The importance of vitamin D in bone health is recognised with rickets in children and osteomalacia and osteoporosis in adults being the traditional clinical conditions associated with vitamin D deficiency( Reference Holick 1 , Reference Sahota 2 ). Furthermore, the expression of vitamin D receptors in different tissues( Reference Wang, Zhu and DeLuca 3 ) indicates additional biological functions of the vitamin. Low vitamin D status has been reported to be associated with a range of health outcomes( Reference Grant 4 , Reference Wacker and Holick 5 ), including an increased risk of cardiometabolic disorders( Reference Parker, Hashmi and Dutton 6 ), some cancers( Reference Gandini, Boniol and Haukka 7 ), autoimmune diseases( Reference Kriegel, Manson and Costenbader 8 ) and mortality( Reference Zittermann, Iodice and Pilz 9 ). However, it is still unclear whether these associations are causal. Studies have found vitamin D status to be associated with sociodemographic and lifestyle-related factors( Reference Thuesen, Husemoen and Fenger 10 , Reference Jaaskelainen, Knekt and Marniemi 11 ); thus, vitamin D status may serve as an indicator of general health and/or lifestyle.
The accepted biomarker of vitamin D status is 25-hydroxyvitamin D (25(OH)D) concentration in the blood( Reference Seamans and Cashman 12 ). In Denmark, the following values are used to define deficient, insufficient and sufficient serum 25(OH)D concentrations: < 25; 25–50; >50 nmol/l( 13 ). The Institute of Medicine (USA) has defined 30 nmol/l as the limit beyond which adverse effects on bone might occur( 14 ). Serum 25(OH)D concentrations of 40 and 50 nmol/l have been considered to represent the estimated average requirement and the recommended daily allowance, which are assumed to meet the requirement in the average population and the majority of the population for bone health. A threshold value >125 nmol/l has been considered to be the at-risk value for harm by the Institute of Medicine (USA).
The use of cut-off values and the comparison of vitamin D status between studies are complicated due to variations in measurements between the methods as well as among the laboratories using the same 25(OH)D assay( Reference Farrell, Martin and McWhinney 15 – Reference Carter 17 ). In 2010, the National Institute of Standards and Technology (NIST) introduced a standard reference material for the measurement of vitamin D concentrations in human serum( Reference Phinney, Bedner and Tai 18 ), which is expected to improve the analytical performance of 25(OH)D measurements and to facilitate harmonisation across 25(OH)D assays( Reference Farrell, Martin and McWhinney 15 ).
Thus, there is a need for comparative data on vitamin D status yielded by standardised and calibrated methods to better compare vitamin D status between population groups and evaluate the current situation of vitamin D deficiency. Vitamin D status has been measured in different population groups in Denmark( Reference Thuesen, Husemoen and Fenger 10 , Reference Brot, Vestergaard and Kolthoff 19 – Reference Andersen, Mølgaard and Skovgaard 23 ). None of the previous studies has assessed a broad range of age and sex groups in both children and adults. Especially, there is a lack of data on vitamin D status in young boys. One of the studies has reported vitamin D status in men, women and girls from Danish immigrant families( Reference Andersen, Mølgaard and Skovgaard 23 ). However, to our knowledge, no previous studies have quantified the impact of the familial relationship on vitamin D status. This knowledge on vitamin D status within families will be helpful when considering strategies to improve vitamin D status.
The objective of the present study was to assess serum 25(OH)D concentration and its determinants in children and adults among families in late summer in Denmark.
Subjects and methods
Study population
The present cross-sectional study used baseline data obtained from the VitmaD study( Reference Madsen, Rasmussen and Andersen 24 ) conducted in Denmark (56°N). Children and adults were recruited as families randomly drawn from the Danish Civil Registration System. Inclusion criteria were age between 4 and 60 years and a permanent address in Gladsaxe Municipality. Exclusion criteria were pregnancy and disease or use of medication influencing vitamin D metabolism (including dietary supplements with vitamin D levels >10 μg/d for children and >5 μg/d for adults, which correspond to the typical levels in multivitamin supplements in Denmark). Of the 782 recruited children and adults, 755 (representing 200 families) had their serum 25(OH)D concentration measured and complete questionnaire data at baseline. The present analyses were conducted on these subjects. Written informed consent was obtained from all the adult subjects and from the guardians of the children. The study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the Research Ethics Committee of the Capital Region of Denmark (record no. H-4-2010-020) and registered in ClinicalTrials.gov (NCT01184716).
