Quantitative Imaging Reveals Steatosis and Fibroinflammation in Multiple Organs in People With Type 2 Diabetes: A Real-World Study

Type 2 diabetes, as a multisystem cardiometabolic disease, causes a significant burden of microvascular and macrovascular disease with substantial end-organ damage. Surveillance of cardiovascular and renal complications is a clinical priority. In >4.5 million individuals with type 2 diabetes, cardiovascular disease (CVD) prevalence was estimated at 32%, accounting for ∼50% of the total deaths (1). Similarly, incident chronic kidney disease (CKD) occurred in 36% of >1.1 million European individuals with type 2 diabetes (2).

The role of liver fat accumulation in the pathophysiology of type 2 diabetes has been widely recognized (3), and more recent clinical focus has shifted to considering metabolic dysfunction–associated steatotic liver disease (MASLD). MASLD prevalence is 56% globally (4), with a low, but inherent risk of severe liver disease (increased incidence of fibrosis [5], hepatic decompensation, or hepatocellular carcinoma [6]) in individuals with type 2 diabetes. This need to assess liver health is reflected in guidelines (7,8). Similarly, individuals with type 2 diabetes are also at higher risk of pancreatic disease (acute pancreatitis and pancreatic ductal adenocarcinoma), but these outcomes are rarer (9).

The current focus of treatment of individuals with type 2 diabetes has shifted away from the traditional glucocentric approach (targeting reductions in HbA_1c, a downstream intervention) to a weight-centric and holistic approach (an upstream intervention). This approach recognizes obesity as a key pathophysiological driver of type 2 diabetes and its associated metabolic complications. The newer drug classes for type 2 diabetes, including sodium-glucose cotransporter 2 inhibitors (SGLT2is) (10) and glucagon-like peptide 1 receptor agonists (GLP-1RAs) (11), aside from their glucose-lowering effects, are associated with significant weight loss, and both these mechanisms contribute to benefits for the heart, kidneys, and liver (12,13). However, little is known about overall health changes, even with optimal management of diabetes.

Multiparametric MRI (mpMRI) serves as a noninvasive, reproducible tool for the quantitative assessment of organ manifestations associated with obesity, prediabetes, or diabetes (or indeed, any multisystem disease), providing quantitative analysis of tissue composition (14,15). Aside from assessment of body composition (differentiating between subcutaneous adipose tissue [SAT] and visceral adipose tissue [VAT]) (16), mpMRI has commonly been used for organ tissue characterization. Proton density fat fraction (PDFF) is a quantitative measure of fat (16,17) used predominantly as a clinical trial end point in the liver and pancreas, including trials with the latest dual and triple agonists and weight loss drugs (18,19), while T1 mapping mpMRI methods have been developed for the liver, pancreas, and kidneys.

Increases in T1 relaxation time acquired without contrast, reflecting increased water content in biological tissues, can be indicative of edema, inflammation, and/or fibrosis (collectively termed fibroinflammation) (20). Iron-corrected T1 (cT1) is a marker of fibroinflammatory change in MASLD guidelines and guidance (7,21,22); cT1 has sensitivity to fat but correlates with liver disease activity from pathology (18,19) and has been shown to predict liver-related outcomes in metabolic dysfunction–associated steatohepatitis (MASH) (23) and CVD outcomes (24). In the pancreas, increased T1 relaxation time discriminates acute pancreatitis, resolving in response to anti-inflammatory treatment (25), can stage chronic pancreatitis (26) and pancreatic fibrosis (27), and correlates with reduced exocrine function in pancreatic ductal adenocarcinoma and chronic or autoimmune pancreatitis (28). In the kidneys, increased T1 correlates with a reduction in renal function (by estimated glomerular filtration rate [eGFR]) (29), is diagnostic of CKD (30), and is elevated in the cortex relative to the medulla in kidneys with interstitial fibrosis (31).

