Has underweight increased among children and adolescents during 2012–2019, and does the prevalence depend on socioeconomic circumstances? A nationwide register-based study from Denmark

Abstract

Background

Childhood underweight, indicated by low body mass index (BMI) for age and sex, is associated with morbidity, mortality and poverty in low-income settings, but is often overlooked in high-income countries, due to unequivocal focus on overweight. This study examined the prevalence and socioeconomic characteristics of underweight among children and adolescents in Denmark during 2012–2019.

Methods

We used nationwide Danish register data on height and weight measured objectively in 4 61 041 6–7 year-olds and 3 65 312 14–15 year-olds during 2012–2019, covering 89.4% and 67.2% of all schoolchildren in these age groups. We calculated BMI z-scores and classified underweight using the International Obesity Task Force age- and sex-specific BMI cut-offs and linked with data on parental education and household income as indicators of socioeconomic circumstances.

Results

The prevalence of underweight remained relatively stable at~8.5% among 6–7-year-old girls during 2012–2019. Among 14–15-year-old girls, it was initially higher but decreased slightly (<1%-point) to 8.6% in 2019. Among boys, the prevalence increased from ~7% to just above 8% in both age groups during the period. Across all years, underweight was more common among 14–15 year-olds whose parents had long higher education (girls: 10.4%–12.1%; boys: 8.7%–10.2%) compared with primary school (girls: 5.6%–7.5%; boys: 6.1%–7.8%), with the most pronounced differences among girls. Similar differences across parental education were observed among 6–7 year-olds, although they were smaller. Less distinct but comparable differences in underweight prevalence were found across household income.

Conclusions

Underweight is prevalent but has remained relatively stable during the 2010s among Danish children and adolescents. Unlike the pattern reported in low-income settings, and that seen for overweight, underweight is associated with socioeconomic advantage in this high-income setting.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Despite extensive research among children with underweight in low-income countries and the U-shaped association between BMI and morbidity in adults, underweight in children is largely overlooked in high-income countries.

WHAT THIS STUDY ADDS

• This study examines time trends in the national prevalence of underweight among 6–7 and 14–15 year-olds in Denmark based on objectively measured anthropometry from 2012 to 2019. We show only minor fluctuations in underweight (~8%) and minimal changes in the BMI z-score distribution across the 2010s. Moreover, we document that children and adolescents from more advantaged socioeconomic circumstances are more likely to have underweight compared with their less advantaged peers.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The non-negligible proportion of children and adolescents with underweight in a high-income setting like Denmark highlights the need for a better understanding of the nutritional and health impacts of underweight throughout childhood and adolescence. Moreover, public health interventions designed to prevent childhood overweight and obesity should ensure they do not simultaneously increase the incidence of underweight until its health impacts are better understood.

Introduction

Underweight in childhood, also known as ‘thinness’, is indicated by low body mass index (BMI) relative to age and sex¹ and is associated with malnutrition, morbidity, mortality and poverty among children in low- and middle-income countries.^{2 3} Even with the rise in childhood obesity during the past decades, underweight has been shown to be consistently more prevalent than obesity among Danish schoolchildren during 1936–2011.⁴ Although 5%–11% of children and adolescents have been reported to have underweight in high-income countries4,8 and a recent meta-analysis indicated increased prevalence during the last 10–15 years in European countries,⁹ the characteristics and consequences of underweight are highly overlooked in the scientific literature. This is plausible due to the unequivocal focus on childhood overweight in these settings and that many studies include children with underweight in the normal weight category. However, a few existing studies in high-income settings, such as the UK and Japan, indicate that childhood underweight is associated with more frequent use of general practitioners,¹⁰ lower bone mineralisation¹¹ as well as poorer academic performance and mental health outcomes^{12 13} but also lower risk of respiratory illnesses.¹⁴

Socioeconomic circumstances (SECs) in childhood, as indicated by factors such as household income, parental education and occupation, have been linked to physical and mental health in various phases of the lifecourse^{15 16} and are therefore important to consider in child health. It is well established that overweight and obesity are associated with more disadvantaged SECs in childhood and adolescence.17,19 However, limited research has investigated the socioeconomic characteristics of underweight and with less consistent findings.20,24 The potential health consequences of underweight in a population may depend on the SECs that it is connected to. As these circumstances link to nutritional, behavioural and health-related factors, it is important to investigate.