Methods
The subjects were examined and their blood samples were collected in September–October 2010 in an authorised laboratory (Copenhagen's General Practitioners Laboratory, Søborg, Denmark). The weight of the subjects was measured to the nearest 0·1 kg in normal clothes without shoes (1 kg was subtracted from the measured weight) with a body composition analyser (Tanita BC-418; Tanita Europe B.V.). Height was measured to the nearest centimetre without shoes with an ultrasonic height measure (Soehnle 5001; Soehnle Professional GmbH & Company). BMI was calculated based on the measured weight and height, and the subjects were categorised into normal-weight, overweight and obese classes according to the standards for children( Reference Cole, Bellizzi and Flegal 25 ) and the WHO International standard for adults( 26 ). Non-fasting venous blood samples were drawn from the subjects, and serum and plasma were collected and stored at − 80°C until analysis.
Serum 25(OH)D concentration was measured at the Clinical Biochemical Department, Holbæk Hospital, Denmark, using isotope dilution liquid chromatography–tandem MS (LC–MS/MS) according to the principles described elsewhere( Reference Blomberg Jensen, Bjerrum and Jessen 27 ). The method was calibrated against the NIST standard for the analysis of vitamin D in human serum (standard reference material 972)( Reference Lai, Lucas and Banks 16 ), and the inter-assay CV for the method used in the present study were 2·2 and 2·8 % at 30 and 180 nmol/l, respectively, for 25(OH)D3 and 7·6 and 4·6 % at 43 and 150 nmol/l, respectively, for 25(OH)D2. The analytical quality of this method was ensured through participation in the Vitamin D External Quality Assessment Scheme (DEQAS). In this validation scheme, the mean bias for our method compared with the mean of the DEQAS LC–MS group during the period the present analyses were carried out was − 3·2 %. Plasma parathyroid hormone (PTH) concentration was measured using an immunology analyser (Cobas e601; Roche Diagnostics), according to the standard procedures of the manufacturer (CV = 3·4 %).
Information on background, health, sun exposure and lifestyle, including the use of multivitamin and vitamin D supplements, of the subjects was obtained using detailed self-administered web-based questionnaires. Dietary vitamin D intakes were recorded using a semi-quantitative FFQ adapted from a FFQ used in the European union project Towards a strategy for Optimal Vitamin D Fortification (OPTIFORD)( Reference Andersen, Mølgaard and Skovgaard 23 ). The vitamin D intakes were calculated based on the reported consumption frequencies and the vitamin D concentrations in the food items given in the Danish Food Composition Databank( 28 ).
Statistical analyses
Data were analysed using the SPSS statistical software (version 20.0, IBM SPSS, Inc.), and statistical significance was evaluated at a level of P< 0·05 (two-sided). Linear mixed models with family as a random variable were used in all the analyses to account for the non-independency of the subjects. Before analysis, serum 25(OH)D and PTH concentrations were logarithmically transformed to meet the model requirements. Trend analyses were carried out to test for linear relationships between PTH concentrations and 25(OH)D groups. Univariate models were used to assess the association between serum 25(OH)D concentration and each of the following sun-related variables: outdoor transport to school/work ( < 15, 15–30, 31–60, or >60 min/d); solarium use at least once a week (yes or no); sun preference (prefer sun, sometimes in sun, or avoid sun); outdoor stay in light clothes (most of the time, often, sometimes, or seldom/never); sunscreen use (always, most of the time, sometimes, or seldom/never); sun vacation the preceding summer in June–September (yes or no). Sun-related variables with P< 0·05 significance in their univariate model were included in a multiple analysis (linear mixed model) together with the following categorical variables: age (4–10, 11–17, 18–40, or 41–60 years); sex (female or male); BMI (normal weight, overweight, or obese); dietary vitamin D intake (quartiles: < 1·7, 1·7–2·4, 2·5–3·3, or >3·3 μg/d); multivitamin use (yes or no). The interaction between age and sex was tested. This multiple analysis was carried out for all subjects and for children and adults separately. The strength of the familial component was considered by calculating the intra-class correlation for each model:
$$\begin{eqnarray} \rho = \frac { \omega ^{2}}{( \omega ^{2} + \sigma ^{2})}, \end{eqnarray}$$
where ρ is the intra-family correlation, σ is the within-family standard deviation and ω is the between-family standard deviation. The lower the variation within the classes, the higher the intra-class correlation, which in this case means that the closer the correlation is to 1, the more alike the subjects are within a family with respect to their vitamin D status.
