Study design and ethics
The study design of the D-pro trial has been described in detail elsewhere . The overall aim of the D-pro study was to investigate effects of vitamin D and dairy protein on children’s growth and health. Effects on the primary outcome, bone mineral density , and the secondary outcomes fat mass index, fat-free mass index (FFMI), and cardiometabolic markers have been published previously . Noteworthy, we found no effects of the intervention on FFMI . Here, we report results on muscle strength and physical function along with LM as a supportive outcome linked to the former two outcomes. LM is highly comparable with FFMI as previously reported, but LM will be used in the present context, because it is a more accurate measure of muscle mass compared to FFMI, since bone mineral content is subtracted .
In brief, D-pro was a 2 × 2-factorial, randomized, controlled trial where 200 healthy, 6–8-year-old white, Danish children were included. They were randomly allocated to (a) ingest blinded tablets with either 20 µg/day vitamin D3 or placebo, and (b) substitute 260 g/day dairy in their diet with high protein (HP) or normal protein (NP) yogurt. The intervention lasted 24 weeks and was conducted at latitude 55°N during extended winter to avoid cutaneous synthesis of vitamin D. The study was conducted at Department of Nutrition, Exercise and Sports, University of Copenhagen, Frederiksberg, Denmark. Baseline and endpoint measurements were performed during August–October 2019 and February–April 2020, respectively . The study was conducted following the Declaration of Helsinki and approved by the Committees on Biomedical Research Ethics for the Capital Region of Denmark (no. H-19008199) and registered at clinicaltrials.gov with the ID: NCT03956732 before recruitment was initiated.
Participants and recruitment
Children, aged 6–8 years, who lived in the Capital Region of Denmark were identified through the Danish Civil Registration System. Children were recruited by postal invitation letters to the parents. Families who responded to the invitation letter were prescreened by telephone and were subsequently invited for an information meeting together with the child. Written informed consent was obtained from all custody holders .
Eligible children had to be of Danish or European origin and have white skin since randomization of children with dark skin to placebo tablets would conflict with the national vitamin D recommendations at the time of the study. Children should also like yogurt and high protein yogurt, “skyr”, and have a habitual intake of dairy products of at least 250 mL/day. Moreover, families should not be planning a stay below a latitude of 50° N during the intervention, and at least one parent had to read and speak Danish to be properly informed about the study procedures. Exclusion criteria were (1) known or suspected allergy or intolerance to dairy products, (2) diseases or medication that could interfere with the study outcomes, and (3) concomitant participation in other studies involving dietary interventions or blood sampling. Children could not use vitamin D-containing supplements > 3 days/week the last two months and any in the month immediately preceding the start of the intervention. Finally, only one child from each household could participate in the study .
A flowchart of the children from recruitment to completion is shown in Fig. 1. Children with test data from both baseline and endpoint visits were included in the present study, resulting in a sample of n = 183 for handgrip, n = 182 for leg press, and n = 122 for squat jump, long jump, and the 30 s sit-to-stand test. Handgrip and leg press data were lost for respectively one and two children, due to an electronic disconnection at the baseline visit. During the last three weeks of the endpoint examinations, the protocol was changed as a result of restrictions due to the COVID-19 pandemic. Consequently, performance in squat jump, long jump, and the 30 s sit-to-stand test were missing from 60 children at endpoint (Fig. 1).
Randomization and blinding
Block randomization with 12 children in each block was used to allocate children equally to each of the four intervention groups throughout the autumn. An impartial staff member produced a computer-generated randomization list, from which sealed, sequentially numbered envelopes containing the corresponding group allocation were produced, as previously described . Children were randomly assigned at the end of the baseline visit. The tablet intervention was blinded to all involved in the study, but due to the taste and nature of the yogurt products, blinding the yogurt intervention was not possible. However, the investigators were blinded for data analysis by recoding the allocation.
Vitamin D (Minisun®) and placebo tablets were provided by Oy Verman Ab, Finland. Tablets were chewable, identical in taste and appearance, and packaged in identical bottles. Parents were supplied with one bottle of tablets (containing 200 tablets) and a 7-day tablet dispenser and were instructed to provide one tablet per day to the child in connection with a meal as described previously [18, 19]. Tablet vitamin D content was analyzed by liquid chromatography–tandem mass spectrometry (LC–MS/MS) after completion of the trial (National Food Institute, Søborg, Denmark). The vitamin D content was 25 ± 2 µg/tablet and non-detectable in the vitamin D and placebo tablets, respectively. Tablet compliance was assessed by count of leftover tablets .