Using nationwide register data of objectively measured height and weight, our aim was to examine time trends in the sex-specific prevalence of underweight and BMI z-score distribution among Danish children and adolescents from 2012 to 2019. We also investigated socioeconomic differences in underweight and whether this changed during the investigated time period.

Materials and methods

Study design and participants

We used data from the Children’s Database, which holds information on individuals participating in the mandatory school health examinations offered to all children at school entry (6–7 year-olds) and graduation (14–15 year-olds) in Denmark. In Denmark, every resident is given a unique civil registration number at birth or immigration, making it possible to link children with individual-level register information on their legal parents.²⁵ We linked data from the child health examinations with parental socioeconomic information from the following Danish registers: (1) the Population Education Register, which holds information about the educational level of all Danish residents²⁶ and (2) the Income Statistics Register, which contains annual data on the income composition of all Danish residents.²⁷

We included school measurements of height and weight, as well as age and sex data, from 2012 to 2019. Recording to the Children’s Database became mandatory in December 2011 and onwards for all municipalities in Denmark, but the national lockdowns implemented to mitigate COVID-19 in 2020 and 2021 affected schools and limited the ability to conduct the measurements. During 2012–2019, a total of n=461 041 6–7 year-olds and n=365 312 14–15 year-olds were registered with height and weight measurements at school entry and graduation, respectively. This covered 89.4% and 67.2% of all children entering and leaving the Danish primary school system, respectively, during the study period.

Study outcomes

Objective measurements of height and weight were conducted by school health nurses or physicians. Height was measured without shoes to the nearest cm, and weight was measured in light clothing to the nearest 0.1 kg using standardised procedures. We calculated BMI as the ratio of bodyweight (kg) to the square of height (m). We also calculated age- and sex-specific BMI z-scores using the formula and method proposed by Cole and Lobstein.²⁸

Weight status was categorised into underweight, normal weight, overweight and obesity based on the widely used age- and sex-specific BMI cut-offs for children, defined by centiles that intersect at adult BMIs of 18.5 kg/m², 25 kg/m² and 30 kg/m², proposed by Cole and Lobstein²⁸ and the International Obesity Task Force (IOTF).²⁹ Thus, 6–7 and 14–15 year-olds with BMI below the age- and sex-specific cut-off corresponding to the adult of 18.5 were categorised as having underweight. We further subdivided underweight into grade 3, grade 2 and grade 1, corresponding to the adult BMI cut-offs of 16 kg/m², 17 kg/m² and 18.5 kg/m²,¹ using sex-specific BMI cut-offs for exact age in months.³⁰ For individuals with multiple measurements at school entry (n=167 830) or graduation (n=35 483), we used the latest recorded measurement from the database to include the most up-to-date height and weight information in the analysis.

Independent variables

We used two indicators of childhood SECs: parental educational level and household income.

Parental educational level

Annual individual-level information on education is collected in November, the year before it is registered in the Population Education Register.²⁶ Therefore, we linked information about the legal parents’ education using the same calendar year as the school measurement of height and weight. Using Statistics Denmark’s education classification system, which is based on the International Standard Classification of Education,³¹ parental education was categorised into primary school, vocational education, high school and short higher education (including short-cycle tertiary education), medium higher education (including bachelor degree or equivalent) and long higher education (including master and doctoral degrees or equivalent). Educational qualifications of the parent with the highest attained level were used in this study. Among the included 6–7 and 14–15 year-olds, 10 054 (2.2%) and 5366 (1.5%), respectively, had missing information on parental education at the year of school measurement and were therefore not included in the socioeconomic analysis (online supplemental figure 1).

Household income

We used information on annual household income from the Income Statistics Register²⁷ recorded in the year prior to the school measurement at the registered home address. Income was based on the equivalised disposable family income, which denotes the total yearly income of a household after taxes and other deductions divided by the household size. In this calculation, family members are equalised by weighing each member according to their age.³² In the present study, household income was categorised into quartiles for every year from 2011 to 2018. The amounts presented in the descriptive analysis were discounted to 2019 prices to account for inflation, using pricing information from Statistics Denmark. In this study, 2078 (0.5%) and 940 (0.3%) of the 6–7 and 14–15 year-olds, respectively, had missing information about household income the year prior to their school measurement and were not included in the socioeconomic analysis (online supplemental figure 1).

