1 INTRODUCTION

Diabetic kidney disease remains a major determinant of morbidity and mortality for people with type 1 diabetes, being the leading cause of end stage renal disease (ESRD) and a key risk factor for cardiovascular disease.^{1, 2} This complication is characterized by progressive increases in urinary albumin excretion associated in some populations with initial hyperfiltration and subsequent declines in glomerular filtration rate (GFR), which can occur following different patterns and at variable rates over the lifetime of people with diabetes.^{3, 4}

In adolescents with type 1 diabetes, advanced stages of diabetes kidney disease, such as proteinuria and clinically relevant reductions in GFR are rare compared to adults, whereas earlier stages, such as hyperfiltration and subclinical increases in urinary albumin excretion, reflecting renal functional and structural changes, can be common.^{3, 5, 6} Previous studies in adolescent cohorts with type 1 diabetes have shown rates of hyperfiltration from 10% to 50%,^{3, 7-9} and identified increased GFR during puberty as an independent predictor of albuminuria and later decline in GFR.^{3, 8, 9} Increased GFR in adolescents reflects the effect of hyperglycemia on renal hemodynamics, but is also strongly influenced by pubertal hormonal changes,^{3, 8} which have also been associated with increases in renal size.³

Biomarkers to identify adolescents at risk of renal complications at an early stage are needed, to guide screening strategies as well as to help in stratifying subjects for future intervention trials.¹⁰ Up to now individual or panels of biomarkers for advanced stages of diabetes kidney disease have been mainly investigated in adult populations.¹⁰ It remains unknown if the same biomarkers could have a similar predictive role in younger population with type 1 diabetes and during earlier stages of renal disease.¹⁰

The aims of the present study were: (a) to explore a set of biomarkers, which we recently assessed in adults with type 1 diabetes,¹¹ in relation to changes in GFR and urinary albumin excretion, in a cohort of adolescents with type 1 diabetes, and (b) to compare the associations between biomarkers and renal outcomes between adolescents and adults with type 1 diabetes.

2 METHODS

3 RESULTS

3.1 Biomarkers

Supplementary Table S1 shows the full list of the biomarkers measured, which passed quality control checks and were included in subsequent analyses, with median levels and IQR. Supplementary Figure S1 shows the correlation matrix of the biomarkers with each other and with eGFR and ACR.

3.1.1 Cross-sectional associations between biomarkers and HbA1c at baseline

In regression models adjusted for age, sex and diabetes duration, significant positive associations (at a P value <.002) were found between HbA1c and nine biomarkers, such as kidney injury molecule-1, fibroblast growth factor-21, alpha-1 microglobulin, fibroblast growth factor-23, tissue inhibitor of metalloproteinases 1, clusterin, growth-regulated alpha-protein, interleukin-18, interleukin-1 receptor type 1, whereas the biomarker osteopontin showed a negative association with HbA1c (Supplementary Table S2).

3.1.2 Cross-sectional associations between biomarkers and eGFR

When modeling baseline eGFR, and after adjusting for basic clinical covariates (age, sex, and diabetes duration), three biomarkers reached the statistically significant threshold of P < .002, and all showed a negative association with eGFR: trefoil factor-3 (B coefficient [95%CI]: −0.19 [−0.27, −0.12], P = 7.0 × 10⁻⁷), cystatin C (−0.18 [−0.26, −0.11], P = 5.1 × 10⁻⁶⁾; and beta-2 microglobulin (−0.12 [−0.20, −0.05], P = 1.6 × 10⁻³) (Table 2). The results remained unchanged in models adjusted for baseline ACR and the full clinical covariates (data not shown).