The prognostic relevance of perirenal and renal sinus fat deposition and its relationship with an increased risk of CKD in individuals with type 2 diabetes has also been shown with mpMRI (30,32). mpMRI has also provided metrics of vascular health that have been extensively validated as independent predictors of incident cardiovascular events and are diagnostic of atherosclerosis and aortic aneurysms (33,34).

The aim of the current study was to determine the underlying burden of ectopic fat and fibroinflammation affecting the liver, pancreas, and kidneys and to assess the relationship with visceral adiposity and vascular health using mpMRI in individuals with type 2 diabetes compared with age- and BMI-matched and normal weight people without type 2 diabetes. Additionally, as a secondary aim, we evaluated the impact of clinical features and changes in weight, glycemic control, or drug therapy on underlying multiorgan health.

The Longitudinal Assessment of Multiple Organs in Patients With Type 2 Diabetes (MODIFY) trial (ClinicalTrials.gov identifier: NCT04114682) was a real-world, multicenter study adopting a prospective, longitudinal, observational cohort study design in individuals with type 2 diabetes. There was no intervention to the standard of care. Adult individuals with type 2 diabetes on glucose-lowering therapy were recruited from secondary care settings (119 of 134 individuals) and the community (15 of 134 individuals) between January 2020 and March 2022. Exclusion criteria were hepatitis, Wilson disease, hemoglobinopathies, known renal tract abnormalities, excessive alcohol intake, and contraindication to MRI scanning. The three participating centers were the University Hospital Aintree, Liverpool; Oxford University Hospitals NHS Foundation Trust; and Royal Free NHS Foundation Trust.

All individuals with type 2 diabetes attended a baseline clinical assessment (blood and urine samples, medical history and anthropometrics, MRI) (Supplementary Fig. 1). All were invited for follow-up, with clinical and MRI data collected at a single visit after baseline, funding permitting.

For comparison, we studied three additional groups: 69 matched control individuals with type 2 diabetes (ICD-10 codes E11.0–E11.9) from the general population (UK Biobank, matched for age, sex, ethnicity, and BMI) (35), 134 matched control individuals without type 2 diabetes (from the UK Biobank), and 92 healthy volunteers (Mapping Organ Health Following COVID-19 Disease Due to SARS-CoV-2 Infection [COVERSCAN]; ClinicalTrials.gov identifier: NCT04369807 [15]). These participants had MRI scanning and clinical data collection, but no prospective plasma or urine biochemistry was performed, although HbA_1c values were imputed from earlier visits, as previously described (24). MRI data for liver and pancreas were available from the UK Biobank but not for body adiposity composition, aorta, or kidneys.

HbA_1c, renal profile, liver function tests, lipids, and N-terminal pro B-type natriuretic peptide were measured through accredited clinical laboratories. Metabolic syndrome was defined as per guidelines (36). CKD was defined as an eGFR <60 mL/min/1.73 m² and/or a urine albumin creatinine ratio ≥30 mg/g (37).

At both visits, all individuals with type 2 diabetes and healthy volunteers had a standardized multiorgan mpMRI scan (15) (CoverScan; Perspectum, Oxford, U.K.), which lasted ∼35 min using methods previously demonstrated for the healthy volunteers and the UK Biobank (14,15). All quantitative multiorgan MRI methods were deployed on standard clinical MRI scanners (Siemens Prisma 3T, Siemens Skyra 3T, Siemens Area 1.5T, or a GE Signa Voyager 1.5T), and data were acquired and processed by trained magnetic resonance technologists and radiographers. Data were centrally curated and quality controlled.

Scan-rescan repeatability of the metrics was evaluated in the healthy volunteers using standardized performance testing criteria to derive repeatability coefficients (15). Incidental findings were reported and reviewed by an expert radiologist.

Normal values/reference ranges for MRI metrics for each organ were defined relative to reference ranges from the 92 healthy volunteers (Supplementary Table 1).