Statistical analysis

Sex-specific characteristics of the study population are presented for each age group separately and include means and SDs for symmetrically distributed variables, medians and IQR for skewed variables and n (%) for categorical variables.

To examine time trends in the sex-specific prevalence of underweight among children and adolescents, we calculated the yearly prevalence for each sex and by age group during 2012–2019. We also did this analysis across the three grades of underweight. However, due to the low number of schoolchildren with grades 2 and 3 underweight (online supplemental table 1), the main analyses were based on the composite measure of underweight (grade 1 or more). In this study, we decided to leave out significance tests and CIs because of the nationwide coverage of the registers and large sample size, which imply high precision in the estimations, whose importance should be interpreted with regard to clinical and public health relevance rather than statistical significance. To investigate trends in the age- and sex-specific BMI distribution, BMI z-score distributions by age group and sex and per year of measurement were presented using density plots. For an alternative visualisation, we also conducted this analysis using cumulative BMI z-score distribution plots. In the density and cumulative plots, we truncated the tails of the distribution curves at −5 SD and +5 SD to comply with the non-disclosure policy and recommendations by Statistics Denmark.

To examine socioeconomic differences in underweight and whether this changed over time, we calculated the proportion of individuals with underweight according to parental education and household income for each year during 2012–2019. We also examined socioeconomic differences in underweight grade 1 versus grades 2 and 3 (combined) in a supplementary analysis. Due to the few children with grade ≥2, this analysis was based on combined data from all measurements conducted during 2012–2019, instead of being conducted over time. The results are presented in graphs for each indicator separately and by age group and sex.

Statistical analyses were conducted using the statistical software Stata V.18.0 and R V.4.3.2.

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this study.

Results

On average, 57 630 6–7 year-olds were measured each year at school entry and 45 664 14–15 year-olds at graduation during 2012–2019 (table 1). Together with 14–15-year-old boys, notably fewer girls in both age groups had underweight compared with overweight, with the highest overweight prevalence among the 14–15 year-olds, whereas 6–7-year-old boys had similar proportions of underweight and overweight (table 1). Across age group and sex, more than 45% of the children came from households with an educational level corresponding to a bachelor’s degree or higher, while few came from households with only primary school education (table 1). Also, household income was generally high compared with other European countries³³ (table 1).

Table 1. Characteristics of the study population.