The HP and NP yogurts that were provided to the children were commercially available in Denmark and contained 9–11 g protein/100 g and 3–4 g protein/100 g, respectively . The HP yogurts were low-fat “skyr”, which is strained yogurt, similar to Greek yogurt, with a similar 20:80 whey:casein ratio as in normal yogurt (NP). All yogurt products were fermented with Streptococcus Thermophilus and Lactobacillus Bulgarius. Yogurts were provided free of charge from Arla Foods amba (Denmark) and came in both unflavored and fruit-flavored variants in original packaging. The target intake was 300 g/day, six days per week, and was to be substituted with the same amount of milk or yogurt in the child’s habitual diet to maintain the child’s usual dietary pattern . Therefore, the protein intake was expected to be around 17 g/day higher in HP compared to NP, corresponding to a total protein intake increase of around 25% . Parents were instructed to record the child's daily yogurt intake in pre-coded recording sheets, which were used to assess compliance and macronutrient intake from the yogurts .
The placebo tablets and NP yogurts were regarded as controls since vitamin D supplementation was not recommended in Denmark at the time the project was conducted , and Danish children typically have a dairy protein intake similar to the targeted intake in the NP groups. Apart from the intervention, parents were encouraged to maintain their child’s habitual dietary habits and participation in physical leisure activities during the study period.
Muscle strength and physical function
A battery of muscle strength and physical function tests were selected to obtain measures of both upper and lower body maximal strength, as well as physical function. This comprised: maximal isometric handgrip strength, isometric leg press, squat jump, long jump, and a 30 s sit-to-stand test. The test procedures have been described elsewhere . We previously found the reliability of these tests to be good-to-excellent for handgrip, leg press, and long jump tests [intraclass correlation coefficient (ICC) = 0.86–0.90], and moderate for the squat jump test (ICC = 0.71), even without familiarization , in Danish 6–9-year-olds. Retest reliability for the sit-to-stand-test was low (ICC = 0.63) without familiarization, yet moderate after familiarization (ICC = 0.78) . Therefore, the children of the present study were familiarized with the sit-to-stand test at the information meeting, approximately 1–2 weeks before the baseline test. All physical tests were led by one of two investigators according to standard operating procedures, and for 95% of the children, the examinations were performed by the same investigator at baseline and endpoint.
As previously described , children's PA was objectively measured by waterproof tri-axis AX3 accelerometers with a temperature sensor (Axivity Ltd., Newcastle upon Tyne, United Kingdom) during seven consecutive days and eight nights before the baseline and endpoint visits. The software OmGUI (v. 184.108.40.206) was used in the set-up and data download. The sensitivity was set to ± 8 g and sampling frequency to 100 Hz. Parents were instructed to attach the accelerometer to the children's non-dominant thigh with Fixomull tape (BSN Medical), as described elsewhere . All acceleration data was processed in Matlab (Version 9.9.0 R2020b, Mathworks Inc., Natick, Massachusetts, US), which included resampling into 30 Hz, generating ActiGraph counts , identification of non-wear periods, and summarizing the time spent in different intensity domains. Non-wear periods were identified from both acceleration and temperature . All periods identified as not worn were marked as missing data. Time spent with sedentary, light, moderate, and vigorous PA was estimated using 10-s epochs accounting for the elevated post oxygen consumption during intermittent PA, using the second-by-second epoch data to improve the assessment of vigorous activity [26, 27]. The cut-points defining the intensity domains were determined using an internal calibration study conducted in 5–13-year-old children . Moderate intensity was determined as the average counts for walking at self-selected speed. The energy expenditure during walking at self-selected speed was set by stature and the intensity corresponds to 30–35% of the children’s maximal aerobic capacity . The vigorous intensity threshold was determined as the count level for which 95% of the intensities measured during running were included. Running was considered a vigorous activity performed at > 60% of the children's aerobic capacity. The thigh cut points were 100 counts for sedentary behavior, 4822 for moderate behavior, and 9143 for vigorous behavior. Only subjects with at least two valid (≥ 90% wear time) weekdays and one valid weekend day  was included in the analysis. All PA outcomes were estimated as an average and adjusted using a 5/7 and 2/7 rule to account for the number of weekdays and weekend-days. Moderate-vigorous physical activity (MVPA) was calculated by adding time spent at moderate- and vigorous intensity. Participation in organized sport, i.e. type of sport, weekly frequency, and duration, was evaluated from an electronic questionnaire filled out by the parents at both visits.