TABLE 2. Cross-sectional associations of each biomarker (considered separately) with baseline eGFR, from liner regression models adjusted for age, sex, and diabetes duration

Biomarker	Coefficient (95% CI)	P-value
Trefoil factor 3	−0.19 (−0.27, −0.12)	7.0 × 10−07
Cystatin-C	−0.18 (−0.26, −0.11)	5.1 × 10−06
Beta-2-microglobulin	−0.12 (−0.20, −0.05)	1.6 × 10−03
CD27 antigen	−0.10 (−0.17, −0.02)	1.2 × 10−02
Latency-associated-peptide	0.09 (0.01, 0.17)	2.4 × 10−02
Growth-regulated alpha protein	0.08 (0.01, 0.16)	3.6 × 10−02
Eotaxin-2	−0.05 (−0.13, 0.02)	1.7 × 10−01
Fibroblast growth factor 21	0.06 (−0.02, 0.14)	1.5 × 10−01
Interleukin-6	−0.06 (−0.13, 0.02)	1.4 × 10−01
Insulin-like growth factor-binding protein 7	−0.04 (−0.12, 0.03)	2.7 × 10−01
Interleukin-1 receptor type 1	0.04 (−0.03, 0.12)	2.5 × 10−01
Lectin-like oxidized LDL receptor 1	−0.05 (−0.12, 0.03)	2.4 × 10−01
Interleukin-2 receptor alpha	0.03 (−0.05, 0.10)	5.3 × 10−01
Neutrophil gelatinase-associated lipocalin	−0.03 (−0.11, 0.04)	3.9 × 10−01
Tamm-horsfall urinary glycoprotein	−0.04 (−0.11, 0.04)	3.6 × 10−01
Alpha-1-microglobulin	0.03 (−0.04, 0.11)	3.8 × 10−01
Clusterin	0.03 (−0.04, 0.11)	4.1 × 10−01
Osteopontin	−0.03 (−0.12, 0.06)	5.1 × 10−01
Tissue inhibitor of metalloproteinases 1	0.03 (−0.04, 0.11)	4.1 × 10−01
Fibroblast growth factor 23	0.01 (−0.06, 0.09)	7.6 × 10−01
Interleukin-1 receptor type 2	−0.02 (−0.09, 0.06)	7.0 × 10−01
Vascular endothelial growth factor	0.02 (−0.05, 0.10)	5.3 × 10−01
Kidney injury molecule-1	−0.01 (−0.08, 0.07)	8.6 × 10−01
Interleukin-18	0.00 (−0.08, 0.07)	9.3 × 10−01
Interleukin-8	0.00 (−0.07, 0.08)	9.4 × 10−01

• Note: Regression coefficients are per unit of SD of gaussianised biomarker.
• Abbreviation: eGFR, estimated glomerular filtration rate.

3.1.3 Biomarkers associated with achieved eGFR and eGFR changes during follow up

Final achieved eGFR was inversely associated with two biomarkers (P < .002): osteopontin (−0.21 [−0.28, −0.14], P = 2.3 × 10⁻⁸) and cystatin C (−0.16 [−0.22, −0.09]), P = 1.6 × 10⁻⁶), after adjustments for the basic clinical covariates (age, sex and diabetes duration), baseline eGFR, and length of follow-up (Table 3). These results remained unchanged in models adjusted for baseline ACR, the full set of clinical covariates and treatment (data not shown).

TABLE 3. Association of each biomarker with achieved eGFR from liner regression models adjusted for age, sex, diabetes duration, baseline eGFR, and length of follow up

