Introduction

Why was the cohort set up?

Diabetes mellitus (DM) remains a leading cause of blindness, kidney failure, lower limb amputation, heart attack, and stroke. Its global prevalence has surged, with over half a billion people currently affected1,2. Despite advances in DM treatment, particularly with the use of sodium-glucose cotransporter-2 (SGLT-2) inhibitors and glucagon-like peptide 1 (GLP-1) agonists for type 2 diabetes (T2D), glucose-centered therapies moderately reduce cardiovascular risk while at the same time increasing the risk for hypoglycemia3. The aim of this cohort is to identify DM subtypes with distinct courses of disease related to complications. Additionally, this cohort seeks to investigate new associations between DM subtypes and diseases previously not attributed to DM. This will aid clinical decision-making and enable the development of personalized treatments to prevent, or potentially reverse diabetes-associated complications.

Complications of DM

The traditional classification of chronic diabetes-associated complications is based on the consequences of vascular damage and divides them into two main groups: microvascular and macrovascular complications4. While this classification is convenient in clinical practice, emerging evidence suggests it is limited, as diabetes-related damage extends beyond the vascular system, impacting nearly all organs, tissues and cells5. Microvascular complications are not solely due to microvascular changes, but other changes beyond the vascular bed, such as retinal neurodegeneration in retinopathy6, tubulointerstitial changes in nephropathy7 and neuronal deficits in neuropathy8, also play key roles. Additionally, impaired angiogenesis of large artery microvessels, a form of microangiopathy, contributes to the development of the diabetic macroangiopathic disease9,10,11. There is also increasing recognition of nonclassical diabetes-associated complications, including abnormal pulmonary function, metabolic dysfunction-associated steatotic liver disease (MASLD), cardiomyopathy, increased risk of carcinogenesis5, increased diabetes-related distress and impairment of psychological stress response12,13. By using a holistic approach, the HEIST-DiC cohort has the potential to investigate both classical and nonclassical diabetes-associated complications.

Diabetes subtypes and risk of complications

Efforts to better understand the heterogeneity of DM have led to a paradigm shift, classifying individuals with newly diagnosed adult-onset DM into five clusters based on six simple clinical variables14,15. This new classification is significant as it stratifies individuals with recent-onset DM (diagnosis of DM within the past 12 months) into subgroups with varying risks for diabetes-associated complications. Particularly, individuals with severe insulin-resistant diabetes (SIRD) are at higher risk for cardiovascular, kidney and fatty liver disease, while those with severe insulin-deficient diabetes (SIDD) are more prone to diabetic neuropathy, often already present at disease onset14,15. Another retrospective cohort study found an increased risk of diabetic retinopathy (DR) in the SIDD cluster and a higher risk of nephropathy in the SIRD cluster16. The concept of new diabetes classifications provides deeper insights into disease heterogeneity but raises several unresolved questions, including: (i) the statistical methods used to define clusters, (ii) the stability of clusters over time, (iii) the use of genotypic versus phenotypic datasets, (iv) generalizability across different ethnic groups, (v) clinical relevance of in some cases only marginal absolute differences between subtypes and (vi) applications in clinical care17.

Early studies used k-means clustering at diabetes onset, assigning individuals to fixed clusters14,15,18,19. This approach broadly validated novel diabetes subgroups across Chinese, US, and non-white Emirati populations, suggesting potential generalizability across ethnicities20,21. However, rigid clusters at diagnosis fail to account for dynamic, longitudinal changes driven by initial pathological mechanisms. A German study found that 23% of individuals switched clusters within five years after diabetes onset, raising concerns about the practical utility of this rigid partitioning in managing ongoing diabetes15.

A novel soft-clustering method for patient stratification in ongoing diabetes identified a subgroup with obesity, insulin resistance, dyslipidemia, impaired β-cell sensitivity, rapid disease progression, and higher need for anti-diabetic therapy22. Another soft-clustering approach using genotypic data identified five robust clusters of T2D pointing to disease mechanisms reflected by clinical traits23. Advanced methods, such as a tree-like graph structure using reversed graph-embedded dimensionality reduction, enable stratification of pathophysiological components and diabetes-related complications throughout the course of diabetes24.

A clustering approach in individuals with prediabetes (PRED), who have an increased risk for developing T2D, identified six subtypes based on pathophysiological parameters. Of these, only two had a high imminent risk for developing T2D, despite impaired glucose metabolism25. Notably, one subtype exhibited an increased risk for kidney disease and all-cause mortality, despite only moderate risk for T2D25. Furthermore, diabetes-associated complications are already present in some individuals with PRED. Similar to findings in T2D, clustering analysis for PRED based on genotypic dataset identified six subtypes with distinct genetic score patterns and metabolic traits. Two of these subtypes exhibited a high risk of progressing to T2D26. These findings suggest that pathophysiological heterogeneity emerges before T2D onset, shaping in this way the course of the disease early on. An important open question for future research is the role of PRED cluster identification in predicting progression to diabetes and development of complications independent of diabetes progression. Other studies have examined disease progression in low- and high-risk PRED subgroups during the disease’s natural course and lifestyle intervention27. In contrast, the HEIST-DiC takes a holistic approach, providing a comprehensive clinical and metabolic assessment of diabetes-related complications.