Characteristics	Girls		Boys
	6–7 year-olds	14–15 year-olds	6–7 year-olds	14–15 year-olds
	(n=225 962)	(n=179 160)	(n=235 079)	(n=186 152)
Age in years, median (IQR)	7.1 (6.6–7.6)	14.9 (14.6–15.3)	7.1 (6.7–7.6)	15.0 (14.6–15.4)
Height in cm, mean (SD)	125.0 (6.2)	165.9 (6.4)	126.2 (6.2)	174.5 (8.1)
Bodyweight in kg, median (IQR)	24.5 (22.0–27.5)	56.9 (51.3–63.7)	25.0 (22.6–27.8)	62.0 (55.0–70.0)
BMI z-score*, mean (SD)	0.26 (0.95)	0.37 (0.99)	0.19 (0.93)	0.39 (1.04)
Weight group*†, n (%)
Underweight	2660 (8.5)	2050 (8.5)	2746 (8.4)	2051 (8.1)
Grade 1	2286 (7.3)	1658 (6.9)	2394 (7.3)	1707 (6.7)
Grades 2 and 3	374 (1.2)	392 (1.6)	352 (1.1)	344 (1.4)
Normal weight	24 016 (76.7)	17 647 (73.1)	26 524 (81.0)	18 702 (73.5)
Overweight	3655 (11.7)	3603 (14.9)	2754 (8.4)	3629 (14.3)
Obesity	998 (3.2)	851 (3.5)	729 (2.2)	1.073 (4.2)
Parental education*, n (%)
Primary education	2068 (6.8)	1493 (6.3)	2013 (6.4)	1609 (6.5)
Vocational education	8154 (27.0)	7879 (33.4)	8671 (27.5)	8207 (33.0)
High school and short higher education	3929 (13.0)	3348 (14.2)	4083 (12.9)	3537 (14.2)
Medium higher education	8806 (29.2)	6604 (28.0)	9148 (29.0)	7147 (28.7)
Long higher education	7233 (24.0)	4293 (18.2)	7628 (24.2)	4405 (17.7)
Yearly disposable household income in Euro*‡§, median (IQR)
Q1	19 685 (16 334–22 667)	19 239 (15 932–22 134)	19 649 (16 449–22 646)	19 200 (15 984–22 041)
Q2	29 664 (27 570–31 616)	29 307 (27 041–31 325)	29 577 (27 541–31 587)	29 263 (27 022–31 428)
Q3	37 639 (35 491–39 986)	37 681 (35 461–40 125)	37 562 (35 518–39 930)	37 572 (35 437–40 036)
Q4	51 089 (46 035–60 584)	52 214 (46 820–63 264)	51 488 (46 273–61 430)	52 217 (46 773–63 558)
Schoolchildren measured per year, n
2012	23 866	17 386	25 314	18 104
2013	29 102	23 426	30 377	23 874
2014	31 858	25 950	32 993	27 015
2015	30 660	24 716	32 257	25 275
2016	31 329	24 151	32 753	25 455
2017	25 849	22 203	26 972	22 888
2018	27 289	21 843	28 252	22 765
2019	26 009	19 485	26 161	20 776

^*Based on individuals with measurements in 2016. Year 2016 was chosen to represent a year in the middle of the investigated time period 2012–2019.

^†Weight group was categorised as underweight, normal weight, overweight and obesity according to the age- and sex-specific cut-offs defined by centiles passing through a BMI of 18.5 kg/m2, 25 kg/m2 and 30 kg/m2 at the age of 18, proposed by Cole and Lobstein28 and the International Obesity Task Force (IOTF) classification.29

^‡Based on yearly equivalised disposable family income, which denotes the total income of a household after taxes and other deductions divided by the household size.

^§Discounted to 2019 prices.

BMI, body mass index.

During 2012–2019, the prevalence of underweight was relatively stable among 6–7-year-old girls, ranging between 8.3% and 8.7% (figure 1). Among 14–15-year-old girls, there was a slightly decreasing trend from 9.4% in 2012 to 8.6% in 2019 (figure 1). In contrast, among boys, the prevalence of underweight increased slightly, that is, from 7.0% in 2012 to 8.1% in 2019 in 14–15-year-old boys and the same trend was seen for 6–7-year-old boys (figure 1). Across all years, age and sex, the majority of individuals with underweight had grade 1, that is, the mildest form of underweight, and only around 1%–2% had grade ≥2 (online supplemental table 1).

Figure 1. Sex-specific prevalence of underweight among Danish schoolchildren aged 6–7 and 14–15 years, 2012–2019.

Overall, no substantial shifts in the BMI z-score distributions were observed across time (figure 2; online supplemental figure 2). However, among 6–7 year-olds, a slightly lower peak of the density curve was observed in 2019 compared with previous years, along with a slightly longer tail towards higher BMI values (more right-skewed) (figure 2). This indicates a higher proportion of children in the heavier weight groups together with fewer in the normal weight category, while the proportion of children with underweight remained largely unchanged. A similar tendency was observed among 14–15 year-olds, but less distinct (figure 2; online supplemental figure 2).

Figure 2. Yearly age- and sex-specific distributions of BMI z-scores, 2012–2019.

In relation to socioeconomic differences, the data showed a gradient with a higher prevalence of underweight with a higher level of parental education across all years, age and sex (figure 3). This was most pronounced among girls, especially the 14–15 year-olds, where the prevalence of underweight was approx. 1½- to 2-fold higher among those from households with long higher education compared with primary school (figure 3). The 14–15-year-old girls from households with the highest level of education had the highest prevalence of underweight consistently across all years, ranging between 10.4% and 12.1% (figure 3). There was no apparent change over time in the socioeconomic pattern of underweight across both sexes and age groups.

