New Finnish growth references for children and adolescents aged 0 to 20 years: Length/height-for-age, weight-for-length/height, and body mass index-for-age

Abstract

Background and objectives. Growth curves require regular updates due to secular trends in linear growth. We constructed contemporary growth curves, assessed secular trends in height, and defined body mass index (BMI) cut-off points for thinness, overweight, and obesity in Finnish children.

Material and methods: Mixed cross-sectional/longitudinal data of 73,659 healthy subjects aged 0–20 years (born 1983–2008) were collected from providers in the primary health care setting. Growth references for length/height-for-age, weight-for-length/height, and BMI-for-age were fitted using generalized additive models for location, scale, and shape (GAMLSS). BMI percentile curves passing through BMIs 30, 25, 18.5, 17, and 16 kg/m² at the age of 18 years were calculated to define limits for obesity, overweight, and various grades of thinness.

Results. Increased length/height-for-age was seen in virtually all age-groups when compared to previous Finnish growth data from 1959 to 1971. Adult height was increased by 1.9 cm in girls and 1.8 cm in boys. The largest increases were seen during the peripubertal years: up to 2.8 cm in girls and 5.6 cm in boys. Median weight-for-length/height had not increased.

Conclusions. New Finnish references for length/height-for-age, weight-for-length/height, and BMI-for-age were constructed and should be implemented to monitor growth of children in Finland.

Key words::

• BMI
• child
• growth reference
• height
• length
• weight

Abbreviations
AIC	=	Akaike information criteria
BCPE	=	Box-Cox power exponential
BIC	=	Bayes information criteria
BMI	=	body mass index
GAMLSS	=	generalized additive models for location, scale, and shape
GAIC	=	generalized AIC
IOTF	=	International Obesity Task Force
LMS	=	method of distribution using skewness (L), median (M), and coefficient of variation (S)
SD	=	standard deviation
SDS	=	standard deviation score
WHO	=	World Health Organization

Key messages

• Contemporary, population based height (length/height-for-age) and weight (weight-for-length/height and body mass index-for-age) references were constructed for Finnish children and adolescents aged 0 to 20 years.

• BMI-for-age cut-off points for thinness, overweight, and obesity were defined for Finnish children.

• Because of significant secular trends in height, growth curves currently used in Finland should be replaced by these new references.

Introduction

Monitoring of growth has been a part of preventive child health care for over a century. In Finland, growth has been monitored systematically for nearly as long. The growth references currently in use in Finland were launched in 1986 and are based on height and weight measurements of children born between 1959 and 1971 (1–3). Even in recent years, growth surveys conducted in both developed and developing countries have reported increases in mean height and earlier sexual maturation, a phenomenon termed a ‘positive secular change’ (4–8). It is generally assumed that the positive secular trend arises from environmental changes in factors inhibiting full expression of the growth potential, e.g. inadequate nutrition and morbidity (9).

Despite the secular influence of environmental factors on height, stature is strongly genetically determined in the general population, with heritability estimated at approximately 80% (10). To date, several genome-wide association studies on stature have been published, mostly on Caucasian-based populations (11–15). These studies have identified more than 50 well replicated loci associated with stature in the general population. Because of the polygenic nature of stature, there are notable differences in the average height between populations with different genetic backgrounds.

Due to secular trends in height, child growth references must be updated periodically. Due to the paramount importance of genetic factors on height, it is imperative that these references are based on a population with the same ethnic and genetic backgrounds as the intended population.

Along with reports on the positive secular trend in height, Western countries have been awakened by an epidemic of obesity (16,17). In Finland, the prevalence of overweight adolescents has almost doubled in the past two decades (18). Overweight and obesity in childhood often persist into adulthood and are known as risk factors for chronic diseases such as cardiovascular disease and type 2 diabetes (19–21). Body mass index (BMI) (weight/height², kg/m²) is considered the best tool to monitor the epidemiology of overweight and obesity in both children and adults because of its independence of height, correlation to body fat, and prediction of mortality (22–24). Due to changing body composition in childhood and adolescence, the reference values for BMI are age- and sex-specific. However, exact knowledge of BMI thresholds related to morbidity in childhood is still lacking (21,25–28). The most widely used adult cut-off points are a BMI of 25 kg/m² for overweight and 30 kg/m² for obesity, values also related to increased health risks (16). An International Obesity Task Force (IOTF) expert panel proposed that BMI-for-age percentile curves passing through adult values of 25 kg/m² and 30 kg/m² could be used for defining childhood overweight and obesity, respectively (29,30). Corresponding multinational curves for grade 3, 2, and 1 thinness were constructed by Cole et al. (30), and these were defined as BMI-for-age percentiles passing through adult values of 16, 17, and 18.5 kg/m², respectively (31).