The relationship between serum 25(OH)D concentration and each of the following lifestyle-related variables was explored in univariate models only in adults: smoking status (current, former, or never); alcohol consumption (never, < 1 time/month, 1–3 times/month, 1 time/week, 2–4 times/week, 5–6 times/week, or daily); leisure-time physical activity (mainly sedentary, light-to-moderate activity, regular sport and exercise, or athletic training); self-rated physical shape (really good, good, fairly good, bad, or really bad); self-rated health (excellent, really good, good, or less good/bad); effort to eat healthily (very often, often, sometimes, or seldom/never); education after state and/or upper secondary school (none or technical education, < 3 years of higher education, 3–4 years of higher education, or >4 years of higher education). Lifestyle-related variables with P< 0·05 significance in their univariate model were included in the multiple model described above.
Results
The characteristics of the study population and the lifestyle-related characteristics of adults in the study population are given in Tables 1 and 2, respectively. The median ages of the youngest children (4–10 years), the oldest children (11–17 years), the youngest adults (18–40 years) and the oldest adults (41–60 years) were 7 (interquartile range (IQR) 6–9), 13 (IQR 12–15), 37 (IQR 33–39) and 45 (IQR 43–48) years, respectively. The sexes were evenly distributed among both the children and adults, and the majority of the subjects were of normal weight (Table 1). The median dietary vitamin D intakes were similar across the age groups (range 2·3–2·6 μg/d). The total median vitamin D intake of the multivitamin users (41 % of the children and 25 % of the adults) was 6·7 (IQR 4·5–10·2) μg/d. Approximately half of the adults had a medium-long- to long-duration higher education, and their lifestyles were generally healthy (Table 2).
Characteristics of the study population (n 755) (Number of subjects and percentages)
* Quartiles (Q) for the whole study population.
Lifestyle-related characteristics of adults in the study population (n 415) (Number of subjects and percentages)
The individual serum 25(OH)D concentrations (sum of 25(OH)D2 and 25(OH)D3 concentrations) ranged from 9·3 to 161·9 nmol/l, with an overall geometric mean of 72·1 (IQR 61·5–86·7) nmol/l. Serum 25(OH)D2 was found in 11 % of the samples in the range of 3–29 nmol/l. The serum 25(OH)D concentration of the different age and sex groups and its distribution across the ranges are summarised in Table 3. The overall distribution of serum 25(OH)D concentrations < 30, < 40 and < 50 nmol/l was 1, 2 and 9 %, with no children being found to have serum 25(OH)D concentration < 30 nmol/l. In the adults, the geometric mean PTH concentrations in the 25(OH)D concentration < 25, 25–49, 50–75 and >75 nmol/l groups were 59·8 (95 % CI 43·2, 82·8), 39·9 (95 % CI 35·8, 44·3), 36·2 (95 % CI 34·3, 38·2) and 32·7 (95 % CI 31·1, 34·4) ng/l, respectively (P for trend < 0·001). The same trend was observed in children aged 4–10 years (P for trend = 0·012), but not in children aged 11–17 years (P for trend = 0·067).
Serum 25-hydroxyvitamin D (25(OH)D) concentration and distribution by age and sex groups among the families in late summer in Denmark (56°N) (Geometric mean values and interquartile ranges; number of subjects and percentages)
* No overall differences in serum 25(OH)D concentration among the age (P= 0·190), sex (P= 0·332), and the combined age and sex groups (P= 0·223) as analysed in the linear mixed model with family as a random variable; sex, age, BMI, dietary vitamin D, multivitamin use, solarium use, outdoor stay in light clothes, sun preference and sun vacation as the categorical variables; and the logarithm of the serum 25(OH)D concentration as the dependent variable.
