Article Text
PDF
Abstract
Introduction We aimed to determine whether caregiver responses to the Strengths and Difficulties Questionnaire (SDQ) are predictive of HbA1c trajectory membership in children and adolescents with type 1 diabetes, when adjusting for covariates.
Research design and methods For a Danish 2009 national cohort of children and adolescents with type 1 diabetes, we analyzed yearly HbA1c follow-up data during 2010–2020 including sociodemographic data from Danish national registries. Using group-based trajectory modeling and multinomial logistic regression, we tested whether caregiver SDQ scores predicted HbA1c trajectory membership when adjusting for sex, age at diabetes diagnosis, diabetes duration, family structure, and caregiver education.
Results In total, 835 children and adolescents (52% females) with a mean (SD) age of 12.5 (3.3) years, and a mean diabetes duration of 5.2 (3.1) years, were included. Based on 7247 HbA1c observations, four HbA1c trajectories were identified: (1) ‘on target, gradual decrease’ (26%), (2) ‘above target, mild increase then decrease’ (41%), (3) ‘above target, moderate increase then decrease’ (24%), and (4) ‘well above target, large increase then decrease’ (9%). Higher SDQ total difficulties scores predicted trajectories 3 and 4 (p=0.0002 and p<0.0001, respectively). Regarding the SDQ subscale scores, emotional symptoms predicted trajectories 3 and 4, and conduct problems and hyperactivity/inattention predicted trajectories 2, 3, and 4. Single-parent family and low caregiver education level both predicted trajectories 3 and 4.
Conclusions Caregiver SDQ responses and sociodemographic information may help detect children and adolescents with type 1 diabetes, who need intensive multidisciplinary medical and psychological interventions.
- diabetes mellitus, type 1
- glycated hemoglobin A
- cohort studies
- psychology, child
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Caregiver-reported Strengths and Difficulties Questionnaire (SDQ) total difficulties scores have been repeatedly found to have significant, cross-sectional associations with HbA1c levels in children and adolescents with type 1 diabetes. However, it is unknown if SDQ scores could predict long-term glycemic control.
WHAT THIS STUDY ADDS
Based on 7247 HbA1c observations, four 11-year HbA1c trajectories could be identified.
Suboptimal HbA1c trajectories could be predicted by (1) higher caregiver-reported SDQ total difficulties scores or subscale scores for emotional symptoms, conduct problems, or hyperactivity/inattention symptoms, and (2) single-parent family and/or low caregiver education level.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Caregiver SDQ responses and sociodemographic information may help detect adolescents and young adults with type 1 diabetes, who need intensive multidisciplinary medical and psychological interventions.
Introduction
The American Diabetes Association (ADA) and the International Society for Pediatric and Adolescent Diabetes (ISPAD) recommend that healthcare providers of children and adolescents with type 1 diabetes periodically screen for psychological problems.1 2 The Strengths and Difficulties Questionnaire (SDQ) is a globally used screening tool with good psychometric properties to detect an array of mental health problems.3 Compared with age-matched norms of children and adolescents without type 1 diabetes, those with diabetes, and in particular subgroups with diabetes like males under 15 years or females 4–7 years,4 have higher caregiver-reported SDQ total difficulties scores,4 5 as well as higher scores on the emotional symptoms,4–6 conduct problems,4 5 and hyperactivity/inattention5 subscales. However, a cross-sectional German study did not find any such associations.7
Few studies have investigated correlations between SDQ and HbA1c in children and adolescents with type 1 diabetes.4 5 8 In cross-sectional studies, higher HbA1c levels were significantly correlated with higher SDQ total difficulties scores,8 and for SDQ’s five subscales, significant correlations with HbA1c were found for the emotional symptoms, conduct problems, and hyperactivity/inattention subscales, but not for peer relationship problems or prosocial behavior subscales.4 5 A 3-year longitudinal study found that caregiver-reported SDQ total difficulties scores were significantly associated with HbA1c in Dutch children and adolescents aged 8–15 years.9
However, no study has investigated whether caregiver SDQ scores can be used to longitudinally predict membership to distinct future HbA1c trajectories in children and adolescents with type 1 diabetes. Using a Danish national cohort of children and adolescents with type 1 diabetes of whom caregivers completed the SDQ in 2009,4 our aims were: (1) to identify 11-year HbA1c trajectories in children and adolescents with type 1 diabetes using group-based trajectory modeling (GBTM) and (2) to determine whether SDQ total difficulties scores and the SDQ subscale scores predict membership to HbA1c trajectories after adjusting for sex, age at diabetes diagnosis, diabetes duration, family structure and caregiver education level. Our hypothesis was that elevated SDQ scores predict membership to less favorable HbA1c trajectories that are above generally recommended HbA1c targets.1 10
Research design and methods
Study design, setting, participants
The study is a longitudinal study, comprising the national Danish 2009 cohort of children and adolescents with type 1 diabetes, followed to 2020. The sample initially included all Danish children and adolescents (2–17 years old) diagnosed with type 1 diabetes whose families had participated in a 2009 nationwide online psychosocial survey.4 11 The present study comprises the subsample of 835 children and adolescents (4–17 years old) who had a diabetes duration over 1 year, a baseline HbA1c value, sociodemographic variables in 2009, and whose caregivers completed the caregiver report version of the SDQ. There was 100% completeness for all baseline variables including the SDQ (figure 1).
