Introduction

A large battery of trace elements and minerals are required by the human body for proper physiological functioning [1]. Apart from being measurable in blood and urine, minerals are also detectable in scalp hair. During the hair growth process, hair is exposed to the blood supply which contains traces of anything consumed by the individual. In this way, minerals are incorporated in the protein structure of scalp hair and are no longer affected by the prevailing metabolic conditions. Hair mineral concentrations have been shown to be related to a number of (patho-)physiological conditions (e.g. Parkinson’s disease) [2], but there is also evidence suggesting that hair mineral levels may reflect dietary habits. Hair mineral analyses might thus be a valuable tool providing information on an individual’s dietary intake of essential elements, particularly for long- or medium-term assessment. After all, bulk hair concentrations present an average over several months and are thus far less affected by daily fluctuations in environmental conditions than, e.g., blood or urine [3].

Hair mineral levels have been associated with, e.g., the consumption of highly processed foods, slimming and laxative preparations, coffee, tea and vegetables and with malnourishment and vegetarian diets [4–11]. In contrast, some other studies did not observe such relationships [12–14]. It is difficult to determine the reliability of these inconsistent results because of the limited number of studies that have been conducted to specifically address the association between hair mineral concentrations and nutritional status (taking into account food consumption patterns or nutrient intake) [5, 7, 9–11], but also because of difficulties in the interpretation of results. For example, hair mineral concentrations may not necessarily represent blood or other endogenous concentrations or ingested mineral doses, as stated by Kempson et al. [2]. Elevated hair mineral concentrations could as well indicate limited use of the mineral by tissues and subsequent elimination in hair as excretory route, as postulated by Chojnaka et al. [5], although these hypotheses need to be investigated.

A first aim of this study was to establish age-specific reference values for the minerals calcium (Ca), copper (Cu), iron (Fe), magnesium (Mg), phosphorus (P), sodium (Na) and zinc (Zn) for a large sample of healthy, Belgian (Flemish) elementary school girls, a previously unexplored study population (N = 218). Also the correlations between these hair minerals were investigated. A next, parallel aim of this study was to examine the relationship between consumption frequencies of different food types and hair mineral concentrations in young children (i.e. elementary school girls), as this has not been studied before. More specifically, the current investigation studied the proportion of variation in hair mineral concentrations that can be explained by the consumption frequency of specific food groups.

Methods

Study Participants

Agreement for participation was obtained through parental written informed consent for 218 elementary-school girls participating to the 2010 baseline survey of the ChiBS project (‘Children’s Body composition and Stress’; mean age = 8.44 year, SD = 1.11 year; mean BMI z-score = −0.05; SD = 1.23). The ChiBS project, designed at Ghent University and embedded within the European IDEFICS study [15], investigates the relationship between chronic psychosocial stress and changing body composition in children aged 5 to 10 years old (at baseline) over a 2-year follow-up period (2010–2012) [16]. The ChiBS project offered the opportunity to study the utility of hair samples as biomarker or diagnostic tool for stress and mineral status in children and to investigate its interest as a simple and non-invasive alternative for biological sampling (such as blood sampling) in young children. In total, 523 healthy children participated to the 2010 baseline ChiBS survey. Participation to this study was however limited to the female participants of the ChiBS project (N = 263/523 or 50.3 %), as hair sampling was only performed in girls to ascertain the required hair length of 6 cm (informed consent for N = 218/263 or 82.8 %). No differences were found between boys and girls for age, BMI, parental education or food consumption frequencies, except for milk and yoghurt consumption which was higher in boys (p = 0.035). Detailed research goals of the ChiBS project and socio-demographic information of the ChiBS study population is described elsewhere [16]. The ChiBS project was conducted according to the guidelines laid down in the Declaration of Helsinki and was approved by the Ethics Committee of the Ghent University Hospital.

