreas in type 2 diabetes. PLoS One. 2015;10:e0126825. doi: 10.1371/journal.pone.0126825 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Martin S, Sorokin EP, Thomas EL, Sattar N, Cule M, Bell JD, Yaghootkar H. Estimating the effect of liver and pancreas volume and fat content on risk of diabetes: a Mendelian randomization study. Diabetes Care. 2022;45:460–468. doi: 10.2337/dc21-1262 [DOI] [PubMed] [Google Scholar]
- 11. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, Downey P, Elliott P, Green J, Landray M, et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12:e1001779. doi: 10.1371/journal.pmed.1001779 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Fry A, Littlejohns TJ, Sudlow C, Doherty N, Adamska L, Sprosen T, Collins R, Allen NE. Comparison of sociodemographic and health‐related characteristics of UK biobank participants with those of the general population. Am J Epidemiol. 2017;186:1026–1034. doi: 10.1093/aje/kwx246 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Littlejohns TJ, Holliday J, Gibson LM, Garratt S, Oesingmann N, Alfaro‐Almagro F, Bell JD, Boultwood C, Collins R, Conroy MC, et al. The UK biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions. Nat Commun. 2020;11:2624. doi: 10.1038/s41467-020-15948-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Langner T, Ostling A, Maldonis L, Karlsson A, Olmo D, Lindgren D, Wallin A, Lundin L, Strand R, Ahlstrom H, et al. Kidney segmentation in neck‐to‐knee body MRI of 40,000 UK biobank participants. Sci Rep. 2020;10:20963. doi: 10.1038/s41598-020-77981-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Liu Y, Basty N, Whitcher B, Bell JD, Sorokin EP, van Bruggen N, Thomas EL, Cule M. Genetic architecture of 11 organ traits derived from abdominal MRI using deep learning. elife. 2021;10:e65554. doi: 10.7554/eLife.65554 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Govindarajulu U, Malloy E, Ganguli B, Spiegelman D, Eisen E. The comparison of alternative smoothing methods for fitting non‐linear exposure‐response relationships with cox models in a simulation study. Int J Biostat. 2009;5:1557–4679. doi: 10.2202/1557-4679.1104 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Mazumdar M, Smith A, Bacik J. Methods for categorizing a prognostic variable in a multivariable setting. Stat Med. 2003;22:559–571. doi: 10.1002/sim.1333 [DOI] [PubMed] [Google Scholar]
- 18. RJA L, Rubin DB. Statistical Analysis with Missing Data. 3rd ed. Wiley; 2019. [Google Scholar]
- 19. Butler AE, Janson J, Bonner‐Weir S, Ritzel R, Rizza RA, Butler PC. β‐Cell deficit and increased β‐cell apoptosis in humans with type 2 diabetes. Diabetes. 2003;52:102–110. doi: 10.2337/diabetes.52.1.102 [DOI] [PubMed] [Google Scholar]
- 20. Sakuraba H, Mizukami H, Yagihashi N, Wada R, Hanyu C, Yagihashi S. Reduced beta‐cell mass and expression of oxidative stress‐related DNA damage in the islet of Japanese type II diabetic patients. Diabetologia. 2002;45:85–96. doi: 10.1007/s125-002-8248-z [DOI] [PubMed] [Google Scholar]
- 21. Saisho Y, Butler AE, Manesso E, Elashoff D, Rizza RA, Butler PC. Beta‐cell mass and turnover in humans: effects of obesity and aging. Diabetes Care. 2013;36:111–117. doi: 10.2337/dc12-0421 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Polonsky KS, Given BD, Van Cauter E. Twenty‐four‐hour profiles and pulsatile patterns of insulin secretion in normal and obese subjects. J Clin Invest. 1988;81:442–448. doi: 10.1172/jci113339 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Chen N, Unnikrishnan IR, Anjana RM, Mohan V, Pitchumoni CS. The complex exocrine–endocrine relationship and secondary diabetes in exocrine pancreatic disorders. J Clin Gastroenterol. 2011;45:850–861. doi: 10.1097/MCG.0b013e31822a2ae5 [DOI] [PubMed] [Google Scholar]