No differences were found in serum 25(OH)D concentration among the age (P= 0·190), sex (P= 0·332) or age, and sex groups (P= 0·223) in the multiple analysis of all subjects (Table 4). When the children were analysed separately, serum 25(OH)D concentration was found to be associated with sex (P= 0·034) and so 25(OH)D concentration was estimated to be 5 % lower in the girls than in the boys. In the univariate models, outdoor transport to work/school (P= 0·972) and sunscreen use (P= 0·154) were not associated with 25(OH)D concentration and thus not included in the multiple models. In the multiple analysis of all subjects, serum 25(OH)D concentration was found to be negatively associated with BMI (P< 0·001) and positively associated with dietary vitamin D intake (P= 0·008), multivitamin use (P= 0·019), solarium use (P= 0·006), outdoor stay in light clothes (P= 0·001), sun preference (P= 0·002) and sun vacation (P< 0·001) (Table 4). When the children and adults were analysed separately, serum 25(OH)D concentration was found to be associated with BMI and sun preference in the children. In the adults, the associations remained the same as in the model with all subjects, but the significance for dietary vitamin D intake, multivitamin use and sun vacation was weakened.
Associations between potential determinants and serum 25-hydroxyvitamin D (25(OH)D) concentration* (Ratio of means and 95 % confidence interval)
* Analysed in the linear mixed models with family as a random variable; sex, age, BMI, dietary vitamin D, multivitamin use, solarium use, outdoor stay in light clothes, sun preference and sun vacation as the categorical variables; and the logarithm of the serum 25(OH)D concentration as the dependent variable.
† The regression coefficients were exponentially transformed (10β).
‡ Quartiles (Q) for the whole study population.
From these multivariate models, it was found that the variations in serum 25(OH)D concentration were higher within the families than between the families with an intra-family correlation of 0·27 in all subjects. The intra-family correlation was higher in the children than in the adults (Table 5).
Standard deviation for the within-family and between-family effects and the intra-family correlation*
* Derived from the linear mixed models with family as a random variable; sex, age, BMI, dietary vitamin D, multivitamin use, solarium use, outdoor stay in light clothes, sun preference, and sun vacation as the categorical variables; and the logarithm of the serum 25-hydroxyvitamin D concentration as the dependent variable.
† Calculated as follows: between-family standard deviation2/(between-family standard deviation2+within-family standard deviation2).
In a further analysis of the adults, serum 25(OH)D concentration was found to be associated with leisure-time physical activity (P< 0·001), self-rated physical shape (P= 0·001) and self-rated health (P= 0·003) in their univariate models. When these variables were included in the multiple analysis together with the vitamin D source-related variables, none remained significant, although leisure-time physical activity was borderline significant (P= 0·054). Serum 25(OH)D concentration was not associated with smoking status (P= 0·722), alcohol consumption (P= 0·070), effort to eat healthily (P= 0·193) or education (P= 0·219).
Discussion
The overall geometric mean serum 25(OH)D concentration among the families in the present study was 72·1 nmol/l, with no differences being observed between the age and sex groups in the analysis of all subjects. The distribution of 25(OH)D concentrations < 25, < 50 and < 75 nmol/l was 1, 8 and 54 %, with no children being found with 25(OH)D concentration < 25 nmol/l. A novelty of the present study was that the familial component was quantitatively assessed by calculating the intra-class correlation. The intra-family correlation for all subjects was 0·27, which indicates that a subject's vitamin D status is not strongly related to the familial relationship. However, to our knowledge, no previous studies have quantified the familial component for vitamin D status and thus we do not have a value for comparison of the familial relationship. The intra-family correlation was almost double in the children (0·42) as in the adults (0·24), indicating that the children within a family were more alike than the adults within a family with respect to their vitamin D status. This might be an indication of the influence of genetic factors on vitamin D status or more similar habits in children than in adults from the same family. It is likely that children within a family share activities and dietary/supplementation habits to a greater extent than adults within a family, e.g. outdoor stay, sun protection, multivitamin use and lunch in school.