Biomarker	Coefficient (95%CI)	P-value
Osteopontin	−0.21 (−0.28, −0.14)	2.3 × 10−08
Cystatin-C	−0.16 (−0.22, −0.09)	1.6 × 10−06
Insulin-like growth factor-binding protein 7	−0.10 (−0.16, −0.03)	3.1 × 10−03
CD27 antigen	−0.08 (−0.15, −0.02)	7.2 × 10−03
Fibroblast growth factor 23	0.08 (0.02, 0.14)	9.6 × 10−03
Growth-regulated alpha protein	0.08 (0.02, 0.15)	9.8 × 10−03
Fibroblast growth factor 21	0.07 (0.01, 0.13)	3.0 × 10−02
Interleukin-2 receptor alpha	−0.07 (−0.13, −0.01)	2.8 × 10−02
Beta-2-microglobulin	−0.06 (−0.12, 0.00)	5.8 × 10−02
Kidney injury molecule-1	0.04 (−0.02, 0.10)	2.3 × 10−01
Trefoil factor 3	−0.04 (−0.10, 0.03)	2.4 × 10−01
Eotaxin-2	−0.03 (−0.09, 0.03)	2.9 × 10−01
Interleukin-6	0.03 (−0.03, 0.09)	3.4 × 10−01
Lectin-like oxidized LDL receptor 1	0.03 (−0.03, 0.10)	3.0 × 10−01
Alpha-1-microglobulin	−0.02 (−0.08, 0.04)	5.5 × 10−01
Interleukin-18	0.01 (−0.05, 0.08)	6.7 × 10−01
Interleukin-1 receptor type 1	−0.02 (−0.09, 0.04)	4.9 × 10−01
Latency-associated-peptide	−0.02 (−0.08, 0.04)	5.5 × 10−01
Tissue inhibitor of metalloproteinases 1	0.02 (−0.05, 0.08)	6.1 × 10−01
ascular endothelial growth factor	0.02 (−0.04, 0.08)	5.0 × 10−01
lusterin	−0.01 (−0.07, 0.06)	8.7 × 10−01
Interleukin-1 receptor type 2	−0.01 (−0.08, 0.05)	7.0 × 10−01
Interleukin-8	0.00 (−0.06, 0.07)	9.5 × 10−01
Neutrophil gelatinase-associated lipocalin	0.00 (−0.06, 0.07)	9.5 × 10−01
Tamm-horsfall urinary glycoprotein	0.00 (−0.06, 0.06)	9.5 × 10−01

• Note: Regression coefficients are per unit of SD of gaussianised biomarker.
• Abbreviation: eGFR, estimated glomerular filtration rate.

In logistic regression models, the rapid decliner phenotype was positively associated to osteopontin with an odds ratio of 1.83 [95% CI 1.42, 2.41] (P = 7.3 × 10⁻⁶), adjusted for the basic clinical covariates, baseline eGFR, and length of follow-up (Table 4). In similar logistic regression models, the rapid increaser phenotype was associated with fibroblast growth factor-23 (1.59 [1.23, 2.04], P = 2.6 × 10⁻⁴) and inversely associated with osteopontin (0.58 [1.42, 0.80], (P = 2.6 × 10⁻⁴)) (Table 4).

TABLE 4. Association of each biomarker with rapid decliner (A) and rapid increaser (B) phenotypes from logistic regression models adjusted for age, sex, diabetes duration, baseline eGFR and length of follow up

A
Biomarker	Odds ratio (95% CI)	P-value
Osteopontin	1.83 (1.42, 2.41)	7.3 × 10−06
Cystatin-C	1.26 (1.02, 1.56)	3.0 × 10−02
Insulin-like growth factor-binding protein 7	1.25 (1.02, 1.55)	3.5 × 10−02
Growth-regulated alpha protein	0.80 (0.65, 0.98)	2.8 × 10−02
eutrophil gelatinase-associated lipocalin	0.82 (0.67, 1.00)	5.6 × 10−02
Fibroblast growth factor 21	0.84 (0.69, 1.02)	8.3 × 10−02
Interleukin-1 receptor type 2	1.19 (0.98, 1.46)	8.1 × 10−02
Lectin-like oxidized LDL receptor 1	0.85 (0.70, 1.04)	1.2 × 10−01
Tissue inhibitor of metalloproteinases 1	0.86 (0.70, 1.04)	1.2 × 10−01
Tamm-horsfall urinary glycoprotein	1.16 (0.95, 1.42)	1.4 × 10−01
Fibroblast growth factor 23	0.87 (0.71, 1.06)	1.6 × 10−01
Interleukin-18	0.89 (0.73, 1.09)	2.7 × 10−01
Interleukin-8	0.89 (0.73, 1.09)	2.6 × 10−01
Interleukin-6	0.91 (0.75, 1.11)	3.6 × 10−01
Interleukin-2 receptor alpha	1.08 (0.88, 1.32)	4.8 × 10−01
Alpha-1-microglobulin	0.93 (0.76, 1.14)	4.9 × 10−01
Beta-2-microglobulin	0.93 (0.75, 1.13)	4.7 × 10−01
Vascular endothelial growth factor	0.95 (0.78, 1.16)	6.0 × 10−01
Trefoil factor 3	0.95 (0.78, 1.16)	6.3 × 10−01
Latency-associated-peptide	1.05 (0.86, 1.28)	6.6 × 10−01
CD27 antigen	1.03 (0.85, 1.26)	7.5 × 10−01
Clusterin	0.99 (0.81, 1.20)	8.8 × 10−01
Eotaxin-2	0.99 (0.81, 1.21)	9.4 × 10−01
Interleukin-1 receptor type 1	1.01 (0.82, 1.23)	9.6 × 10−01
Kidney injury molecule-1	1.00 (0.82, 1.22)	9.9 × 10−01