Hair Mineral Analysis

Hair samples were obtained from the vertex posterior region of the scalp by trained researchers. The hair samples were cut as close to the scalp as possible using clean, stainless steel scissors and tied together with a little cord to mark the proximal side. To guarantee that the same time period was investigated in all children, only the most proximal 6 cm of the hair strands was analysed. None of the hair samples was artificially coloured. The samples were stored in a folded piece of paper in individual zip-lock bags in a dark, dry place and at constant temperature until analysis in the Department of Analytical Chemistry of Ghent University. The hair contents of Ca, Cu, Fe, Na, Mg, P and Zn were quantitatively determined via inductively coupled plasma–mass spectrometry (ICP-MS), after microwave-assisted acid digestion of the samples. Approximately 0.1 g of each hair sample was subjected to a washing procedure prior to the digestion in order to remove any external contamination, such as grease, sweat, dust etc. This cleaning stage consisted in stirring the samples, first in acetone and subsequently in ultrapure water (obtained from a Direct Q3 water purification system from Millipore, fed by distilled water) inside an ultrasonic bath (Branson 5510) followed by further rinsing of the samples with Milli-Q water. Immediately afterwards, samples were allowed to dry in an oven at a temperature of 60–70 °C until complete dryness. Once dried, the samples were weighed, transferred into microwave TFM vessels along with 1 mL of 14 M HNO3 (Chemlab Belgium, pro analysis, further purified by means of sub-boiling) and 9.8 M of H2O2 (Fluka Analytical, Sigma Aldrich Belgium, for trace analysis) and subjected to the following microwave power program for digestion: 2 min at 250 W, 2 min at 0 W, 6 min at 250 W, 5 min at 400 W and 5 min at 600 W (Milestone mls 1200 mega microwave labstation). The digests thus obtained were then allowed to cool down to room temperature before further dilution with Milli-Q water in pre-cleaned PP tubes. Samples were analysed by means of sector field ICP-MS (Thermo Element XR). Simultaneous monitoring of Ca, Cu, Fe, Na, Mg, P and Zn and the internal standard Ge (added to a final concentration of 25 μg L−1) was accomplished using the instrument settings and data acquisition parameters summarized in Annex 1. Sector field ICP-MS was selected for its superior detection power and capability to avoid spectral overlap by measurement at higher mass resolution. The internal standard was relied on to correct for matrix effects, instrument instability and signal drift. External calibration was accomplished relying on a series of standard solutions prepared from commercially available 1 g L−1 stock solutions. Validation of the analytical method developed was carried out by analysing BCR 397 powdered human hair certified reference material. However, certified reference values were available only for Zn and for the rest of the micronutrients considered (with the exception of Na) only informative values were provided. Recovery assays were also carried out for all the target elements as an alternative validation procedure. Results for both accuracy checks are shown in Annex 2. While most of the results are in a good agreement with the certified/informative values for the reference material, the value obtained for P attracts attention. However, results for the recovery assays were satisfactory and further analysis by ICP-OES provided a similar result. Moreover, the measured value also seems to be in accordance with the typical values reported in the literature for a normal P hair content. Analytical parameters such as precision and limits of detection for all the elements of interest were also estimated. With this aim, one sample available in a sufficient amount was selected for evaluating the homogeneity of the hair and method reproducibility. Five different locks of that sample were analysed for the same target elements and the relative variation of the results was taken as an indicator for the precision of the method. The detection limit for each element of interest was calculated as three times the standard deviation on the analysis of ten different procedural blanks, divided by the slope of the external calibration curve. These figures are gathered in Annex 3.

Food Consumption Frequencies

Children’s dietary habits were assessed using the self-administered parental questionnaire ‘Children’s Eating Habits Questionnaire—Food Frequency Questionnaire’ (CEHQ-FFQ), which was fully completed for 109 children. The CEHQ-FFQ is a 43 food-item-containing questionnaire developed and validated within the IDEFICS project [17–19] and is used as a screening instrument to investigate dietary habits and food consumption frequency in children [20]. Parents were asked to report on their child’s consumption frequency of selected food items (e.g. fruits, vegetables, drinks, breads and cereals, milk products, meat, fish, fats, snacks and desserts) in a typical week during the preceding 4 weeks, outside the school canteen or childcare meal provision settings, using the following response options: ‘never/less than once a week’, ‘1–3 times a week’, ‘4–6 times a week’, ‘1 time per day’, ‘2 times per day’, ‘3 times per day’, ‘4 or more times per day’ or ‘I have no idea’. These frequency categories were converted to consumption frequencies per week. Frequencies of intake were assessed without quantifying portion sizes. Additionally, a questionnaire on the use of vitamin supplements was included (type of supplements, formulation, frequency of intake; completed for N = 187).

