Lifetime antimicrobial use is associated with weight status in early adolescence—A register‐based cohort study

Summary Background Antimicrobial agents (AMs) are the most prescribed drugs to children. Early and repeated exposure to AMs in infancy is associated with increased risk of childhood overweight and obesity. Aims We extended the investigation of AMs use, from birth to early adolescence, and evaluated their association with weight status. Materials & Methods A total of 10093 children from Finnish Health in Teens cohort (Fin‐HIT) with register‐based data on AMs purchases and measured weight status at the mean age of 11.2 y (SD 0.82) were included in the study. The key exposures were the number AM purchases at a given age or the sum of these during the entire follow‐up time to describe lifetime exposure / use. Outcome was weight status in early adolescence defined with International Obesity Task Force cut‐offs for the age‐ and sex‐specific body mass index. Odds Ratio (OR) and 95% confidence intervals (CI) were estimated using Multinomial Logistic Regression. Results Of children, 73.7% were normal weight, 11.1% thin and 15.2% overweight/obese. AMs use was highest during the second year of life, when 65% of all children used AMs, but thereafter decreased with age. The highest mean purchases and prevalence at any given age along with the highest lifetime use were consistently seen among overweight children. Each episode of AMs use throughout life increased the risk of being overweight in adolescence [OR = 1.02 (1.02–1.03)]. However, there was an inverse association between AMs use and thinness [OR = 0.98 (0.97–0.99)]. Discussion Despite a high prevalence of AMs use during the early years, lifetime‐use was associated with weight status in early adolescence in a dose response manner. Conclusion Future studies should address mechanisms underlying the relationship between AM use and weight.


| INTRODUCTION
Elderly adults (≥60 years) and children up to 9 years old are exposed to the highest amount of antimicrobial agents (AMs). 1 In many countries, AMs are among of the most prescribed drugs in young children and adolescents, [2][3][4] especially in the early stage of life. 5 For instance, 40% of Danish children are prescribed AM during the first year of life. 6 In many low-and middle-income countries, AMs are frequently purchased over-the-counter, which may increase the risk of inappropriate use. 7 Moreover, in low-resource settings, there is a considerable research interest to study mass antibiotic treatment to pregnant women or children in order to reduce child mortality and improve growth and development. 8,9 Repeated perturbations caused by the use of AMs are likely to shift the microbiota's community structure by eliminating some strains and allowing new strains to enter the community. This may affect the microbiome's capacity and host metabolism. 10,11 It is suggested that AMinduced microbiome alterations may increase susceptibility to intestinal infections, immune-mediated diseases and dysregulation of metabolism, likely contributing to weight gain especially in children. 10 Overweight and obesity are global health problems and, to date, in Finland, the combined prevalence in children is around 15%. 12,13 The contribution of accumulated lifetime AM use for the obesity epidemic is yet to be determined.
In 2015, World Health Organization (WHO) introduced a global action plan to improve the quality of AM use and to decrease the number of unnecessary prescriptions. 14 In European countries, where national actions to improve the rational use of antibiotics have been applied, a decreasing trend in AM use has been reported, likewise in Finland. 4 Still differences in AM use between countries appear: in 2018, the mean total consumption of AMs for systemic use in community and hospital sector was 15.5 defined daily doses per 1000 inhabitants per day (DDD) in Finland, while the mean was 20.1 DDD in the EU/EEA, ranging from 9.7 in the Netherlands to 34 in Greece. 4 The primary purpose of the present study was to evaluate if an exposure to AMs at any given age, or the total/lifetime exposure from birth to early adolescence, is associated with weight status at early adolescence. As a secondary aim, the trends in AM use and exposure during pregnancy were described and their relevance to weight status were evaluated. Previously, early and repeated exposure to AMs in infancy have been associated with increased risk of childhood overweight and obesity. [14][15][16] In several studies, the follow-up times have been relatively short 15,17,18 and they have paid less or no attention to accumulated exposure. We extended the investigations by utilizing the nationwide data on AM purchases from well-defined Finnish national registries as an objective measure of lifetime AMs exposure 19 and evaluated its association with weight status in early adolescence.

| METHODS
This study was conducted among participants of the Finnish Health in Teens cohort (Fin-HIT), which is a school-based cohort of mainly 9 to 12 years old children without specific exclusion criteria. In total, 11 407 children and one of their parents (n = 9935, mostly mothers) participated the baseline study in 2011 to 2014. Details on the Fin-HIT cohort are described elsewhere. 13 The Coordinating Ethics Committee of the Hospital District of Helsinki and Uusimaa has approved the study protocol (169/13/03/00/10) including linkage to national health registers, and written informed consent was obtained from all participants and their parents.

| Variables in the study
Age in years, sex and language spoken at home (Finnish, Swedish and others) for children and mothers were obtained from the consent form and were confirmed by record linkages with the National Population Information System at the Population Register Centre. Mothers' anthropometric measures and educational level were obtained using a questionnaire at baseline data collection. Children's height (in m) and weight (in kg) were measured in a standardized way by trained field workers. Children's BMI (weight in kg divided by height in m 2 ) to describe weight status was categorized as thin, normal weight, overweight and obese according to the International Obesity Task Force. 20 However, we combined children with overweight and obesity (denoted as overweight) in the analyses because the number of children with obesity was small (2.4%). Not having weight status is an exclusion criterion for the sample size in the present study.