As opposed to most Western countries, weight in Finnish children has been assessed as percentage of the median weight-for-length/height, instead of BMI-for-age percentiles (1–3). Data correlating BMI and weight-for-length/height in childhood are scarce. A single study from the United States showed a poor correlation indicating that these two end-points are not interchangeable (32). Clinically, there is a need for a Finnish BMI reference to monitor weight in children and adolescents.

The objectives of this study were:

• To assess the timing and magnitude of the secular trend in height in Finland by comparing the growth of a contemporary healthy Finnish child population born between 1983 and 2008 to the current Finnish growth reference which includes subjects born between 1959 and 1971.

• To construct new Finnish growth references for length/height-for-age, weight-for-length/height, and BMI-for-age.

• To formulate cut-off curves for defining thinness, overweight, and obesity for Finnish children and adolescents based on BMI values of 16, 17, 18.5, 25, and 30 kg/m² in adults.

Materials and methods

Population and measurements

Data for the present study were collected from providers of public primary care in the city of Espoo, Finland's second largest city with a population of 241,600 inhabitants. With a significant net migration from all parts of Finland, its population has grown 10.6-fold over the past 60 years. The majority of the population (94.4%) is of Finnish origin, which mirrors the whole of Finland (97.3%).

The Finnish social security system provides regular, free-of-charge visits to public primary care child health clinics and school health care to permanent residents of Finland regardless of social status or income level. Primary care nurses in Finland are specially trained in child health care and health prevention, and their duties include assessment of health and development at scheduled visits including standardized weight and height measurements. Children in Espoo have regular visits at child health clinics at the ages of 1 to 2 weeks; 3 to 6 weeks; 6 to 8 weeks; at 2, 3, 4, 5, 6, 8, 10, 12, 18, and 24 months, and then at 3.5, 5, and 6 years of age. Visits with school nurses are scheduled annually. Children may have also extra visits if special health concerns are suspected.

At each scheduled visit, length/height and weight are measured using standardized techniques and calibrated equipment. In infants up to age 20 to 24 months, length is measured to the nearest 0.1 cm in a fully extended supine position with heels in contact with a baseboard. Infants are weighed without clothing on calibrated baby scales, and the weight is rounded to nearest 0.005 kg. From the age of 24 months, standing height is measured to nearest 0.1 cm using standardized stadiometers, and standing weight is measured on calibrated mechanical or electronic step scales and rounded to nearest 0.1 kg. Since 2003, all measurements in the Espoo area had been captured in a networked electronic patient management system known as Effica (Tieto Ltd, Finland). The weight and length measured in the birth hospital, as well as data on premature birth, were also recorded in the Effica program at the first visit to the child health clinic.

Permission for the present study was obtained from Espoo Municipality Institutional Review Board. No contact was made with the study subjects since the data were handled anonymously.

Growth reference data collection and database cleaning

For the weight-for-length/height and BMI-for-age references, we used the auxological data from Espoo recorded in the Effica system between 1 January 2003 and 12 May 2009. All potentially false measurements, typing errors, missing values, or duplicate recordings were corrected or excluded. The initial database for weight-for-length/height and BMI-for-age reference contained 561,392 length/height and weight measurements from 75,810 subjects aged 0 to 20 years. Data cleaning occurred in two phases (Figure 1).

Figure 1. Data cleaning procedure for the weight-for-length/height and BMI-for-age reference database.

