BMI Z-Score (SDS) versus Calculated Body Fat Percentage: Association with Cardiometabolic Risk Factors in Obese Children and Adolescents

Abstract

Introduction

BMI or BMI-standardized deviation score (SDS) in children and adolescents is still the standard for weight classification. [BMJ. 2019;366:4293] developed a formula to calculate body fat percentage (%BF) based on age, sex, height, weight, and ethnicity. Using data from the German/Austrian APV registry, we investigated whether the calculated %BF is superior to BMI-SDS in predicting arterial hypertension, dyslipidaemia, and impaired glucose metabolism.

Methods

94,586 children and adolescents were included (12.5 years, 48.3% male). Parental birth country (BC) was used to depict ethnicity (15.8% migration background); 95.67% were assigned to the ethnicity “white.” %BF was calculated based on the Hudda formula. The relationship between BMI-SDS or %BF quartiles and outcome variables was investigated by logistic regression models, adjusted for age, sex, and migration background. Vuong test was applied to analyse predictive power.

Results

58.4% had arterial hypertension, 33.5% had dyslipidaemia, and 11.6% had impaired glucose metabolism. Boys were significantly more often affected, although girls had higher calculated %BF (each p < 0.05). After adjustment, both models revealed significant differences between the quartiles (all p < 0.001). The predictive power of BMI-SDS was superior to %BF for all three comorbidities (all p < 0.05).

Discussion

The prediction of cardiometabolic comorbidities by calculated %BF was not superior to BMI-SDS. This formula developed in a British population may not be suitable for a central European population, which is applicable to this possibly less heterogeneous collective. Additional parameters, especially puberty status, should be taken into account. However, objective determinations such as bioimpedance analysis may possibly be superior to assess fat mass and cardiometabolic risk than calculated %BF.

Introduction

Rates of obesity in children and adolescents are climbing worldwide [1]. The prevalence remains high despite recently the reported stagnation in Europe [2]. In Germany, about one million children and adolescents are currently obese [3]. However, the situation worsened due to the COVID-19 pandemic [4]. Several comorbidities are well described in obese children and adolescents, including cardiometabolic risk factors, non-alcoholic fatty liver disease, orthopaedic abnormalities, psychosocial disorders, and impaired quality of life [5, 6]. The development of cardiometabolic diseases is mainly attributed to body composition, particularly an increased proportion of body fat [7]. Body mass index (BMI) or the BMI-SDS (standard deviation score), and the respective z-score, which is corrected for age and sex, are often used as surrogate markers for nutritional status. However, this purely arithmetical value lacks the ability to differentiate between lean (fat-free) mass and fat mass. Additionally, there are considerable differences between ethnic groups. Studies conducted in Great Britain and the USA showed that BMI overestimates body fat percentage in black African children and underestimates it in Asian children [8]. Other studies showed that BMI underestimated body fat percentage in South Asian girls and overestimated it in Pacific Islander girls from New Zealand. Hudda et al. [8] were the first to develop a predictive model to improve fat mass estimation in UK children and adolescents aged 4–15 years based on age, sex, and weight, with particular reference to ethnicity. This formula was validated in further 19 countries [9]. For this purpose, the data of 5,715 healthy children and adolescents aged 4–15 years coming from Australia, Austria, Bangladesh, Brazil, China, Mexico, Namibia, Nepal, The Netherlands, New Zealand, the Philippines, Peru, Poland, Russia, South Africa, Spain, Sri Lanka, Tunisia, and the USA were used. The prediction model provides reliable predictions of childhood fat-free mass, and hence fat mass, in the UK and a range of non-UK settings. Additionally, the model demonstrates good generalizability in both low-middle-income and high-income populations of healthy children and adolescents aged 4–15 years. So far, this formula has not been validated for German children and adolescents. However, there is a considerable need for more straightforward methods to assess body fat (percentage) and the associated estimation of cardiometabolic comorbidities in the context of juvenile obesity. Therefore, we tested whether calculating fat mass with this formula is more informative than BMI-SDS with regard to the presence of cardiometabolic risk factors in subjects from the German-Speaking Obese Patients Follow-up Registry (German “Adipositas Patienten Verlaufsdokumentation”; APV).