In individuals with a cardiometabolic risk factor, advanced MASLD and metabolic dysfunction–associated and alcohol-related liver disease (MetALD) were imaging-based definitions for both liver fat ≥5% and liver disease activity (cT1) ≥800 ms, previously shown to be diagnostic of steatohepatitis in biopsy-paired data sets (18), in the absence or presence, respectively, of high consistent alcohol intake as per multiple society recommendations (38). An additional threshold of liver disease activity ≥875 ms was also applied, which is associated with risk of liver fibrosis in biopsy-paired data sets (18). Liver volumes were also assessed but not included in the definition of liver MRI abnormality (15).

Metrics of pancreas fat (PDFF) and fibroinflammation scanner-referenced T1 (srT1) were collected, and elevation in either defined pancreatic abnormality, with disease defined as both steatosis and fibroinflammation. Scanner referencing to derive srT1 includes 1) a scanner normalization step, which involves referencing to a specific MRI scanner of the same field strength, and 2) a field strength adjustment step to 3T, when applicable.

Metrics of renal sinus fat volume and fibroinflammation in the renal cortex (cortical T1) were collected for both kidneys, and elevation in either defined renal abnormality, with disease defined as both steatosis and fibroinflammation.

Distensibility was determined at three positions: proximal ascending, proximal descending, and abdominal aorta (34,39), and reduction at any position was considered a stiff, unhealthy aorta. The diameter lumen at systole was measured at the abdominal position; >3 cm defined aortic abdominal aneurysm, as per guidelines.

Cross-sectional areas of SAT and VAT were determined from a single two-dimensional section positioned at the third lumbar vertebra (this region has been shown to be strongly associated with whole-body skeletal muscle distribution and to accurately estimate total SAT and VAT volumes [16]). Elevation in either defined abnormal body composition.

From recent guidelines (40,41) in type 2 diabetes, we considered a relative change of 10% body weight or a return to HbA_1c levels of <7% (53 mmol/mol) as a clinically meaningful outcome, although a relative reduction of 5% in weight and absolute reduction of 0.9% in HbA_1c were also investigated, as these were the usual indicators in our real-world hospital settings.

The study was powered for the primary end point: to evaluate liver disease activity (measured by cT1) in individuals with type 2 diabetes compared with matched control individuals without type 2 diabetes. A priori, we performed a power calculation for a group difference in liver cT1 between the baseline type 2 diabetes cohort and healthy control individuals matched for sex, BMI, and age at 90% power and α of 0.05. At the final sample size of 134 per group, this enabled a minimum detectable group difference of 33 ms.

The descriptive statistics for continuous and categorical variables are expressed using the mean (SD) and frequency (percentage prevalence), respectively. For groupwise comparisons, Wilcoxon rank sum test was applied for continuous variables and Fisher exact test for categorical variables. For groupwise comparisons between the type 2 diabetes group and unmatched healthy volunteers, linear and logistic regression models were used to statistically control for differences in age, sex, ethnicity, and BMI in order to evaluate differences in continuous and categorical variables, respectively. Statistical significance was defined by a P < 0.05 (two-sided) threshold. All statistical analyses were conducted using R 4.2.1 software.

All data relevant to the study are included in the article or uploaded in the Supplementary Materials.

One hundred thirty-four individuals with type 2 diabetes mainly from secondary care underwent a baseline evaluation (mean age 61 years, 41% female, 87% Caucasian, mean BMI 32 kg/m², 5% smokers) (Table 1). More than one-half (55%) had a duration of type 2 diabetes of >10 years, and 22% had a duration of <5 years (Supplementary Table 2).

We compared metabolic comorbidities and MRI organ metrics with 134 matched control individuals without type 2 diabetes (mean age 61 years, 46% female, 83% Caucasian, mean BMI 32 kg/m², 6% smokers) and 69 matched individuals with type 2 diabetes (mean age 62 years, 45% female, 94% Caucasian, mean BMI 31 kg/m², 6% smokers, 51% with hypertension, 46% obese) from the general population. The latter presented with acceptable glycemic control (mean HbA_1c 6.9% [SD 0.9], mean diabetes duration 5 [SD 4] years) (Supplementary Table 2). We also compared with 92 healthy volunteers of normal weight (mean age 44 years, 66% female, 92% Caucasian, mean BMI 23 kg/m², 3% smokers).

