Growth and subcutaneous fat during the first five years of insulin-dependent diabetes in children

Harnessing machine learning models for non-invasive pre-diabetes screening in children and adolescents

BACKGROUND AND OBJECTIVES

Pre-diabetes has been identified as an intermediate diagnosis and a sign of a relatively high chance of developing diabetes in the future. Diabetes has become one of the most frequent chronic disorders in children and adolescents around the world; therefore, predicting the onset of pre-diabetes allows a person at risk to make efforts to avoid or restrict disease progression. This research aims to create and implement a cross-validated machine learning model that can predict pre-diabetes using non-invasive methods.

METHODS

We have analysed the national representative dataset of children and adolescents (5-19 years) to develop a machine learning model for non-invasive pre-diabetes screening. Based on HbA1c levels the data (n = 26,567) was segregated into normal (n = 23,777) and pre-diabetes (n = 2790). We have considered eight features, six hyper-tuned machine learning models and different metrics for model evaluation. The final model was selected based on the area under the receiver operator curve (AUC), Cohen's kappa and cross-validation score. The selected model was integrated into the screening tool for automated pre-diabetes prediction.

RESULTS

The XG boost classifier was the best model, including all eight features. The 10-fold cross-validation score was highest for the XG boost model (90.13%) and least for the support vector machine (61.17%). The AUC was highest for RF (0.970), followed by GB (0.968), XGB (0.959), ETC (0.918), DT (0.908), and SVM (0.574) models. The XGB model was used to develop the screening tool.

CONCLUSION

We have developed and deployed a machine learning model for automated real-time pre-diabetes screening. The screening tool can be used over computers and can be transformed into software for easy usage. The detection of pre-diabetes in the pediatric age may help avoid its enhancement. Machine learning can also show great competence in determining important features in pre-diabetes.

Good glycemic control without exceeding the BMI trajectory during the first 5 years of treatment in children and adolescents with type 1 diabetes

OBJECTIVE

To study BMI changes and glycemic control in children and adolescents during the first 5 years following diagnosis of type 1 diabetes.

RESEARCH DESIGN AND METHODS

The 295 children and adolescents (<18 years) diagnosed with type 1 diabetes started on multiple injection treatment and were followed during the first 5 years of treatment with respect to glycemic control and weight change. Growth curves preceding the onset of diabetes were obtained from the school health services and child care centers. BMI was recalculated into BMI SD scores (BMISDS).

RESULTS

Prior to the onset of diabetes, the BMISDS was 0.46 ± 1.24 (mean ± SD), which decreased to -0.61 ± 1.36 (p < 0.001) at presentation. At 1 year, BMISDS was 0.59 ± 0.99 (p > 0.05) and increased to 0.80 ± 1.03 at 5 years; 0.97 ± 0.93 in females versus 0.68 ± 1.08 in males (p < 0.001). BMISDS at 1 year and 5 years were directly proportional to and highly predicted by BMISDS prior to the onset of type 1 diabetes, (r = 0.76; p < 0.001) vs. (r = 0.58; p < 0.001). HbA1c at 1 year was 50 ± 10 mmol/mol, which increased to 58 ± 12 mmol/mol (p < 0.001) at 5 years; females had HbA1c 60 ± 14 mmol/mol versus males 56 ± 11 mmol/mol (r = 0.35, p < 0.001). There was a correlation, irrespective of gender, between HbA1c and BMISDS at 1 year (r = 0.18, p < 0.003), but not at 5 years (r = 0.036, (p > 0.5).

CONCLUSION

During the first 5 years of treatment of type 1 diabetes in children and adolescents it is possible to achieve good glycemic control without excess weight gain.

Characteristics of Selected Somatic and Motor Abilities of Youth Soccer Players with Diabetes Type 1 Treated with Insulin Pump Therapy

Long-term insulin treatment can slow the growth process and decrease physical fitness level in children. In diabetic children, these two developments should be constantly monitored. The aim of the present study was to examine differences in somatic and physical fitness characteristics between soccer-training boys with type 1 diabetes and healthy boys of the same age (reference values based on Polish population norms for somatic and motor parameters). The participants were 94 boys (8–17 years), diagnosed with diabetes, who participated in soccer training on a regular basis and received routine medical care. The study involved (a) anthropometric and body composition measurements, (b) general motor ability assessments, and (c) comparison of those characteristics with the healthy Polish population. The diabetic boys were found to have lower levels of almost all somatic traits and motor abilities as compared with the healthy boys (p ≤ 0.05). Handgrip strength was a variable with the smallest difference between the two groups. The observed differences indicate the necessity to design an appropriate control and assessment system based on simple medical and fitness field tests for diabetic children and adolescents. It will allow optimizing advanced training as well as minimize health risks before, during, or after exercise.