Flow chart for the inclusion of the participants. SDQ, Strengths and Difficulties Questionnaire; T1D, type 1 diabetes.
Predictor variables
Clinical and sociodemographic variables
Sex, age at diabetes diagnosis, and diabetes duration were obtained at the time of answering the SDQ, or from the Danish Registry for Children and Adolescents (DanDiabKids).12 Sociodemographic variables, including family/household structure, that is, single-parent versus two-parent family, and caregiver education, were obtained from registers held by Statistics Denmark from the year 2009, the same year the SDQ was completed. Highest caregiver education level was categorized into three groups (low, medium, or high education level) using the International Standard Classification of Education system.13 Apart from the SDQ data, all the above data were also tracked on non-participating individuals from DanDiabKids and Statistics Denmark.
The SDQ
The SDQ was completed by caregivers of children/adolescents aged 4–17 years.4 The SDQ is a brief, 25-item behavioral screening instrument with item responses of ‘not true’, ‘somewhat true’ and ‘certainly true’. It consists of a total difficulties scale (score range=0–40) and five separate 5-item subscales (score range=0–10) that generate scores for emotional symptoms, conduct problems, hyperactivity/inattention, peer relationship problems, and prosocial behavior, respectively.14 The SDQ has been widely used, and satisfactory psychometric properties of the instrument, including reliability and validity, have been reported for the Danish translation.15 For the present study, Cronbach’s alpha was 0.83 for the SDQ total difficulties scale, and was between 0.60 (peer relationship problems) and 0.78 (hyperactivity/inattention) for the SDQ subscales.
Outcome variable
HbA1c
HbA1c data comprised a baseline HbA1c obtained from a centrally analyzed blood sample at the time of answering the SDQ.4 These baseline HbA1c data were merged with annual HbA1c data from 2010 to 2020 obtained from the DanDiabKids register12 to the age of 18, and thereafter from the Danish Adult Diabetes Registry.16 Most of the longitudinal registry-based HbA1c data were analyzed locally. Likewise, longitudinal HbA1c data were tracked on non-participants. HbA1c values under 4.9% (30 mmol/mol) or over 20.4% (200 mmol/mol) were dropped (n=8) as they were considered outliers. Only one HbA1c value per year per participant was entered into the sample, and if there were more HbA1c per year, the one nearest the birthday of the participant was used. All reported HbA1c values were measured in accordance with the International Federation of Clinical Chemistry in mmol/mol.17 Corresponding HbA1c in National Glycohemoglobin Standardization Program units (%) is reported.