Almost one-half (63 of 133 [47%]) of all individuals were taking hypertension medications. Hypertension was prevalent, with 52% of individuals exhibiting a systolic blood pressure ≥140 mmHg (Fig. 1), of whom 35 of 69 were taking hypertension medications. Diastolic blood pressure levels were lower (14% of individuals had ≥90 mmHg). Hypertension prevalence was similar whether individuals with type 2 diabetes were from the hospital or general population (Supplementary Table 2).

The mean HbA_1c was 8% (63 mmol/mol), with 83% of the cohort at >6.5% (48 mmol/mol), with marginally worse glycemic control evident with longer disease duration (mean 7.5% if the duration was <10 years vs. 8.2% if ≥10 years, P < 0.001). Glycemic control was worse in the hospital setting compared with individuals from the general population (Supplementary Table 2). A preexisting diagnosis of metabolic syndrome was common in this population (77%), and 12 individuals (9%) had mild to moderate (GFR stage ≥3) CKD, of whom three had concomitant albuminuria.

Patient management comprised 30 different stable combinations of glucose-modifying drugs (Fig. 1). One-third of the cohort was taking metformin alone, and metformin was used in combination with other drug classes in 56%. Almost one-half (49%) of the individuals with type 2 diabetes were taking either an SGLT2i or a GLP-1RA or both. Treatment allocation to SGLT2i and/or GLP-1RA was more frequent in those with longer type 2 diabetes duration (mean 14 vs. 8 years duration, P < 0.001) and worse glycemic control (mean HbA_1c 8.2% vs. 7.5% [66 vs. 58 mmol/mol], P = 0.006) compared with allocation to metformin alone.

Based on imaging, hepatic steatosis was present in 94 of 134 (70%) individuals, and liver disease activity at thresholds diagnostic of steatohepatitis (advanced MASLD/MetALD) (18) was present in 53 of 128 (41%) individuals. These proportions were higher compared with healthy volunteers (0 of 92 [0%]) and matched control individuals without type 2 diabetes (17 of 134 [13%]) (Table 1). Liver disease activity at thresholds diagnostic of steatohepatitis and significant fibrosis (18) was prevalent in 23% of individuals. Advanced MASLD/MetALD was more frequent in individuals with obesity but not in those with longer diabetes duration or poor glycemic control (Fig. 2). The advanced MASLD/MetALD group had a higher BMI (33 vs. 30 kg/m²) and liver biomarkers AST and ALT outside normal ranges, but significant elevation in Fibrosis-4 index was not observed (Supplementary Table 3 and Fig. 3). Advanced MASLD/MetALD prevalence was similar with and without hypertension (40% with vs. 44% without, P = 0.594). Separate analysis of individuals with type 2 diabetes from the general population indicated that advanced MASLD/MetALD was slightly less prevalent (32% vs. 41% in our prospective hospital cohort, P = 0.22) (Supplementary Table 2).

Abnormal organ characteristics in the pancreas were also very common (73 of 109 [67%]), with proportions significantly higher than in healthy volunteers (12 of 92 [13%]) (Table 1). Pancreatic steatosis was frequent (70 of 112 [63%]) and was independent of BMI status, glycemic control, or duration of diabetes (Fig. 2). Fibroinflammation (20 of 99 [20%]) occurred where glycemic control was poor (Fig. 2) and was more frequent than in healthy volunteers (2 of 93 [2%]) or matched control individuals without diabetes (17 of 134 [13%]). Pancreatic disease with both steatosis and fibroinflammation was prevalent in 17% (17 of 102), of whom all (100%) had poor glycemic control (HbA_1c >6.5%). These findings were significantly more prevalent in our prospective hospital setting compared with type 2 diabetes in the general population (6% pancreatic disease, P = 0.035) (Supplementary Table 2). Hypertension did not discriminate pancreatic disease (18 vs. 16%, P > 0.99).

