Quantitative β-Cell Mass Imaging Redefines Disease Staging and Glycemic Control in Type 1 Diabetes

Abstract

Noninvasive measurement of pancreatic β-cell mass remains an important unmet need in type 1 diabetes because conventional surrogate markers, such as C-peptide, often lack sensitivity in advanced disease. This study evaluated the glucagon-like peptide 1 receptor–targeted positron emission tomography tracer, ¹⁸F-labeled exendin-4–based probe conjugated with polyethylene glycol, [¹⁸F]FB(ePEG12)12-exendin-4 (¹⁸F-exendin-4), to determine its ability to visualize pancreatic β-cell mass. Positron emission tomography/computed tomography performed at 60 and 120 min after tracer injection in individuals with type 1 diabetes was compared with data from healthy control participants. No serious adverse events occurred. Pancreatic uptake was consistently lower in individuals with type 1 diabetes and showed clear separation between individuals with insulin-dependent diabetes and healthy control participants at 120 min. Pancreatic uptake at 120 min correlated with fasting C-peptide index and inversely with hemoglobin A_1c and daily insulin dose per body weight. These findings support [¹⁸F]FB(ePEG12)12-exendin-4 positron emission tomography/computed tomography as a noninvasive approach for assessing β-cell mass and disease status.

Article Highlights

• We undertook this study to address the persistent need for noninvasive assessment of β-cell mass in type 1 diabetes.

• We aimed to determine whether ¹⁸F-exendin positron emission tomography/computed tomography can reliably visualize residual β-cell mass and distinguish stages of disease.

• We found that pancreatic tracer uptake was consistently reduced in type 1 diabetes, differentiated insulin-dependent patients from control participants, and aligned with markers of β-cell function and glycemic status.

• Our findings suggest that ¹⁸F-exendin imaging may offer fundamental platform for disease staging, therapeutic monitoring, and individualized care

Graphical Abstract

Introduction

Type 1 diabetes (T1D) is defined by progressive destruction of pancreatic β-cells, ultimately leading to absolute insulin dependence. T1D is clinically heterogeneous, encompassing acute onset, fulminant, slowly progressive variants, such as latent autoimmune diabetes in adults, and antibody-negative idiopathic diabetes (1). Regardless of etiology, progressive loss of β-cell mass (BCM) remains the central determinant of disease course. A noninvasive, accurate method to assess BCM would advance understanding of disease heterogeneity and support individualized therapy and evaluation of emerging interventions (2,3), particularly in T1D, where residual BCM influences glycemic stability yet is difficult to measure in vivo.

Currently, clinicians rely on surrogate markers, such as C-peptide, to infer β-cell function. However, C-peptide cannot directly visualize islet cells or their spatial distribution. C-peptide secretion is strongly influenced by glycemia, and although glycemia-adjusted indices have been proposed, interpretation of these measures remains difficult (4), limiting their utility for longitudinal or cross-sectional comparisons. In addition, C-peptide levels often fall below detection in long-standing T1D, despite persistence of scattered insulin-positive cells demonstrated in histopathological studies (5,6).

Moreover, because C-peptide reflects the product of BCM and per-cell secretory function, impaired secretion can mask preserved BCM. This is suggested by some cases in the early stages of type 2 diabetes (T2D), in which insulin secretion is impaired but later recovers, suggesting that BCM is relatively preserved despite severe functional impairment (7), and by human embryonic stem cell–derived pancreatic endoderm transplantation, where C-peptide remains undetectable for weeks to months before becoming measurable (8). Such limitations hinder both disease staging and the development of innovative therapies aimed at expanding or restoring BCM.

The anti-CD3 antibody teplizumab can delay the onset of T1D and preserve β-cell function, but responses are heterogeneous and appear to depend on residual BCM (9,10). Noninvasive BCM quantification could stratify and monitor responses to such immunotherapies. More broadly, it could support regenerative/replacement, including induced pluripotent stem cell–derived β-cell therapy.

In this context, glucagon-like peptide 1 receptor (GLP-1R)–targeted exendin-based imaging has emerged as a promising approach for direct visualization of β-cells (3). Radiolabeled exendin tracers using ⁶⁸Ga for positron emission tomography (PET) or ¹¹¹In for single photon emission computed tomography (SPECT) have demonstrated pancreatic uptake in humans (11,12). However, these studies showed substantial variability and overlap in pancreatic uptake between T1D and healthy individuals and lacked incorporation with the clinical parameters relevant to diabetes (12). Because PET should provide higher spatial resolution and quantitative accuracy than SPECT (13), and ¹⁸F should offer higher image resolution than ⁶⁸Ga, owing to its shorter positron range (14), we developed a ¹⁸F-labeled exendin-4-based probe conjugated with polyethylene glycol [¹⁸F]FB(ePEG12)12-exendin-4 (¹⁸F-exendin-4) (15,16).

Although ¹⁸F-exendin-4 PET/CT has been shown to safely and reproducibly visualize pancreatic uptake in healthy volunteers and in a transplanted pancreas graft (16,17), and has demonstrated clinical utility in insulinoma detection, supporting its β-cell specificity (18–22), its potential to quantify residual BCM in T1D has not been systematically investigated. We therefore conducted the first T1D study, examining associations between pancreatic uptake, biochemical BCM surrogates, and glycemic control.