Statistics
Participant characteristics were examined using descriptive statistics. Comparisons between study participants and non-participants were made using independent t-tests. A sample of 835 participants with 7247 HbA1c observations were analyzed using GBTM, which, based on longitudinal data, identifies a finite number of distinct trajectory clusters and assigns likely membership of individuals into these clusters.18 19 The GBTM procedures were similar to models previously used20–22 with no new adaptations. Participants 4–17 years old at baseline were included, but their HbA1c observations were dropped from further analyses when under age 8 or over 27 years when building the trajectories as the number of measurements outside this age range was deemed too small. Each group was examined with cubic and quadratic trends. In line with previous studies,20–22 the optimal number of groups with homogenous trajectories was based on: (1) model fit using Bayesian information criterion (BIC), (2) visual inspection of the data for all participants and the meaningfulness (ie, clinical context) of the distinctions in the trajectories, and (3) a minimum percent membership in each group of >5%. Having >5% group membership ensures: (1) statistical stability and robustness of the estimated trajectories, (2) meaningful interpretation of distinct subpopulations with more adequate group representation, (3) generalizability because the identified trajectories are more representative of the larger population, and (4) precision of the parameter estimates with narrower CIs.18 19 23 A larger percent decrease in BIC signifies an improvement in model fit.18 19 23 The improvement/decrease in the BIC values was large (10.9%) going from a one-trajectory solution to a two-trajectory solution. However, the decrease going from a four-trajectory solution to a five-trajectory solution was less than 1% and there was an even smaller decrease for more elaborate models. Although BIC was lower using the five-group and six-group models, both of these models yielded one or more groups with a lower number of individuals. The four-group model was chosen due to an amalgamation of its model fit, simplicity, clinical relevance, and the minimum percent membership in each group was over 5%.
When labeling the trajectories/groups, the first part of the label represents the starting point for glycemic control compared with guidelines.1 10 The second part represents the trajectories form or direction over time. After the optimal number of groups was selected, we examined the relationship between probabilistic HbA1c group membership and SDQ scores, sex, age at diabetes onset, diabetes duration, family structure, and caregiver education level. Multinomial logistic regression models were used to assess which variables were predictive of membership in the respective groups.20–22 Procedure trajectories were done using Stata V.17. The reported estimates are changes in log OR of being in one trajectory versus a reference trajectory, given a one-unit increase in the risk factor.21 22 Coefficients >0 indicate an increased probability of membership in the trajectory group under investigation compared with the reference group as the risk factor increases, and vice versa for coefficients <0.21 P values <0.05 were considered statistically significant.
Results
Participant characteristics
The final sample included 835 participating children and adolescents (48% males) aged 4–17 years with type 1 diabetes. All caregivers had completed the caregiver-reported SDQ at baseline, and all had their HbA1c tracked for 11 years from childhood to young adulthood (figure 1). At baseline, mean (SD) age was 12.5 (3.3) years, mean (SD) diabetes duration was 5.2 (3.1) years, and mean (SD) HbA1c was 8.0 (1.3)% (64 (12) mmol/mol). Fifteen percent lived in a single-parent family, and 11% had caregivers with a low education level (table 1). The mean age of the responding caregiver was 42.7 (SD=5.4) years and 82.5% (689/835) of the caregiver respondents were female. HbA1c at baseline was significantly and positively correlated with the SDQ total difficulties scale (r=0.18, p<0.0001), emotional symptoms subscale (r=0.19, p<0.0001), hyperactivity/inattention subscale (r=0.13, p=0.0001), and conduct problems subscale (r=0.20, p<0.0001), but baseline HbA1c was not correlated with the peer relationship problems (r=0.04, p=0.20) or prosocial behavior subscales (r=−02, p=0.61).
Baseline data and longitudinal patient characteristics
HbA1c trajectories
Participants (n=835) with 7247 longitudinal HbA1c values had a mean of 9 (median=10, range=2–12) HbA1c measurements per participant, and over 99.2% of participants had 3–12 HbA1c measurements over an 11-year period. Compared with the percentage of males in the whole sample, there was not a significant difference in the percentage of males in each HbA1c trajectory group.