B
Biomarker	Odds ratio (95% CI)	P-value
Osteopontin	0.58 (0.42, 0.80)	9.5 × 10−04
Fibroblast growth factor 23	1.59 (1.23, 2.04)	2.6 × 10−04
Lectin-like oxidized LDL receptor 1	1.55 (1.11, 2.16)	9.6 × 10−03
otaxin-2	0.66 (0.47, 0.91)	1.4 × 10−02
Cystatin-C	0.69 (0.49, 0.97)	3.2 × 10−02
Growth-regulated alpha protein	1.42 (1.03, 1.95)	3.2 × 10−02
Interleukin-18	1.36 (0.99, 1.89)	6.0 × 10−02
Neutrophil gelatinase-associated lipocalin	1.38 (0.99, 1.92)	5.6 × 10−02
Alpha-1-microglobulin	1.31 (0.92, 1.93)	1.5 × 10−01
Fibroblast growth factor 21	1.25 (0.89, 1.78)	2.0 × 10−01
Vascular endothelial growth factor	1.26 (0.92, 1.75)	1.5 × 10−01
Clusterin	1.26 (0.91, 1.74)	1.6 × 10−01
Interleukin-1 receptor type 1	1.22 (0.88, 1.72)	2.5 × 10−01
Interleukin-2 receptor alpha	0.81 (0.57, 1.13)	2.2 × 10−01
Insulin-like growth factor-binding protein 7	0.84 (0.61, 1.13)	2.6 × 10−01
Interleukin-6	1.21 (0.89, 1.54)	1.5 × 10−01
CD27 antigen	0.89 (0.64, 1.22)	4.8 × 10−01
Latency-associated-peptide	0.91 (0.70, 1.24)	5.3 × 10−01
Kidney injury molecule-1 simoa	1.09 (0.79, 1.47)	5.8 × 10−01
Trefoil factor 3	1.09 (0.78, 1.52)	6.0 × 10−01
Interleukin-8	1.07 (0.75, 1.46)	6.9 × 10−01
Tissue inhibitor of metalloproteinases 1	0.97 (0.69, 1.35)	8.8 × 10−01
Tamm-horsfall urinary glycoprotein	1.03 (0.76, 1.42)	8.3 × 10−01
Beta-2-microglobulin	0.98 (0.70, 1.39)	9.2 × 10−01
Interleukin-1 receptor type 2	0.99 (0.73, 1.37)	9.4 × 10−01

• Note: Odds ratios are per unit of SD of gaussianised biomarker.

When considering the rapid decliners phenotype as outcome, the forward selected panel only included osteopontin and improved the AUC beyond basic clinical covariates from 0.689 to 0.717. For the rapid increaser phenotype, forward selection chose in order osteopontin, interleukin-18, eotaxin-2 and fibroblast growth factor-23, thus marginally improving the AUC from 0.645 to 0.679.

