ObjectiveTo identify the factors associated with residual C peptide production at least 10 years after diagnosis in children and adolescents with type 1 diabetes.Research design and methods73 children and adolescents (<25 years), born in 1988–2005, diagnosed with type 1 diabetes were included during the 4-year study period (2013–2016). At least 10 years after diagnosis, we measured any remaining C peptide concentration using an ultrasensitive C peptide ELISA (≥1.17 pmol/L). The average hemoglobin A1c (HbA1c) was calculated during each of the 10 years after diagnosis and further grand average was calculated for the entire study period.ResultsC peptide was detectable in 38% of participants. The C peptide concentration was 4.3±5.3 pmol/L. At onset of type 1 diabetes, participants were on average approximately 5 years of age, and their average HbA1c was 9.4% (79 mmol/mol). During the first 3 years after diagnosis, HbA1c was lower in the group with detectable C peptide at follow-up ≥10 years later. Moreover, detectable C peptide was more common among female participants. Body mass index SD scores had not increased since the 1-year follow-up, but were higher in patients with measurable C peptide. Nine participants (12%) had been diagnosed with celiac disease and two (3%) with hypothyreosis. Eighteen (25%) participants had retinopathy.ConclusionsChildren and adolescents with detectable C peptide after more than 10 years of diabetes duration were predominantly female and had better HbA1c than others during the first 3 years after diagnosis.
The development of ultrasensitive methods for C peptide analysis has led to findings that show glucose-responsive residual insulin production may occur for decades after diagnosis of type 1 diabetes.1–3 Although most patients 10 years after onset of disease do not have levels of residual insulin secretion high enough to lower mean blood glucose concentration and provide less blood glucose fluctuations,4 surviving immune-resistant beta cells are an attractive target for new protocols aiming to expand beta cell mass.
There is little knowledge on the factors that correlate to a long-term preserved residual beta cell mass in type 1 diabetes. However, we recently identified immunological factors, particularly higher production of interleukin-35 by immune cells, in patients with residual C peptide for more than 10 years.5 Also the higher frequency of the HLADR3 genotype1 and the level of ZnT8 autoantibodies2 have been correlated with C peptide concentrations remaining for decades.
The present study took advantage of the unique longitudinal follow-up data of the Swedish Childhood Diabetes Registry SWEDIABKIDS and combined these with participants’ records, aiming to identify other potential factors, aside from immunological factors, that are associated with long-standing preservation of beta cell function in type 1 diabetes.
Adolescents and young adults with at least 10 years’ duration of type 1 diabetes were recruited from the Uppsala University Children’s Hospital (n=51) and the Department of Endocrinology and Diabetology, Uppsala University Hospital (n=22) in 2013–2016. All included patients were born in 1988–2005 and presented with type 1 diabetes in 1991–2007. All participants had been hospitalized at the Children’s Hospital at onset of disease and were started on multiple insulin injection therapy (MIT), usually with human insulin (Actrapid and Insulatard). During the course of treatment, most of the participants had been switched to insulin aspart (NovoRapid) and insulin detemir (Levemir), or alternatively to insulin pump treatment. Follow-up was done at the Children’s Hospital and according to the framework of the Swedish national program for pediatric diabetes.6 At the age of 18, participants were transferred to the diabetes service of the Department of Endocrinology and Diabetology. The clinical diagnosis of type 1 diabetes did not need to be revised in any of the participants. Any presence of retinopathy was assessed every second year, with assessment becoming more frequent on occurrence of changes and with participants being 10 years of age.
Non-fasting blood samples were obtained during regular follow-up visits to respective diabetes services. Hemoglobin A1c (HbA1c), serum creatinine, blood glucose, glutamic acid decarboxylase antibodies (GADA; detection limit 5 IU/mL) and islet antigen-2 antibodies (IA-2A; detection limit 7.5 kU/L) were analyzed according to standards and routine at the Department of Clinical Chemistry at Uppsala University Hospital. The laboratory is certified by a Swedish government authority (Swedac). To analyze C peptide concentrations, plasma was separated from samples of EDTA blood and stored frozen at −80°C until analysis. Plasma C peptide concentrations were analyzed by Mercodia (Uppsala, Sweden) using an ultrasensitive ELISA (catalog no 10-1141-01; Mercodia). The assay was calibrated against the international reference reagent for C peptide, C peptide 84/510, which is the WHO standard and is categorized by the US Food and Drug Administration as class I in vitro diagnostic device. The detection limit of the assay is set to 1.17 pmol/L, with interassay and intra-assay coefficients of variation of 5.5% and 3.8% at 37 pmol/L.
