Background

Vitamin D functions as a hormone and its importance in the immunomodulatory effects on disease is being increasingly recognised [1], [2]. In the UK, there is evidence of a resurgence of vitamin D deficiency in children [3], [4], [5], [6]. The National Diet and Nutrition Survey (NDNS) for young people [7] (4–18 years old) in Great Britain has already reported that approximately 8% of children had 25-hydroxy Vitamin D (25[OH]D) levels <25 nmol/L. Data published from the NDNS has also shown that there is evidence of Vitamin D deficiency (<25 nmol/L) in most population age groups including in older children and young adults [8], [9]. In a US study of vitamin D levels of children aged 1–11 years old, 18% had levels of <25 nmol/L [10]. Factors known to influence vitamin D metabolism include skin pigmentation, age, ethnicity, sunshine exposure, body mass index, and season. Dietary factors include intake of vitamin D rich food, vitamin D supplementation, and the association of obesity with lower vitamin D levels [8], [11]. Genetic factors have also recently been implicated in the regulation of vitamin D status [12], and its association with chronic diseases such as multiple sclerosis, hyperglycaemia and the metabolic syndrome, hypertension, type 1 diabetes, and malignancy [2], [13], [14]. Recent UK government published guidelines [15] recommend vitamin D supplementation (7 micrograms/day) only to children younger than 5 years of age (Healthy Start initiative). However, recommendations have generally not been implemented as vitamins have only been made freely available to those receiving specific government benefits [8], [16]. There are no UK Reference Nutrient Intake values for vitamin D or government supplementation recommendations for older children; although the 2007 Scientific Advisory Committee on Nutrition report [8] does recommend that individuals who are at risk of inadequate sunshine exposure should receive supplementation (10 micrograms/day).” Additionally the government recommendations advise regular, short periods of sunlight exposure without sunscreen during the summer months, they also highlight the risk of skin cancer and recommend the use of sunscreen (with a high UVB factor) for the majority of time spent outside.

There is increasing consensus that vitamin D insufficiency (VDI) should be defined as levels of <50 nmol/L [16], [17], [18]. There is a paucity of data in the UK and indeed worldwide on the prevalence, predictors and associations of VDI. There have been three recent and relatively large cross-sectional studies exploring associations with vitamin D deficiency and associations: (1) National Health and Nutrition Examination Survey (NHANES) in the USA [19] (n = 6275), Philadelphia [20] (n = 382) and New Zealand [21] (n = 144). However, the studies were either single site with Vitamin D data available for a subset of cases (Philadelphia), included a narrow age range (12–22 months) of patients (New Zealand) or unable to evaluate specific important predictors (and or determinants) of VDI such as season of measurement of 25(OH)D levels, the latitude of the participants' residents, and information on outdoor play and sports from the NHANES study.

Additionally, there is also a paucity of data on the relative importance of proposed predictors for vitamin D status in UK children [16]. There is a need to identify factors that need to be adequately controlled for when conducting case control studies involving vitamin D and its association with autoimmune conditions. Furthermore, clinicians need to be better informed as to when to recommend vitamin D supplementation for children. Finally, it is important to characterise the relative importance of modifiable risk factors, as this will inform wider recommendations concerning the prevention of VDI.

Objective

The aim of this study was to describe vitamin D status and comprehensively evaluate important currently accepted predictors of VDI in Great Britain, in children aged 4–18 years of age by examining data obtained from the NDNS [22] of young people.

Study Design

The survey plan and procedures have been described in detail in the NDNS report [7] (Gregory et al, 2000). Briefly, a nationally representative sample of young people aged 4 to 18 years living in private households was acquired. Fieldwork covered a 12 month period (January 1997–1998), to cover any seasonality of factors related to nutritional status. Where there was more than one young person living in the same household, only one was randomly selected to take part in the survey. The sample was selected using a multi-stage random probability design with postal sectors as first stage units. The elements of the survey included:

• a detailed interview to provide general information about dietary habits and background information about lifestyle and socio-demographic characteristics;
• a seven-day weighed intake dietary record of food and drink consumed;
• a seven-day physical activity diary for young people aged 7–18 years;
• physical measurements — height and weight;
• a fasting blood sample (with written consent) analysed. This sample was subdivided and used for the measurement of a wide range of biochemical status analyses in three laboratories.

Children aged 11 years and older were expected to keep their own dietary and physical activity record. Younger children needed varying levels of help from parents, teachers and other carers. To help the nutritionists evaluate the quality of the dietary records completed by the young people, interviewers completed a quality assessment questionnaire. This included information on how accurate the interviewers thought the weighing and recording of items eaten had been and whether the diary was an accurate reflection of the young person's actual diet. In a sub-group of a feasibility study sample the validity of the dietary recording methodology was tested using the doubly-labelled water methodology to compare energy expenditure against reported energy intake [7]. For the same sub-group the physical activity information collected in the diary was validated by directly measuring the young person's activity level using a motion sensor.