3.1.4 Comparison with an adult cohort with type 1 diabetes

When comparing the AdDIT biomarker results with those obtained in the adult SDRNT1BIO cohort (Supplementary Table S3), the negative associations between baseline eGFR and trefoil factor-3, cystatin C, and beta-2 microglobulin were also present in the adult cohort. Similarly, when comparing the results for the final achieved eGFR, ostepontin and cystatin C were also negatively associated with this outcome in the adult cohort. However, in the adult cohort several additional biomarkers emerged to be associated with eGFR, among them alpha 1-microglobulin, CD27 antigen, kidney injury molecule-1, which were not identified in the younger AdDIT cohort (Figure 1A,B).

No associations were found between biomarkers and ACR at baseline or during follow-up in the AdDIT cohort, whereas several biomarkers were associated with ACR in the adult cohort (Figure 1C).

4 DISCUSSION

In this population of adolescents with type 1 diabetes, predominantly normoalbuminuric at baseline, and showing high rates of hyperfiltration, trefoil factor-3, cystatin C and beta-2 microglobulin were negatively associated with eGFR at baseline, whereas osteopontin and cystatin C were associated with faster eGFR declines, and fibroblast growth factor-23 with more rapid increases in eGFR over a median follow up period of 3.9 years.

In adolescents with type 1 diabetes, changes in GFR, mainly in the form of hyperfiltration, are common, and they are partly related to the diabetes milieu, with hyperglycemia being a main driver for hyperfiltration, together with the effect of pubertal physical and hormonal changes.^{3, 8} In the present study, high rates of hyperfiltration were evident at baseline, and they persisted during the study period. Decreased levels of trefoil factor-3, cystatin C, and beta-2 microglobulin were associated with higher eGFR levels at baseline, likely reflecting that these are renal biomarkers of filtration.

Beta-2 microglobulin is considered a surrogate marker for eGFR being freely filtered by the glomerulus and then almost completely reabsorbed by the proximal tubule.¹⁵ In adults with type 2 diabetes, beta-2 microglobulin has emerged as a predictor of renal function decline^{16, 17} and ESRD.¹⁸ In the present study, we could not confirm an association between beta-2 microglobulin and changes in eGFR during follow-up, and this might be due to differences in the age range and/or diabetes duration of our study population and the relatively preserved eGFR during follow-up compared to participants recruited in previous studies, or alternatively differences between subjects with type 1 and type 2 diabetes.

The other biomarker associated with eGFR at baseline was trefoil factor-3, but again this biomarker was not associated with prospective changes in eGFR. In a recent study, based on a Mendelian randomization approach, trefoil factor-3 emerged as a valuable diagnostic marker for early stages of kidney diseases in individuals with type 2 diabetes as well as in those with prediabetes, and levels of this biomarker increased as a result of decreased eGFR.¹⁹ Trefoil factor-3 also improved the discrimination for chronic kidney disease in addition to eGFR alone, highlighting its potential as a predictive biomarker.¹⁹

In the present study, some other interesting associations were found between some of the biomarkers measured at baseline and prospective changes in eGFR. One biomarker associated with changes in eGFR during follow-up was osteopontin, with increased levels associated with greater declines in eGFR and a lower eGFR by the end of the follow-up period. Osteopontin is a phosphorylated sialic acid-rich non-collagenous matricellular protein, named for its function as a bridge between cells and minerals in the bones.²⁰ This protein has been associated with cardiovascular disease, cancer, diabetes, and kidney stone diseases and is implicated in the process of inflammation, biomineralization, cell viability, and wound healing.²⁰ Circulating osteopontin levels are increased in patients with diabetes compared to healthy controls and are related to insulin resistance, inflammation, vascular calcification.²⁰ Studies in adults with type 2 diabetes have reported an association with retinopathy as well as with progression of diabetic nephropathy and cardiovascular disease.^{21, 22} In type 1 diabetes, data from the large FinnDiane study have shown that osteopontin is an independent predictor of incident microalbuminuria and cardiovascular events.²³ In a small cross-sectional pediatric study, osteopontin was found to be elevated in children and adolescents with type 1 diabetes compared to healthy controls, and to be related to subclinical signs of vascular disease, such as increased intima-media thickness, as well as to urinary albumin excretion.²⁴ These data suggest that osteopontin could be a potential common biomarker for renal and cardiovascular complications, reflecting similar mechanisms underlining these complications.