Other Variables

Within the ChiBS and the IDEFICS project, information was collected on the child’s age, body mass index (BMI) (categorisation according to the International Obesity Task Force guidelines [21]), physical activity (i.e. hours of playing outdoors and in sports clubs), parental education (i.e. International Standard Classification of Education—ISCED [22]) and parental income (relative to the average national net household income).

Statistical Methods

Statistical analyses were performed with PASW Statistical software version 19.0.0 (SPSS Inc, IBM, USA) and SAS software version 9.3 (SAS Institute Inc., USA). p values <0.05 were considered statistically significant for all tests. Non-normally distributed data were presented by their median and interquartile range (Kolmogorov–Smirnov, Shapiro–Wilk). Spearman correlation analyses were performed to examine the correlation between hair mineral concentrations and continuous variables (age, physical activity, diet score, etc), while Kruskal–Wallis tests were applied to investigate differences in terms of hair minerals between (categorical) groups.

As hair minerals were previously shown to be influenced by age [2, 23–27], we reported age-specific reference values, using the LMS method of Cole [28] in which each year of age was considered as one age group (age 6 to 10). Only one child had the age of 5 and was therefore excluded from this analysis. Age-specific, smoothed L (skewness), M (median), and S (coefficient of variation) curves and percentile reference values (3rd, 10th, 25th, 50th, 75th, 90th and 97th percentile) were obtained for all minerals, except for Na, as the number of samples that exceeded the LOD was too low (N = 50; LMS Chartmaker Pro Software (version 2.54)) [29]. The degrees of freedom for L, M and S, respectively, were: 2, 2, 2 for Ca; 2, 3, 2 for Cu; 4, 4, 4 for Fe; 2, 4, 4 for Mg; 2, 3, 3 for P and 3, 4, 4 for Zn. Q tests and detrended Q-Q plots were used to asses goodness-of-fit and normality.

To identify food groups that explain the largest proportion of variation in the hair mineral concentrations, reduced rank regression (RRR) was applied using the partial least square procedure in SAS. RRR determines linear functions of predictors (i.e. food groups) and extracts so-called factors by maximizing the explained variation in responses (i.e. hair mineral concentrations) [30]. Factor loadings indicate the relationship between the food groups and the derived factor and thus indicate which food groups load highly onto the factor which explains variation in hair minerals. Data on food consumption frequencies and hair minerals were firstly adjusted for several covariates (children’s age, BMI z-score, physical activity level, hair colour and parental income) using linear regression and then entered as residuals into the RRR analyses. As RRR implies that the number of extracted factors equals the number of selected responses, one factor was obtained for each hair mineral. For clarity, only food groups with factor loadings ≥0.15 were reported. RRR analyses were not performed for hair Na as the sample size was too small for these analyses (N = 26).

Results

Hair Mineral Reference Values

Population characteristics of the participating girls and age-specific reference values for Ca, Cu, Fe, Mg, Na, P and Zn are presented in Tables 1 and 2, respectively.

Table 1 Population characteristics of the studied girls (N = 218)
Full size table
Table 2 Age-specific L, M, S and percentile values for hair minerals
Full size table

These reference values refer to healthy, female and predominantly Caucasian girls between 6 and 10 years old living in Flanders (Belgium). Table 2 also presents previously reported reference ranges for other childhood and adolescent populations. The child’s age and hair Fe concentrations were negatively correlated (Spearman’s rho = −0.277, p < 0.001). For the other minerals, no significant correlation with age was observed (results not shown).

Table 3 presents hair mineral contents in relation to parental income and natural hair colour, as Cu and P concentrations in hair were shown to differ significantly according to parental income and natural hair colour, respectively (Kruskal–Wallis p = 0.047 and p < 0.001, respectively). Parental education and physical activity were not associated with the hair mineral concentrations (data not shown).

Table 3 Hair mineral concentrations and population characteristics
Full size table

Inter-mineral Correlations

As shown in Table 4, hair minerals are strongly positively intercorrelated. More particularly, Ca and Cu, Cu and Na, Cu and Mg, Mg and Zn and P and Zn showed correlation coefficients >0.3, while Ca and Mg showed an even higher correlation coefficient (r = 0.881).