First, the primary health care nurses retrospectively evaluated the eligibility of each subject for the new growth reference. These nurses were specifically trained by a pediatric endocrinologist (L.D.) to exclude medical diagnoses or medications that could have possibly interfered with growth. Second, any potential bias as a result of repeatedly measuring the same subject was eliminated by grouping the measurements into 29 time slots according to the official visiting schedule in the Espoo child health clinics and school health care. Only one measurement per child per time slot was selected (the closest visit to the correct visiting age). The weight-for-length/height and BMI-for-age references were constructed from the resulting mixed cross-sectional and longitudinal database comprising 73,459 subjects aged 0 to 20 years (35,890 girls and 37,569 boys) with 428,326 length/height and weight measurements. The subjects were born between 9 May 1983 and 5 May 2009.

For the length/height-for-age reference, the initial database comprised subjects from the weight reference population who had at least one scheduled visit between 10 March 2008 and 9 March 2009 (40,655 subjects/285,488 measurements) (Figure 2). The past and present growth data of each measured individual was evaluated systematically by automated screening according to the existing Finnish growth screening rules (Figure 2) (2,3). Children with faltering growth curves identified by the automated screening method were referred to a pediatric endocrinologist who excluded children with suspected abnormal growth curves from the database. Male data were supplemented by 200 measurements of healthy, consecutive army conscripts aged 18 to 20 years from the Garrison of Santahamina (Helsinki, Finland) and fulfilling the above general inclusion criteria for the database. The military service is compulsory for all healthy young male (aged over 18 years) adults in Finland, and the subjects chosen for the present study originated from the same area of Southern Finland as the whole study population.

Figure 2. Data cleaning procedure for the length/height-for-age reference database. *Growth screen: height SDS deviation outside ±2.7 SD from the mean, or outside ±2.3 SD from the target height, or change in height SD in the preceding 1, 3, or 5 years exceeding the allowed age-specific cut-offs; ^†Underweight: weight-for-height below −15% of the median in height of 50–129 cm, below −20% in height of 130–159 cm, and below −25% in height of 160–180 cm; ^‡Obesity: weight-for-height above 20% or 40% of the median weightfor- height in children less than 7 years or 7 years or more, respectively (2,3).

Furthermore, subjects (and all their measurements) with conditions possibly affecting growth or significantly differing from the average (i.e. those with prematurity, low or unknown birth-weight, height below −4 SD or over +4 SD, and underweight or obesity) (Figure 2) were excluded from the height reference database.

The influence of overweight on height in the height reference 1983–2008 population at the age-range of 0–12 years was assessed in the subgroup of boys with measurements at the age of 6.5–7.5 years. Overweight was defined according to IOTF criteria (BMI percentile passing through BMI 25 at the age of 18 years) (30). Overweight boys were slightly taller than normal weight boys (maximum difference of 2.9 cm at the age of 10 years). The difference between overweight boys and final curve model was 0.2–3.1 cm (maximum at the age-range of 10.5 to 11 years.)

The final database for the length/height-for-age reference included mixed cross-sectional and longitudinal data, including 181,785 length/height and weight measurements of 26,636 full-term, healthy subjects (12,895 girls, 13,741 boys) born between 18 December 1983 and 23 December 2008. The number of measurements per subject varied from 1 to 17.

Curve construction and statistical analysis

In the construction of growth references (length/height-for-age, weight-for-length/height, BMI-for-age), the distribution of response variables and smoothing techniques of the distribution parameter curves over the explanatory variables were chosen by closely following the guidelines provided by the World Health Organization (WHO) (33). Generalized additive models for location, scale and shape (GAMLSS) were used choosing the distribution of a response variable from the flexible Box-Cox power exponential (BCPE) distribution family and using cubic splines as a smoothing technique. The BCPE distribution can be described in terms of four parameters: M for median, S for coefficient of variation, L for Box-Cox transformation power, and T as a parameter related to kurtosis.