Abnormal tissue characteristics in the kidneys were present in 90 of 122 (74%) individuals and more frequent than in healthy volunteers (5 of 81 [6%]) (Table 1). This was due to renal fibroinflammation (26 of 133 [20%]) or steatosis (77 of 121 [64%]). Obesity was associated with steatosis in the kidneys (Fig. 2). Thirteen of 132 (10%) individuals with type 2 diabetes had renal disease, with both steatosis and fibroinflammation, compared with none of the healthy volunteers, but eGFR was low in only 2. Hypertension did not discriminate renal disease, renal steatosis, or fibroinflammation (all P > 0.5).

No individuals had abdominal aortic aneurysms. Stiffness of the aorta (low distensibility) was very frequent (103 of 125 [82%]) in the individuals with type 2 diabetes, particularly in the proximal position and in those with longer disease duration (Table 1 and Fig. 2), compared with healthy volunteers (5 of 75 [7%]). One-half (62 of 124) of individuals with a stiff aorta also had hypertension, and hypertension did not discriminate those with aortic stiffness (P = 0.101).

High VAT or SAT were very frequent (109 of 134 [81%]), particularly in those with obesity, and for SAT only with shorter disease duration (Table 1 and Fig. 2), compared with only 6% of healthy volunteers. Hypertension did not discriminate prevalence of high VAT or SAT (P = 0.053 and P > 0.5, respectively).

Overall, imaging showed that all 134 (100%) individuals with type 2 diabetes had abnormal tissue characteristics in at least one organ at baseline, and 126 of 134 (94%) had abnormal tissue characteristics in at least two organs (Fig. 3). Abnormal tissue characteristics were found in 109 of 134 (81%) individuals in at least three organs, particularly when these individuals were also living with obesity (Fig. 2). In contrast, routine biomarkers informing on the same organs without MRI indicated that only 58 of 134 (43%) had abnormal values in at least three organs, even if almost all (132 of 134 [99%]) had an abnormal value in at least one organ (Fig. 3).

Having advanced MASLD was always associated with abnormal tissue characteristics in at least one other organ site (Fig. 3). Most common was advanced MASLD with elevated renal steatosis in the right kidney (69% with advanced MASLD vs. 48% without, P = 0.036) and higher VAT (281 cm² with advanced MASLD vs. 225 cm² without, P = 0.002) (Supplementary Table 3). Renal and pancreatic fibroinflammation overlapped with advanced MASLD in 19% and 15% of individuals with advanced MASLD, respectively. Only 5 of 17 individuals with pancreatic disease also had advanced MASLD, and 6 of 13 individuals with renal disease had advanced MASLD.

Forty-six participants had at least one incidental finding (kidney and liver involvement were most common), of whom at least nine (7%) were recommended for further targeted clinical assessment, resulting in one partial nephrectomy for renal cancer with complete recovery. Eight of nine individuals (89%) had a stiff aorta and either a coprevalent liver or kidney abnormality or both; one (11%) had coprevalent advanced MASLD.

A total of 93 individuals (69%) identified at baseline returned for follow-up evaluation (with similar clinical characteristics overall) (Supplementary Table 4). The mean time from baseline to follow-up assessment was 218 (SD 44) days. In the follow-up group, treatment allocation did not substantially change, with 13 (14%) individuals on different treatments at follow-up. Anthropometrics and routine biomarkers showed negligible differences between baseline and follow-up. Few individuals with type 2 diabetes showed clinically meaningful outcomes in weight (3 of 93 [3%] lost 10% of body weight) or blood glucose (6 of 82 [7%] returned to an HbA_1c of 7%), although more individuals displayed meaningful change (lost 5% of body weight or changed by 0.9% HbA_1c) (Fig. 4).