Figure 2 illustrates the four-group model of HbA1c trajectories. HbA1c group 1 ‘on target, gradual decrease’ had a significant quadratic trajectory starting with a mean (SD) HbA1c of 7.2 (0.7)% (55 (6) mmol/mol), then slightly increasing until peaking at an HbA1c of 7.3 (0.7)% (56 (8) mmol/mol) at age 19 years, and then gradually decreasing to an HbA1c of 6.7 (0.5)% (49 (6) mmol/mol) at 27 years. This group, HbA1c group 1, was the reference group. HbA1c group 2 ‘above target, mild increase then decrease’ had a significant cubic trajectory starting with an HbA1c of 8.1 (0.6)% (65 (7) mmol/mol), then slightly increasing until peaking at an HbA1c of 8.3 (0.8)% (67 (9) mmol/mol) at age 19 years, and then decreasing to an HbA1c of 7.6 (0.9)% (60 (10) mmol/mol) at 27 years. HbA1c group 3 ‘above target, moderate increase then decrease’ had a significant cubic trajectory starting with an HbA1c of 7.9 (0.5)% (63 (5) mmol/mol), then increasing until peaking at an HbA1c of 9.8 (1.2)% (84 (13) mmol/mol) at age 21 years, and then decreasing to an HbA1c of 8.4 (1.1)% (68 (12) mmol/mol) at 27 years. HbA1c group 4 ‘well above target, large increase then decrease’ had a significant cubic trajectory starting with an HbA1c of 8.1 (0.8)% (65 (9) mmol/mol), then increasing until peaking at an HbA1c of 11.8 (1.2)% (105 (13) mmol/mol) at the age of 19 years, and then decreasing to an HbA1c of 9.2 (1.9)% (77 (21) mmol/mol) at 27 years.
HbA1c versus time trajectories spanning over childhood, adolescence and young adulthood. Group 1 ‘on target, gradual decrease’ (blue: 26%), Group 2 ‘above target, mild increase then decrease’ (red: 41%), Group 3 ‘well above target, moderate increase then decrease’ (green: 24%), and Group 4 ‘well above target, large increase then decrease’ (orange: 9%).
Predictors of membership to the HbA1c trajectories
In the main multivariate model 1 (table 2), higher scores on the SDQ total difficulties scale were significantly predictive of group 3 or 4 membership after adjusting for sex, diabetes-specific, and sociodemographic variables (p=0.0002 and p<0.0001, respectively). Caregiver reports of increased emotional symptoms were significantly predictive of group 3 or 4 membership, when adjusting for the same covariates. Increased difficulties reported on the conduct problems and hyperactivity/inattention subscales were significantly predictive of group 2, 3 or 4 membership, when adjusting for the same covariates. Single-parent family, as compared with two-parent family status, was significantly predictive of group 3 and 4 membership (p=0.032 and p<0.0001, respectively) and low caregiver education status, as compared with high education status, and was significantly predictive of group 3 or 4 membership (p<0.001 and p=0.014, respectively) (table 2). In models 2–6, single-parent family status or low caregiver education consistently predicted membership to groups 3 and 4 (online supplemental table S1). In table 2 and online supplemental table S1, the clinical relevance of many of the coefficients can be described using a single example: a one-unit increase in the emotional subscale score corresponds to a 100*(exp(0.257)–1)=29.3% increase in the odds of membership to group 4.
Supplemental material
Predictors of membership to the HbA1c trajectories
Comparisons between participants and non-participants
The mean age of participants (n=835) at the time of SDQ completion was significantly lower compared with non-participants (n=881) (12.5 vs 13.9 years), p<0.0001. The percentage of male participants (48%) was significantly lower than the percentage of male non-participants (52.7%), p=0.04. At baseline, the percentage of single-parent families and caregivers with a low education level in non-participants was higher than participants (27.7% vs 14.5%, p<0.0001; 22.7% vs 11.1%, p<0.0001 respectively). The mean (SD) baseline HbA1c of 8.0 (1.1)% (64 (12) mmol/mol) was significantly lower in participants compared with non-participants at 8.7 (1.4)% (71 (16) mmol/mol) (p<0.0001). The longitudinal mean HbA1c of 8.3 (1.3)% (67 (15) mmol/mol) from 2009 to 2020 in participants was significantly lower compared with the HbA1c of non-participants at 8.7 (1.6)% (72 (18) mmol/mol) (p<0.0001).