The HEIST-DiC cohort provides a unique opportunity to assess individuals as usually seen in clinical practice. This study enables early identification of high-risk subgroups before complications manifest and investigates whether mechanisms driving (pre)diabetes onset contribute to the progression and development of complications. Most importantly, the cohort’s inclusion of individuals with long-term diabetes allows for the identification of factors distinguishing those with specific diabetes-associated complications from those without, beyond the presence and duration of diabetes. Leveraging the study’s intensive clinical-experimental characterization, HEIST-DiC enables the investigation of pathophysiological cellular mechanisms underlying different diabetes subgroups.

Personalized intervention strategies for DM and its complications

Previous attempts to prevent or slow the progression of diabetes-associated complications have achieved only moderate success in reducing absolute risk for diabetes-associated complications in both type 1 diabetes (T1D)28,29,30 and T2D31. This limited success is likely due, in part, to the glucose-centered focus of these interventions, as well as the lack of consideration of DM subtypes. A post hoc analysis of longitudinal data from intensive lifestyle interventions in T2D, utilizing age at diabetes diagnosis and k-means clustering, identified a cluster with poor glucose control associated with increased cardiovascular risk after the intervention32. In a randomized controlled trial using a fasting-mimicking diet in a T2D subgroup from HEIST-DiC, assessing vulnerability to fasting required deep phenotyping and longitudinal observation to distinguish individuals who might benefit from the intervention from those at risk of adverse effects33. Data-driven cluster analysis demonstrated that individuals with SIRD benefit with better glycemic control from thiazolidinediones, while glycemic control in those with mild age-related diabetes is better controlled with sulfonylurea18. These findings emphasize that use of novel diabetes subtypes is justified only if they demonstrate clinical relevance, specifically through subtype-specific responses to stratified intervention strategies34. Focusing on pathophysiology-based therapeutic approaches is key to delivering personalized treatment in T2D, and subtype-specific randomized clinical trials will be critical in assessing the clinical relevance of these new DM subtypes17,34.

The HEIST-DiC cohort stands out for its comprehensive inclusion of individuals across the full spectrum of glucose metabolism—from normal glucose tolerance to prediabetes, recent-onset diabetes, and long-standing diabetes with and without complications. This breadth provides an unparalleled opportunity to address critical unanswered questions in diabetes research, including the validation and refinement of clustering methods, the stability of subtypes over time, and the identification of novel biomarkers and high-risk individuals. By enabling the stratification of participants into subtypes and focusing on their pathophysiological mechanisms, the HEIST-DiC cohort bridges the gap between large-scale epidemiological studies and clinical-experimental research. This approach will not only validate existing clustering methodologies but also drive the development of new frameworks for understanding the dynamic and heterogeneous nature of diabetes. Ultimately, the cohort’s design supports the testing of tailored therapeutic strategies and interventions, advancing the clinical relevance of subtype-specific approaches and paving the way for precision medicine in diabetes care35.

Study design and methods

The Heist-DiC is an ongoing monocentric, prospective longitudinal observational study involving intensive metabolic and clinical phenotyping conducted annually for at least 11 years. Participants are recruited from the Clinic of Endocrinology, Diabetology, Metabolic Diseases and Clinical Chemistry (Internal Medicine 1) and are evaluated at the Study Center for Diabetes and Metabolism at the University Hospital of Heidelberg in Germany.

The aims of the HEIST-DiC cohort are to identify: (i) DM subtypes with different development patterns of diabetes-associated complications, (ii) predictors of diabetes-associated complications at early and late disease stages, (iii) pathophysiological mechanisms beyond hyperglycemia that contribute to the onset of diabetes-associated complications, (iv) novel nonclassical diabetes-associated complications, (v) personalized interventions for the prevention, improvement or remission of diabetes-associated complications.

The HEIST-DiC cohort is a hypothesis-generating study that will validate established clustering methods for diabetes classification, develop and apply new approaches – particularly for complications-specific clustering – and identify individuals from specific subtypes for targeted clinical-experimental studies.

Composition of the cohort

To capture the various stages of disease development and progression, the study imposes no restrictions on disease stage or DM duration. Therefore, the study includes normal glucose tolerant individuals (NGT) and individuals with PRED, defined by oral glucose tolerance test (oGTT) and/or glycated hemoglobin (HbA1c) according to current ADA guidelines36, as well as individuals already diagnosed with DM (T1D or T2D) aged 18 to 85 years. Inclusion of older individuals in the cohort allows for a more accurate representation and assessment of age-related variability, while enabling identification of short- to medium-term trends that remain clinically relevant, particularly with respect to the mild age-related diabetes subtype14,37. Individuals with an initial HbA1c ≥ 9.5% are excluded, to minimize the risk of acute glycemic deterioration during study visits and its potential impact on diabetes-related complications. This also ensures participant safety and the accuracy of metabolic phenotyping under stable medical conditions; however, later inclusion is possible after appropriate treatment lowers their HbA1c to < 9.5%. Individuals with type 3 (e.g. pancreatogenic) or type 4 (gestational) diabetes, as well as pregnant women, are also excluded. A complete list of inclusion and exclusion criteria is detailed in Table 1.

Table 1 Inclusion and exclusion criteria.
Full size table

Study participants are recruited through advertisements in local newspapers, public events, and information shared via the institutional website or health practitioners. Potential participants undergo a prescreening telephone interview to assess inclusion and exclusion criteria. Suitable candidates are then invited to the first study visit, which includes medical history, physical examination, blood withdrawal, questionnaires, and assessment of diabetes-associated complications. Decisions regarding participation in advanced examinations, such as the euglycemic hyperinsulinemic clamp test and tissue biopsies, are made and scheduled afterward. All participants provide written informed consent to the study protocol.