Research Design and Methods

Study Design and Participants

This prospective single-center study was conducted at Kyoto University Hospital from September 2020 to March 2025. We enrolled individuals aged ≥20 years who 1) were clinically recognized as having insulin-dependent diabetes mellitus (IDDM) with fasting C-peptide immunoreactivity <0.6 ng/mL, or 2) had diabetes with one or more islet-related autoantibody: anti-GAD antibody, anti–IA-2 antibody, anti–zinc transporter 8 (ZnT8) antibody, islet cell antibody, or anti-insulin antibodies (before exogenous insulin therapy). This definition was intended to capture a broad spectrum of T1D.

Individuals were excluded from the study if they had severe hepatic or renal dysfunction, were receiving GLP-1R agonists or related medications that could not be safely discontinued, had known contraindications to exenatide administration, or had documented alcohol intolerance. In addition, any individual deemed unsuitable for study participation by the investigators was also excluded.

The study protocol was approved by The Kyoto University Graduate School and Faculty of Medicine, Kyoto, Japan, Ethics Committee (CRB5180002) and conducted according to the Declaration of Helsinki. Written informed consent was obtained from all participants before enrollment.

Clinical Assessment

Baseline clinical characteristics (age, sex, height, weight, BMI, disease duration, and insulin treatment history) were collected for all participants. HbA_1c, fasting plasma glucose, fasting serum C-peptide levels, islet autoantibody status (as defined in the inclusion criteria), and blood chemistry data, including renal function parameters, such as serum creatinine and estimated glomerular filtration rate (eGFR), were also recorded. Glucagon-stimulated C-peptide response (ΔCPR), 24-h urinary C-peptide excretion, and daily insulin dose per body weight were analyzed if available from clinical records.

GLP-1R PET/CT Protocol

The synthesis method of ¹⁸F-exendin-4 was as previously reported (16,17). Each participant received 37.6–48.7 MBq ¹⁸F-exendin-4 intravenously over 5 min. PET/CT scans were performed at 60 and 120 min after the start of the injection using an integrated PET/CT scanner (Discovery IQ; GE Healthcare, Waukesha, WI) with a bismuth germinate scintillator. The participants were required to fast from the start of the study until the completion of imaging and were administered intravenous saline throughout the procedure. Water intake was permitted during the fasting period. To prevent nausea and vomiting potentially induced by exendin-4, 10 mg metoclopramide was added to the continuous saline infusion. Blood glucose was measured at 10, 30, 60, and 120 min after the start of the injection, and vital signs and symptoms were continuously measured and recorded.

Image Analysis

Quantitative analysis of PET/CT images was conducted using the Advanced Workstation (AW) Server 3.2 (GE Healthcare). PET images were reconstructed using an ordered-subsets expectation maximization algorithm (12 subsets, four iterations) with point spread function correction. The field of view in the axial direction was 600 mm, the matrix size was 128 × 128, and the voxel size was 4.68 × 4.68 mm.

In all organs, regions of interest (ROIs) were placed with reference to PET/CT fused images to match anatomical structures and were set as spherical volumes entirely contained within the respective organs, with a target size of at least 1 cm³ when feasible. For subdiaphragmatic organs subject to respiratory motion (liver, pancreas, kidneys, spleen), only areas evidencing clear PET-CT overlap were included for ROI measurement. In the pituitary gland, gastric wall, and left ventricular wall, ROI volumes occasionally fell below 1 cm³ due to anatomical size constraints.

For each participant, the mean standardized uptake value (SUVmean) of the pancreas was determined by placing ROIs separately on the pancreatic head, body, and tail. The pancreatic head was defined as including the uncinate process and the parenchyma located to the right of the left border of the superior mesenteric vein, the tail was defined as the portion to the left of the left lateral margin of the aorta, and the body was defined as the remaining part. The average of these three regional SUVmean results was calculated to represent the overall pancreatic uptake.

These values were used to calculate pancreas-to-organ SUV ratios. To ensure consistency across all participants, PET images from a previous phase I study involving healthy adults (16) were reanalyzed using the same protocol as in the current study. These healthy volunteer data were included in selected analyses as a comparative reference for baseline uptake in individuals without diabetes.

Statistical Analysis

Statistical analyses were conducted using JMP Pro 18.0.2 software (JMP Statistical Discovery LLC, Cary, NC). For the primary analysis, participants were classified into two groups according to recorded fasting C-peptide levels: those with C-peptide ≥0.6 ng/mL were assigned to the the non–insulin-dependent diabetes mellitus (NIDDM) group, and those with <0.6 ng/mL were assigned to the IDDM group. Three participants who had undergone islet or pancreas transplantation were excluded from group analyses.

Correlations between PET-derived SUVmean and clinical/biochemical markers were assessed using the Spearman rank correlation coefficient, due to the nonnormal and potentially nonlinear nature of these variables, particularly those involving C-peptide. The Spearman correlation coefficient (ρ) and corresponding P values were both reported for all correlation analyses. To evaluate discrimination between clinical classifications, we performed receiver operating characteristic (ROC) curve analysis using pancreatic SUVmean. The 95% CI for the area under the curve was estimated using nonparametric bootstrap resampling (2,500 iterations).

For comparisons between two independent groups, the Welch t test was used as the default method. For three-group comparisons involving PET-derived parameters such as SUVmean, one-way ANOVA, followed by the Tukey-Kramer post hoc test was selected. Assessment of statistical assumptions and additional statistical details are provided in the Supplementary Material. All statistical tests were two-tailed, and P values < 0.05 were considered statistically significant.