Data were collected from patient records and the SWEDIABKIDS, a national incidence and quality control register.7
From SWEDIABKIDS, baseline data for each patient at the time of diagnosis of type 1 diabetes were collected. These included age, gender and comorbid diseases diagnosed before the onset of diabetes, weight, height, HbA1c and blood pH at the time of admission, and weight, height, HbA1c, insulin dose (units/kg body weight/day) and use of insulin pump registered at follow-up visits. Body mass index (BMI) (kg/m2) and BMI SD scores (BMI SDS) are generated automatically by the SWEDIABKIDS register.8 The presence of retinopathy and persistent microalbuminuria are also registered.
From October 2010, HbA1c has been analyzed according to the International Federation of Clinical Chemistry (IFCC) standard and expressed as mmol/mol. Prior to this, analyses were according to the Mono S standard and expressed in percent. Analyses performed with the Mono S standard were recalculated using the following expression: HbA1c (IFCC; mmol/mol)=10.45×HbA1c (Mono S; %) – 10.62. Mono S standard was also recalculated according to the Diabetes Control and Complications Trial (DCCT) standard. HbA1c was finally expressed following both DCCT and IFCC standards (http://www.diabetes.co.uk).
For the first 10 years of follow-up, the average HbA1c was calculated for each individual year. For each year, one to six (median 3) HbA1c measurements were usually available. HbA1c grand average was calculated as the average of the 10 individual 1-year averages. A further grand average was calculated for the entire study period when participants had been followed up for more than 10 years.
Statistical analyses were performed using SPSS V.20.0.0. Data are presented as mean±SD. Student’s t-test for independent samples and χ2 test were used to analyze continuous and categorical data, respectively. Predictors of detectable C peptide secretion as a binary outcome (detectable or not) were analyzed using univariate and multivariate logistic regression analyses, expressed with OR and 95% CI. The multivariate models included variables that were statistically significant at the univariate level (gender and celiac disease). BMI SDS was also significant, but due to the high proportion of missing data this was not included in the final model.
At diagnosis of type 1 diabetes, participants were on average approximately 5 years old (table 1) and 55% were male. The average HbA1c at diagnosis was as high as 9.4%±3.8% (79±18 mmol/mol), and six (8%) participants presented with ketoacidosis (pH <7.3). Seven (10%) participants started treatment in an intensive care unit. Three participants had previously been diagnosed with celiac disease and one with hypothyreosis.
|Age (years)||5.1±3.0||0.6 to 12.5|
|BMI SDS||−0.81±1.44||−4.17 to 2.57|
|Blood glucose (mmol/L)||25.1±8.7||7.9 to 58|
|Hemoglobin A1c (%)||9.4±3.8||6.0 to 13.6|
|Hemoglobin A1c (mmol/mol)||79±18||42 to 125|
|Ketoacidosis (pH <7.30)||6 (8%)|
|Intensive care||7 (10%)|
|Previously diagnosed celiac disease||3 (4%)|
|Previously diagnosed hypothyreosis||1 (1%)|
|At 1-year follow-up|
|Age (years)||6.2±3.0||1.6 to 13.7|
|Duration of diabetes (years)||1.0±0.1||0.8 to 1.4|
|BMI SDS||0.60±1.1||−2.35 to 2.71|
|Hemoglobin A1c (%)||6.6±2.9||5.3 to 9.8|
|Hemoglobin (mmol/mol)||49±9||34 to 84|
|Celiac disease||4 (5%)|
|Multiple injection therapy||72 (99%)|
|Insulin pump||1 (1%)|
|Daily insulin dose (U/kg)||0.7±0.2||0.2 to 1.1|
|At long-term follow-up|
|Age (years)||16.9±3.3||10.6 to 24.0|
|Duration of type 1 diabetes (years)||12.3±2.4||10 to 20|
|BMI SDS||0.52±1.06||−2.40 to 2.57|
|Hemoglobin A1c (%)||8.1±3.1||6.1 to 11.2|
|Hemoglobin A1c (mmol/mol)||65±11||43 to 99|
|Celiac disease||9 (12%)|
|Multiple injection therapy||41 (56%)|
|Insulin pump||32 (44%)|
|Daily insulin dose (U/kg)||0.9±0.3||0.5 to 1.7|
|Creatinine (µmol/L)||61±12||37 to 96|
|Blood glucose (mmol/L)||10.6±5.4||2.8 to 21.9|
|C peptide (pmol/L)||1.17±5.31||1.17 to 22.7|
|GADA detectable||32 (44%)|
|IA-2A detectable||23 (32%)|
|GADA and IA-2A antibodies detectable||10 (14%)|
At 1-year follow-up, glycemic control improved, with HbA1c of 6.6%±2.9% (49±9 mmol/mol). All participants but one were on MIT with a daily dose of insulin of 0.7±0.2 (U/kg).