As of October 2023, the study has enrolled 552 participants, including 68 individuals with NGT, 119 with PRED, 83 with T1D and 282 with T2D. DM diagnosis and type are confirmed according to current ADA guidelines36 only for participants diagnosed at their first study visit, including those with latent autoimmune diabetes of the adult (LADA). No genetic analysis is conducted to assess for maturity onset diabetes of the young (MODY).

Duration of the study

As of October 1 2023, all study participants reported here have completed the full study program at baseline and annually thereafter. Electrophysiological examination, by nerve conduction velocity (NCV), and quantitative sensory testing (QST) were not performed at year 3 and 5. Starting October 2023, participants undergo the complete study program at baseline and at years 4, 8, and 12, with a reduced study program in the years in between. A detailed list of examinations for each study visit is provided in Table 2. Recruitment is ongoing, making the precise dropout rate difficult to determine, particularly as participants may miss visits due to illness or scheduling problems. By beginning of October 2023, 19% of participants were lost to follow-up, with 11% of these being deceased (total mortality-rate 2%), while others cited reasons such as lack of time, transportation issues, or were simply inaccessible. Death cause was acute myocardial infarction (4 participants), hemorrhagic stroke (1 participant), respiratory failure due to fungal pneumonia (1 participant), urothelial carcinoma (1 participant), sarcoma (1 participant) and unknown (7 participants). 92% of the deceased participants were T2D individuals, predominantly male (69%) and 8 years older than the ongoing participants. The rest of the dropouts (non-deceased participants) were mostly females, 5 years older than the ongoing participants, with the group-distribution being comparable to the ongoing study participants. Nearly half (40%) of dropouts occurred in the first year following the baseline visit. Participants can pause their involvement for any reason and resume later. Most missed appointments were due to cancellations related to the Covid-19 pandemic, with visit 3 being the most affected (18% missed the visit, half due to Covid-19-related restrictions). Overall, the yearly follow-up visits (from visit 1 to visit 5), had an average response rate of 82%, including participants who attended or are still eligible to attend. As many participants are recruited through our university outpatient clinic, routine visits as part of standard care allow us to re-establish contact with those participants lost to follow-up, re-inform them about the study, and offer the opportunity to resume participation.

Table 2 Questionnaires and examinations for all visits as of October 2023.
Full size table

Measurements

Demographic data and diabetes-specific medical history, including current medication, are recorded at baseline and updated during each follow-up visit (Table 2). Clinical and metabolic variables are also documented at both baseline and follow-up visits (Tables 3 and 4).

To assess chronic kidney disease (CKD), the study utilizes albuminuria, estimated glomerular filtration rate (eGFR), and kidney ultrasound33,38. Distal sensorimotor polyneuropathy (DSPN) is evaluated using the neuropathy symptoms score (NSS), neuropathy disability score (NDS), Purdue pegboard test, NCV, transcutaneous electrical nerve fiber stimulation, QST in accordance with the DFNS (German Research Network on Neuropathic Pain) protocol, and high-resolution magnetic resonance neurography (MRN)39,40,41,42,43,44,45,46. Heart rate variability (HRV) is measured to assess cardiovascular autonomic neuropathy (CAN)12. Anamnestic assessment of CAN is performed using the SAS questionnaire47. Ophthalmological assessment include a funduscopic examination of the undilated pupil for DR48. Diabetes related restrictive lung diseases (RLD) is evaluated through the 6-minute walk test, spirometry, body plethysmography, and carbon monoxide-based diffusing capacity measurements (DLCO)49,50. Transient elastography and hepatic ultrasound are conducted to evaluate MASLD and liver fibrosis51. The ankle-brachial index and carotid intima-media thickness are used to examine peripheral atherosclerosis52. Hand grip strength and bioelectrical impedance analysis (BIA) are utilized to assess muscle strength and body composition52,53.

OGTT and euglycemic hyperinsulinemic clamp test are conducted to assess beta-cell function and insulin sensitivity54. Hepatic insulin sensitivity is measured through the co-infusion of [6,6-2H2]glucose55. Additionally, energy expenditure and substrate oxidation during fasting and hyperinsulinemia are evaluated using indirect calorimetry56.

At each visit, whole-blood, erythrocytes, leucocytes, plasma (EDTA), serum and urine samples are stored at −80 °C for biomarker analysis57,58. Beginning in 2022, skeletal muscle, subcutaneous adipose tissue, and skin samples can also be collected for future analyses. The collected biospecimens support a broad spectrum of molecular and cellular analyses, including biomarker discovery, tissue-specific investigations of insulin resistance, inter-tissue communication and metabolic dysregulation, as well as mechanistic studies using patient-derived specimens.

Health-related quality of life and physical health are assessed using the 12-item short-form health survey (SF-12), while somatic and depression symptoms are evaluated with the patient health questionnaire (PHQ)59,60. A detailed overview of the measurements is provided in Table 2, and the experimental protocols are described in Supplementary file 1.

Definition of diabetes-associated complications

CKD was defined by an increased urinary albumin-to-creatinine ratio (uACR ≥ 30 mg/g) and/or a decreased eGFR (< 60 ml/min/1.73 m2), following the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines for DM management of CKD61.