Data and Resource Availability

Deidentified data generated in this study will be shared by the corresponding author upon reasonable request, subject to applicable ethical and legal restrictions.

Results

Patient Characteristics

The study enrolled 23 participants with T1D. Three participants had previously undergone islet or pancreas transplantation and were therefore excluded from group-based analyses. Among the remaining 20 participants, 13 were classified as IDDM and 7 as NIDDM. The study flow diagram is presented in Supplementary Fig. 1. Image data of phase I healthy male control participants were reanalyzed (n = 6) (16).

Clinical characteristics of all participants of this study and of phase I participants are summarized in Table 1. Details of islet-related autoantibody positivity are provided in Supplementary Table 1. No participants were on incretin-based agents at imaging.

Table 1.

Clinical characteristics of study participants

	IDDM	NIDDM	Phase I	IDDM + NIDDM	IDDM + NIDDM + phase I
Parameter	(n = 13)	(n = 7)	(n = 6)	(n = 20)	(n = 26)
Age (years)	54.3 ± 17.6	45.9 ± 15.6	22.0 ± 0.9	51.4 ± 17.0	44.6 ± 19.5
Sex, n
Male	4	1	6	5	11
Female	9	6	0	15	15
Duration of diabetes (years)	19.6 ± 11.5	4.3 ± 7.8	—	14.3 ± 12.6	—
Any islet-related autoantibody (positive), n	8	7	—	15	—
BMI (kg/m2)	23.2 ± 3.3	23.4 ± 2.3	17.9 ± 1.0	23.3 ± 2.9	22.5 ± 2.9
HbA1c (%)	8.2 ± 0.66	7.6 ± 1.63	5.2 ± 0.15	7.9 ± 1.1	7.2 ± 1.5
HbA1c (mmol/mol)	66 ± 7.2	60 ± 17.8	33 ± 1.6	63 ± 12.0	55 ± 16.4
Serum creatinine (mg/dL)	0.73 ± 0.14	0.63 ± 0.09	0.72 ± 0.08	0.68 ± 0.22	0.72 ± 0.20
eGFR (mL/min/1.73 m²)	83.3 ± 25.1	91.2 ± 9.6	122.4 ± 10.1	84.7 ± 24.0	87.5 ± 21.9
Fasting plasma glucose (mg/dL)	214.7 ± 107.0	126.7 ± 35.1	76.2 ± 7.1	183.9 ± 97.3	159.0 ± 96.7
Fasting serum C-peptide immunoreactivity (ng/mL)	0.00 ± 0.01	1.21 ± 0.46	1.03 ± 0.18	0.43 ± 0.65	0.57 ± 0.63
Fasting CPI	0.00 ± 0.00	2.49 ± 1.66	3.13 ± 1.33	0.35 ± 0.52	0.58 ± 0.65
ΔCPR (ng/mL)	0.00 ± 0.01a	1.73 ± 1.43	—	0.73 ± 1.25	—
Urinary C-peptide (µg/day)	0.2 ± 0.6b	40.9 ± 22.5	—	15.2 ± 24.0	—
Daily insulin dosage per body weight (units/kg/day)	0.60 ± 0.19	0.36 ± 0.31	—	0.52 ± 0.26	—

Baseline characteristics of all participants included in the present analysis: individuals with insulin-dependent diabetes mellitus (IDDM), non–insulin-dependent diabetes mellitus (NIDDM), and healthy male volunteers from a phase I trial (16). Values are mean ± SD. “–” indicates unavailable or not applicable data.

^aΔCPR was unavailable for three IDDM participants. ^bUrinary C-peptide excretion was unavailable for one IDDM participant.

Safety and Tolerability

All 23 participants completed ¹⁸F-exendin-4 PET/CT imaging. No serious adverse events occurred. One participant in the IDDM group developed hypoglycemia (38 mg/dL) at 114 min postinjection, which resolved promptly after 20 mL of 50% intravenous dextrose; serum C-peptide immunoreactivity at the time was below the detection limit, suggesting prolonged fasting rather than a direct effect of the tracer. Transient nausea/vomiting occurred in three participants and resolved after 10 mg of intravenous metoclopramide; these symptoms are consistent with known effects of GLP-1R agonists (23). No other adverse events were observed.

Biodistribution of ¹⁸F-Exendin-4 in Cases of T1D

To assess the potential impact of T1D on systemic distribution of ¹⁸F-exendin-4, we compared organ-specific SUVmean between phase I healthy volunteers (n = 6) and the participants with T1D analyzed in this study (n = 20). Biodistribution was broadly similar across groups (Table 2), with minor differences in blood-pool/urinary signals consistent with renal clearance. Pancreatic uptake was significantly lower in participants with T1D than in healthy control participants, both at 60 min (2.47 ± 0.38 vs. 3.74 ± 0.61, P = 0.0026) and 120 min (2.25 ± 0.73 vs. 3.43 ± 0.36, P < 0.0001) postinjection, supporting the use of pancreatic SUVmean as a disease-specific imaging marker.

Table 2.