Materials and Methods

Study Design

We obtained anonymized data from the APV obesity register, which continuously collects relevant data from affiliated obesity programs. In 1999, computer software was developed for the standardized prospective documentation of children and adolescents with obesity. It was based on German guidelines for the diagnosis and therapy of overweight children and adolescents (www.a-p-v.de) [10, 11]. Anthropometric parameters, sociodemographic data, and possible comorbidities (including but not limited to hypertension, dyslipidaemia, and impaired glucose metabolism) are documented per patient visit. The software enables standardized patient reports, local aggregation of data, and patient selection. 224 centres specializing in paediatric obesity care in Germany, Austria, and Switzerland currently submit anonymized data twice yearly. To ensure validity of the analyses, inconsistent data are reported back to the centres for correction. The Ethics Committee of the University of Ulm approved the analysis of anonymized longitudinal APV data (approval 133/22). Each participating centre also adheres to its local ethical and data protection guidelines (see online suppl. File; for all online suppl. material, see https://doi.org/10.1159/000535216).

Study Population and Anthropometric Data

All patients aged 4–15 years presenting at participating institutions between January 2000 and June 2021 were included in this analysis. According to the Federal Statistical Office of Germany (Destatis, see www.destatis.de), Europe (including Turkey) was assigned to the “white” ethnicity, Africa and the Caribbean to “black,” India, Pakistan, Bangladesh, and Sri Lanka to “South Asian,” the rest of Asia to “Asian,” and remaining countries to “other.” If the parents’ countries of birth belonged to different groups, the child was assigned to the “other” group. If only one parent’s country of birth was known, the child was classified according to that country. If no country of birth was known for both parents, a European origin (“white” ethnicity) was assumed.

Height (in cm) and weight (in kg) were given as standard deviation score (online suppl. Table 1). BMI, expressed as weight in kilograms/squared height in metre (kg/m²).

Calculation of BMI-SDS

To adjust BMI values for age and gender, standard deviation score (SDS) was calculated using reference data from a nationally representative sample of German adolescents provided by the Robert-Koch Institute [12]. Subjects were categorized into quartiles based on BMI‐SDS.

Calculation of Body Fat Percentage (%BF) According to Hudda et al.

%BF was calculated based on following formula, according to Hudda et al. [8]:

If child is of black (Africa), South Asian (SA), other Asian (AO), or other (O) ethnic origins:

• $\%\text{BF}=\text{weight}–\exp \left[0.3073\times {\text{height}}^2–10.0155\times \text{weight}–1+0.004571\times \text{weight}+0.01408\times \text{birth}\text{country}\text{Africa}–0.9180\times \ln \left(\text{age}\right)+0.6488\times {\text{age}}^{0.5}+0.04723\times \text{male}+2.8055\right]$

• $\%\text{BF}=\text{weight}–\exp \left[0.3073\times {\text{height}}^2–10.0155\times \text{weight}–1+0.004571\times \text{weight}–0.06509\times \text{birth}\text{country}\text{SA}–0.9180\times \ln \left(\text{age}\right)+0.6488\times {\text{age}}^{0.5}+0.04723\times \text{male}+2.8055\right]$

• $\%\text{BF}=\text{weight}–\exp \left[0.3073\times {\text{height}}^2–10.0155\times \text{weight}–1+0.004571\times \text{weight}–0.02624\times \text{birth}\text{country}\text{AO}–0.9180\times \ln \left(\text{age}\right)+0.6488\times {\text{age}}^{0.5}+0.04723\times \text{male}+2.8055\right]$