- 24. Saisho Y. Pancreas volume and fat deposition in diabetes and Normal physiology: consideration of the interplay between endocrine and exocrine pancreas. Rev Diabet Stud. 2016;13:132–147. doi: 10.1900/RDS.2016.13.132 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Schrader H, Menge BA, Schneider S, Belyaev O, Tannapfel A, Uhl W, Schmidt WE, Meier JJ. Reduced pancreatic volume and beta‐cell area in patients with chronic pancreatitis. Gastroenterology. 2009;136:513–522. doi: 10.1053/j.gastro.2008.10.083 [DOI] [PubMed] [Google Scholar]
- 26. Girdhar K, Soto M, Huang Q, Orliaguet L, Cederquist C, Sundaresh B, Hu J, Figura M, Raisingani A, Canfora EE, et al. Gut microbiota regulate pancreatic growth, exocrine function, and gut hormones. Diabetes. 2022;71:945–960. doi: 10.2337/db21-0382 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Pietzner M, Budde K, Ruhlemann M, Volzke H, Homuth G, Weiss FU, Lerch MM, Frost F. Exocrine pancreatic function modulates plasma metabolites through changes in gut microbiota composition. J Clin Endocrinol Metab. 2021;106:e2290–e2298. doi: 10.1210/clinem/dgaa961 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Lee Y, Nemet I, Wang Z, Lai HTM, de Oliveira Otto MC, Lemaitre RN, Fretts AM, Sotoodehnia N, Budoff M, DiDonato JA, et al. Longitudinal plasma measures of trimethylamine N‐oxide and risk of atherosclerotic cardiovascular disease events in community‐based older adults. J Am Heart Assoc. 2021;10:e020646. doi: 10.1161/JAHA.120.020646 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Kurilshikov A, van den Munckhof ICL, Chen L, Bonder MJ, Schraa K, Rutten JHW, Riksen NP, de Graaf J, Oosting M, Sanna S, et al. Gut microbial associations to plasma metabolites linked to cardiovascular phenotypes and risk. Circ Res. 2019;124:1808–1820. doi: 10.1161/CIRCRESAHA.118.314642 [DOI] [PubMed] [Google Scholar]
- 30. Tang WHW, Hazen SL. Unraveling the complex relationship between gut microbiome and cardiovascular diseases. Circulation. 2024;149:1543–1545. doi: 10.1161/CIRCULATIONAHA.123.067547 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Saisho Y, Butler AE, Meier JJ, Monchamp T, Allen‐Auerbach M, Rizza RA, Butler PC. Pancreas volumes in humans from birth to age one hundred taking into account sex, obesity, and presence of type‐2 diabetes. Clin Anat. 2007;20:933–942. doi: 10.1002/ca.20543 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Caglar V, Songur A, Yagmurca M, Acar M, Toktas M, Gonul Y. Age‐related volumetric changes in pancreas: a stereological study on computed tomography. Surg Radiol Anat. 2012;34:935–941. doi: 10.1007/s00276-012-0988-x [DOI] [PubMed] [Google Scholar]
- 33. Stuart CEA, Singh RG, Alarcon Ramos GC, Priya S, Ko J, DeSouza SV, Cho J, Petrov MS. Relationship of pancreas volume to tobacco smoking and alcohol consumption following pancreatitis. Pancreatology. 2020;20:60–67. doi: 10.1016/j.pan.2019.10.009 [DOI] [PubMed] [Google Scholar]
- 34. Hokenek UD, Aydiner O, Kart JS, Arslan G, Saracoglu KT. Evaluation of the effect of pancreatic volume on mortality in patients with acute pancreatitis. Am J Emerg Med. 2023;63:38–43. doi: 10.1016/j.ajem.2022.10.032 [DOI] [PubMed] [Google Scholar]
- 35. Fukumoto T, Watanabe T, Hirai I, Kimura W. Pancreatic volume is one of the independent prognostic factors for resectable pancreatic ductal adenocarcinomas. J Hepatobiliary Pancreat Sci. 2016;23:472–479. doi: 10.1002/jhbp.365 [DOI] [PubMed] [Google Scholar]
- 36. Virostko J, Wright JJ, Williams JM, Hilmes MA, Triolo TM, Broncucia H, Du L, Kang H, Nallaparaju S, Valencia LG, et al. Longitudinal assessment of pancreas volume by MRI predicts progression to stage 3 type 1 diabetes. Diabetes Care. 2024;47:393–400. doi: 10.2337/dc23-1681 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Leffondré K, Abrahamowicz M, Xiao Y, Siemiatycki J. Modelling smoking history using a comprehensive smoking index: application to lung cancer. Stat Med. 2006;25:4132–4146. doi: 10.1002/sim.2680 [DOI] [PubMed] [Google Scholar]
- 38. Lafata KJ, Wang Y, Konkel B, Yin FF, Bashir MR. Radiomics: a primer on high‐throughput image phenotyping. Abdom Radiol (NY). 2022;47:2986–3002. doi: 10.1007/s00261-021-03254-x [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1
Tables S1–S10
Figures S1–S2