Recovery of premorbid BMI trajectory without overshoot during the first year of treatment of children with type 1 diabetes

OBJECTIVE

To study body mass index (BMI) changes and metabolic control in children and adolescents during the first year following the diagnosis of type 1 diabetes.

RESEARCH DESIGN AND METHODS

200 children and adolescents (<18 years) diagnosed with type 1 diabetes, started on multiple injection treatment and followed up for 1 year were studied with respect to metabolic control and weight change. Growth curves preceding the onset of diabetes were procured from the school health services. BMI was recalculated into BMI SD scores (BMISDS).

RESULTS

Glycated hemoglobin (HbA1c) at 1 year was 6.7±1.3% (50±10 mmol/mol). HbA1c was positively correlated with daily insulin dose (R(2)=0.13; p<0.001), negatively correlated with age (R(2)=0.03; p<0.05) but not related to gender, BMISDS at 1 year, HbA1c at presentation, or ketoacidosis at presentation. Prior to the onset of diabetes, BMISDS was 0.41±1.20 and decreased to -0.63±1.25 at presentation. BMISDS at 1 year was 0.54±0.97 and not different from the premorbid value (p>0.05). In a multiple regression analysis, BMISDS at 1 year was directly proportional to and highly predicted by BMISDS prior to onset of diabetes (R(2)=0.57; p<0.001). BMISDS at 1 year was also inversely correlated with age (R(2)=0.03; p<0.001) but could not be predicted by gender, daily insulin dose, HbA1c at 1 year, HbA1c at presentation, or by ketoacidosis at presentation.

CONCLUSIONS

During the first year of treatment of type 1 diabetes in children and adolescents, it is possible to achieve good metabolic control without excess weight gain.

Body mass index changes in youth in the first year after type 1 diabetes diagnosis

OBJECTIVES

To describe changes in weight and body mass index (BMI) during the first year following diagnosis of type 1 diabetes (T1D) and associations with demographic and clinical characteristics.

STUDY DESIGN

The Pediatric Diabetes Consortium includes 7 US centers with prospective longitudinal data from initial T1D diagnosis. This analysis includes 530 youth with diabetes duration of ≥1 year and measures of BMI at 3 and 12 months after diagnosis. BMI trajectory of participants and relationships between the change in BMI z-score from baseline (3 months) to 12 months with demographic characteristics, hemoglobin A1c at baseline, and insulin delivery mode at baseline were evaluated.

RESULTS

As a group, BMI z-scores increased sharply from diagnosis for 1-3 months but remained relatively stable from +0.51 at 3 months to +0.48 at 12 months. Children aged 2-<5 years experienced a significant positive change in BMI z-score between 3 and 12 months, and there was a similar trend among girls that did not reach statistical significance. No significant differences were found for race, socioeconomic status, or insulin delivery mode.

CONCLUSIONS

These data suggest that increased BMI during the first year of treatment of most youth with T1D reflects regain of weight lost before diagnosis. There is, however, a propensity toward additional weight gain in younger children and girls.

Changes in weight and BMI following the diagnosis of type 1 diabetes in children and adolescents

The aim of this study was to study weight and body mass index (BMI) before, at, and after diagnosis of type 1 diabetes (T1D) and to identify factors associated with weight gain. Studied retrospectively were 209 children <18 years with T1D followed for 6 years. Data collected included clinical and laboratory data before diagnosis, at diagnosis, and during 6 years of follow-up. Anthropometric parameters of patients were compared along follow-up and with those of their parents and siblings. Mean BMI-standard deviation score (SDS) was below average at diagnosis (-0.66 ± 1.27), had increased to 0.37 ± 0.93 at 3 months, and decreased to a nadir at 6 months in females and 12 months in males; between 1 and 3 years, there was a slight increase and between 3 and 6 years a further increase only in the females. BMI-SDS at 6 years was significantly higher than pre-diabetes BMI-SDS (0.35 ± 0.83 vs. -0.04 ± 1.23, p < 0.001). Patients' BMI-SDS at 6 years was similar to that of their parents and siblings, was higher in the females (0.53 ± 0.74 vs. 0.27 ± 0.82, p = 0.02) and in those keeping diabetes a secret (0.66 ± 0.82 vs. 0.33 ± 0.78, p = 0.027), and was not associated with age or pubertal stage at diagnosis, ethnicity, or metabolic control. A longer duration of insulin pump therapy was associated with a lower BMI-SDS (r = -0.2375, p < 0.025). BMI-SDS increased during the 6 years following diagnosis of T1D in pediatric patients, especially in the females, but remained in the normal range and was similar to that of other family members.