DSPN was defined by the presence of either at least moderate neuropathic deficits (NDS ≥ 6) or the combination of mild deficits with moderate neuropathic symptoms (NSS ≥ 5 together with NDS 3–5)62, or according to the Toronto consensus definition of confirmed DSPN: abnormal results of the nerve conduction study of the sural nerve and additional at least one abnormal parameter from nerve conduction studies of the common peroneal or tibial nerve (parameters below the 2.5 percentile values of the cohort) together with an NSS and/or NDS of ≥ 341,58,63,64. Cases with bilaterally non-recordable sural nerve were assessed as having abnormal sensory function and values were calculated as lowest value of the cohort.

CAN was defined by two pathological values for age-dependent beat-to-beat variation with deep breathing (E/I-quotient), 30/15 heart rate ratio with standing, or orthostatic hypotension (systolic or diastolic change below the 2.5 percentile values of the cohort)12,65,66.

DR was defined by criteria indicating at least stage I of retinopathy based on funduscopic examination of the undilated pupil, as previously reported67. In this study, we do not analyze non-proliferative or proliferative DR.

Diabetes related RLD was defined by reduced forced vital capacity (FVC < 80%), reduced total lung capacity measured with body plethysmography (TLC-B < 80%), or reduced single-breath diffusing capacity of the lung for carbon monoxide (DLCO < 80%), in the presence of a normal forced expiratory volume in 1 s adjusted to vital capacity (FEV1/VC > 70%), as previously reported49,68.

MASLD was defined by an increased controlled attenuation parameter (CAP) (M-probe: CAP ≥ 248 dB/m, XL-probe: CAP ≥ 302 dB/m) measured by transient elastography. Liver fibrosis was defined by increased liver stiffness (≥ 8 kPa) measured by transient elastography41.

Discussion

What has it found? Key findings and publications

Baseline characteristics

Baseline characteristics of the cohort recruited between September 2016 and October 2023 are presented in Tables 3 and 4. The percentage of individuals diagnosed with early-onset T2D (diagnosis at age < 45 years) in our cohort aligns with findings from previous studies based on U.S. registry data (Table 3)69. Additionally, the prevalence of PRED at an early age (< 45 years) is consistent with global prevalence rates of PRED (Table 3)70. In our cohort, the percentage of males with T2D is higher than that of males with T1D, reflecting trends seen in a previously reported German cohort of individuals with recent-onset T1D and T2D (Table 3)71. Overall, our cohort includes more male participants than female participants with PRED, T1D, and T2D. The increased proportion of males in the pathological DM groups aligns with findings from population-based studies72,73 and likely reflects the known underrepresentation of females in clinical research74. There is a higher percentage of female participants than male participants in the NGT group (Table 3), potentially attributed to the greater motivation of healthy women to participate in time-consuming examinations. To address this limitation and ensure a more balanced representation, we are planning to recruit a higher proportion of male participants in the future through public events, as well as through information shared on the institutional website and by healthcare practitioners. For the analysis of different subtypes, study participants will be matched for sex, age at diabetes diagnosis, BMI, glycemia, and homoeostasis model estimates calculated using c-peptide (HOMA-IR).

Table 3 Baseline characteristics in individuals with NGT, PRED, T1D and T2D.
Full size table

Family history of DM was more frequent in the T2D group (Table 3). The cohort has not been screened for genetic defects that could indicate MODY. Diagnosis of DM and its respective type is confirmed only for participants diagnosed with DM at the first study visit, which means individuals with MODY may still be included in the cohort.

Most participants with T1D (76%) had a DM duration exceeding 5 years (Table 3), and therefore a negligible residual beta-cell function (Table 4, Panel b). As expected, participants with T2D exhibited low insulin sensitivity and high whole-body and adipose tissue insulin resistance (Table 4, Panel b). In our cohort glycemic control at baseline appears to be slightly better in T2D participants compared to those with T1D (Table 4, Panel b). All but one individual with T1D were treated with insulin and only 5% with metformin, which was the most common medication in individuals with T2D (62%). Notably metformin was also taken by 2 participants with NGT (one with a history of polycystic ovary syndrome and one diagnosed with insulin resistance) and 1 with PRED (indication of metformin therapy unknown), Additionally, 2% of participants with T1D were treated with SGLT-2 inhibitors (Table 3).

Table 4 Baseline anthropometric and metabolic characteristics in individuals with (a) NGT/PRED, and (b) T1D/T2D.
Full size table

While our cohort is not population-based and primarily focuses on diabetes-associated complications and distinct pathophysiological mechanisms, the mean age and body mass index (BMI) of participants align closely with other German and European cohorts (Table 4, Panel a and Panel b)71,75,76,77. As anticipated, participants with T1D are younger and leaner compared to those with T2D (Table 4, Panel b) and PRED. However, T1D participants exhibit similar age and BMI to those with NGT (Table 4, Panel a).