• $\%\text{BF}=\text{weight}–\exp \left[0.3073\times {\text{height}}^2–10.0155\times \text{weight}–1+0.004571\times \text{weight}–0.01745\times \mathrm{O}–0.9180\times \ln \left(\text{age}\right)+0.6488\times {\text{age}}^{0.5}+0.04723\times \text{male}+2.8055\right]$

If child is of not black, South Asian, other Asian, or other (other) ethnic origins and if child is of unknown ethnic group, treat as of white ethnic origins:

• $\%\text{BF}=\text{weight}–\exp \left[0.3073\times {\text{height}}^2–10.0155\times {\text{weight}}^{‐1}+0.004571\times \text{weight}–0.9180\times \ln \left(\text{age}\right)+0.6488\times {\text{age}}^{0.5}+0.04723\times \text{male}+2.8055\right]$

Analogous to BMI-SDS, subjects were also categorized into quartiles based on calculated %BF.

Blood Pressure Values

Systolic and diastolic blood pressure values were expressed in mm Hg. Standard deviation scores (SDS) based on German reference data were calculated using national reference data for blood pressure [12]. Arterial hypertension was defined by blood pressure values above the 95th percentile or >140/90 mm Hg [13, 14].

Laboratory Parameters

To account for differences between individual laboratory methods, we used the multiple of the mean to mathematically standardize HbA1c measurements to the Diabetes Control and Complications Trial reference range (4.05–6.05%; 20.7–42.6 mol/mol) [15]. Oral glucose tolerance tests (oGTT) were performed according to international guidelines, and the results were assessed according to national and international guidelines [16]. Impaired glucose metabolism was present if either fasting blood glucose ≥100 mg/dL and/or 2-h glucose ≥140 mg/dL and/or HbA1c ≥ 5.7%.

Lipids were expressed in mg/dL. According to the German guidelines, dyslipidaemia was assumed if total cholesterol levels were >200 mg/dL and/or HDL <35 mg/dL and/or LDL >130 mg/dL and/or triglycerides >150 mg/dL [10].

Data Analysis

For statistical analysis, SAS version 9.4 TS1M7 (SAS Inst. Inc., Cary, NC, USA) was used on a Windows 2019 server. Results for continuous variables are presented as median with interquartile range, results for binary and categorical values as percentage. Logistic regression models were constructed to examine the association between the BMI-SDS quartile or the %BF quartile and cardiometabolic risk factors. Models were adjusted for age, sex, and migration background. The Vuong test was used to determine which predictor (BMI-SDS or %BF) displayed a closer relationship with cardiometabolic risk factors. Two-tailed p values were adjusted with the Bonferroni-Holm correction for repeated testing, and p < 0.05 was considered significant.

Results

94,586 children and adolescents aged 4–15 years were included in the analysis. The median age was 12.5 years (IQR: 10.6; 13.9) and 48.3% were male. 40.1% of the children and adolescents were overweight (BMI 90th–97th percentile), 49.5% were obese (BMI 97th–99.5th percentile), and 10.4% were severely obese (BMI >99.5th percentile).

15.8% had a migration background. In total, 90,490 (95.7%) of the sample was assigned to the ethnicity “white” based on their parents’ country of birth. 533 (0.6%) were assigned to “black,” 126 (0.1%) were assigned to “South Asian,” 1,527 (1.6%) were assigned to “Other Asian” and 1,910 (2.0%) were assigned to the category “other”.

The median BMI-SDS was 1.99 (IQR: 1.70; 2.30). %BF calculated according to Hudda et al. [8] was 39.4% (IQR: 36.4; 42.6) and fat mass was 27.5 kg (IQR: 22.2; 34.2). Girls were significantly younger, had lower height- and BMI-SDS, and their relative and absolute calculated body fat percentage was higher (see Table 1). Weight-SDS and absolute BMI did not differ between boys and girls. Table 2 shows the BMI-SDS quartiles and %BF quartiles.

Table 1.