R statistical software (GAMLSS package) was used in the analysis (34). First, optimal power transformation was calculated by non-linear regression for the explanatory variable in relation to the response variable as it was found to improve goodness of fit. Second, optimal degrees of freedom for parameter curves were defined using optim function and the Akaike information criteria (AIC), the generalized AIC (GAIC), and the Bayes information criteria (BIC) (which have penalty h of 2, 3 and log(n) in the formula -2L-hp, where L is the maximized likelihood, p is the number of parameters in the model, and n is the number of observations). Modeling was started from the normal distribution (BCPE with L = 1 and T = 2) and extended to the Box-Cox normal (BCPE with T = 2, also called LMS method of distribution where L is the skewness, M is the median, and S is the coefficient of variation) or full BCPE, if needed. In choosing the ‘best’ model, the main emphasis was settled to the plot of fitted percentiles with observed percentiles. Usually, the BIC information criterion was found more applicable than AIC or GAIC, possibly due to large numbers of observations.

We ended up with a normal distribution assumption for height in the height-for-age analysis. Growth curves for height-for-age were depicted as the mean (M) and mean plus or minus two standard deviations (mean (2 SD)) which covered the 2.3rd, 50th, and 97.7th percentiles. Response variables (BMI and weight) assumed a normal Box-Cox distribution. BMI, in BMI-for-age analysis, was described in terms of the 3rd, 10th, 50th, 90th, and 97th percentiles and the percentiles passing through BMI values of 16, 17, 18.5, 25, and 30 at age 18 years, the wide-spread method by Cole to define various grade of thinness, overweight, and obesity for Finnish children (30,31). Median (the 50th percentile) curves were reported from the analysis of weight-for-length/height. Fitted curves are presented together with corresponding observed percentiles (Supplemental Figures 1–3) to illustrate empirical data and goodness of fit.

Results

Length/height-for-age reference

Figure 3 shows the new Finnish growth reference 1983–2008 for length/height-for-age in comparison with the old growth reference 1959–1971. The mean birth lengths in the reference 1983–2008 population were 50.3 cm in girls and 51.1 cm in boys compared to 50.2 cm and 50.7 cm in the old reference, respectively. The mean birth length of full-term babies in the Finnish birth cohort from the year 2000 to 2008 is 50.0 cm in girls and 50.9 cm in boys (unpublished observation), suggesting that the population selected for the new length/height reference is a fair representation of the contemporary Finnish population. The mean adult height of girls increased by 1.9 cm (from 165.3 cm in the reference 1959–1971 to 167.2 cm in the reference 1983–2008 population); the mean adult height for the cohorts for boys increased by 1.8 cm from 178.9 cm to 180.7 cm, respectively.

Figure 3. The new Finnish length/height-for-age reference (mean ± 2 SD, solid lines). Curves based on 181,785 measurements from 26,636 full-term healthy subjects born between 1983 and 2008 compared to the current Finnish growth reference based on subjects born between 1959 and 1971 (mean ± 2 SD, dashed lines). A: girls aged 0–2 years; B: girls aged 2–19 years; C: boys aged 0–2 years; and D: boys aged 2–20 years.

In the reference 1983–2008, girls and boys at the age of 18 years were 1.4 and 1.1 cm taller, respectively, than girls and boys in the reference 1959–1971 at the age of 20 years. This indicates that there still is a positive secular change also in adult heights in Finland.

Secular trends in height between the growth reference populations 1959–1971 and 1983–2008

The growth pattern resulting in the increased adult height had distinct age- and sex-specific features (Figure 4). Between the ages of 0 to 20 years, subjects in the reference 1983–2008 were generally taller than subjects in the reference 1959–1971; however, the difference was almost non-existent in boys and negative in girls during infancy (approximately 5 to 11 months of age). In that age-range, girls from the 1983–2008 population were, on average, 0.2 cm/0.07 SDS (standard deviation score) shorter than girls from the reference 1959–1971. The positive secular change was more substantial in boys at all ages even though the adult height increase was similar in boys and girls (1.8 cm/0.27 SDS and 1.9 cm/0.32 SDS, respectively). The maximum difference in height in boys was seen around the age of 13 years (a 5.6 cm/0.70 SDS increase when compared to the reference 1959–1971 population). In girls, the maximum difference between the two references was seen earlier, around at the age of 11.5 years, and was less than that observed in boys (2.8 cm/0.40 SDS).