Figure 3. Sex-specific prevalence of underweight by age group and according to parental educational level, 2012–2019.

When using household income as an indicator of SECs, we found a similar gradient with the most pronounced socioeconomic differences among girls, especially the oldest girls, compared with boys (figure 4). However, the differences in underweight were generally smaller between income groups than observed between education groups, especially among boys (figures3 4).

Figure 4. Sex-specific prevalence of underweight by age group and according to household income, 2012–2019.

In the analysis of SECs associated with underweight grade 1 versus grades 2 and 3, we found the same pattern for grade 1 as for the composite underweight measure: higher parental education and household income were associated with higher underweight prevalence across both sexes and age groups (online supplemental figures 3 and 4). Likewise, these differences were most pronounced across parental education and among 14–15-year-old girls, with 5.2% versus 9.3% and 5.7% versus 8.5% between the lowest and highest levels of parental education and household income, respectively (online supplemental figures 3 and 4). In contrast, no evident socioeconomic differences were found in underweight grades 2 and 3, with differences in prevalence of only up to 0.3%-point across SECs (online supplemental figures 3 and 4).

Discussion

This nationwide register-based study showed that the prevalence of underweight was substantial (~8%) among Danish children and adolescents during the 2010s, but also that only 1%–2% was severely underweight, corresponding to an adult BMI below 17. We furthermore showed that the prevalence of underweight was relatively stable from 2012 to 2019, with the exception of a slight increase among boys and a slight decrease among adolescent girls. This was also reflected in a relatively stable BMI z-score distribution, which showed a small increase in the proportion of children with higher BMI but not fewer with underweight over the period. Finally, we found that underweight was most prevalent among children and adolescents from households with the most advantaged SECs. The socioeconomic gradient was stable over time, being most pronounced among girls, especially those aged 14–15 years and for underweight grade 1.

Previous studies on the prevalence and trends of underweight among children and adolescents in high-income settings during overlapping time periods have shown mixed results. In Swedish children and adolescents objectively measured during 2004–2015, Eriksson et al³⁴ found a relatively stable underweight prevalence, which is comparable to our findings. Likewise, Holstein et al³⁵ reported a fairly stable underweight prevalence among Danish schoolchildren based on self-reported height and weight data from 1998 to 2018. However, both studies defined childhood underweight as grade ≥2.^{34 35} In the Danish study, the overall underweight prevalence was 3% in boys and 5% in girls, which is in contrast to the prevalences of grades 2 and 3 underweight in our study at around ~1%–2%. Likewise, the Swedish study also reported lower prevalences of grades 2 and 3, ranging from 0% to 2%,³⁴ and similar results were observed in Norwegian children, with <2% having these more extreme grades of underweight.⁶ Like our findings, but investigating an earlier time period, Aarestrup et al⁴ also showed that underweight Danish children primarily fell into grade 1. In contrast to our findings, the US National Health and Nutrition Examination Survey indicated a downward trend in underweight among American children aged 6–11 years between 2009–2010 and 2017–2018 and a slight increase among those aged 12–19 years during the same period.³⁶ However, their sex-specific analyses, like ours, documented a slight increase in underweight among boys, while the prevalence was largely stable among girls.³⁶ Furthermore, a meta-analysis based on European children and adolescents showed an overall upward trend in underweight from 7% during 2000–2006 to 9% during 2007–2017 among 2–13 year-olds and from 6% to 8% among 14–18 year-olds.⁹ However, the authors mainly found increases in Eastern, Northern and Southern Europe, whereas in Western Europe childhood underweight tended to decrease.⁹

Using data from the decades leading up to our study period, a recently published Danish study analysed school measurements from 1936 to 2011 and found that the prevalence of underweight decreased from 10% to 7% among 7-year-old girls and from 9% to 6% in boys.⁴ Among 13-year-olds, underweight decreased from 11% to 10% in girls and from 10% to 6% among boys.⁴ When combined with our results, this suggests that the decreasing trend in underweight among Danish schoolchildren may have plateaued in the 2010s, especially among girls at school entry, and may even be slightly increasing among boys. Furthermore, similar to our findings, the study documented an increasingly right-skewed BMI distribution over time; however, these shifts were more significant than our findings and most pronounced among the oldest children aged 13 years.⁴