| Register data on antimicrobials
Children's and mothers' information on AM purchases were obtained from the Drug Prescription Register maintained by the Social Insurance Institution of Finland (KELA). The data utilizes Anatomical Therapeutic Chemical classification system held by WHO, 22 and date of each purchase. The registry does not include medicines that are not reimbursable nor medicines that are inexpensive and cost less than the set reimbursement level. 19 The linkage to the register data has been made through the participants' unique social security numbers.
Medications prescribed for the child are registered with his/her social security number, and those prescribed to the mother are registered with the mother's social security number. Medical purchase history was available from 11 212 children and 7508 mothers from January Date of purchase is considered to reflect the use of AM. Together with children's birth date, we created following variables: for each participant, we summarized AM use for the first 6 months, for each full year of age, and total use from birth to early adolescence (eg, here referred as lifetime use) as main exposures. At each full year of age, the number of AM purchases was dichotomized (1/0) to describe the prevalence of AM use. Mothers' AM use during pregnancy reflecting prenatal exposure was studied as a secondary aim, while postnatal maternal AM use was considered as a confounder. Other medical purchases excluding AMs were summarized and used as a proxy of overall morbidity of the child in further analyses.
Data on Medical Birth Registry held by the National Institute for Health and Welfare were used to confirm that the children were born in Finland (n = 10 944) and to exclude extremely preterm born children (<28 weeks, n = 12), as these might affect the results. Thus, records were considered for 10 772 children and 7220 mothers ( Figure S1).

| Statistical analysis
Categorical variables between weight status were compared using chi-square tests. Continuous variables were compared between two groups using t-test. Comparisons of continuous variables between more than two groups were carried out using ANOVA, and it was adjusted by Brown-Forsythe test when appropriate. Prevalence of AMs at different ages were compared between weight status using chi-square test, and residual analysis was carried out to identify differences between the categories. In these analyses, Bonferroni correction was used to adjust for multiple tests. Association between variables related to the use of AM and the children's weight status was estimated through odds ratio (OR) and 95% confidence intervals (95% CI) by multinomial logistic regression, as weight status was evaluated in three categories. Multivariate models were created for children with information on weight status (ie, BMI, gender and age) (n = 10 093) and were adjusted for age, gender, language spoken at home and number of medical purchases other than AMs ( Figure S1).  When mother's use of AM during pregnancy was included, models were estimated for those who had mother's information (n = 6708) available and were adjusted for age, gender, language spoken at home and mother's age.
All statistical analyses were carried out using SPSS version 25, and a 5% statistical significance level was adopted.

| RESULTS
The average age for children in baseline study visit was 11.  Figure 1B). Figure 2 shows that the penicillins with extended spectrum, macrolides, and combinations of penicillins were the most commonly prescribed AMs during the first years of life. Penicillin with extended spectrum had the highest frequency of use up to 13 years of age.
We observed no association between maternal and child AM purchases during the follow-up time ( Figure 3A). Children with overweight had a higher number of AM purchases compared with other weight status at least until 12 years of age ( Figure 3B), while no differences in children's AM purchases were observed by mothers' weight status ( Figure 3C). Interestingly, the number of purchases for children with foreign home language was lower especially during early childhood compared with Finnish and Swedish speaking children ( Figure 3D).
The prevalence of AM purchases was compared between children's weight status (Table S1): higher prevalence in overweight children and lower prevalence in thin children are observed in most ages.
In Table 2, we further compared baseline characteristics between children's weight status. The higher the weight status, the higher percentage of children and mothers used AMs during the first 6 months of child's life. In addition, the first exposure to AMs occurred at earlier stage in overweight group than with other weight status. On the other hand, mother's AMs use during pregnancy was not associated with child's weight status. Table 3 shows the results for several multivariate models evaluated in this study. In all models, the total number of AM purchases was positively associated with overweight and inversely associated with thinness (P < 0.05). We found that children's AMs use during the first 6 months (OR = 0.80; 95% CI: 0.67-0.95) and when they have completed 2 years (OR = 0.83; 95% CI: 0.72-0.96) were inversely associated with thinness. However, these variables did not remain statistically significant in the models after adjusting for total number of prescriptions (models 5 and 6), pointing out that the total number of prescriptions had a greater impact on BMI than the early use. Furthermore, the association of lifetime AM purchases with preadolescent's BMI was not affected by intrauterine exposure to AM. In the models evaluating number of purchases separately for the four main AM types, we observed that penicillin with extended spectrum and first generation cephalosporins were consistently positively associated with overweight and inversely associated with thinness (P < 0.05).

| Multinomial regression models
This was not the case for macrolides.

| DISCUSSION
In this cohort study with registry-based medical purchase history, the highest prevalence of AMs use occurred during the first 2 years of life and thereafter decreased with age in children. This is the most extensive report on the association between AM exposure from the fetal period to early adolescence and weight status in Finnish children. We The main strength of our study is its large data size with a long follow-up time: we have collected AM use from the national Drug Prescription Register from birth to early adolescence for 10 000 children in Finland. The records were available for 98% of the participants, and these were linked with measured BMI at the mean age of 11 years, which gives us a strong, objective data on the exposure and outcome. These we have supplemented with data from Medical Birth Registry which allowed to limit the analyses to children born in Finland, and exclude those with extremely preterm birth, and with comprehensive background information supporting and allowing more complete modelling. Unfortunately, we do not have data on the disease or indication for AMs use, but since it is prescription-based and assumed rational in terms of dosing and duration. It has been speculated, for example, 18 that the association between high AMs use and overweight might reflect a higher tendency for people with overweight to seek for medical care. We assumed that the sum of other medical purchases might mirror overall morbidity, and those with high number of other medications may have a chronic disease or seek medical care more often, which we wanted to consider in the analysis. However, this was not the case, since the associations remained at the same level. One of the study limitations is that we lacked information on whether AMs were taken as prescribed as we