Prevalence of abnormal organ features did not substantially change; for example, advanced MASLD prevalence in the follow-up group was 42% at baseline and 45% after 7 months (Supplementary Table 4). While overall changes were small in most individuals, those with ≥5% weight loss showed improvements in VAT and SAT (P < 0.001) and liver steatosis (P = 0.011) (Fig. 4). Similarly, better glycemic control (HbA_1c ≥0.9%) improved mainly liver steatosis (P < 0.001) but also renal steatosis (P = 0.01) and VAT (P = 0.01). Small, statistically significant improvements in aortic stiffness, VAT, liver disease activity, and liver size were observed in individuals with type 2 diabetes taking SGLT2i and/or GLP-1RA medications, compared with all other treatments (mainly those taking metformin alone) (Fig. 4). Of 14 individuals who had an abnormality in all organs at baseline, all abnormalities remained at follow-up in all except four taking an SGLT2i/GLP1-RA in whom liver disease activity returned to normal levels (Fig. 5).

In this quantitative, real-world, multiorgan MRI assessment study of liver, kidney, and pancreatic tissue composition; aortic distensibility; and visceral/subcutaneous adiposity in individuals living with diabetes and obesity, we demonstrate a significant cumulative burden of multimorbidity with multiorgan abnormalities. These were more pronounced than in age- and BMI-matched individuals and healthy volunteers, highlighting the deleterious and widespread impact of poor underlying metabolic health. The multiorgan accumulation of fat and associated fibroinflammation, occurring secondary to obesity and poor glycemic control, highlight a likely mechanism for long-term organ damage, e.g., CKD, liver fibrosis/cirrhosis, and CVD, and for the increased risk of hepatobiliary and renal malignancy.

Two recent studies have suggested that SGLT2is and GLP-1RAs in individuals with type 2 diabetes and MASLD have protective effects against adverse liver (42) and cardiovascular and mortality outcomes (43). In our study, multiorgan abnormalities were evident in almost every individual (94%), despite multiple glucose-lowering treatments, recognizing a predictable cluster of multiorgan risk factors with a more global, whole-body assessment. A noninvasive, comprehensive approach using mpMRI to simultaneously examine the health of multiple organs could avoid the need for multiple outpatient visits, organ-specific imaging (e.g., separate renal and liver ultrasonography), and the potential need for (liver/renal) biopsy with their inherent risks.

We previously showed that both liver and pancreatic fat are elevated in people with type 2 diabetes (16). Furthermore, liver disease activity, measured by cT1, is more widely representative of multiorgan health, including in the heart and brain, and can predict cardiovascular outcomes (14,24). In our study, liver involvement was frequently indicative of widespread abnormality. Advanced MASLD/MetALD was present in 41% of individuals and universally accompanied by abnormal organ tissue characteristics elsewhere (e.g., kidneys, pancreas); in contrast, steatosis and fibroinflammatory disease activity rarely co-occurred in the kidneys and pancreas. This work argues for multidisciplinary management, already cost-effective in diabetes (44). Prioritization of modifiable risk factors that mediate liver disease has been shown to be cost-effective in the management of MASLD and poorly controlled type 2 diabetes in some health care settings (45) but is infrequently used (46).

Obesity and hypertension, both widespread in our cohort, provide synergistic risk factors for development and progression of CKD in individuals with type 2 diabetes (47). Detection, early prevention, and treatment of CKD are critical to prevent progression to end-stage renal disease.

Reduction in pancreatic fat with weight loss interventions has been observed with improvements in insulin secretion and glucose homeostasis (48), but we did not observe significant weight change in our cohort over the 7-month follow-up period in routine care. Of interest is the association between poor glycemic control and pancreatic fibroinflammation in our study, a finding reproduced recently by Mak et al. (49) in individuals with MASLD from the Amsterdam NAFLD‐NASH cohort study (ANCHOR). Pancreatic fibroinflammation may have a more critical role than pancreatic steatosis in established type 2 diabetes.

