Exocrine Pancreatic Enzymes Are a Serological Biomarker for Type 1 Diabetes Staging and Pancreas Size

Abstract

Exocrine pancreas abnormalities are increasingly recognized as features of type 1 diabetes. We previously reported reduced serum trypsinogen levels and in a separate study, smaller pancreata at and before disease onset. We hypothesized that three pancreas enzymes (amylase, lipase, and trypsinogen) might serve as serological biomarkers of pancreas volume and risk for type 1 diabetes. Amylase, lipase, and trypsinogen were measured from two independent cohorts, together comprising 800 serum samples from single-autoantibody-positive (1AAb+) and multiple-AAb+ (≥2AAb+) subjects, individuals with recent-onset or established type 1 diabetes, their AAb-negative (AAb-) first-degree relatives, and AAb- control subjects. Lipase and trypsinogen were significantly reduced in ≥2AAb+, recent-onset, and established type 1 diabetes subjects versus control subjects and 1AAb+, while amylase was reduced only in established type 1 diabetes. Logistic regression models demonstrated trypsinogen plus lipase (area under the receiver operating characteristic curve [AUROC] = 81.4%) performed equivalently to all three enzymes (AUROC = 81.4%) in categorizing ≥2AAb+ versus 1AAb+ subjects. For cohort 2 (n = 246), linear regression demonstrated lipase and trypsinogen levels could individually and collectively serve as indicators of BMI-normalized relative pancreas volume (RPVBMI, P < 0.001), previously measured by MRI. Serum lipase and trypsinogen levels together provide the most sensitive serological biomarker of RPVBMI and may improve disease staging in pretype 1 diabetes.

Trial registration: ClinicalTrials.gov NCT02234947.

© 2021 by the American Diabetes Association.

Figures

Figure 2

Exocrine pancreatic enzyme levels pre- and posttype 1 diabetes (T1D) onset. Serum trypsinogen (A), lipase (B), and amylase (C) levels in controls, AAb− FDR, subjects with 1AAb+, subjects with ≥2AAb+, subjects with recent-onset T1D (T1D Recent), and subjects with established T1D (T1D Established). Log-transformed data were converted via back transformation for graphical presentation such that individual dots are unadjusted raw data. Black lines are median (95% CI) which enable simultaneous visualization of data adjusted for age, sex, race, and CDC-standardized BMI-for-age percentile; one-way ANCOVA. Data are shown in the scatter plots for all subjects, including those where BMI data were missing, which excluded them from analysis by ANCOVA. *P < 0.05; **P < 0.01; ***P < 0.001. Dotted horizontal lines show the clinical reference range. Subjects who had serum levels of trypsinogen (D), lipase (E), or amylase (F) below, within, or above the clinical reference range are shown. Data were analyzed by χ2 test (trypsinogen, P < 0.001; lipase, P < 0.001; and amylase, P < 0.02). G: The first two PCs assigned to each subject are displayed. Component 1 increases with larger values of the three exocrine pancreatic enzymes (amylase, lipase, and trypsinogen). Component 2 increases with larger values of amylase alone. Light blue circles: AAb− control subjects, AAb− FDR, and 1AAb+ subjects; red circles: ≥2AAb+ subjects and patients with recent-onset or established T1D. The P value is <0.001 for the mean difference in component 1 for low risk vs. stage 1–3 T1D is <0.001. The P value is 0.052 for the mean difference in component 2 for low risk vs. stage 1–3 T1D.

Figure 1

Exocrine pancreatic enzyme levels vs. age. Partial effect plots show the relationship (adjusted R2 and P value) between age at blood draw and serum levels of trypsinogen (A), lipase (B), or amylase (C) in islet AAb− subjects, as well as the relationship (adjusted R2 and P value) between age at diagnosis and serum levels of trypsinogen (D), lipase (E), or amylase (F) in individuals with new-onset type 1 diabetes, while accounting for sex, BMI, race, and cohort. Trypsinogen (R2 = 0.31, P < 0.001) and lipase (R2 = 0.11, P < 0.001) were significantly correlated with age at draw, while amylase was not (R2 = −0.002, P > 0.05). Trypsinogen (R2 = 0.06, P < 0.01), lipase (R2 = −0.003, P > 0.05), and amylase (R2 = −0.003, P > 0.05) were not significantly correlated with age at type 1 diabetes onset. Log-transformed data are shown as scatter plots plus trend lines with 95% CIs.

Figure 3

Receiver operating characteristic (ROC) curve showing sensitivity and specificity for classifying subjects according to type 1 diabetes (T1D) risk or status. Logistic regression models categorizing subjects as low risk (AAb− and 1AAb+) vs. stage 1–3 T1D (≥2AAb+, recent-onset, and established T1D) are shown for no enzymes, amylase, lipase, trypsinogen, or all three enzymes as covariates (A), or for no enzymes, trypsinogen + amylase, lipase + trypsinogen, lipase + amylase, or all three enzymes (as indicated on the graph) (B) with age, sex, BMI percentile, and cohort 1 or 2 in our study included as additional covariates. Logistic regression models categorizing subjects as 1AAb+ vs. ≥2AAb+ are shown for no enzymes, amylase, lipase, trypsinogen, lipase + trypsinogen, or all three enzymes as covariates (as indicated on the graph), with age, sex, BMI percentile, and cohort 1 or 2 in our study included as additional covariates (C). A repeated k-fold cross-validation was performed (k = 10), and the AUROC was calculated for each model (as indicated on the graph).

Figure 4

Partial effect plots for each single-enzyme linear regression model. The relationship with RPVBMI is shown for amylase (A), lipase (B), and trypsinogen (C) after controlling for the other covariates in the models (age, sex, race, and levels of the other two enzymes). Amylase had the smallest effect (adjusted R2 = 0.12, P < 0.05) compared with lipase (adjusted R2 = 0.18, P < 0.001) and trypsinogen (adjusted R2 = 0.23, P < 0.001).