Figure 4. Age- and sex-specific features of the secular change in mean length/height in Finland. Comparison between growth reference 1959–1971 population subjects (those born between 1959 and 1971) and reference 1983–2008 population subjects (those born between 1983 and 2008). Curves indicate differences from the reference 1959–1971 population in: A: mean height in cm from birth to age 2 years; B: mean height in cm from age 2 to 20 years; C: mean height in SD units from birth to age 2 years; and D: mean height in SD units from age 2 to 20 years. Dashed line = girls; solid line = boys.

The total number of length/height measurements from different age-groups of the height reference 1983–2008 population and the percentage of measurements outside the mean ±2 SD of the existing height reference 1959–1971 are shown in Table I. On average, measurements in girls (3.4%, range 1.2%–5.6%) and boys (4.3%, range 1.9%–8.9%) were above +2 SD, which was clearly different from the expected 2.3% of normally distributed height). The proportion of measurements above +2 SD was less than 2.3% in girls aged 4 to 18 months and in boys aged 5 to 10 months. The highest proportions of measurements above +2 SD were seen in girls aged 11 to 12 years and in boys aged 12 to 13 years. Only 0.9% of measurements in girls (range 0.4%–1.8%) and 0.5% (range 0.1%–1.0%) of measurements in boys were below -2 SD of the reference 1959–1971.

Table I. Number of height measurements in the growth reference 1983–2008 population and percentage of measurements ≤ −2 SD and ≥ +2 SD when compared to the growth reference 1959–1971.

	Growth reference 1983–2008 population, girls	Girls outside the ±2 SD range of growth reference 1959–1971 population		Growth reference 1983–2008 population, boys	Boys outside the ±2 SD range of growth reference 1959–1971 population
Age-group	Number of measurements	Percentage ≤ −2 SDS	Percentage ≥ +2 SDS	Number of measurements	Percentage ≤ −2 SDS	Percentage ≥ +2 SDS
<1 mo	3969	0.7	5.3	4178	0.3	6.5
1–2 mo	4596	0.4	5.4	4943	0.3	6.0
2–3 mo	3161	0.8	3.6	3285	0.1	4.5
3–4 mo	5262	0.8	3.1	5658	0.3	3.7
4–5 mo	4016	1.1	2.2	4230	0.5	3.0
5–6 mo	4310	1.4	1.2	4526	0.8	2.1
6–8 mo	4805	1.7	1.3	5105	0.7	1.9
8–10 mo	4236	1.7	1.4	4500	0.8	1.9
10–12 mo	3233	1.2	1.5	3268	0.9	2.4
12–18 mo	4491	1.0	2.0	4764	0.6	3.0
18–24 mo	3794	0.5	3.6	4010	0.5	4.2
2-3 y	3872	0.4	3.4	4195	0.4	4.4
3–4 y	3757	0.5	3.9	4004	0.3	4.5
4–5 y	1339	0.8	4.3	1473	0.4	4.5
5–6 y	3369	0.5	3.9	3609	0.3	5.2
6–7 y	3290	0.6	4.3	3386	0.2	4.8
7–8 y	2141	0.7	4.4	2405	0.3	5.5
8–9 y	2608	0.6	4.4	2689	0.2	6.0
9–10 y	2459	0.5	4.5	2648	0.3	5.8
10–11 y	1444	1.2	4.8	1504	0.4	4.8
11–12 y	2646	0.8	5.6	2621	0.4	5.4
12–13 y	2010	1.4	4.2	1993	0.4	8.9
13–14 y	2503	1.8	4.3	2443	0.3	8.1
14–15 y	1999	1.3	4.3	2071	0.8	4.0
15–16 y	1551	1.2	3.6	1566	1.0	3.2
16–17 y	1816	1.4	3.9	1824	0.8	2.7
>17 y	661	1.2	4.4	735	1.0	3.0
Total	83338	0.9	3.4	87633	0.5	4.3

Total number of girls = 12,895 girls; boys = 13,741 boys. Birth lengths were excluded. Growth reference 1983–2008 population includes subjects born between 1983 and 2008; growth reference 1959–1971 population includes subjects born between 1959 and 1971.