Organ-specific biodistribution of ¹⁸F-exendin-4 at 60 and 120 min postinjection in healthy volunteers (phase I) and patients with T1D

		Phase I	Participants of this study
SUVmean of each organ	Postinjection time (min)	(n = 6)	(n = 20)	P value
Pancreas	60	3.74 ± 0.61	2.47 ± 0.38	0.0026
	120	3.43 ± 0.36	2.25 ± 0.73	<0.0001
Liver	60	0.89 ± 0.20	1.02 ± 0.32	0.2553
	120	0.62 ± 0.08	0.67 ± 0.05	0.4732
Spleen	60	0.88 ± 0.12	0.87 ± 0.06	0.9212
	120	0.60 ± 0.09	0.59 ± 0.05	0.8766
Kidney	60	58.57 ± 6.82	50.53 ± 15.40	0.0847
	120	47.23 ± 8.22	38.70 ± 13.30	0.0782
Gastric wall	60	1.02 ± 0.12	0.82 ± 0.06	0.1983
	120	0.70 ± 0.14	0.61 ± 0.35	0.3581
Small intestine	60	0.96 ± 0.25	0.88 ± 0.29	0.5300
	120	0.61 ± 0.14	0.57 ± 0.28	0.7229
Colon	60	0.77 ± 0.23	0.69 ± 0.22	0.4281
	120	0.56 ± 0.26	0.50 ± 0.21	0.6292
Parotid gland	60	1.75 ± 0.62	1.18 ± 0.41	0.0778
	120	1.34 ± 1.13	0.94 ± 0.39	0.4317
Myocardial wall	60	0.77 ± 0.16	0.83 ± 0.27	0.4616
	120	0.47 ± 0.11	0.54 ± 0.22	0.2722
Left ventricular cavity	60	1.28 ± 0.29	1.67 ± 0.46	0.0255
	120	0.69 ± 0.12	1.09 ± 0.42	0.0008
Urinary bladder	60	31.92 ± 11.49	79.18 ± 56.68	0.0019
	120	79.19 ± 58.63	99.48 ± 66.16	0.4888
Bone marrow	60	0.54 ± 0.26	0.53 ± 0.25	0.9750
	120	0.36 ± 0.12	0.41 ± 0.24	0.4358
Lung	60	0.45 ± 0.16	0.35 ± 0.12	0.2343
	120	0.34 ± 0.14	0.20 ± 0.10	0.0551
Thyroid	60	0.65 ± 0.09	0.74 ± 0.21	0.1527
	120	0.39 ± 0.07	0.47 ± 0.22	0.1382

Values are expressed as mean ± SD. Organ-specific mean standardized uptake value (SUVmean) was measured at 60 and 120 min after ¹⁸F-exendin-4 injection. Comparisons were made between healthy volunteers from the phase I study (n = 6) and a subset of participants with T1D (n = 20), derived from the 23 patients enrolled in the present clinical trial after excluding 3 individuals who had previously undergone islet or pancreas transplantation. P values were calculated using Welch t test.

Regional ¹⁸F-Exendin-4 Uptake Patterns in the Pancreas Across the Course of T1D

Regional SUVmean (head/body/tail) and region-to-whole pancreas ratios were compared across IDDM, NIDDM, and phase I healthy control participants at 60 and 120 min. Absolute regional SUVmean showed group differences and generally tended to be lower in the IDDM and NIDDM subgroups than in phase I healthy control participants at both time points, across head, body, and tail, whereas region-to-whole pancreas ratios did not differ by group at either time point (Supplementary Table 2). These results indicate that disease status reduces overall uptake without altering the relative intrapancreatic distribution pattern.

Pancreatic ¹⁸F-Exendin-4 Uptake by Time Postinjection Across Study Groups

We next examined whether pancreatic SUVmean at 60 and 120 min postinjection varied systematically with disease status. As shown in Fig. 1A, the 60-min SUVmean was significantly lower in the IDDM and NIDDM groups than in phase I healthy control participants (one-way ANOVA, P < 0.0001; post hoc Tukey-Kramer test demonstrated significant differences between the phase I group and the IDDM and NIDDM groups). Representative images are shown in Fig. 1B–D.

Figure 1.

Pancreatic uptake of ¹⁸F-exendin-4 PET/CT across study groups and time points. Representative images of ¹⁸F-exendin-4 PET/CT. Dot-and-box plots showing pancreatic SUVmean at 60 min (A) and 120 min (E) postinjection in phase I control participants and in IDDM, and NIDDM groups. Each dot represents one participant. Boxes indicate interquartile range (IQR), horizontal bars show medians, and whiskers extend to 1.5 × IQR. Outliers are plotted individually. Group comparisons were assessed using one-way ANOVA, followed by Tukey-Kramer post hoc tests. For panels with a significant one-way ANOVA (P < 0.05), brackets and asterisks denote significant Tukey-Kramer pairwise comparisons (P < 0.05). Mean ± SD: 60 min—phase I 3.74 ± 0.61, IDDM 2.41 ± 0.38, NIDDM 2.57 ± 0.40; 120 min—phase I 3.43 ± 0.36, IDDM 2.04 ± 0.43, NIDDM 2.65 ± 1.02. Representative 60-min PET/CT images. Arrows indicate pancreas (green), kidneys (blue), and gallbladder (gray). B: A 23-year-old healthy man from phase I. HbA_1c: 5.3% (34 mmol/mol); injected dose: 64.2 MBq; SUVmean: 4.32. C: A 51-year-old woman with IDDM (GAD⁺; disease duration: 35 years; fasting C-peptide undetectable; HbA_1c: 9.3% (78 mmol/mol); 38.7 MBq; SUVmean: 2.07). D: 53-year-old man with NIDDM (positive for GAD, islet cell antibody, and anti–insulin antibody at diagnosis; disease duration: 2.5 years; fasting C-peptide: 2.17 ng/mL; HbA_1c: 5.5% (37 mmol/mol); 39.5 MBq; SUVmean: 3.01). F–H: Corresponding 120-min images from the same individuals in B–D. F: Healthy control individual; SUVmean: 3.74. G: IDDM participant; SUVmean: 2.04. H: NIDDM participant; SUVmean: 2.81.