Anthropometric data: entire group, girls versus boys

	Total	Girls	Boys	p value
N	94,586	48,930	45,656
Age, years	12.5 (10.6; 13.9)	12.46 (10.5; 13.9)	12.54 (10.8; 13.8)	0.00031*
Migration background, %	15.8	15.2	16.5	<0.001**
Ethnicity “white”, %	95.67	95.74	95.59	0.0041**
Ethnicity “black”	0.56	0.61	0.51
Ethnicity “South Asian”	0.13	0.10	0.16
Ethnicity “South Asian”	1.61	1.53	1.70
Ethnicity “others”	2.02	2.01	2.03
Weight-SDS	2.0 (1.6; 2.4)	1.97 (1.6; 2.4)	1.97 (1.6; 2.4)	0.14540*
Height-SDS	0.7 (0.0; 1.4)	0.63 –(0.06; 1.36)	0.71 (0.04; 1.42)	<0.001*
BMI, kg/m2	28.3 (25.8; 31.6)	28.3 (25.7; 31.8)	28.3 (25.9; 31.5)	0.27434*
BMI-SDS	1.99 (1.70; 2.30)	1.97 (1.68; 2.29)	2.01 (1.72; 2.31)	<0.001*
%BF	39.4 (36.4; 42.6)	40.9 (38.3; 43.9)	37.5 (34.7; 40.8)	<0.001*
Fat mass, kg	27.5 (22.2; 34.2)	28.5 (22.8; 35.3)	26.5 (21.8; 32.8)	<0.001*
Arterial hypertension, %	58.4	57.2	59.6	<0.001*
Dyslipidaemia, %	33.5	32.7	34.3	0.00009**
Impaired glucose tolerance, %	11.7	11.2	12.3	0.00052**

*p values with Bonferroni-Holm correction for repeated testing.

**χ² test. SDS, standard deviation score.

Table 2.

Anthropometric data: BMI-SDS quartiles and %BF quartiles

	BMI-SDS				%BF
	Q1 (IQR)	Q2 (IQR)	Q3 (IQR)	Q4 (IQR)	Q1 (IQR)	Q2 (IQR)	Q3 (IQR)	Q4 (IQR)
N	23,646	23,647	23,646	23,647	23,647	23,646	23,647	23,646
Age, years	12.7 (11.3; 13.8)	12.6 (11.0; 13.9)	12.6* (10.7; 13.9)	11.9* (8.8; 13.9)	13.0 (10.3; 14.2)	12.5* (10.7; 13.8)	12.3* (10.7; 13.7)	12.3* (10.6; 13.7)
Male, %	45.5*	47.9*	49.8*	49.9*	77.6*	48.5*	38.9*	28.2*
Migration background, %	11.4*	14.0*	17.5*	20.5	11.8	13.9*	16.8*	21.1*
Weight-SDS	1.4 (1.2–1.7)	1.8* (1.6–2.0)	2.1* (1.9–2.3)	2.6* (2.4–2.9)	1.7 (1.4–2.1)	1.8* (1.4–2.1)	2.0* (1.7–2.3)	2.4* (2.1–2.7)
Height-SDS	0.4 (−0.2 to 1.1)	0.6* (−0.1 to 1.3)	0.7* (0.1–1.4)	1.0* (0.2–1.7)	0.9 (0.2–1.6)	0.9* (0.0–1.3)	0.6* (−0.1 to 1.3)	0.6* (−0.2 to 1.3)
BMI, kg/m2	25.7 (24.2; 26.9)	28.0 (26.1; 29.4)	30.8 (28.2; 32.3)	34.2 (28.7; 37.1)	26.0 (23.4; 28.0)	26.8 (24.8; 29.1)	28.8 (26.6; 31.3)	32.9 (30.2; 36.3)
BMI-SDS	1.54 (1.70; 2.30)	1.85* (1.70; 2.30)	2.14* (1.70; 2.30)	2.52* (1.70; 2.30)	1.65 (1.50; 1.90)	1.81* (1.61; 2.02)	2.05* (1.86; 2.23)	2.40* (2.22; 2.60)
%BF	36.0 (33.9; 37.9)	38.5* (36.5; 40.5)	41.1* (39.0; 43.1)	44.1* (41.0; 46.6)	34.3 (32.6; 35.5)	38.0* (37.2; 38.7)	40.9* (40.1; 41.7)	44.9* (43.6; 46.7)
Fat mass, kg	23.0 (19.9; 25.5)	27.3* (23.0; 30.3)	32.3* (26.5; 35.9)	38.3* (24.3; 44.9)	23.3 (17.1; 27.4)	25.2* (20.3; 29.9)	28.7* (23.5; 34.2)	36.1* (29.8; 43.1)