Weight-for-length/height reference

New weight-for-length/height percentile (3rd, 10th, 50th, 90th, 97th) curves for girls and boys in the weight reference 1983–2008 population are compared to median (50th percentile) curve of the reference 1959–1971 population (Figure 5). The median weight-for-length/height curves for both populations were nearly superimposable up to a height of 140 cm (corresponding to height age of 10 years) in boys and 155 cm (height age of 12 years) in girls. Subsequently, a slight increase in median weight-for-length/height was observed in reference 1983–2008 population subjects.

Figure 5. The new Finnish weight-for-length/height percentile (3rd, 10th, 50th, 90th, 97th) curves. Curves based on 428,526 length/height and weight measurements taken from 73,659 healthy subjects in the weight reference 1983–2008 population (those born between 1983 and 2008; solid lines) compared to the median weight-for-length/height curve of the reference 1959–71 population (those born between 1959–1971; dashed line). A: girls 45–90 cm; B: girls 90–180 cm; C: boys 45–90 cm; D: boys 90–190 cm.

BMI-for-age reference

BMI percentile curves (3rd, 10th, 50th, 90th, and 97th) for Finnish children and adolescents aged 2 to 18 years are shown in Figure 6. BMI percentile curves for Finnish children have not been previously published. The median (50th percentile curve) BMI-for-age decreases slightly from the age of 2 years until approximately 6 years in both girls and boys and then starts to increase. The point at which BMI values increase is called ‘adiposity rebound’ (35). This took place later in the lower percentiles compared to the higher ones (Figure 6). Median BMI-for-age was slightly higher in boys than in girls between the ages of 2 to 5 years and 8 to 13 years (up to 0.3 kg/m²). After the age of 15 years, the difference between boys and girls increased rapidly, with boys having up to a 1.1 kg/m² higher median BMI than girls by the age of 18 years.

Figure 6. The new Finnish BMI-for-age reference percentile curves. The 3rd, 10th, 50th, 90th, and 97th curves are shown for children and adolescents aged 2–18 years. A: girls; B: boys.

BMI-for-age cut-offs for thinness, overweight, and obesity

Finnish cut-off curves passing through BMIs of 30, 25, 18.5, 17, and 16 kg/m² at the age of 18 years (corresponding to obesity, overweight, and grade 1, 2, and 3 thinness, respectively) were calculated for Finnish children (Figure 7). A comparison of the Finnish ‘normal area BMI’ (i.e. between grade 2 thinness and overweight) with respect to international data is also provided (30,31).

Figure 7. Finnish BMI-for-age percentile curves for grade 3 and 2 thinness, overweight, and obesity. Shown are the percentiles passing through BMIs of 16, 17, 25, and 30 kg/m² at the age of 18 years and the ‘normal BMI area’ between percentiles of grade 2 thinness and overweight (shaded area). Corresponding BMI-for-age percentile curves from multi-ethnic data (dashed lines) indicate grade 2 thinness and overweight (International Obesity Task Force international reference) (30,31). A: girls; B: boys.

Finnish and international (IOTF) percentile curves defining overweight were nearly superimposable in girls with the maximum difference of only 0.7 kg/m². In boys, the largest difference (up to −1.2 kg/m²) between the two overweight curves was seen between the ages of 2 to 6 years, with the Finnish overweight curve lower than the IOTF curve.

The cut-off curves for overweight and obesity (i.e. a BMI of 30 kg/m² at the age of 18 years) were the 87.8th percentile and 98.2nd percentile in girls and the 78.2nd percentile and 95.6th percentile for boys, respectively.

The cut-off curves for grade 1, 2, and 3 thinness (i.e. BMIs of 18.5, 17, and 16, respectively) were at the 17.9th, 5.0th, and 1.3rd percentiles in girls and the 12.1st, 3.3rd, and 1.0th percentiles in boys, respectively.

Discussion

We developed new length/height-for-age, weight-for-length/height, and BMI-for-age growth curves for assessment of the growth of children in Finland. The advantages of these curves are that they include a large, contemporary population-based sample of healthy subjects; careful and prospective screenings and exclusion of subjects with potentially faltering growth; up-to-date statistical methods in curve construction; correction of a substantial secular change in the stature of Finnish children; and validation of the constructed BMI-for-age curves against the international standard. We showed that misclassification of the height of contemporary Finnish children is common when using the existing height reference based on subjects born between 1959 and 1971, which further supported the need for an updated growth reference. These curves, applicable across the population, should be immediately implemented to monitor the growth of children and adolescents in Finland.