As shown in Fig. 1E, the 120-min SUVmean was significantly lower in the IDDM group than in phase I healthy group (one-way ANOVA, P = 0.0006; post hoc Tukey-Kramer test indicated a significant difference between the phase I and IDDM groups). The 120-min pancreatic SUVmean showed clear separation between the healthy control group and the IDDM group, with no overlap in the distributions (Fig. 1E). Representative images are shown in Fig. 1F–H.

In addition, pancreatic SUVmean normalized to reference organs (pancreas-to-organ ratios) was also compared (Fig. 2), as such ratios may reduce interindividual variability and improve BCM estimation (16,22). At 60 min postinjection, the pancreas-to-liver SUVmean ratio was significantly lower in the IDDM group than in phase I healthy control participants (ANOVA, P = 0.0171; Tukey-Kramer: phase I vs. IDDM, P = 0.0128), the pancreas-to-spleen ratio was not significantly different among groups (ANOVA, P = 0.0857), and pancreas-to-left ventricular cavity ratio was significantly lower in the IDDM and NIDDM groups than in phase I healthy control participants (ANOVA, P = 0.0001; Tukey-Kramer: phase I vs. IDDM, P = 0.0001; phase I vs. T2D, P = 0.0009) (Fig. 2A–C). At 120 min postinjection, pancreas-to-liver and pancreas-to-spleen ratios were not significant (ANOVA, P = 0.3557 and P = 0.6640, respectively), whereas the pancreas-to-left ventricular cavity ratio remained significantly lower in the IDDM and NIDDM groups than in phase I healthy control participants (ANOVA, P < 0.0001; Tukey-Kramer: phase I vs. IDDM, P < 0.0001; phase I vs. NIDDM, P = 0.0008) (Fig. 2D–F).

Figure 2.

Pancreas-to-organ SUVmean ratios of ¹⁸F-exendin-4 PET/CT across study groups. Dot and box plots indicate pancreas-to-organ SUVmean ratios for individual participants. Each dot represents one participant. Boxes indicate the interquartile range (IQR: 25th–75th percentile), horizontal bars indicate the median, and whiskers extend to 1.5 × IQR. Outliers beyond this range are plotted individually. Pancreas-to-organ SUVmean ratios at 60 min postinjection: pancreas-to-liver (A), pancreas-to-spleen (B), and pancreas-to-left ventricular cavity (C). Pancreas-to-organ SUVmean ratios at 120 min postinjection: pancreas-to-liver (D), pancreas-to-spleen (E), pancreas-to-left ventricular cavity (F). Group differences were evaluated using one-way ANOVA, followed by Tukey-Kramer post hoc tests where applicable. For panels with a significant one-way ANOVA (P < 0.05), brackets and asterisks denote significant Tukey-Kramer pairwise comparisons (P < 0.05).

Relationship of PET-Derived Indices With Fasting C-Peptide Index

¹⁸F-exendin-4 PET/CT provided several quantitative parameters, including pancreatic SUVmean at 60 and 120 min postinjection and pancreas-to-organ SUV ratios. These indices were compared with fasting C-peptide index (CPI), a surrogate marker of BCM defined as fasting C-peptide immunoreactivity (ng/mL) × 100/fasting plasma glucose (mg/dL), which has been believed to reflect β-cell function or mass (24,25). Among the PET-derived parameters, the SUVmean of the pancreas at 120 min and the pancreas-to-left ventricular cavity ratio at 120 min demonstrated strong correlation with fasting CPI across the groups (Table 3).

Table 3.

Correlation between PET-derived parameters and the CPI

Correlation with CPI	Postinjection time (min)	ρ	P value
Pancreatic SUVmean	60	0.64	0.0005
	120	0.73	<0.0001
SUVmean ratio
Pancreas-to-liver	60	0.57	0.0026
	120	0.27	0.1753
Pancreas-to-spleen	60	0.37	0.0609
	120	0.48	0.0121
Pancreas-to-left ventricular cavity	60	0.37	0.0609
	120	0.62	0.0008

Spearman rank correlation coefficients (ρ) and corresponding P values are presented for each imaging parameter. Mean standardized uptake values (SUVmean) of pancreas were measured at 60 and 120 min postinjection (n = 26, phase I, IDDM and NIDDM groups combined). Pancreas-to-organ SUV ratios were calculated using liver, spleen, and left ventricular cavity as reference tissues at both time points.

We further evaluated the influence of renal function on probe distribution. Because eGFR depends on sex and muscle mass and all phase I volunteers were young men with preserved renal function, eGFR-based analyses were restricted to participants with T1D. Some reference organ signals (liver, spleen, and left ventricular cavity) and derived ratios were associated with renal function, whereas pancreatic SUVmean was not (Supplementary Table 3).

Based on these results, absolute (non–ratio-based) pancreatic SUVmean appears to better reflect BCM than pancreas-to-organ SUV ratios, combining strong correlations with fasting CPI and lack of association with renal function.