Participants in the T2D group exhibited a high prevalence of cardiovascular disease at baseline (Table 3). Hypertension was prevalent among T2D participants, with 77% reporting the condition, corresponding to a higher usage of antihypertensive therapy (69%) in this group compared to T1D (43%). Despite higher use of antihypertensive medication, the mean blood pressure values are higher in T2D participants than in those with T1D (Table 4, Panel b). Among PRED participants, 44% reported hypertension, with 40% receiving antihypertensive therapy. In the NGT group, 18% reported hypertension, and a similar percentage was on hypertensive therapy at baseline (Table 3). Statin use at baseline was reported by 6% of participants with NGT, 17% with PRED, 28% with T1D, and 45% with T2D (Table 3). While overall statin use was low across all groups, the undertreatment was particularly notable in T1D and T2D participants, with 73% of T1D participants and 86% of T2D participants classified as having high to very high cardiovascular risk according to the ESC-Score 2, independent of DM as a risk-factor (Table 3). Although low-density lipoprotein (LDL) levels for both T1D and T2D participants (Table 4, Panel b) are lower than in individuals with NGT and PRED (Table 4, Panel a), this poor utilization of statins is still reflected in the small proportion of participants achieving LDL-cholesterol goal levels, with only 17% of T1D and 30% of T2D participants meeting these targets.

Liver transaminases and lipoprotein (a) [Lp(a)] levels were comparable across all study groups (Table 4, Panel a and Panel b). However, levels of high-sensitivity C-reactive protein (hsCRP) were elevated in all groups except the NGT group, indicating a possible inflammatory response in participants with DM and PRED. N-terminal pro–B-type natriuretic peptide (proBNP) was elevated above the upper limit (125 ng/L < 71 year old; 450 ng/L for ≥ 71 year old) in 35% of T1D and 22% of T2D, and levels of high-sensitivity cardiac troponin T (hsTNT) were elevated above the upper limit (> 14 ng/L) in 14% of T1D and 27% of T2D (Table 4, Panel b), as markers indicating cardiac dysfunction and damage of heart muscle, respectively.

The primary focus of this cohort is the assessment of diabetes-associated complications, with findings summarized in Fig. 1; Table 5. At baseline, CKD was identified in 19% of individuals with T1D and 28% of those with T2D (Fig. 1). Additionally, uACR levels were elevated in both the T1D and T2D group compared to the NGT and PRED group, with T2D participants showing higher uACR levels than those of the T1D group (Table 5 Panel a and Panel b). The increased prevalence of CKD in T2D is supported by a greater usage of renin-angiotensin system blockers (Table 3). According to previous European cohort studies, CKD is more common in T1D than in T2D78,79. The higher prevalence of CKD in T2D within our cohort may be attributed to a greater number of T2D participants in older age groups despite longer diabetes duration, whereas the T1D group consists mainly of younger individuals (Table 2). This is partly due to the initial active recruitment of individuals with DM and established diabetic complications, such as increased albuminuria. A few participants from the non-DM groups also met criteria for diagnosing CKD at baseline, specifically four from the NGT group and seven from the PRED group (Fig. 1). This may be related to other underlying conditions contributing to nephropathy, such as hypertension or undiagnosed renal diseases. In terms of DSPN, baseline prevalence was found to be 11% in PRED participants, 23% in T1D, and 46% in T2D (Fig. 1). The higher prevalence of DSPN in T2D aligns with previous intervention and cohort studies, highlighting the increasing burden of DSPN in older populations80,81,82. A systematic review on DSPN prevalence in adults with PRED indicated that 72% of studies reported a prevalence of ≥ 10%, consistent with our findings83. CAN was identified in 11% of T1D and 17% of T2D participants (Fig. 1), with only 2% of PRED participants and 2% of NGT participants meeting the criteria for CAN at baseline. The observed prevalence of CAN in PRED in our cohort is notably lower than previously reported estimates65,84, likely due to variability in the definitions and methodologies used in earlier studies84. For DR, our cohort reported a higher prevalence in T1D (30%) compared to T2D (22%), consistent with findings from previous studies85,86, with only one case identified in the NGT group and three cases in the PRED group. Diabetes related RLD phenotype was found in 28% of NGT participants, 35% of PRED participants, 43% of T1D participants, and 44% of T2D participants (Fig. 1). These numbers exceed those previously reported49, indicating a potential underdiagnosis of diabetes related RLD that warrants further investigation in future studies, especially in terms of treatment responses compared to non-diabetic cases with RLD50. The prevalence of MASLD was reported as 43% in the NGT group, 56% in the PRED group, 23% in the T1D group, and 73% in the T2D group, whereas the prevalence for liver fibrosis was 6% in NGT, 13% in PRED, 5% in T1D and 23% in T2D (Fig. 1). The higher prevalence in PRED and T2D was also reflected by elevated liver enzymes and triglycerides in the PRED and T2D groups, with T2D expressing the highest levels (Table 4 Panel a and Panel b). While the fatty-liver-index (FLI), as an index of MASLD, follows the same pattern, the fibrosis indices - fibrosis-4 index (FIB-4) and metabolic dysfunction-associated steatotic liver disease fibrosis score (MASLDS) - were only elevated in T2D (Table 4 Panel a and Panel b). The high prevalence of MASLD in NGT, PRED and T2D is likely due to high BMI in our cohort (Table 4 Panel a and Panel b), since BMI was previously reported to be the strongest covariate in predicting MASLD when using transient elastography87. The prevalence of MASLD in T1D in our cohort is slightly lower than in previous reported studies88,89. The higher prevalence for liver fibrosis, measured by elastography, in individuals with PRED and T2D and the fibrosis indices indicating a higher risk for advanced liver fibrosis for T2D, highlight the important role of impaired glucose metabolism in progression to liver fibrosis.