Similar to the socioeconomic differences found in the present study, de Ruiter et al²³ reported higher underweight prevalence among children from more advantaged SECs, especially among girls, using data from Spanish children in the 2010s. In contrast, studies among British,²⁴ Australian²⁰ and Dutch²² children and adolescents showed no clear socioeconomic differences in underweight. When examining SECs across different grades of underweight, our findings align with those of Holstein et al,³⁵ who also found no overall association between grade ≥2 underweight and SECs among Danish schoolchildren. Similarly, no association between SECs and grade ≥2 was reported in British schoolchildren born between 2000 and 2006.³⁷ In contrast, Pearce et al²¹ reported a higher likelihood of underweight in British children from more disadvantaged SECs, with the greatest socioeconomic disparity among those with grade ≥2. Taken together with our findings, these results suggest that the socioeconomic pattern of childhood underweight may vary with its severity, but further research on this is needed.

Our findings indicate that factors related to SECs among Danish schoolchildren are linked to the prevalence of underweight, particularly grade 1 underweight. One possible explanation for this association could be that more advantaged families have higher health literacy and adopt healthier lifestyles, which may also lead to a strong focus on healthy living and weight management resulting in lower bodyweight in these households. A study of British children aged 2–15 years found parental weight status to be a strong predictor of childhood underweight;³⁸ however, it remains unclear to what extent this was due to genetic versus environmental and cultural factors. Among Australian schoolchildren, those with underweight, compared with normal weight, were less likely to desire getting fitter or be encouraged by parents to exercise but were also less likely to perceive their weight as appropriate.²⁰ This is a concern, as negative perception of weight and body image are predictors of later disordered eating behaviours and other weight-related issues.³⁹ Underweight was most prevalent among 14–15-year-old girls across the study period, despite a slight decline in prevalence over time (<1% point). However, we also observed a small increase (~1% point) in underweight among both age groups of boys. This increase could indicate changing societal expectations, with growing pressures on boys to achieve a leaner physique, but further research is needed to better understand this change.

A major strength of this study is the use of nationwide height and weight data on Danish schoolchildren over multiple years, minimising selection bias and allowing us to examine trends in underweight and BMI distribution in a representative population of Danish schoolchildren. Furthermore, information on height and weight was derived from objective measurements carried out by health professionals and thus not subject to bias due to self-report.⁴⁰ Due to the unique civil registration number assigned to all residents in Denmark, we were able to link the study population with official individual-level data on parental education and household income. The children included in this study were representative of Danish children in general in terms of parental education.⁴¹ We used the international IOTF BMI cut-offs,^{28 29} making results from this study comparable with other studies in the literature. However, the low number of children with underweight grades 2 and 3 limits some of the results. Due to this, it was not possible to analyse SECs associated with these grades separately. Additionally, to prevent data disclosure, the analysis of SECS across underweight grade 1 versus grades 2 and 3 (combined) could not be conducted across time.

In conclusion, we found that underweight among Danish schoolchildren is prevalent (~8%) and has remained relatively stable during the 2010s, with a minor decline in 14–15-year-old girls and a minor increase among boys aged 6–7 and 14–15 years. Further, grades 2 and 3 underweight is rare in this population and has been stable during 2012–2019. Similarly, the BMI distribution has been relatively stable during this period, but with a slight increase in children with higher BMI, and not fewer with underweight. Furthermore, underweight was consistently most prevalent among children and adolescents from households with the most advantaged SECs, across all years investigated, and particularly for grade 1. These findings do not indicate that childhood underweight is a growing public health issue among Danish schoolchildren; however, the slight increase among boys and the social gradient in the prevalence warrant some attention. Additionally, our findings show that schoolchildren with underweight constitute a non-ignorable proportion of the population. Public health strategies aimed at preventing childhood overweight and obesity should ensure that they do not simultaneously increase the burden on the lower end of the BMI spectrum until we have a better understanding of the health impacts of childhood underweight.

Data availability statement

Data may be obtained from a third party and are not publicly available.