Cystatin C was also found to be a predictor of decline in eGFR during the study period as already well described in adults.²⁵ In addition, higher levels of fibroblast growth factor-23 were associated with a higher eGFR increases over time. Fibroblast growth factor-23 is a known key regulator of phosphorus homeostasis and it has also been suggested to be implicated in inflammation, immunosuppression, and cardiovascular disease.²⁶ Of interest, it has also emerged as a new risk factor in chronic kidney disease, being an independent predictor of progression of chronic kidney disease in people without diabetes.²⁷ In patients with type 2 diabetes and macroalbuminuria, fibroblast growth factor-23 has been associated with renal outcomes,^{27, 28} although there is a need for additional studies in type 1 diabetes to explore its longitudinal changes and potential value as a biomarker of renal function.

Some interesting findings of the present study emerged from the comparisons of the results of the same set of biomarkers in an adult cohort with type 1 diabetes, the SDRNT1BIO.^{11, 29} Osteopontin and cystatin C were also found to be negatively associated with eGFR in the adult cohort. However, in the same adult cohort, additional biomarkers were associated with eGFR at baseline and predict its prospective changes, and some biomarkers even showed an opposite association compared with the adolescent cohort. These discordant findings between adults and adolescents with type 1 diabetes could be related to differences in the age and diabetes duration between the two study populations, or in the ontology of changes in renal function through adolescence into adult life. Regarding decline in eGFR, the fall detected in our adolescent cohort mainly reflected a reduction of the initial hyperfiltration and therefore a return of GFR to within a normal range, as confirmed by the higher eGFR at baseline in the “rapid decliner” group compared to the rest of the study population. This could be considered a physiological decline, occurring mainly once pubertal development has been completed,³ and could reflect a different underlining process than the one leading to the decline occurring in adult cohorts. The present cohort of adolescents also differed from adult cohorts with type 1 diabetes for the presence of a subgroup with rapid increasing GFR during follow up. This subgroup was characterized by a greater number of females, higher HbA1c at baseline, longer diabetes duration and higher percentages of smokers. This group confirms that hyperfiltration is a common feature in adolescents with type 1 diabetes whereas it has less often observed in adult cohorts.⁹

In this young cohort of adolescents with type 1 diabetes, there were no significant associations between the assessed biomarkers and ACR. This was in contrast with the results from the adult cohort where the same panel of biomarkers was evaluated. These differences are likely explained by the fact that the AdDIT cohort was predominantly normoalbuminuric at baseline.

It needs to be acknowledged that, in the present study, we assessed changes in renal function during a short period of follow up and, therefore power to detect associations with eGFR decline was limited. Only longer-term follow-up will provide further insights on the potential predictive values of the assessed biomarkers in relation to hard renal outcomes and allow better comparisons with adult cohorts.

Another potential study limitation was the use of an estimate of GFR to assess renal function. However, this was based on a creatinine formula derived from our own laboratory from a population of adolescents with type 1 diabetes. This formula can be easily implemented into clinical practice as well as for repeated measurements of GFR, being based on the routinely measured creatinine. However, this formula differs from that used in SDRT1BIO adult cohort, which was one of the standard formulas recommended in adults. This highlights a common issue in longitudinal studies, when pediatric participants transition into adulthood, with the need of changing the formula used to estimate GFR and a potential bias due to differences in formulas performance.