Validation of Pancreatic SUV as a Marker of BCM: Correlations With Surrogate Indices and Clinical Parameters in T1D

In participants with T1D only (n = 20; IDDM and NIDDM combined), correlations were examined for surrogate markers of β-cell function or mass. These analyses included indices available only in T1D participants, such as ΔCPR and urinary C-peptide excretion, which were not assessed in the phase I study. At 120 min postinjection, pancreatic SUVmean showed significant positive correlations with fasting CPI (ρ = 0.48, P = 0.0310), fasting serum C-peptide immunoreactivity (ρ = 0.50, P = 0.0259), and urinary C-peptide excretion (ρ = 0.47, P = 0.045) (Fig. 3A–D and Table 4). The correlation with ΔCPR did not reach statistical significance (ρ = 0.45, P = 0.069).

Figure 3.

Correlation between ¹⁸F-exendin-4 pancreatic uptake and clinical parameters in T1D. Scatter plots show the relationship between pancreatic SUVmean at 120 min postinjection and selected clinical indices in participants with T1D (n = 20). Fasting CPI (A), fasting serum C-peptide immunoreactivity (B), 24-h urinary C-peptide (C), glucagon-stimulated ΔCPR (D), HbA_1c (E), daily insulin dosage per body weight (F), duration of diabetes (G), and BMI (H). Spearman rank correlation coefficients (ρ) and associated two-tailed P values are reported to reflect monotonic associations. For panels with statistically significant correlations (Spearman correlation, P < 0.05), a simple least-squares regression line with 95% CI (shaded area) is overlaid for illustrative purposes. The 24-h urinary C-peptide was unavailable in one participant, and ΔCPR was unavailable in three participants.

Table 4.

Spearman’s rank correlation between pancreatic SUVmean at 120 min postinjection and clinical parameters in participants with T1D

Correlation with Pancreatic SUVmean 120 min postinjection	ρ	P value
Fasting CPI	0.48	0.0310
Fasting serum C-peptide immunoreactivity (ng/mL)	0.50	0.0259
Urinary C-peptide (µg/day)	0.46	0.0450
ΔCPR (ng/mL)	0.45	0.0691
HbA1c (%)	−0.51	0.0217
Daily insulin dosage per body weight (units/kg/day)	−0.47	0.0362
Duration of diabetes (years)	−0.42	0.0645
BMI (kg/m2)	−0.037	0.8774
Age (years)	−0.28	0.2401
Fasting plasma glucose (mg/dL)	−0.31	0.1816
Preprobe-injection plasma glucose (mg/dL)	0.14	0.5563
Serum creatinine (mg/dL)	0.22	0.3597
eGFR (mL/min/1.73 m²)	0.014	0.9548
Injected probe (exendin-4 dose) (μg)	0.23	0.3357
Injected probe (radioactivity) (MBq)	0.097	0.6840

Spearman rank correlation coefficients (ρ) and P values are shown for the relationship between pancreas standardized uptake value mean (SUVmean) at 120 min postinjection and clinical or physiological parameters (n = 20; IDDM and NIDDM groups combined). β-Cell–related markers include fasting CPI, fasting serum C-peptide, urinary C-peptide excretion, and ΔCPR. Additional variables include HbA_1c, daily insulin dosage per kilogram body weight, duration of diabetes, BMI, age, glucose levels, renal function markers (serum creatinine and eGFR, and injected probe characteristics. The 24-h urinary C-peptide was unavailable in 1 participant. ΔCPR was unavailable in 3 participants.

Additionally, in T1D, residual BCM is considered to contribute to glycemic control (11,26). If pancreatic ¹⁸F-exendin-4 uptake accurately reflects BCM, it may correlate not only with established surrogate markers of β-cell function but also with clinical parameters such as glycemic control, insulin requirements, and disease duration. To explore this possibility, we analyzed correlations between pancreatic SUVmean and clinical parameters in participants with T1D. Pancreatic SUVmean exhibited significant inverse correlations with HbA_1c (ρ = −0.51, P = 0.0217) (Fig. 3E) and daily insulin dosage per body weight (ρ = −0.47, P = 0.0362) (Fig. 3F). In contrast, none of the conventional β-cell function indices—including fasting CPI, fasting C-peptide immunoreactivity, urinary C-peptide, or ΔCPR—were significantly correlated with HbA_1c (ρ = −0.33,−0.34,−0.25, and −0.39, P = 0.16, 0.14, 0.31, and 0.13, respectively, for each index) in our study group. A modest inverse correlation was observed between pancreatic SUVmean 120 min postinjection and diabetes duration (ρ = −0.42, P = 0.0645) (Fig. 3G), while no correlation was seen with BMI (ρ = −0.037, P = 0.8774) (Fig. 3H).

Age, fasting, and preprobe injection plasma glucose levels, renal function, and injected exendin-4 dose/radioactivity showed no significant associations with pancreatic SUVmean at 120 min (Table 4). At 60 min, overall correlations showed trends similar to those at 120 min, but correlations were weaker (Supplementary Table 4).

Consistency of PET-Derived BCM Estimation With Clinical Classifications

To evaluate consistency between PET-based BCM estimation and clinical classifications, we performed ROC analysis using 120-min pancreatic SUVmean. The area under the curve was 0.95 (95% CI 0.82–1.00) for T1D (IDDM + NIDDM) versus phase I, 0.85 (0.65–0.98) for IDDM versus non-IDDM (NIDDM + phase I), and 0.73 (0.46–0.96) for IDDM versus NIDDM (Supplementary Fig. 2A–C).