Fig. 1
figure 1

Diabetes-associated complications in individuals with NGT, PRED, T1D and T2D. Percentage (x-axis) and total numbers (in bars) of individuals without (grey) and with (red) respective diabetes-associated complications. CKD: chronic kidney disease, DSPN: diabetic sensorimotor peripheral neuropathy, CAN: cardiovascular autonomic neuropathy, DR: diabetic retinopathy, RLD: diabetes related restrictive lung disease, MASLD: metabolic dysfunction-associated steatotic liver disease.

Full size image
Table 5 Parameters of diabetes-associated complications in individuals with (a) NGT/PRED, and (b) T1D/T2D.
Full size table

Summary of the results obtained so Far

By mid-2024, based on the HEIST-DiC study 44 articles were published in peer-reviewed journals, reflecting its ongoing contributions to the understanding of diabetes-associated complications. The distribution of these publications across the years is as follows: 1 article in 201790, 4 articles in 201849,91,92,93, 3 articles in 201994,95,96, 5 articles in 202097,98,99,100,101, 6 articles in 202150,57,102,103,104,105, 7 articles in 202212,33,46,106,107,108,109, 7 articles in 202338,44,45,58,110,111,112, and 11 articles by mid-202413,39,40,41,42,43,52,53,113,114,115.

Assessment of classical diabetes-associated complications

  • Using MRN, our previous studies have successfully identified fascicle lesions in the sciatic nerve trunk of individuals with T2D and DSPN116,117. These lesions were characterized by hyperintense mono- or multifocal patterns, predominantly located at the thigh level. In the HEIST-DiC cohort, we further established that the nerve lesion loads observed through MRN correlated with the severity of clinical symptoms associated with DSPN, as well as with impaired nerve conduction and sensory loss97 and with HbA1c118. Specifically, MRN-derived parameters, such as fractional anisotropy, which indicate nerve fiber integrity, demonstrated strong correlations with increased NDS and decreased nerve conduction velocity, independent of factors like age, sex, BMI, and HbA1c58. These findings suggest that MRN can serve as a valuable non-invasive diagnostic tool for evaluating nerve function in DSPN and DM. This method offers significant advantages over traditional assessment methods, such as standardized questionnaires and clinical examinations, which are often reliant on subjective evaluations and individuals’ cooperation.

  • Another significant finding from our research is the demonstration that QST can effectively identify early sensory deficits in individuals with and without T2D. These deficits were found to be associated with markers of insulin resistance, metabolic syndrome, and glycation end-products44. Based on QST data analysis, we confirmed four sensory phenotypes: healthy, thermal hyperalgesia, mechanical hyperalgesia, and sensory loss. Longitudinal analysis of these QST-based sensory phenotypes, provided valuable insights into the natural progression of DSPN, with the sensory loss phenotype being the most strongly correlated with DSPN43. Additionally, MRN of T2D individuals who exhibited the most severe sensory phenotypes (mechanical hyperalgesia and sensory loss) revealed diminished structural integrity of the sciatic nerve. This structural decline appears to precede the sensory loss observed in peripheral nerves, highlighting the potential of MRN as a tool for early detection and monitoring of nerve integrity in the context of diabetes-associated complications41.

  • Regarding CKD, our longitudinal analysis indicated a progression of CKD over a 4-year period. Specifically, we observed that 6% of individuals with PRED experienced CKD progression, while 12% of individuals with T2D without previously known CKD also showed similar progression. Notably, 46% of individuals with T2D who had existing CKD experienced a worsening of their condition38. Furthermore, we found that albuminuria could be temporarily improved through dietary interventions in individuals with T2D and known CKD. This improvement in albuminuria was associated with significant changes in parameters related to fatty acid oxidation and was not sustained after the end of dietary intervention. These findings indicate that dietary management could play a role in the treatment of renal complications in DM33.

  • Regarding MASLD, a 4-year longitudinal analysis showed that 24% of individuals with PRED experienced progression, indicated by worsening of liver stiffness. This was also observed in 19% of participants with T2D without increased liver stiffness at baseline and 15% of those with T2D and increased liver stiffness at baseline. Interestingly, 36% of the participants without known diabetes-associated complications also exhibited increased liver stiffness38.

Assessment of non-classical diabetes-associated complications

  • Our study identified a significant increase in breathlessness and a notable prevalence of RLD among individuals with long-term T2D (27%). In comparison, 21% of individuals with recent-onset T2D and 9% of individuals with PRED showed similar patterns of lung disease49. To confirm the presence of interstitial lung disease, we utilized multidetector computed tomography, and histological analysis revealed evidence of fibrosis in the lungs of individuals with T2D49. These findings align with previous studies that have linked a decline in pulmonary function with T2D119,120. Our results underscore the importance of recognizing diabetic RLD as a critical component in the standard care for managing diabetes-associated complications50.

  • We showed that that DSPN does not solely affect the lower limbs; it also significantly impacts the upper limbs, making them susceptible to diabetes-induced damage. The sensory phenotype observed in individuals with upper limb neuropathy closely resembles that of lower limb neuropathy, primarily characterized by loss of sensory function107.