*p value <0.05 compared to Q1; p values with Bonferroni-Holm correction for repeated testing. SDS, standard deviation score.

Correlation with Comorbidities

58.4% of the children and adolescents had arterial hypertension (n = 46,107), 33.5% had dyslipidaemia (n = 20,926), and 11.6% had impaired glucose metabolism (n = 6,546). The prevalence of arterial hypertension, dyslipidaemia, and impaired glucose tolerance was higher in boys (see Table 1).

The proportion of risk factors in relation to the four quartiles based on BMI-SDS or %BF is presented in Figure 1a–c, adjusted for age, sex, and migration background in the total group and stratified by gender. For both BMI-SDS and %BF, highly significant associations with all 3 cardiometabolic risk factors existed in the trend test within the total group and for both sexes (all p < 0.001).

Fig. 1.

Prevalence of cardiometabolic risk factors (a–c) stratified by quartiles of BMI or %BF and gender; arterial hypertension (a), dyslipidemia (b), and impaired glucose tolerance (c).

In the total population, the model with BMI-SDS as explanatory variable was closer related to the prevalence of arterial hypertension (adjusted Vuong statistics z = −2.5884; p = 0.0096), dyslipidaemia (adjusted Vuong statistics z = −3.0701; p = 0.0021), and impaired glucose metabolism (adjusted Vuong z = −4.1096; p < 0.0001) compared to the model with %BF.

Comparing boys and girls, there were inconsistent results. In girls, BMI-SDS was closer related to arterial hypertension (adjusted Vuong statistics z = −5.2812; p < 0.0001) and impaired glucose tolerance (adjusted Vuong z = −2.0067; p < 0.0448). For dyslipidaemia the association with %BF was closer (adjusted Vuong statistics z = 4.3636; p < 0.0001).

In boys, there was no significant difference in arterial hypertension (adjusted Vuong statistics z = 0.2621; p = 0.7933), but BMI-SDS was associated more closely to dyslipidaemia (adjusted Vuong statistics z = −8.2691; p < 0.0001) and impaired glucose tolerance (adjusted Vuong z = −4.6472; p < 0.0001) compared to the model with %BF.