Discussion

Summary of Major Findings

This study evaluated ¹⁸F-exendin-4 PET/CT as a noninvasive method to assess residual BCM in individuals with T1D. Biodistribution was consistent with prior reports (16) and preclinical mouse studies (15), and pancreatic uptake was reduced in T1D (Table 2). Pancreatic SUVmean at 120 min postinjection correlated strongly with multiple surrogate indices of BCM and showed inverse correlations with HbA_1c levels and daily insulin dosage, suggesting potential relevance to overall glycemic management.

Regional Uptake Patterns Within the Pancreas

Some prior autopsy and histological studies have inconsistently suggested regional differences in β-cell loss (tail-predominant in both T1D and T2D [27], or head-predominant in T2D [28]). In contrast, our analyses did not support preferential loss (Supplementary Table 2), making whole-pancreas SUVmean an appropriate summary metric for BCM assessment.

Imaging-Based Differentiation of Residual BCM

A notable observation was the nonoverlapping separation of pancreatic SUVmean at 120 min postinjection between healthy control participants and participants with T1D (Fig. 1E). This supports ¹⁸F-exendin-4 PET/CT as a single-metric, noninvasive discriminator of markedly reduced BCM.

This clear separation contrasts with prior ¹¹¹In-labeled exendin SPECT reports showing some unreasonable overlap between individuals with IDDM and healthy control participants (12). In that study, the T1D group met inclusion criteria of not stimulated C-peptide levels being unmeasurable and disease duration of >5 years, broadly corresponding to our IDDM group. The absence of such overlap in the present data set suggests that ¹⁸F-exendin-4 PET/CT may enable more distinct discrimination of residual BCM.

The NIDDM group exhibited a broader range of SUVmean, overlapping with both the IDDM and control groups. This likely reflects heterogeneity in residual β-cell preservation across the spectrum of T1D.

Consistent with this heterogeneity in SUVmean, ROC analysis based on the 120-min pancreatic SUVmean showed modest discrimination between IDDM and NIDDM. In contrast, discrimination was higher for T1D versus control participants and for IDDM versus non-IDDM (NIDDM + phase I) (Supplementary Fig. 2).

Interpretation of Imaging and Biochemical Correlations

Histological assessment of pancreatic tissue (autopsy or biopsy) is the gold standard for BCM quantification but is too invasive for routine or longitudinal evaluation. Conceptually, probe uptake is the second most direct indicator of BCM, next to histology.

In our study, histological assessment could not be performed in each participant for ethical reasons. Accordingly, correlations with fasting CPI, an indirect but commonly used BCM surrogate, were evaluated (24,25). Notably, previous autopsy data showed β-cell area correlates with the fasting C-peptide–to–glucose ratio, supporting the rationale for CPI as a surrogate marker of BCM (25).

Among PET-derived parameters, the absolute pancreatic SUVmean at both 60 and 120 min showed strong correlations with the fasting CPI, suggesting that they may better reflect BCM compared with ratio-based parameters. The 120-min SUVmean exhibited the highest correlation coefficient (ρ = 0.73, P < 0.0001). The superior performance at the delayed time point likely reflects exendin probe pharmacokinetics: clearance of circulating tracer with sustained, high-affinity GLP-1R binding in β-cells, increasing target-to-background contrast (15,16).

Consistent with this, preclinical in vivo and in vitro data support time-dependent GLP-1R–mediated internalization/retention of the probe, increasing measured uptake at later time points (15), and prior first-in-human data also favor later imaging (16). Thus, the absolute pancreatic SUVmean, especially at 120 min postinjection, may serve as an appropriate parameter for BCM quantification.

Moreover, reference-organ ratios are renal-function sensitive and may be less reliable (Table 3 and Supplementary Table 3). This may reflect specific binding to β-cells, whereas signals in nontarget organs largely reflect circulating/nonspecific probe.

PET-Derived Marker of BCM in Relation to Clinical Features, Including Glycemic Control

Pancreatic SUVmean 120 min postinjection showed positive correlations with fasting CPI, fasting serum C-peptide immunoreactivity, and 24-h urinary C-peptide excretion. The association with ΔCPR was weaker and did not reach statistical significance (Fig. 3 and Table 4), potentially due to the smaller number of participants with available data of glucagon stimulation tests.

Notably, pancreatic SUVmean 120 min postinjection correlated inversely with HbA_1c, a metric reflecting long-term glycemic control (Fig. 3 and Table 4), but showed no significant association with plasma glucose levels measured immediately prior to probe injection (Table 4). This contrasts with CPI, which did not correlate significantly with HbA_1c in our study group. Unlike C-peptide–based surrogates, which cannot directly visualize BCM and fluctuate with glycemic conditions, ¹⁸F-exendin-4 PET may therefore provide a more direct, glycemia-independent measure of BCM that can track longitudinal changes even when biochemical surrogates lose sensitivity.

Individuals with T1D often face challenges in achieving stable glycemic control despite exogenous insulin therapy (29). Histological studies have shown that β-cells can persist even in long-standing T1D (5), and emerging evidence suggests that even trace endogenous insulin secretion that is undetectable by fasting serum C-peptide immunoreactivity can contribute meaningfully to glycemic stability, as reflected in urinary or postprandial C-peptide measurements (26,30).