Vitamin D assay

The NDNS currently uses the DiaSorin Liaison Total assay (formerly Incstar Minnesota, USA) for the measurement of 25(OH)D. There were 800 micro-litres of serum available for 25(OH)D analysis. The radioimmunoassay kit which was used was based on the developmental work by Hollis et al [23]. The Vitamin D assay was conducted at the Medical Research Council (MRC) Dunn Nutrition Unit (DNU) in Cambridge, UK (now the MRC Human Nutrition Research). Analysis of samples was conducted in a sequence so as to avoid bunching within fieldwork areas, during each batch analysis. Each assay was performed in duplicate, and if agreement between duplicates failed to meet quality controls repeat assays were performed. Percentage coefficient of variation reported for the 4 fieldworks was 15.38–21.11% [7]. In addition, 20 independent samples for inter-laboratory quality assurance comparisons were analysed yielding a mean deviation of 0.02 for sample concentration 12.9–81.4 nM.

Participants and sample size

Data from the NDNS [22] of young people aged 4–18 years were utilised as the survey is the largest and most detailed ever undertaken of the diet and nutritional status of young people in Britain. The NDNS programme aims to provide a comprehensive, cross-sectional picture of the dietary habits and nutritional status of the population of Great Britain living in private households. The multivariable nature of prognostic research makes it difficult to estimate required sample size, and ideally require several hundred outcome events [24]. The NDNS provides a suitable sample as a total of 1102 Vitamin D (25[OH]D) blood sample results (from a total of 1701 subjects interviewed) were available for analysis, and of those 917 had detailed interview, and physical activity diary records.

Weighting factor adjustments were used to adjust for known socio-demographic differences, and differential probability for having blood tests between the composition of the survey sample and that of the entire (census) population of Great Britain.

Variables used as predictors

Proposed candidate predictors which have previously been reported as prognostic were used. The following variables were derived from the dataset and investigated as predictors:

• Age, gender and ethnicity
• Whether on income Support
• Region residing
• Month blood was taken
• Amount spent per week in sports involving outdoor exercise or play
• Amount spent a day per week watching TV
• Vitamin D dietary intake and supplemental Vitamin D intake
• Body Mass Index (BMI [kg/m²])

To classify overweight and obesity in young people of 2–18 years of age, cut offs described by Cole et al (on behalf of the International Obesity Task Force) were used [25], [26]. Cole et al developed a series of age and sex-specific BMI cut-off points based on data collected from UK, Brazil, Hong Kong, Singapore, the Netherlands, and the USA. These BMI cut-offs were derived from sex specific BMI age curves that pass through a BMI of 25 and 30 (the cut-off points used in adults to define overweight and obesity) at 18 years of age, respectively.

Outcomes measured

For the purposes of this study, Vitamin D status was defined [16] as insufficient when 25(OH)D levels were <50 nmol/L).

Statistical methods

Statistical analysis was carried out with SPSS version 17.0 and included descriptive statistics, univariate regression and binary loglinear regression. In the different models, Vitamin D was used as a continuous variable, and as a binary outcome for insufficiency (<50 nmol/L). Chi squared and t-tests were also used to investigate significance of proposed predictors for VDI and vitamin D status respectively. For the binary logistic regression (for VDI), the dataset was split into two parts (70% and 30%) and the model developed on the larger sample called the ‘training dataset.’ A backward elimination approach was used for the logistic regression, with entry at p value of <0.05 and removal at p<0.1. Performance of the model was assessed with classification plots, Hosmer and Lemeshow test, concordance index, ROC analysis, and Nagelkerke R Square. Internal validation was conducted by assessing the predictive accuracy of the model on the second portion (30%) of the dataset, called the ‘validation dataset.’

Ethical Approval

The NDNS survey was approved by the National Health Service Local Research Ethics Committees (LRECs) in the areas representing each of the 132 postcode sectors where fieldwork took place. The National Diet and Nutrition Survey (NDNS) was carried out by the Medical Research Council (MRC), Office for National Statistics, and the Dental Schools of the Universities of Birmingham, Newcastle, Dundee and Wales. Novel data analysis and interpretation was carried out by the authors independently of the funding sources based on the available dataset provided by the UK Data Archive. Written consent was obtained by the parents or guardians of children involved in the study. Appropriate permission was gained from the UK Data Archive to publish data from the NDNS.