The serum 25(OH)D concentrations found in the present study were higher and the distribution of 25(OH)D concentrations < 50 nmol/l was lower than that reported in previous studies among similar age and ethnicity groups in Denmark( Reference Thuesen, Husemoen and Fenger 10 , Reference Mølgaard, Larnkjær and Cashman 20 – Reference Andersen, Mølgaard and Skovgaard 23 ), other European countries( Reference Jaaskelainen, Knekt and Marniemi 11 , Reference Andersen, Mølgaard and Skovgaard 23 , Reference Gonzalez-Gross, Valtuena and Breidenassel 29 – Reference Hill, Cotter and Mitchell 33 ), the USA( Reference Forrest and Stuhldreher 34 , Reference Weng, Shults and Leonard 35 ) and Canada( Reference Whiting, Langlois and Vatanparast 36 ). Most of these studies measured vitamin D status across different seasons or in the winter as opposed to late summer in the present study; however, the distribution of 25(OH)D concentrations < 50 nmol/l found in the present study was also lower than that found during summer. Most of the participants of the present study were of normal weight, which could favourably affect the serum 25(OH)D concentration compared with, for example, the high rate of obesity in the USA( 37 ). Nevertheless, the studies should be compared with caution as differences may also depend on the laboratory and method used for the measurement of serum 25(OH)D concentrations( Reference Carter 17 ). In the present study, we used the LC–MS/MS method that might be considered the gold standard( Reference Fraser and Milan 38 , Reference Roth, Schmidt-Gayk and Weber 39 ), and our method was standardised and calibrated against the international reference material of the NIST( Reference Phinney, Bedner and Tai 18 ). The chromatographic methods are more specific compared with the frequently used immunoassays that are limited in their ability to detect vitamin D2 ( Reference Fraser and Milan 38 , Reference van den Ouweland, Vogeser and Baecher 40 ). In the DEQAS, the LC–MS method is positively biased for the all-laboratory trimmed mean, whereas the immunoassays are mostly negatively biased( Reference Carter 17 ). Several studies have also found the LC–MS method to yield better results than some other 25(OH)D assays( Reference Farrell, Martin and McWhinney 15 , Reference Lai, Lucas and Banks 16 , Reference Roth, Schmidt-Gayk and Weber 39 – Reference Moon, Cho and Hur 41 ). In a study of the German population, standardisation to the LC–MS/MS method has been found to reduce the prevalence of vitamin D deficiency ( < 30 nmol/l) from approximately 48 to 16 %( Reference Perna, Haug and Schoettker 42 ). This might partly explain the higher serum 25(OH)D concentrations found in the present study compared with previous studies in similar population groups.
Similar to the common finding of an inverse relationship between 25(OH)D and PTH concentrations( Reference Sai, Walters and Fang 43 ), we observed a negative trend between PTH and 25(OH)D groups in the adults and in children aged 4–10 years. In present study, relatively few subjects had a low serum 25(OH)D concentration ( < 25 nmol/l), but the PTH concentration was markedly higher in this group than in the other 25(OH)D groups. Elevated PTH concentrations may result in increased bone resorption in adults, whereas the implication for bone health in children is unclear( Reference Cashman 44 ).
For all subjects, outdoor stay in light clothes and sun vacation were major determinants of serum 25(OH)D concentration in late summer. In the adults, sun preference, solarium use and BMI were also strong determinants. We expected vitamin D status to be related to sun exposure as cutaneous vitamin D synthesis is considered to be the major source of vitamin D during summer( Reference Macdonald 45 ). It is interesting though that several expressions of sun exposure were related to serum 25(OH)D concentrations at the same time. Most suggestive may be that vitamin D status was associated with sun vacation (abroad), despite that the hours of sunshine the preceding summer (2010) in Denmark were only slightly less than (3 %) the average hours of sunshine in the preceding 10 years( 46 ). To our knowledge, this association between vitamin D status and sun vacation during the summer season has not been shown previously. The dietary vitamin D intake was also associated with serum 25(OH)D concentration in the present study, despite the median intake (2·5 μg/d) being much lower than the new recommended intake proposed by the Nordic Nutrition Recommendation. The recommended intake has recently been increased from 7·5 to 10 μg/d for 2- to 60-year-olds( 47 ). This makes room for an even greater improvement in vitamin D intake, and our finding suggests that dietary vitamin D intake is also important during summer even in a group of children and adults frequently staying outdoors.