At long-term follow-up, with diabetes duration of 12.3±2.4 years and age of 16.9±3.3 years, HbA1c had increased to 8.1%±3.1% (65±11 mmol/mol). The BMI SDS had not increased since the 1-year follow-up. Nine participants (12%) had now been diagnosed with celiac disease and two (3%) with hypothyreosis. The majority (56%) were on MIT, usually with insulin aspart (NovoRapid) and insulin detemir (Levemir). The daily dose of insulin was 0.9±0.3 (U/kg). Eighteen (25%) participants had retinopathy, with 16 classified as having mild retinopathy and 2 as having moderate retinopathy. Serum creatinine was within the reference range for young adults (men: 60–105 µmol/L; women: 45–90 µmol/L). Urine albumin secretion had not been consistently monitored during the follow-up. GADA was detectable in 32 (44%) participants and IA-2A in 23 (32%), and both GADA and IA-2A were detectable in 10 (14%).
At long-term follow-up, C peptide was detectable (≥1.17 pmol/L) in 28 (38%) participants. In these participants, the C peptide concentration was 4.3±5.3 pmol/L, ranging from 1.17 to 22.7 pmol/L (table 2). C peptide was detectable in a larger proportion of female participants compared with male participants, while age, BMI SDS, HbA1c, presence of ketoacidosis, or GADA and IA-2A at diagnosis were not related to whether C peptide was detectable ≥10 years later (table 2). The daily insulin needs at 1-year follow-up were similar in the groups of participants with or without detectable C peptide at long-term follow-up (0.7±0.2 U insulin/kg vs 0.7±0.2 U insulin/kg). During the course of treatment, HbA1c increased in both groups, but the participants with detectable C peptide at long-term follow-up had lower HbA1c during the first 3 years of treatment (table 2). The participants with detectable C peptide also had higher BMI SDS and celiac disease was more prevalent. There was a tendency of an increased prevalence of retinopathy in those without detectable C peptide (p=0.105). The two participants with retinopathy classified as moderate were found in the latter group.
|Ultrasensitive C peptide (pmol/L)|
|C peptide detectable|
|C peptide not detectable|
|Male/female||11/17 (39%/61%)||29/16 (64%/36%)||0.036*|
|Hemoglobin A1c % (mmol/mol)||9.4±4.0 (79±20)||9.4±3.7 (79±17)||0.917|
|Ketoacidosis (pH <7.30)||1 (4%)||5 (11%)||0.244|
|Average hemoglobin A1c % (mmol/mol)|
|First year||6.1±2.6 (43±6)||6.5±2.8 (49±8)||0.003**|
|Second year||6.6±2.7 (49±7)||7.0±2.8 (53±8)||0.029*|
|Third year||6.8±2.8 (51±8)||7.2±2.9 (56±9)||0.026*|
|Fourth year||7.1±2.8 (54±8)||7.5±3.0 (58±10)||0.065|
|Fifth year||7.5±2.9 (58±9)||7.5±3.1 (59±11)||0.505|
|Sixth year||7.5±2.8 (59±8)||7.7±3.1 (61±11)||0.575|
|Seventh year||7.8±2.7 (62±7)||7.7±2.9 (61±9)||0.560|
|Eighth year||8.0±2.7 (64±7)||7.7±2.9 (61±9)||0.247|
|Ninth year||7.9±2.9 (63±9)||7.8±2.9 (62±9)||0.519|
|Tenth year||7.9±3.0 (63±10)||8.1±2.9 (65±9)||0.401|
|Ten-year grand average||7.4±2.6 (57±6)||7.5±2.7 (58±7)||0.260|
|Grand average for the entire study period||7.5±2.6 (58±6)||7.6±2.7 (60±7)||0.218|
|At long-term follow-up|
|Duration of type 1 diabetes (years)||11.8±2.0||12.7±2.6||0.098|
|BMI SDS (n=50)||0.90±0.75||0.28±1.17||0.043*|
|Hemoglobin A1c % (mmol/mol)||8.0±3.3 (64±13)||8.1±2.9 (65±9)||0.774|
|Treatment with insulin pump||11 (39%)||21 (47%)||0.537|
|Celiac disease||7 (25%)||2 (4%)||0.041*|
|Hypothyreosis||0 (0%)||2 (4%)||0.527|
|Retinopathy||4 (14%)||14 (31%)||0.105|
|Blood glucose (mmol/L)||10±6||11±5||0.460|
|GADA detectable||15 (54%)||17 (38%)||0.186|
|IA-2A detectable||10 (36%)||13 (29%)||0.542|