Discussion

Based on a large sample of juvenile patients with obesity from Germany, Austria, and Switzerland, we analysed the predictive power for arterial hypertension, dyslipidaemia, and impaired glucose metabolism of the %BF calculated according to Hudda et al. [8] compared to the well-established BMI-SDS. BMI-SDS provided superior prediction for all three comorbidities in the total population. Possible reasons to explain these findings are manifold. First of all, Hudda et al.’s [8] formula is based on the observation that the BMI-SDS differs by ethnicity. Hudda et al. [8] defined “ethnicity” based on parents’ self-reported origin; the categories were based on the 2001 UK Census. We defined ethnicity or migration background as the parental country of birth based on the categories mentioned in the official German population statistics, Destatis. Results did not differ when the allocation was chosen analogously to the UN statistics, which classify Turkey as a part of Asia. It is possible that the ethnic groups described in the formula represent the British society and are much less common in the German-speaking countries participating in the APV registry as almost 96% of the ethnicity was assigned to “white.” Alternatively, ethnicity may not reliably be derived from the parental country of birth. Whether a formula adapted to German conditions would produce superior results compared to the formula developed in England remains speculative at present. However, there may also be differences by gender, which could explain the inconsistent results between girls and boys. This is not surprising, as in the age group between 5 and 14 years, enormous developmental steps take place that also influence body composition, especially in the context of puberty. However, there may also be differences by gender, which could explain the inconsistent results between girls and boys. This is not surprising, as in the age group between 5 and 14 years, enormous developmental steps take place that also influence body composition, especially in the context of puberty. Hence, notably during the pubertal phase, a surge in lean mass and fat mass is observed, with the latter being more pronounced in females (Veldhuis et al. [17]). Alongside sex hormones, the adipokine leptin also assumes significance. Prior to the onset of puberty, leptin increases in both sexes [18]. In females, this trend persists throughout the pubertal phase into adulthood, while in males, it tends to diminish [19, 20]. Furthermore, Stumper et al. [21] demonstrated that the elevation of sex hormones during puberty exerts an immunomodulatory influence on inflammatory biomarkers, with noticeable gender-related disparities. Consequently, the inclusion of pubertal status assumes a pivotal role in both developmental assessments and registries.

In general, the calculation of %BF based on weight, height, age, and sex must be viewed with caution. Direct determinations of fat mass, for example, measured body composition, body fat percentage, or abdominal circumference, might be better predictors of cardiometabolic risk factors [22]. However, Bohn et al. [23] analysed 3,327 children and adolescents from the APV participants (BMI >90th percentile) and reported no difference between BMI and measured body fat in terms of correlation with cardiovascular risk factors (hypertension, dyslipidaemia, elevated liver enzymes, abnormal carbohydrate metabolism). A recent study of almost 9,000 children and adolescents conducted by Xiao et al. [24] showed, on the other hand, that an increased proportion of “lean mass” had a protective effect on total cholesterol, LDL levels, and insulin resistance, independent of BMI and fat mass. In a multi-ethnic Asian cohort, 377 six-year-olds in the “Growing Up in Singapore towards Healthy Outcomes” study had their fat or lean mass determined by magnetic resonance imaging [25]. In this population, fat mass was more strongly associated with the occurrence of cardiometabolic risk factors. Additionally, Ali et al. [26] showed that subcutaneous obesity based on MRI and DEXA (dual-energy X-ray absorptiometry) measurements was the strongest predictor of metabolic syndrome. Other body fat parameters such as skinfold thickness measurements did not add value in predicting insulin resistance in preschool children in the IDEFICS study [27]. However, assessing muscle mass and muscle-to-fat ratio in terms of sarcopenic obesity is also useful for assessing health risk [28].

In addition, when calculating the %BF in our analysis, a larger number of girls were found in the upper quartiles; for example, the proportion in Q1 was 22.4% and in Q4 71.8%. In contrast, for BMI-SDS, the sex ratio was balanced in all quartiles. As mentioned above, a higher body fat percentage in girls compared to boys is not surprising [29]; however, these differences seem not to be adequately modelled by the %BF formula used.

Strengths and Limitations

The strengths of this analysis are the large sample size and the standardized collection of data. The data can therefore be considered representative of children and adolescents with obesity in Germany, Austria, and Switzerland. However, participation in the obesity register is voluntarily for participating institutions such as outpatient or inpatient therapy facilities, hospitals, or paediatric practices. One main limitation is the recording of ethnic background: it is not uncommon for the parents of children born in Germany to have a migration background, which is not assessed if only the country of birth is queried, not the country of origin of previous generations. Additionally, the concept of “ethnicity” is primarily based on skin colour and originates from England and the USA, and this procedure is hardly transferrable to the German context. The largest group of people with a migration background in Germany comes from Turkey. They are usually classified as “Caucasian” or “white.” Even if UN statistics is used and Turkey is attributed to Asia, no differences were found (data not shown). People with an African-American or Asian background, on the other hand, are still rare in Germany. Another significant limitation lies in the absence of pubertal status information in both the formula and the APV register. It would be advantageous to incorporate puberty scales, such as those proposed by Tanner and Whitehouse in 1976 [29], for developmental classification. Given the substantial interindividual variability, taking into account the hormonal status within the process of design such formulae may provide a more meaningful approach than considering age alone.