Discussion
In a national Danish cohort of children and adolescents with type 1 diabetes from 2009 we aimed to determine whether caregiver responses to the SDQ were predictive of HbA1c trajectory membership for the period of 2010–2020. We identified four distinct HbA1c trajectories: (1) ‘on target, gradual decrease’, (2) ‘above target, mild increase then decrease’, (3) ‘above target, moderate increase then decrease’, and (4) ‘well above target, large increase then decrease’. These results for the first time showed how higher scores on the SDQ total difficulties scale or subscale for emotional symptoms were predictive of the two least favorable HbA1c trajectories (groups 3 and 4), meanwhile higher scores on the conduct behavior problems and hyperactivity/inattention subscales predicted membership to any less favorable HbA1c trajectory (group 2, 3 or 4). Likewise, family sociodemographic risk factors—in particular, single-parent family and/or low caregiver education status—were predictive of membership to the two least favorable HbA1c trajectories.
The cohort of 835 children and adolescents could be split in four subgroups with four distinct HbA1c trajectories over the follow-up period of 11 years until ages 15–28 years. In two larger multicountry studies in which youths with type 1 diabetes were followed 10–19 years20 or 18 years,24 respectively, the cohorts were split into five HbA1c trajectories. Five HbA1c trajectories were also used in a smaller US study for adolescents who were 11–13 years at baseline and tracked over 11 years.21 Two less comparable studies identified three HbA1c trajectories, that is, one study involved US adolescents 13–18 years old at baseline who were tracked over only 18–24 months,25 and one study involved children 9–11 years at baseline who were tracked over 3 years.26 Compared with these studies,20 21 24 25 the baseline age range of our participants was wider, our starting age was younger, and our end age was older. A common feature of the studies is that a large percentage of children and adolescents with type 1 diabetes follow an optimal HbA1c trajectory on target, a large percentage follow the moderate-risk HbA1c trajectories, and too many follow the higher risk HbA1c trajectories. Another common feature is worsening HbA1c levels in a large percentage of youths with type 1 diabetes throughout adolescence up to 19 years,20 21 24 and then decreasing HbA1c levels after age 22.21 22
We found that higher caregiver proxy SDQ total difficulties scores and higher scores in three SDQ subscales at baseline could predict membership to the less favorable HbA1c trajectories over an 11-year follow-up period. The predictive value of SDQ for future HbA1c trajectories has never been investigated, although Helgeson et al21 demonstrated that higher ‘psychological distress’, as assessed by three separate questionnaires that measured children’s depression, anxiety, and anger symptoms, was predictive of the least favorable of five 11-year HbA1c trajectories, and cross-sectional studies have shown associations between SDQ total difficulties scores and HbA1c.4–6 9 We found that higher scores on the SDQ emotional symptoms subscale were predictive of membership to the two highest risk HbA1c trajectories. This finding is consistent with Helgeson et al’s longitudinal study21 and cross-sectional studies showing positive associations between depressive symptoms and elevated HbA1c levels27–29 and positive associations between anxiety symptoms and HbA1c.5 30 We demonstrated that higher scores in the SDQ’s conduct problems or hyperactivity/inattention subscales predicted membership to HbA1c trajectory groups 2, 3, and 4. In accordance with our results, Helgeson et al’s longitudinal study21 and a Dutch cross-sectional study found that externalization symptoms (ie, hyperactivity, inattention, impulsivity, oppositional behavior, disruptive conduct behaviors) associated positively with HbA1c.8 Our observation is also in accordance with a national Swedish and a large German cross-sectional study of children and youths with type 1 diabetes, showing the least favorable metabolic control in patients with attention deficit hyperactivity disorders.31 32 Taken together, cross-sectional studies have shown positive associations between emotional symptoms, conduct behavior problems, symptoms of hyperactivity and inattention with metabolic control. We now have demonstrated the ability of caregiver-reported SDQ scores to predict future HbA1c trajectories. The clinical implications are that identifiable, and early-onset psychological problems persist, and the SDQ may help with early detection of children and adolescents with type 1 diabetes, who need more intensive help from the multidisciplinary diabetes team.