Table 4 Inter-mineral correlations in hair of Belgian elementary school girls (N = 217a)
Full size table

Food Consumption Frequencies and Hair Mineral Concentrations

Table 5 presents the amount of variation in hair mineral concentrations explained by the retained RRR factor, as well as the loadings of the food groups (based on consumption frequency) for the respective factors (i.e. loadings ≥ 0.15). Relatively large percentages (i.e. at least 40 %) of hair minerals were explained by the retained RRR factors.

Table 5 Percentage of variation in hair mineral concentration explained by reduced rank regression analyses and factor loadings of food groups with value ≥0.15
Full size table

Examination of food groups that contribute to these factors reveals that the factor explaining Ca mainly consists of milk desserts (ice cream, milk bars), meat, eggs and snacks; the situation is similar for Mg (milk desserts, eggs, sweet snacks). The factor explaining Cu primarily covers snacks, fruits and white bread products. The factor for Fe loads highly on sweetened milk products (sweetened milk, yoghurt, fermented beverages), sweet snacks, chocolate or nut-based spreads and whole meal bread products. Similarly, the factor explaining P mainly consists of sweetened milk products, sweet snacks, white bread products and fish. To end, the Zn factor covers fruits, fish and sweet snacks. Table 5 indicates the top ten of contributing food groups for each hair mineral factor separately. No relationship was observed between hair mineral concentrations and vitamin or mineral supplement use (N = 25/187; data not shown).

Discussion

As few studies have addressed hair mineral concentrations in relation to specific dietary habits and the significance of this topic has remained controversial, we analysed the relationship between hair minerals (Ca, Cu, Fe, Mg, P and Zn) and food group consumption frequency in a large sample of young girls and established age-specific reference values for Belgian elementary school girls, a previously unexplored study population. In addition, we developed reference values for hair minerals in our study population.

Hair Mineral Reference Values

The data presented are of significance as up to now no reference ranges for hair mineral concentrations were available for Flemish children. The presented reference values are generally in line with previously reported data for other childhood and adolescent populations, although our concentrations of Ca, Mg and Zn were slightly higher, while the observed Fe and Na concentrations were lower than those previously reported (Table 2) [23, 24, 26, 31, 32]. However, comparison of our observations with female specific reference ranges for girls aged 6 to 11 and 7 to 10, published by Perrone et al. and Senofonte et al., respectively, showed that our findings on Ca, Mg and Fe are similar, with even higher concentrations for Mg and lower concentrations for Fe reported in Senofonte et al. and Perrone et al., respectively [23, 26].

Apart from analytical variability among laboratories [33], some inter-individually differing characteristics may account for these varying reference ranges such as ethnicity, geographical and environmental location (urban area, rural area etc.) [34], age, gender, personal habits (use of shampoos, dietary habits etc.) and genetics [2, 35]. As our study exclusively consisted of female participants and hair minerals have been shown to differ between genders [23–27, 32, 36, 37], comparison of the reference ranges obtained with those for other general childhood and adolescent populations may not be straightforward. Gender differences in hair minerals have been attributed to differences in metabolism, hormonal balance and physiological role of minerals between males and females [24, 36]. Nevertheless, as our study population was homogenous (i.e. only girls), the presented reference values were thus indirectly normalized for gender.

We observed significant age-specific variations in hair Fe concentrations, although Table 2 demonstrates that also Ca and Mg showed a decreasing (but not significant) trend with increasing age [27]. This is in agreement with findings in adult populations [37, 38] but in contrast with other childhood findings [26]: Sakai et al. and Perrone et al. observed age-specific variations for the elements Zn and Cu in hair from birth over childhood to adolescence, representing variations in (hair) minerals with different physiological, developmental periods [23, 24]. As the age range included in our study was small (i.e. a 5-year range), it was not possible to observe a clear, significant age trend for each mineral.

Moreover, we recorded a significant influence of hair colour on hair mineral concentrations, more specifically for P with an increasing trend towards darker hair colours [2]. This is in contrast to findings reported by Chojnacka et al. who did not detect an influence of natural hair colour on the P level, whereas they showed the Cu concentration to be highest in dark hair and those of Ca and Mg in blond hair [27], findings we could not confirm. No relationship was observed between parental education or physical activity level and hair mineral levels, although parental income was significantly associated with hair Cu concentrations. This may be due to a dietary pattern that varies with socio-economic status [39, 40]; a more elaborate study of this issue is beyond the scope of this manuscript.