In the present study, no association between vitamin D status and age was found. Some previous studies have shown an association between vitamin D status and age in both children( Reference Weng, Shults and Leonard 35 ) and adults( Reference Jaaskelainen, Knekt and Marniemi 11 ), whereas others did not find an association between 25(OH)D concentration and age( Reference Thuesen, Husemoen and Fenger 10 , Reference Hirani, Mosdol and Mishra 31 , Reference Hintzpeter, Mensink and Thierfelder 32 ). The observed higher vitamin D status in boys than in girls has been reported previously among similar age groups( Reference Rockell, Green and Skeaff 48 ). In the present study, this sex difference was not attributable to differences in dietary vitamin D intakes or multivitamin use. An explanation might be the higher level of physical activity in the boys than in the girls (56 % of the boys compared with 35 % of the girls reported to be involved in sports and physically active play activities in their leisure time; results not shown), assuming that these activities were primarily outdoor activities. Serum 25(OH)D concentration was strongly inversely related to BMI in the adults, whereas there was no association in the children. This might be because the children were not categorised as obese and only a few were overweight. Another study in healthy-weight children with a broad age span did not find an association between 25(OH)D concentration and BMI or fat mass either( Reference Weng, Shults and Leonard 35 ). The association between 25(OH)D concentration and obesity is a common finding( Reference Jaaskelainen, Knekt and Marniemi 11 , Reference Forrest and Stuhldreher 34 , Reference Bertrand, Giovannucci and Liu 49 ), and one explanation could be the sequestration of vitamin D in fat tissue( Reference Wortsman, Matsuoka and Chen 50 ). In a cross-sectional study of 686 adults, adjusting for body weight eliminated the obesity-related component of variability in serum 25(OH)D concentrations( Reference Drincic, Armas and Van Diest 51 ). This indicates that the lower 25(OH)D concentration could be due to dilution in the large fat mass of obese subjects rather than sequestration and that vitamin D requirements could be based on body weight.
In the adults, serum 25(OH)D concentration was not related to lifestyle when assessed together with the influence of vitamin D source-related factors, except for a borderline relationship with leisure-time physical activity. It might be that physical activity acts like a surrogate marker for sun exposure, assuming that the activities are mainly outdoor activities. Previous studies have found an association between 25(OH)D concentration and various lifestyle-related factors( Reference Thuesen, Husemoen and Fenger 10 , Reference Jaaskelainen, Knekt and Marniemi 11 ). In one study, an overall lifestyle index was used and vitamin D concentrations were found to be substantially higher in those with the healthiest lifestyle than in those with a less-healthy lifestyle and this difference was found be substantially higher than that between the single components of the lifestyle index( Reference Jaaskelainen, Knekt and Marniemi 11 ). This suggests that a high vitamin D concentration may serve as an indicator of a generally healthy lifestyle.
The strength of the present study was the random and population-based inclusion of families, which made it possible to compare vitamin D status across the age and sex groups. This has not been done previously in studies of the Danish population. Another strength was the use of detailed information on vitamin D sources including several variables for sun exposure, dietary vitamin D intake and supplement use. A limitation of the present study was that it was conducted at a single site in Denmark. However, the sample size of the present study was large and the subjects were randomly selected with few exclusion criteria, and we believe that the results are likely to be generalisable.
We assessed vitamin D status in a representative sample of Danish families using a standardised and calibrated method, and thus the results are useful for future comparisons of vitamin D status between populations. In conclusion, the majority of children and adults among the families had serum 25(OH)D concentrations >50 nmol/l in late summer in Denmark. Vitamin D status was associated with BMI, dietary vitamin D intake, multivitamin use, solarium use, outdoor stay in light clothes, sun preference and sun vacation, but was not associated with lifestyle-related factors in the adults when these were assessed together with the other determinants. Children within a family appeared to be more alike than the adults within a family with respect to their vitamin D status.
Acknowledgements
The authors are grateful to the families for their active and enthusiastic participation. They thank the students of the Metropolitan University College and the University of Copenhagen for assisting in the examination of the participants and the staff at Copenhagen's General Practitioners Laboratory for collecting the blood samples. They cordially thank the VitmaD project team and other colleagues at the Division of Nutrition, National Food Institute, Technical University of Denmark, for their assistance during the planning and implementation of the study and for helpful professional discussions.
The present study was supported by grants from the Danish Dairy Research Fund, Centre for Advanced Food Studies, and the European Regional Development Fund. The funders had no role in the design and analysis of the study or in the writing of this article.
The authors' contributions are as follows: I. T., R. A., H. M., K. H. M., L. B. R. and C. M. designed the study; H. M., J. N., K. H. M., L. B. R. and R. A. conducted the study; E. W. A., K. H. M., L. B. R. and R. A. analysed the data; K. H. M. wrote the first draft of the manuscript, which was critically reviewed and approved by all authors.
None of the authors has any conflicts of interest to declare.