5 CONCLUSION

In this young population with type 1 diabetes and high rates of hyperfiltration, osteopontin was the most consistent biomarker associated with prospective changes in eGFR, although there was only a modest increment in predicting future eGFR changes over clinical variables and baseline eGFR and ACR. Fibroblast growth factor-23 was associated with rapid eGFR increases, whereas trefoil factor-3, cystatin C, and beta-2 microglobulin were only associated with eGFR at baseline.

Further follow up of the AdDIT cohort will allow us to explore the potential predictive value of these biomarkers on long-term changes in eGFR and ACR.

STUDY GROUPS

Adolescent type 1 Diabetes cardio-renal Intervention Trial (AdDIT) study group.

Carlo L. Acerini (University of Cambridge, Cambridge, UK), Frank Ackland (Northampton General Hospital, Northampton, UK), Binu Anand (West Suffolk Hospital, NHS Foundation Trust, Bury St Edmunds, UK), Tim Barrett (Birmingham Children's Hospital and University of Birmingham, Birmingham, UK), Virginia Birrell (James Cook Hospital, South Tees Hospitals NHS Foundation Trust, Middlesbrough, UK), Fiona Campbell (Leeds General Infirmary, The Leeds Teaching Hospitals NHS Trust, Leeds, UK), Marietta Charakida (King's College London, London, UK), Tim Cheetham (Royal Victoria Infirmary, Newcastle upon Tyne Hospitals NHS Foundation Trust and Newcastle University, Newcastle, UK), Scott T. Chiesa (University College London, London, UK), John E. Deanfield (University College London, London, UK), Chris Cooper (Stepping Hill Hospital, Stockport NHS Foundation Trust, Stockport, UK), Ian Doughty (Royal Manchester Children's Hospital, Manchester, UK), Atanu Dutta (Stoke Mandeville Hospital, Aylesbury, UK), Julie Edge (John Radcliffe Hospital, Oxford, UK), Alastair Gray (University of Oxford, Oxford, UK), Julian Hamilton-Shield (University of Bristol and University Hospitals Bristol NHS Foundation Trust, Bristol, UK), Nick Mann (Royal Berkshire Hospital, Reading, UK), M. Loredana Marcovecchio (University of Cambridge, Cambridge, UK), Sally M. Marshall (Newcastle University), H. Andrew W. Neil (University of Oxford, Oxford, UK), Gerry Rayman (Ipswich Hospitals NHS Trust, Ipswich, UK), Jonathon M. Robinson (Royal Albert Edward Infirmary, Wrightington, Wigan and Leigh NHS Foundation Trust, Wigan, UK), Michelle Russell-Taylor (Wycombe Hospital, Buckingham Healthcare NHS Trust, High Wycombe, UK), Vengudi Sankar (Royal Bolton Hospital, Bolton NHS Foundation Trust, Bolton, UK), Anne Smith (Northampton General Hospital, Northampton, UK), Nandu Thalange (Norfolk and Norwich University Hospitals NHS Foundation Trust, Norwich, UK), Chandan Yaliwal (Royal Berkshire Hospital, Reading, UK); Paul Benitez-Aguirre (The Children's Hospital at Westmead and University of Sydney, Sydney, NSW, Australia), Fergus Cameron (Royal Children's Hospital, Murdoch Children's Research Institute and The University of Melbourne, Melbourne, VIC, Australia), Andrew Cotterill (University of Queensland, Brisbane, QLD, Australia), Jennifer Couper (Women's and Children's Hospital and University of Adelaide, Adelaide, SA, Australia), Maria Craig (The Children's Hospital at Westmead, University of Sydney, and University of New South Wales, Sydney, NSW, Australia), Elizabeth Davis (Perth Children's Hospital for Children and University of Western Australia, Perth, WA, Australia), Kim Donaghue (The Children's Hospital at Westmead and University of Sydney, Sydney, NSW, Australia), Timothy W. Jones (Princess Margaret Hospital for Children and University of Western Australia, Perth, WA, Australia), Bruce King (University of Newcastle, Newcastle, NSW, Australia), Charles Verge (Sydney Children's Hospital and University of New South Wales, Sydney, NSW, Australia), Phil Bergman (Monash Children's Hospital, Clayton, VIC, Australia), Christine Rodda (University of Melbourne, Melbourne, VIC, Australia); Cheril Clarson (London Health Sciences Centre and Western University, London, ON, Canada); Jacqueline Curtis (The Hospital for Sick Children and University of Toronto, Toronto, ON, Canada), Denis Daneman (The Hospital for Sick Children and University of Toronto, Toronto. ON, Canada), Farid H. Mahmud (The Hospital for Sick Children and University of Toronto, Toronto, ON, Canada); Etienne Sochett (The Hospital for Sick Children and University of Toronto, Toronto, ON, Canada).