  • We found that diabetes-related distress was significantly associated with lower glycemic control, higher insulin resistance, and longer DM duration in individuals with T2D13. Additionally, we demonstrated an impairment in the autonomic nervous system’s response to stress, which is partially reflected in the psychological stress response12. Importantly, we found a notable association between moderate to severe childhood neglect and an intensified psychological stress response in individuals with T2D. This connection highlights the interplay between psychological factors, physiological stress responses, and childhood experiences in the context of DM management, suggesting the need for comprehensive approaches addressing both emotional and physical health in T2D individuals105.

Identification of biomarkers for diabetes-associated complications

  • In our study, we identified circulating mRNA levels of myelin protein zero as a novel potential biomarker associated with DSPN, which together with the already reported biomarker for DSPN, neurofilament light chain protein, were found changed in our cohort; specifically, decreased levels of myelin protein zero were predictive of hypoalgesia, while increased levels of neurofilament light chain were linked to a hyperalgesia phenotype57. Furthermore, neurofilament light chain protein levels correlated with sensorimotor deficits in both the upper and lower limbs in individuals with T2D58. These findings suggest that these biomarkers could be valuable in assessing neuropathic pain and sensory function in diabetic individuals.

  • We demonstrated that the bioelectrical phase angle derived from BIA can serve as a straightforward tool for assessing cardiovascular risk and detecting DSPN in individuals with and without DM52,53. This finding highlights the phase angle’s potential utility in clinical settings for early identification of individuals at risk for cardiovascular complications and neuropathy, aiding in timely interventions and management strategies.

Novel cellular mechanisms linked to diabetes-associated complications and precise interventions

The traditional glucose-centric approach to diabetes management, which focused on lowering glycemia, has been increasingly challenged, leading to a shift toward a holistic, patient-centered approach aimed at reducing the risks of diabetes-related complications121. This framework acknowledges the complexity of diabetes as a chronic disease and emphasizes the importance of a pathogenesis-centric strategy, targeting the underlying pathophysiological mechanisms that both cause and complicate diabetes and its associated complications122.

  • Cross-sectional analysis of individuals with PRED and T2D showed a strong association between the presence of CKD, RLD phenotype, and increased liver stiffness with elevated markers for DNA damage, senescence, and senescence-associated secretory phenotype (SASP)38. Furthermore, longitudinal analysis over four years revealed that the progression of CKD was significantly predicted by these markers of DNA damage, senescence, and SASP. In contrast, the progression of RLD was primarily associated with increased DNA damage and elevated levels of interleukin-6 (IL-6)38. These findings suggest that cellular aging plays a critical role in the complications observed in individuals with diabetes.

  • In the HEIST-DiC cohort, we successfully validated a novel p21-dependent mechanism of tubular senescence previously identified in animal models, which contributes to hyperglycemic memory in the context of CKD and T2D109,115. We observed that tubular and urinary p21-levels from individuals with T2D were significantly associated with the severity of CKD. Notably, these p21 levels remained elevated even when blood glucose levels improved through treatment with SGLT-2 inhibitors or dietary interventions33,109,115. This persistence of p21 elevation, despite better glycemic control, underscores the potential for tubular senescence to play a pivotal role in the progression of CKD.

  • We discovered elevated hydroxyacetone levels and increased activity of aldo-keto-reductase in red blood cells of T2D individuals, as markers associated with compensatory mechanisms for methylglyoxal detoxification. These markers may serve as useful indicators to distinguish between T2D individuals with and without complications. Specifically, individuals with T2D who do not exhibit complications appear to retain protective alternative detoxification pathways for methylglyoxal, which likely contributes to their better health profile. Conversely, those with T2D complications show a loss of these protective mechanisms, suggesting that impaired detoxification processes may contribute to the progression of diabetes-associated complications91.

  • In a subgroup of individuals with T2D and CKD we conducted a randomized-controlled trial focusing on the effects of a diet intervention involving periodic fasting over 6 months. This study aimed to evaluate outcomes related to diabetes-associated complications, particularly the impact on microalbuminuria and somatosensory nerve function33,112. Our findings indicated that improvement of microalbuminuria under periodic fasting was linked to specific changes in acylcarnitine profile33, and had no effect on somatosensory nerve function112. Additionally, the effectiveness of periodic fasting on weight loss and maintenance appeared to be influenced by a genetic polymorphism of p53113. In a following proof-of-concept study we demonstrated that glucose intake during refeeding after periodic fasting led to an increased oxidative stress response in T2D individuals with complications114. T2D individuals without complications seemed to be unaffected by the glucose-induced changes, whereas individuals with NGT experienced enhanced cellular resistance to oxidative stress114. These results suggest that while periodic fasting may offer benefits for managing microalbuminuria in T2D individuals with CKD, the oxidative stress response post-refeeding could vary significantly based on the presence of diabetes-associated complications.

Taken together, our findings are significant as they shift the focus from the traditional glucose-centered hypothesis to a broader understanding of the various mechanisms influencing the development and progression of diabetes-associated complications. By identifying these novel mechanisms, we open avenues for designing targeted interventions aimed at preventing, improving, or even achieving remission of these complications.

What are the main strengths and limitations?