Growth, puberty, and final height in children with Type 1 diabetes

OBJECTIVE

The aims of this study were to assess the physical growth and pubertal development in a group of diabetic children and to evaluate the effect of height at diagnosis, duration of illness, and degree of glycemic control on final height and sexual maturation.

RESEARCH DESIGN

A cohort of 72 Sudanese diabetic children, 7-13 years of age at diagnosis, was followed longitudinally from the onset of diabetes until the attainment of final height.

RESULTS

The mean height standard deviation scores (SDS) at diagnosis were 0.04 in boys and -0.15 in girls, which was greater than their genetic target height (GTH). The growth velocity between diagnosis and final height was slow, with significant reduction in pubertal growth spurt. The mean final height attained by these children was lower than their GTH, a finding that contradicts most of the recently published reports. The average age at menarche in girls (15.1 years) and the mean age of full sexual maturation in boys (17.2 years) were significantly delayed in this group of diabetic patients. This retardation in physical growth and pubertal development was positively correlated with the duration of diabetes before the onset of puberty and glycated haemoglobin (HbA1c) concentration. The majority of these patients were thin at diagnosis of diabetes, with median body mass index (BMI) <22, but showed a remarkable, progressive weight gain during puberty, which was more evident in girls. The weight gain was independent of weight at diagnosis and duration of diabetes, but was positively correlated with the daily dose of insulin and HbA1c concentration.

CONCLUSION

Conventional therapy of diabetic children is associated with impairment of physical growth and delayed sexual maturation.

Diabetes control deteriorates in girls at cessation of growth: relationship with body mass index

AIM

Diabetic patients, particularly girls, often experience poor metabolic control during puberty and adolescence. The aim of this study was to investigate metabolic control during adolescence, especially in relation to pubertal stages, growth, insulin treatment and body mass index (BMI).

METHODS

We studied the records of 38 (consecutive) girls with prepubertal onset of Type 1 diabetes mellitus. Data from the age of 10 to 18-20 years were obtained with regard to glycaemic control, growth, age at menarche, final height and BMI, and analysed in relation to both chronological age and age at menarche.

RESULTS

HbA1c was lowest 3 years before menarche; mean (+/- sd) 7.6 (+/- 1.2). After the pubertal growth spurt, there was a marked impairment of metabolic control, the highest level of HbA1c occurring 3 years after menarche. Mean age at menarche was 13.3 (+/- 1.1) years and mean linear growth after menarche only 4.7 cm, giving a final height of 164.9 (+/- 5.3) cm which is 2.7 cm below the Swedish mean. During adolescence the degree of correlation between BMI and HbA1c continuously increased, pointing out the effect of body fat on metabolic control in this age group. The level of HbA1c at 10 years of age could not predict the metabolic control after cessation of puberty, but prepubertal BMI appears to be a risk factor for both obesity and poor glycaemic control in late adolescence.

CONCLUSIONS

The highest HbA1c was found after cessation of growth. Prepubertal BMI is a possible predictor of metabolic control in adolescent diabetic girls.

Serum lipoproteins and apolipoproteins in children during the first five years of diabetes

Serum lipoproteins and apolipoproteins were followed in 34 children during a period of 5 years from the onset of diabetes. The group did not differ in these respects from a healthy control group after 5 years of disease. The variation in serum triglycerides and very-low-density lipoprotein (VLDL) triglycerides was more pronounced, some patients having high values. Serum triglycerides and VLDL lipids were significantly correlated to subcutaneous fat, measured as triceps and subscapular skinfolds. None of the patients had albuminuria, so lipid levels could not be related to renal albumin excretion. There was no significant correlation between any serum lipid and haemoglobin A1c.