SDRN TYPE 1 BIORESOURCE INVESTIGATORS

John Chalmers (Diabetes Centre, Victoria Hospital, U.K), Andrew Collier (Diabetic Day Centre University Hospital, UK), Colin Fischbacher (NHS, UK), Fiona Green (Research & Development Support Unit, Dumfries & Galloway Royal Infirmary, UK), Robert Lindsay (British Heart Foundation Glasgow Cardiovascular Research Centre, University of Glasgow, UK), John McKnight (Western General Hospital, NHS, UK), Sandra MacRury (Highland Diabetes Institute, Raigmore Hospital, NHS Highland, Inverness, UK), Colin Palmer (Cardiovascular and Diabetes Medicine, University of Dundee, UK), Alan Patrick (Royal Infirmary of Edinburgh, NHS Lothian, UK, Donald Pearson (JJR Macleod Centre for Diabetes, Endocrinology and Metabolism, Aberdeen Royal Infirmary, UK), John Petrie (Institute of Cardiovascular & Medical Sciences, University of Glasgow, UK) and Sandeep Thekkepat (David Matthews Diabetes Centre, Monklands Hospital, UK).

CONFLICT OF INTEREST

H. M. C. reports grants and personal fees from Eli Lily and Company, during the conduct of the study. Grants from AstraZeneca LP, other from Novartis Pharmaceuticals, grants from Regeneron, grants from Pfizer Inc, other from Roche Pharmaceuticals, other from Sanofi Aventis, grants and personal fees from Novo Nordisk, outside the submitted work. The other coauthors do not have any duality of interest associated with their contribution to this manuscript.

AUTHOR CONTRIBUTIONS

Maria Loredana Marcovecchio researched and analyzed data, wrote the first draft of the manuscript, contributed to discussion; Marco Colombo analyzed the data, contributed to critical review of the manuscript; David B. Dunger, Helen M. Colhoun, Raymond Neil Dalton conceived the study and obtained funding for the study; Paul Benitez-Aguirre, Fergus J. Cameron, Scott T. Chiesa, Jennifer J. Couper, Maria E. Craig, Denis Daneman, Elizabeth A. Davis, John E. Deanfield, Kim Donaghue, Timothy W. Jones, Farid H. Mahmud, contributed to data acquisition, critical review of the manuscript; Sally M. Marshall and Andrew Neil contributed to data interpretation and critical review of the manuscript; David B. Dunger, Helen M. Colhoun, Raymond. Neil Dalton, Paul M. McKeigue contributed to the study concept and design, data interpretation and critical review of the manuscript. All authors reviewed and edited the drafts and approved the final manuscript.

FUNDING

This study was supported by funding from Juvenile Diabetes Research Foundation (Ref. 1-SRA-2016-333-M-R). The AdDIT study was funded by Diabetes UK, Juvenile Diabetes Research Foundation, the British Heart Foundation and in Canada the JDRF-Canadian Clinical Trial Network (CCTN), the Canadian Diabetes Association and the Heart and Stroke Foundation Canada. The SDRNT1BIO cohort was funded by Chief Scientists Office Scotland and Diabetes UK.