The HEIST-DiC cohort includes individuals across the full spectrum of glucose metabolism, offering a unique opportunity to address critical gaps in diabetes research, such as refining clustering methods, assessing subtype stability, and identifying biomarkers and high-risk individuals. By stratifying participants based on pathophysiological mechanisms, the cohort bridges epidemiological studies and clinical research. This approach will validate existing methods and drive new frameworks, supporting tailored therapeutic strategies and advancing precision medicine in diabetes care. The main strengths of this cohort include: (i) the inclusion of individuals exhibiting a broad diversity in metabolic and clinical stages, providing in this way a comprehensive representation of clinical trajectories of PRED, T1D and T2D; (ii) extensive metabolic and clinical phenotyping for classical and non-classical diabetes-associated complications; and (iii) an extended follow-up period of 11 years. Consequently, the HEIST-DiC cohort presents the opportunity to examine the natural course of DM while simultaneously investigating the onset and progression of diabetes-associated complications and differentiating between fast and slow progressors.

The comprehensive metabolic and clinical phenotyping of diabetes-associated complications has enabled us to implement novel methods by flexibly adapting the study protocol to current findings and refining the focus based on these insights. For instance, in December 2016, we incorporated the 6-minute walk test to assess emerging lung complications related to diabetes. Hand nerve conduction velocity was added in February 2021 in response to emerging evidence of impaired nerve function in the upper extremities. In June 2023, transcutaneous electrical nerve fiber stimulation was introduced to evaluate small nerve fibers (Table 2). The method we employ selectively assesses C-nociceptor excitability through slow, long-duration pulses at low frequency123,124. Alternative novel techniques utilizing short, high-frequency rectangular pulses, primarily assessing A-fiber function, should be considered in future protocol adaptations to allow for more comprehensive assessment of peripheral sensory function125.

This thorough characterization allows to design precise intervention studies aimed at prevention, improvement or ideally, remission of diabetes-associated complications.

Another notable strength is provision and explanation of detailed results to study participants, which they can share with their healthcare providers. This practice keeps participants informed about their health status and enhances their compliance, further supported by regular educational events organized by the study team on DM awareness and information. On the other hand, the communication of the detailed results to the healthcare providers may prompt adjustments in treatment strategy. Additionally, the annual study visits may increase personal motivation and adherence to therapeutic recommendations, thereby incorporating intervention elements compared to study participants.

HEIST-DiC provides more intensive assessments with annual clinical visits, compared to the German Diabetes Study (GDS), which conducts clinical evaluations every five years71. Unlike the Diabetes Prospective Follow-up Registry (DPV)—a registry collecting demographic, clinical, and treatment data for T1D and T2D since 1995—HEIST-DiC offers detailed insights into diabetes-related complications using advanced methods available only in specialized university research centers126. While the Verona Diabetes Study and the Diabetes Education and Self-Management for Ongoing and Newly Diagnosed (DESMOND) program focus on mortality and self-management, their follow-up periods are shorter, at five years and three years respectively127,128. While the current follow-up period is set at 11 years, future extensions will be implemented in response to emerging scientific questions or the integration of novel methodologies.

While predicting long-term disease trajectories in individuals aged 85 and older may have limited applicability, choosing a high upper age limit ensures inclusivity and better reflects the real-world population seen in clinical practice. Older adults are frequently underrepresented in clinical research, despite bearing a significant burden of chronic diseases such as T2D. Including them allows to capture age-related variability and potentially identify important trends within this population. Furthermore, with regard to diabetes subtypes, individuals with mild age-related diabetes are typically older and present with milder metabolic disturbances compared to other subtypes14,37. This approach opens the possibility of identifying a subgroup of individuals with diabetes who are less susceptible to complications and may achieve healthy aging despite the disease. Establishing and further characterizing this subgroup using advanced analytical tools represents an important goal for future research.

As the majority of participants are recruited from our outpatient university clinic, a potential source of bias may arise from the inclusion of individuals with a longer duration of diabetes and the presence of pre-existing diabetes-associated complications. However, these individuals provide valuable insights into the development of DM and the associated complications, without the beforementioned intervention elements inflicted by the monitoring during a study.

One limitation of the HEIST-DiC study is the insufficient representation of diverse ethnicities, as it exclusively includes individuals residing in Germany and does not record ethnicity data. Additionally, our cohort comprises more female participants in the NGT group, which is currently addressed by strategic recruitment of more male participants.

Given that the HEIST-DiC study is a hypothesis-generating cohort study, the initially conducted sample size calculations are based on a logistic regression model for a binary outcome with a single binary or continuous covariate. Typically, cohort studies incorporate additional co-variables that may act as confounders, which should be integrated into the regression model. This oversight reduces the statistical power of the analysis, necessitating adjustments to the initially calculated sample sizes using the variance inflation factor and a factor accounting for expected dropout rates129. As Heist-DiC adapts its study protocol in response to emerging findings, the inclusion of new target groups and outcomes will be continuously considered throughout the course of the study. Consequently, the final sample sizes may need to be larger than originally calculated. Heist-DiC was intentionally designed as an open-end hypothesis-generating cohort study, without a fixed participant target or end date, to maintain flexibility. This design also supports the potential inclusion of new target groups, correction of imbalances in existing groups, and replacement of drop-outs—ensuring an active cohort and supporting the study’s function as a recruitment platform for intervention studies.

Study investigators are not required to evaluate clinical findings beyond the parameters of the research question, an approach that is approved by the ethics committee and is clearly explained to participants prior to their inclusion in the study. Assessment of preclinical manifestation of diabetes-associated complications or other comorbidities may not result in immediate changes in clinical management. In case any relevant incidental abnormal finding is detected during magnetic resonance imaging, ultrasound or in blood chemistry, participants are informed and, if requested, the information is reported to their healthcare providers for further evaluation.