|GADA and IA-2A detectable||5 (18%)||5 (11%)||0.415|
An additional analysis was performed where participants were dichotomously divided into those with HbA1c <7.0% (53 mmol/mol) and those with HbA1c ≥7.0% (53 mmol/mol) during the first years in relation to the presence or absence of detectable C peptide at long-term follow-up. During the first 2 years, but not the third year, HbA1c <7.0% (53 mmol/mol) predicted retained C peptide secretion at long-term follow-up (X2 (1, n=73)=6.6, p=0.01 for the first year; X2 (1, n=73)=7.5, p=0.006 for the second year) (table 3).
To analyze predictors of retained C peptide secretion, the presence of detectable C peptide was used as the dependent variable in a logistic regression analysis. A 10-unit change in the first-year, second-year or third-year average HbA1c was used as independent variable and was for all these years highly predictive of detectable C peptide, such that lower HbA1c was associated with detectable C peptide (table 4). Female gender, but not celiac disease, independently predicted detectable C peptide when added to the model. Data on BMI SDS at 10-year follow-up were available for only 50 participants. When a one-unit change in BMI SDS at 10-year follow-up was added to the model in this subgroup, it independently predicted detectable C peptide, such that higher BMI SDS was associated with detectable C peptide (OR 3.00, 95% CI 1.15 to 7.85, p=0.025). Adding age at diagnosis, ketoacidosis at presentation, BMI SDS at 1-year follow-up, or the presence of GADA and/or IA-2A at 10-year follow-up to the models did not improve prediction.
|Dependent variable||Independent variable||OR||95% CI||P value|
|Detectable C peptide||First-year average HbA1c||0.35||0.17 to 0.74||0.005|
|Detectable C peptide||First-year average HbA1c||0.3||0.13 to 0.68||0.004|
|Gender||3.85||1.26 to 11.82||0.018|
|Celiac disease||1.45||0.87 to 2.44||0.157|
|Detectable C peptide||Second-year average HbA1c||0.48||0.24 to 0.95||0.034|
|Detectable C peptide||Third-year average HbA1c||0.48||0.25 to 0.94||0.032|
Retinopathy was predicted by a 10-unit change in the 10-year grand average HbA1c (OR 4.42, 95% CI 1.48 to 13.2, p=0.008). Adding first-year average HbA1c or presence of detectable C peptide did not improve prediction.
Using an ultrasensitive C peptide assay with a detection limit in the low picomolar range, we have confirmed that a substantial proportion of individuals with pediatric type 1 diabetes have preserved C peptide secretions, that is, beta cells, many years after the onset of diabetes.1–3 5 The present results show that tight blood glucose control during the first years of treatment, rather than average long-term control, is associated with preserved C peptide production. Since C peptide secretion declines by each year after diagnosis,9–12 preservation of some remaining secretion would depend on secretion remaining at presentation and the rate of decline. C peptide concentrations at diagnosis were not available in our study, and we can therefore not exclude that the better HbA1c during the first years and the long-term preservation of C peptide both reflected higher remaining residual insulin production at diagnosis. However, insulin dosing per kilogram during the first year after diagnosis did not differ between the groups of participants with detectable and not detectable C peptide at long-term follow-up. There may also be other yet unidentified factors, aside from intensive management, that may have contributed to a lower HbA1c during the first years after diagnosis, eventually leading to residual C peptide 10 years later.