Conclusion

In summary, BMI-SDS was superior to calculated %BF to predict all three comorbidities. For a precise evaluation of health risks associated with juvenile obesity, it is imperative to account for additional factors such as Tanner stage, pubertal development, physical fitness, or the muscle-to-fat ratio [30]. Furthermore, the incorporation of these factors becomes essential in registry-based investigations aimed at deducing potential therapeutic interventions. Thus far, objective methodologies like bioimpedance analysis appear to offer superior suitability for assessing fat mass compared to calculated %BF.

Acknowledgments

The first version of the statistical analysis was implemented by Anna Wagner, BS, Ulm. We thank all institutions participating in the APV registry. A detailed list can be found in the online supplementary material. Further thanks to R. Ranz and A. Hungele for their support and the development of the APV documentation software (both clinical data managers, Ulm University).

Statement of Ethics

The APV initiative and the joint analysis of anonymized data have been approved by the Ethics Committee at Ulm University (No. 133/22). The need for informed consent was waived by the Ethics Committee at Ulm University.

Conflict of Interest Statement

The authors Nicole Prinz, Reinhard W. Holl, Paula Moliterno, Gabriel Torbahn, Stefanie Wessely, Susanna Wiegand, Sabine Kaiser, Hagen Wulff, Bettina Gohlke, and Jens Nielinger have no conflicts of interest to declare. C. Joisten received lecture fees from Berlin Chemie, MSD (Merck Sharp & Dohme), Novartis, Abbvie, Pfizer, Janssen, Lilly, Chiesi, Chugai, Novo Nordisk, Daiichi-Sankyo, Sanofi, and Pharmacosmos. There are no relations to this article.

Funding Sources

This manuscript is part of a project (www.imisophia.eu) that has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No. 875534. This joint undertaking receives support from the European Union’s Horizon 2020 research and innovation programme EFPIA and T1D Exchange, JDRF, and Obesity Action Coalition. Funders were not involved in the analysis and interpretation of data, the writing of the report, or the decision to submit the article for publication. Further funding has been received by the German Research Foundation (DFG) and the Federal Ministry of Education and Research within the German Center for Diabetes Research (DZD; Grant No. 82DZD14E03).

Author Contributions

All authors participated in the generation of the data and had final approval of the submitted version. Christine Joisten, Nicole Prinz, and Reinhard Holl were responsible for conceptualization of the analyses. Reinhard Holl and Nicole Prinz analysed the data. Christine Joisten and Stefanie Wessely wrote the first draft. Susanna Wiegand, Bettina Gohlke, Sabine Keiser, Paula Moliterno, Jens Nielinger, Gabriel Torbahn, and Hagen Wulff collected data, reviewed, and edited the manuscript.

Funding Statement

This manuscript is part of a project (www.imisophia.eu) that has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No. 875534. This joint undertaking receives support from the European Union’s Horizon 2020 research and innovation programme EFPIA and T1D Exchange, JDRF, and Obesity Action Coalition. Funders were not involved in the analysis and interpretation of data, the writing of the report, or the decision to submit the article for publication. Further funding has been received by the German Research Foundation (DFG) and the Federal Ministry of Education and Research within the German Center for Diabetes Research (DZD; Grant No. 82DZD14E03).

Data Availability Statement

To ensure compliance with patient consent, no patient-level data can be shared with research groups outside of Ulm University. However, aggregated data can be made available upon reasonable request via email to the senior author. Remote data access and joint research projects are also possible. Further enquiries can be directed to the corresponding author.