Prior to the baseline and endpoint visit, parents completed a weighed, four-day dietary record in the web-based software Madlog (Madlog Aps, Kolding, Denmark)  from which the children’s energy and macronutrient intake was assessed, as previously described . A study investigator looked through the dietary registrations, and if any abnormalities were present, clarification was obtained by dialogue with the parents at the following examination visit. Vitamin D and calcium intake the preceding month was obtained from an electronic, previously validated food frequency questionnaire  at both visits .
Anthropometry and body composition
Bodyweight, height, BMI z-score, weight categories, and pubertal stage were obtained as previously described . LM was obtained from dual-energy X-ray absorptiometry (DXA) scans using a GE Lunar Prodigy (GM Healthcare, Chicago, Illinois, USA) with GE Healthcare software version 17, SP1) [18, 19], and was the result of total mass minus fat mass and bone mass. The scans were performed after a standardized breakfast and emptying of the bladder with the child wearing light clothes without metal. The same investigator checked the lines defining regions of interest on both baseline and endpoint scans and corrected these when necessary. Daily and weekly quality assurance (QA) was conducted with a QA-phantom and a Spine Phantom no. 17466, respectively, and coefficients of variation (CV) for LM was 1.5%.
To assess vitamin D status , an overnight fasting venous blood sample was drawn from the forearm of the child at both baseline and endpoint, after a 10 min supine rest. Local anesthetic patches (EMLA, AstraZeneca, Sweden) were provided beforehand to relieve the discomfort of blood sampling. Serum 25(OH)D2 and 25(OH)D3 were analyzed by LC–MS/MS at University College Cork, Ireland , and vitamin D deficiency and insufficiency were defined as total serum 25(OH)D < 30 nmol/L and < 50 nmol/L, respectively . The intra- and interassay CV were < 6% and < 5%.
The original sample size calculation of the D-pro trial was based on the primary outcome, bone mineral density. Assuming a power of 0.80 and a significance level of 0.05, n = 49 completers were required in each group to detect an accretion difference of 0.0085 g/cm2 or 0.8 SD between any two groups during the intervention .
Data is presented as mean ± SD, median [25th–75th percentile], or n (%), as appropriate. All analyses were conducted using Stata/IC 17 and included children with available data from both baseline and endpoint for the specific outcome (complete cases). Statistical significance and trends towards significance were established at P values below 0.05 and 0.1, respectively. Normal distribution of variables and model assumptions were evaluated visually by residual histograms and QQ-plots. All models met the model assumptions.
Differences in baseline characteristics between dropouts and completers were tested with Pearson's Chi-squared, Wilcoxon rank-sum, or two-sample t test, as appropriate. Within-group changes from baseline to endpoint were tested with paired t test. Differences in dietary intake and PA changes between the four study groups were analyzed with one-way ANCOVA models with study group as fixed factor, adjusted for baseline. Combined effects of the vitamin D and yogurt interventions on the outcomes were evaluated as between-group change differences adjusted for baseline with two-way ANCOVA with the interaction term vitamin D × yogurt, and main effects of vitamin D and yogurts were tested in an additive model. If any interaction was found, subsequent Bonferroni-corrected pairwise comparisons were performed.
To account for biological variation, secondary models were adjusted for age and sex. Further, we investigated whether the observed effects were mediated by change in energy intake or MVPA. In subsequent analyses, we also adjusted for calcium intake due to the interplay of vitamin D and calcium , and the potential association between calcium and muscle function . To check for consistency, per-protocol analyses were conducted with exclusion of those who reported breach of the protocol, either by commencement of non-study-related vitamin D supplementation (n = 4) or reporting that they had traveled south of 50° N during the intervention (n = 4). Finally, to assess whether sex influenced potential effects, the interaction terms vitamin D × sex and yogurt × sex were included in the additive models. We assessed the Bonferroni-adjusted pairwise associations between LM and muscle strength and physical function, respectively, at baseline and endpoint as well as for change from baseline to endpoint.