The Finnish social security and primary health care systems provide for free, regular check-ups performed by well trained nurses with standardized methods during the growth period. These visits are available to all permanent residents of Finland regardless of social status or income level; therefore, virtually every child is included. As a result, population-based growth studies in a primary health care setting, such as the present one, are possible. Genetically, the population of the city of Espoo mirrors the whole of Finland and has grown in size 10-fold over the past 60 years due to net migration from other parts of the country. The proportion of immigrants in Espoo was slightly higher than in the whole of Finland (7% versus 4%); however, this proportion is continuously changing in other parts of the country as well.

The initial database included growth data of the entire child population; however, a series of exclusions were made to create height curves that reflected optimal, healthy, linear growth. Cleaning procedures followed the widely accepted methods described for example by authors of WHO Growth Standard Study (33). Therefore, careful database cleaning was conducted to remove factors possibly influencing linear growth such as prematurity, low birth-weight, childhood underweight, and obesity. Furthermore, special attention was paid to excluding children with any disorder or medication thought to possibly affect linear growth. Subsequently, the past and present growth data of each subject visiting either a child health clinic or school health care during the period of 1 year was assessed systematically according to a specific set of growth screening rules (Figure 2) (2,3). Violation of the screening rules resulted in a visual inspection by a pediatric endocrinologist. As a result of these processes, we believe that our length/height-for-age curves are an estimate of optimal growth.

Optimal weight gain for children and adolescents is much more difficult to define than optimal linear growth. Weight gain is dependent on age and even more strongly on an increase in height, but only very poor or very excessive weight gain is associated with morbidity. Therefore, the original data set for weight-for-length/height and BMI-for-age references was not cleaned as extensively. Only children with chronic diseases and medications affecting growth were excluded.

For curve construction, we chose the GAMLSS method, the same method recommended for mixed cross-sectional and longitudinal data (33) by the WHO Multicentre Growth Study Group.

We found that secular changes in linear growth had distinct age- and sex-specific features. Growth of the infants was more rapid from birth to 1 month of age (up to 0.5 SDS for girls and 0.6 SDS for boys) in the 1983–2008 reference population compared to the 1959–1971 reference population. A change in infant feeding regimen may explain this difference. Currently, about 60% of infants are breast-fed for at least 6 months (36), whereas breast-feeding was relatively uncommon (5% for 6 months) for infants in the reference 1959–1971 population (37). It is known that predominantly breast-fed infants grow faster than formula-fed infants during first months of life. Prolonged exclusive breast-feeding will, however, attenuate linear growth after 3 months of age (38). Indeed, infants in the reference 1983–2008 population grew at a much lower rate than infants in the reference 1959–1971 population between the ages of 3 and 6 months. Some of the differences in height in infancy may also be due to different statistical methods used in the construction of the two references.

From about the age of 1 year up to 11.5 years in girls and 13 years in boys there was a steady increase in mean height of children in the reference 1983–2008 population as compared to the reference 1959–1971 population. Boys between the ages of 12 and 13 years, in particular, grew up to 5.6 cm (0.70 SDS). At all ages, secular increases in height were more marked in boys than in girls. In Finland, overweight is currently much more common in boys than in girls (18), suggesting that excessive energy gain is one factor resulting in the increased growth rate of boys.

Data on pubertal stages were not included in our sample, but secular increases in height appeared to be linked to earlier maturation and earlier timing of the pubertal growth spurt compared to the reference 1959–1971 population as evidenced by a greater increase in height during childhood than during adulthood. Earlier timing of puberty has been recently reported in Danish girls and boys (39,40). In Danish children, early puberty was associated with higher BMI before the onset of puberty (39,40); however, there was a downward trend in the age at which puberty was attained in both girls and boys, regardless of the BMI, that suggests that the obesity epidemic is not solely responsible for the trend.