We noted no significant longitudinal changes in either routine biomarkers or imaging over a 7-month follow-up. However, some interesting trends in liver, vascular health, and body composition were observed in those taking SGLT2is and GLP-1RAs, with weight loss or improvement in glycemic control. In contrast, there was no change in such metrics with a conventional glucose-lowering therapy approach or when weight or glycemic control did not improve. This insight supports the treatment stratification of individuals with type 2 diabetes according to multiorgan phenotypes (e.g., underlying liver disease or CKD) when adopting this technology as a facilitating tool.

We acknowledge some limitations to our study. Future studies, assessing renal size and cardiac structural and functional changes using cardiac MRI may add additional prognostic value. We acknowledge the shorter follow-up interval between successive MRI scans was suboptimal for sufficient temporal resolution of changes with disease progression or treatment. Imaging data sets with greater scale, longer duration follow-up, and greater ethnic diversity may provide further clarity on the prognostic value of these imaging metrics in obesity- and diabetes-related complications. The impact of individual drug classes on these organ metrics remains unknown, and the global multiorgan impact of newer agents, with double-digit weight loss, such as semaglutide and tirzepatide (50), is of interest.

In summary, using comprehensive, multiorgan, mpMRI for the first time in individuals living with obesity and type 2 diabetes without previously diagnosed comorbidities, we demonstrate significant evidence of underlying multiorgan dysfunction involving the liver, pancreas, kidneys, and cardiovascular system, which were more pronounced than expected based on age or BMI. Detailed multiorgan imaging would enhance risk and therapeutic stratification of this high-risk group.

Acknowledgments. The authors thank Jinny Brown and Rob Suriano (Perspectum, Ltd.) for project management during the study. They also thank the multisite clinical team for recruitment and electronic data capture completion: Andrew Williams (University Hospital Aintree [UHA]), Andrew Irwin (UHA), Melissa Day (UHA), Angela Dodd (UHA), Kate Sullivan (UHA), Lia Anguelova (Oxford University Hospitals [OUH]), Victoria Murphy (Perspectum, Ltd.), Soubera Rymell (Perspectum, Ltd.), Karyna Gibbons (OUH), Nicky McRobert (OUH), Layla Hassani (OUH), Vivien Thornton-Jones (OUH), Rajeshwar Ramkhelawon (Royal Free London [RFL]), and Vashist Deelchand (RFL). Finally, the authors thank Rajarshi Banerjee (Perspectum) and Kenneth Cusi (University of Florida) for critical review of the manuscript.

Funding. This study was supported by a joint Innovate UK award (Digital Health Technology Catalyst round 3) to S.N.A., G.J.K., G.T., H.T.B., and D.J.C. UK Biobank data were accessed through application 9914.

Duality of Interest. C.D., P.P., and H.T.B. are employees at Perspectum, Ltd., the company that developed CoverScan. H.T.B. is also a shareholder at Perspectum, Ltd. M.P. and N.E. are former employees at Perspectum, Ltd. M.P. is a consultant for Perspectum. D.J.C. has received funding for conference attendance from Perspectum and has investigator-initiated research grants from Novo Nordisk and AstraZeneca. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.D., N.E., H.T.B., and D.J.C. were involved in the analysis. C.D., P.P., H.T.B., and D.J.C. were involved in the interpretation of the results. M.P., A.H., S.N.A., G.J.K., G.T., and D.J.C. were involved in the conduct of the study. S.N.A., G.J.K., G.T., H.T.B., and D.J.C. were involved in the conception and design of the study. H.T.B. wrote the first draft of the manuscript. All authors edited, reviewed, and approved the final version of the manuscript. D.J.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented orally at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022; in the article “Research Study Scan Detects Grandfather’s Cancer,” https://local.nihr.ac.uk/news/research-study-scan-detects-grandfathers-cancer/33156, 14 April 2023; and as a webinar for World Diabetes Day, https://youtu.be/-4AT3yTd3bU, 9 November 2022.