The observed inverse correlation between pancreatic SUVmean and HbA_1c in our study may thus reflect imaging-based detection of remnant BCM that supports glycemic management, even when functional indices such as CPI fall below their detection thresholds, consistent with prior reports that preserved BCM is associated with reduced glycemic variability (11).

These findings highlight the potential utility of ¹⁸F-exendin-4 PET/CT not only as a correlate of β-cell function but also as an alternative marker where traditional biochemical surrogates lack sensitivity. The visual and quantitative readout provided by PET imaging may represent a new axis for evaluating β-cell preservation in clinical and research settings.

Limitations of the Study

Several limitations of this study should be acknowledged. First, the sample size was relatively small (n = 20), limiting the statistical power for subgroup analyses and multivariable modeling and reducing the generalizability of the findings. Future studies with larger cohorts and/or focused subgroup designs are warranted to examine the relationship between PET-derived estimates of BCM and more detailed clinical disease features. Continuous glucose monitoring data were not collected, precluding a detailed assessment of glycemic variability. The relationship between PET-derived estimates of BCM and continuous glucose monitoring–based glycemic variability should be examined in future studies.

Second, although one of the major advantages of noninvasive PET imaging lies in its potential for repeated, longitudinal assessment within the same individual, this study was cross-sectional in design and did not include follow-up data including cases with within-individual progression such as NIDDM-to-IDDM transition. Further longitudinal studies are needed to confirm whether pancreatic SUVmean can track changes in residual BCM over time.

Third, demographic differences between the T1D group and the phase I healthy control cohort—including age, sex, and renal function—may have influenced quantitative SUV measurements. Importantly, however, the key correlations between pancreatic SUVmean and surrogate or clinical markers of β-cell function—such as fasting CPI, urinary C-peptide, and HbA_1c—were consistently observed even when analyses were restricted to the T1D group alone, excluding the phase I control participants (Fig. 3 and Table 4).

Fourth, given the exploratory nature of this study, no correction for multiple comparisons was applied.

In addition, outside the pancreas and excretory pathways, organ-specific signal was generally low-level in clinical PET/CT with this tracer. In our T1D cohort, we did not observe notable differences in these extrapancreatic signals compared with control participants (Table 2). Future studies in other diseases or physiological settings may clarify the clinical interpretations of these extrapancreatic signals.

Clinical Implications and Future Utility

Our findings suggest that ¹⁸F-exendin-4 PET/CT may provide additional information to conventional biochemical surrogates, especially in advanced disease stages where these indices become less reliable. Furthermore, the observed inverse correlation between pancreatic SUVmean and HbA_1c suggests that PET-derived markers may offer unique insight into real-world glycemic stability, particularly in cases where biochemical indices such as C-peptide fall below detection thresholds (5,11,26,30).

While our study captured only a single time point, the strength of the observed correlations and the clear separation in pancreatic SUVmean between phase I healthy control participants and IDDM participants provide a strong rationale for future longitudinal evaluations of β-cell preservation using ¹⁸F-exendin-4 PET/CT.

Moreover, quantitative PET metrics may provide a foundation for disease staging. Such staging could stratify patients into clinically meaningful categories — for example, those with substantial BCM who may benefit from β-cell–preserving or immunomodulatory strategies versus those with minimal BCM for whom regenerative or replacement approaches (including pluripotent stem cell–derived islet transplantation) may be more appropriate. Although the current study focused on T1D, similar BCM imaging may also help capture the heterogeneity of β-cell loss and treatment responsiveness in T2D.

In addition to assessing residual BCM in established disease, noninvasive quantification of BCM offers the possibility to optimize antidiabetes therapy on an individual basis, thereby opening the door to more personalized precision medicine in diabetes, including the development of curative strategies aimed at expanding residual β-cells or supporting the survival of transplanted or pluripotent stem cell–derived β-cells (3,31).

Beyond monitoring BCM preservation or restoration following β-cell-preserving or regenerative therapies, this approach could also complement immunological staging by identifying likely responders and providing objective efficacy readouts for immunotherapies such as anti-CD3 (9,10,32). This versatility suggests that ¹⁸F-exendin-4 PET/CT may contribute to a paradigm shift toward individualized and stage-specific disease modification across the full spectrum of T1D, from preclinical risk to long-standing disease.

Conclusion

¹⁸F-exendin-4 PET/CT enables noninvasive assessment of residual BCM in T1D, providing additional information to conventional biochemical surrogates. The 120-min acquisition showed the strongest correlations with surrogate markers and the clearest group separation, suggesting that late imaging is useful for BCM evaluation. ¹⁸F-exendin-4 PET/CT may offer a fundamental platform and serve as a reliable imaging biomarker for disease staging, therapeutic monitoring, and clinical decision-making in T1D.

This article contains supplementary material online at https://doi.org/10.2337/figshare.31231369.

Funding Statement

This study was supported by grants from Japan Society for the Promotion of Science (JSPS) Grants-in-Aid for Scientific Research (KAKENHI) (grant 24K02359), the Moriya Scholarship Foundation, Suzuki Manpei Diabetes Foundation, Suzuken Memorial Foundation, Japan Foundation of Applied Enzymology, Advanced Science, Technology & Management Research Institute of KYOTO, The Japan Health Foundation, Foundation of Future Research Support, Japan Diabetes Foundation, Fujiwara Memorial Foundation, and Japan Association for Diabetes Education and Care.