Recent growth studies have reported that secular increases in adult height are gradually leveling off in several Northern European countries (41). In Finnish children, however, a marked increase in adult height of 1.9 cm in girls and 1.8 cm in boys was observed. The general belief is that the secular increase in height continues until the genetic potential of the population is reached. However, our current knowledge of the genetics of stature is insufficient to make an estimation of height potential of a population.

Growth curves are used for screening of growth-related disorders, and a secular trend in linear growth has a direct influence on this process. Through this study we demonstrated how outdated cut-off points may lead to misclassification of children. Theoretically, an uncorrected secular trend of +0.4 SDS in mean height without a change in SD for height would mean that only 0.8% of children remained below the lower −2 SDS limit and as many as 5.5% were above the upper +2 SDS limit, instead of the expected 2.3% at both ends. Our observations were consistent with such a rate of misclassification by the mean ±2 SD limits. Incorrect cut-off points at ±2 SD by 0.4 SDS would results in a 3.4% reduction in specificity and up to a 32.6% reduction in sensitivity of screening of short and tall stature, which is unacceptable.

Monitoring of weight is important for detection of nutritional changes such as risk of malnutrition or obesity, or of chronic diseases affecting weight. Weight reference curves should include the definition of a normal weight range between thinness and overweight. In adults, we use BMI cut-off points that have been agreed upon internationally: underweight BMI is < 18.5 kg/m²); normal weight BMI is 18.5–24.9 kg/m²; overweight BMI is 25.0–29.9 kg/m²; and obesity BMI is ≥ 30 kg/m² (16). It is important to remember that in children, body composition varies according to age; therefore, factors affecting weight include not only height and sex, but also, to some extent, age of the child. Furthermore, body composition and build also differ between populations (42). Thus, BMI reference curves need to be adjusted at least for sex and age and possibly also for ethnicity.

We first constructed BMI-for-age percentile curves for children older than 2 years. Then, by using a wide-spread method by Cole, which uses BMI-for-age percentiles passing through various adult BMI cut-offs, we calculated BMI percentiles to define various grade of thinness, overweight, and obesity for Finnish children (30,31).

The prevalence of grade 1 thinness (percentile passing through a BMI of 18.5 at the age of 18 years) in our sample was as high as 17.9% for girls and 12.1% for boys. As it would be both impractical and confusing to define such a high number of children and adolescents having grade 1 thinness, we recommend using only grade 2 and grade 3 thinness percentile curves (passing through BMIs of 17 and 16 at the age of 18 years, respectively).

For Finnish boys, the prevalence of obesity was approximately the same as that recently reported by Vuorela et al. (4.4% versus 3.2%), but for girls the prevalence was about half in our sample (1.8% versus 4.7%) (18). These results are not directly comparable, however, because we used BMI percentiles specific for Finnish children, while the IOTF cut-offs were used in the other study.

Compared with weight-for-length/height, BMI-for-age has the advantage of being able to capture the changes in the weight-length/height relation with age and of providing a measure that can be used continuously up to adulthood, as well as compared internationally. However, due to significant changes in BMI-for-age during the age of 0 to 2 years (first rapid increase and then decrease) we still recommend using weight-for-length/height in that age-group instead of BMI-for-age. Use of weight-for-length/height has been criticized for several reasons. As reported by Cole (43), it is not independent of age, and therefore it should be used only over a narrow age-range. Furthermore, there are no cut-off values using weight-for-length/height to define underweight and obesity. Therefore, for older children the use of BMI-for-age rather than weight-for-length/height seems generally more advisable. For the screening of obesity in children and adolescents, the United States Preventive Services Task Force recommends the use of BMI- for-age beginning at 6 years of age (44).

Conclusions

The new Finnish growth references for length/height-for-age, weight-for-length/height, and BMI-for-age created in this study using a contemporary, large, Finnish population-based sample are now available for primary health care providers and clinicians for updated growth monitoring. There are several advantages in replacing the growth curves currently in use by these updated and new references. Use of these new references should result in fewer misclassifications of normal growth in children, thereby preventing unnecessary examinations in children with suspected growth failure.

Declaration of interest: The authors state no conflict of interest and have received no payment in the preparation of this manuscript.