Single-parent or low caregiver education status predicted membership to the two most unfavorable HbA1c trajectories. This finding is in line with a study showing the least optimal metabolic control in children and adolescents growing up in single-parent families,25 and two studies showing the poorest metabolic control in families with the lowest caregiver education status.33 34 What’s novel is that our study predicted membership to specific HbA1c trajectories longitudinally over an 11-year period. Growing up in a home with either a single parent or less educated parent may result in insufficient support in relation to diabetes management, possibly leading to suboptimal adherence to diabetes management, and therefore suboptimal glycemic control. Clinicians should avoid stigmatizing young people who come from disadvantaged backgrounds; however, they should simultaneously consider that families with these social determinants of health may need extra support.
Strengths and limitations
One major strength of this study is the use of data from a population-based national Danish cohort of children and adolescents. A second strength is the follow-up period over a decade. A third strength is the SDQ’s generalizability as it has been standardized and validated for use in numerous countries and healthcare settings. A fourth strength is data completeness regarding all the independent variables, and especially that the models controlled for sex, age at diabetes diagnosis, diabetes duration, family structure, and caregiver education.25 33 The study also has limitations. First, differences were found between participants and non-participants indicating selection bias. The mean HbA1c levels of non-participants were significantly higher than those of participants at baseline and during the 11-year follow-up period. Psychological problems have previously been associated with higher HbA1c levels,4 which is why it is likely that non-participants also had higher rates of psychological problems. Second, we identified a significantly higher percentage of single-parent and low-caregiver-education families in non-participants compared with participants. Therefore, the mean HbA1c levels in the unfavorable trajectories may have been even higher if there were higher participation rates. Third, the SDQ was only completed by caregivers and not by adolescents. Using a multi-informant approach may be valuable when assessing the psychosocial status of adolescents with type 1 diabetes.11 22 Fourth, possibly confounding factors like family income, the number of children in the household, migration background, language barriers, family support/conflict, and peer communication were not adjusted for in our models. Fifth, the HbA1c trajectories may have been influenced by changes in the generally recommended HbA1c targets from less than 7.5% (58 mmol/mol) to less than 7.0% (53 mmol/mol) during the observation period.1 10 35 Finally, a limitation is that diabetes technology has changed substantially36 during the period of 2009–2020.
Conclusions
In summary, this study reaffirms that in populations of children and adolescents with type 1 diabetes followed to young adulthood, individuals follow different HbA1c risk trajectories: low, moderate, or higher risk groups. The SDQ total difficulties scores, and the SDQ subscale scores on emotional symptoms, conduct problems, or hyperactivity/inattention, along with family sociodemographic factors, predict membership to distinct HbA1c trajectories over an 11-year period. The SDQ scores, combined with information on caregiver education and family structure, seem to be valuable tools to predict which children and adolescents with type 1 diabetes and their families need more intensive multidisciplinary intervention from the diabetes team. The data add yet more evidence that generally supports the ADA1 and ISPAD2 recommendations but emphasizes the need to periodically screen children and adolescents with type 1 diabetes for a wider array of psychological problems.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the Danish Society of Childhood and Adolescent Diabetes and by the Regional and Danish Ethical Committees under paragraph 10. The study was registered under Aarhus University’s journal reference number 2016-051-000001. Written and informed consent was obtained from all study participants or their caregivers through the online version of the questionnaires or on paper. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
The authors extend their gratitude to all the healthcare providers at the pediatric diabetes units throughout Denmark who supported this research and are most grateful to all the participants and their families who took the time to participate in our survey. The authors also thank the Danish Clinical Registry of Childhood and Adolescent Diabetes, the Danish Adult Diabetes Registry, and the Danish Clinical Quality Program–National Clinical Registries for giving permission to use clinical data for this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
MT and NHB contributed equally as joint last author.
Contributors All authors contributed to the design of the study and collection of data. MBJ was responsible for the main statistical analysis. KPM, MT, MBJ, and NHB wrote the manuscript. All authors critically revised the manuscript and approved the final version. MT, MBJ and NHB are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Funding KPM’s PhD scholarship and work was supported by a research grant from the Danish Diabetes Academy, which is funded by the Novo Nordisk Foundation (grant number NNF17SA0031406), with additional financial support from Steno Diabetes Center Aarhus (SDCA). NHB received a research grant from Poul and Erna Sehested’s Fond.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Author note MT and NHB are joint last authors.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.