Several factors such as low age at presentation, ketoacidosis at presentation, high HbA1c at presentation and BMI have been related to C peptide concentrations measured at diagnosis9 12–16 and to the decline of C peptide secretion during the first years of treatment.9 17 18 These factors were not related to the long-term preservation of C peptide secretion in the present study. It is noteworthy that age at diagnosis, which is an important predictor in many studies,9 12 13 17 did not presently emerge as a predictor. However, the present sample represents young children in a narrow age range, which makes it difficult to detect age-related differences. The young age at diagnosis of our studied participants also distinguishes them from the populations investigated in other studies, which may at least partially explain the observed differences in findings. In this aspect, the present study adds to the existing literature that factors associated with C peptide decline may differ pending on age group.
GADA and IA-2A have also been associated with a more rapid decline in C peptide secretion.18 In the present study, autoantibody titers were not available at presentation or during the first years of treatment, only at long-term follow-up. Therefore, the role of persisting antibodies in the preservation of C peptide secretion cannot be inferred, but their presence or absence at long-term follow-up did not differ. However, autoantibodies may not be the most important predictors of decline in C peptide secretion.9
Female gender was an independent factor for preserving beta cell mass and function for at least 10 years in the present study. Although we lack data on C peptide levels of the included children at diagnosis, a registry of Swedish juvenile type 1 diabetes (Better Diabetes Diagnosis) has previously reported higher C peptide concentrations in female individuals at diagnosis, providing a potential explanation for this finding.19 Practically, however, all patients with type 1 diabetes have measurable C peptide secretion at diagnosis, and the vast majority have retained some after 1 or 2 years later.12 It is therefore evident that almost all patients initially have the prerequisite for C peptide secretion in the long term. In the short term, good glycemic control helps to preserve C peptide secretion, although a progressive decline appears inevitable.20 21 The present results extend this observation to the long-term preservation of C peptide secretion following initial good glycemic control and show the first 3 years’ glycemic control to be an independent predictor. However, a limitation of the present study was that there was low variability in glycemic control was low, with few participants having poorly controlled diabetes. It is notable that HbA1c beyond the first years of treatment did not influence the presence of detectable C peptide at long-term follow-up.
Interestingly, while BMI SDS at diagnosis did not predict residual C peptide concentrations at long-term follow-up, BMI SDS at follow-up was for unknown reasons an independent predictor of this favorable outcome. Insulin is an anabolic hormone, but the recorded low beta cell function is unlikely to have been contributed by its systemic effects. A possibility, however, may be local trophic effects of residual insulin in the pancreas, since atrophy of the exocrine pancreas relates to loss of beta cell function, with exocrine pancreatic insufficiency and potential malnutrition as a consequence.22
Celiac disease was also more prevalent in patients with detectable C peptide concentrations at long-term follow-up. Previous studies have indicated that gluten-free diet may impact on the immune system, and at least after a 1-year follow-up, stimulated C peptide levels were higher in children with type 1 diabetes treated with a gluten-free diet when compared with matched children with type 1 diabetes on a regular diet.23 Our study suggests that gluten-free diet may also have beneficial effects in a longer time frame, although an intervention study in individuals with type 1 diabetes without concomitant celiac disease would be needed to corroborate this matter.
At long-term follow-up, detectable C peptide concentrations, if any, were in the lower range, with a mean of 4.3 pmol/L (range 1.7–22.7 pmol/L). While of potential importance as target for beta cell expansion therapies, HbA1c in participants with detectable C peptide concentrations was similar to those without detectable C peptide, indicating the recorded very low residual beta cell function had no impact on glycemic control. This is consistent with our previous study on older subjects with type 1 diabetes.5 In the present study we also did not observe an association between the low remaining C peptide concentrations and protection from retinopathy, while retinopathy as expected was predicted by a 10-unit change in the 10-year grand average HbA1c. C peptide levels >10 pmol/L have previously been associated with protection from microvascular and macrovascular complications,17 while lower concentrations similar to those in the present study were not clearly associated.24
In our study, non-fasting C peptide measurements were performed at the long-term follow-up and C peptide was detectable in 38% of the participants. A previous study, using a C peptide assay with similar sensitivity to ours, has shown that even more patients (up to 19% of patients with undetectable non-fasting C peptide) may be identified using a mixed meal tolerance test.25 Whether inclusion of such patients would have also impacted on the findings we obtained in the present study is unknown.
The skilled assistance of research nurses Maria Krügel, Marie Olsén Söderstedt, Susanne Rudenholm Elow, Sofia Sjödin and Violeta Armijo del Valle are gratefully acknowledged. We are also deeply grateful to the late Associate Professor Ingemar Swenne, who contributed extensively to this study but passed away during the finalization of the work.