Inter-mineral Correlations

Strong positive correlations between hair minerals were observed [2, 25]. In particular, the strong correlation between Ca and Mg (Table 4) has been demonstrated previously [36, 37, 41, 42] and may indicate common chemical properties, common occurrence in nature or common dietary sources [25, 36]. Indeed, as observed in Table 5, the same food items load on the RRR factors for Ca and Mg (except for water and meat alternatives which are only present for the Ca factor), indicating that consumption frequency of the same food items contribute to the variability in hair Ca and Mg levels. These findings were less apparent for the other inter-mineral correlations. Kempson et al. suggested that mineral relationships may originate or occur at different levels or stages, i.e. at ingestion (e.g. Mg dependence of Ca absorption [36]), during transport, during hair incorporation or after hair formation by way of contamination or substitution [2].

Food Consumption Frequencies and Hair Mineral Concentrations

Few investigators have studied hair minerals in relation to the dietary habits or diet quality. Gonzalez-Reimers et al. analysed the association between hair Fe, Cu and Zn and diets rich in meat, fish or vegetables based on a dietary recall of the last 2 weeks in healthy adults from the Canary Islands. In contrast to our findings, they could not indicate a relationship between hair Zn and fish, or hair Cu and vegetables and judged the relationships between hair minerals and type of diet to be poor (except for their observed relationship between hair Fe and vegetable consumption) [9]. Hong et al. investigated nutrient intakes based on a FFQ in Korean female adults and found that the hair micronutrient content was not directly influenced by each mineral/micronutrient intake: only hair Na correlated positively to dietary potassium and vitamin C intake [7]. The relationship between hair Ca and dietary habits in Polish young women has been investigated by Jeruszka-Bielak et al., using 4-day dietary records and information on fortified foods and food supplements [10]. Hair Ca was found to weakly relate to the diet quality (i.e. meat and dairy consumption), but more importantly synergistic interactions with vitamin D were observed (i.e. higher vitamin D intake from supplements correlated positively with hair Ca). This is in contrast to our study showing no relationship with supplement use or vitamin D supplements in particular, although this could be due to the low number of children taking supplements (N = 25/187). The influence of an omnivorous or semi-vegetarian diet on hair minerals and blood pressure has been studied in Spanish postmenopausal women [11], using a 14-day precise weighing method. Rodenas and colleagues found distinct differences in hair mineral contents between omnivorous and semi-vegetarian participants, suggesting the influence of diet on hair minerals; results which are however in contradiction to Wojciak et al. [13]. Last, reduced hair calcium, magnesium, zinc and copper levels were observed in Polish obese children (10–14 years old) after slimming therapy, indicating an influence of decreased food intake on mineral metabolism and accumulation in hair [43].

Our observations are in agreement with the above-discussed findings in other subject groups, indicating a relationship between diet and hair minerals. However, studies in young children are lacking (apart from the study of Wojciak et al. in obese children from 10 to 14 years old [43]), limiting detailed comparison of our observations. We did not intend to link hair minerals to healthy or unhealthy diets, or to represent the most important mineral contributors to the child’s diet as the administered FFQ was designed to investigate the children’s dietary habits only and not to investigate mineral intakes. In contrast, we examined to what level variation in hair minerals may be explained by consumption frequencies of different food groups in elementary school girls, thereby investigating whether or not hair minerals might to a certain extent be related to dietary habits. As expected, we observed that a significant proportion of variation in hair minerals was explained by the consumption frequency of specific food groups, independently from the child’s age, BMI, hair colour, physical activity and parental income. This may indicate that high consumption frequencies of specific foods may stimulate the excretion and incorporation of minerals into hair, although further fundamental research into mineral metabolism is needed (e.g. use of minerals by tissues, elimination of minerals into scalp hair as excretory route etc.).

A number of food groups known to be ‘rich’ sources of minerals did not emerge as contributors to certain hair minerals: e.g. milk did not load to the Ca and Mg factor; and meat products did not appear in the Zn and Fe factor, despite the known high nutrient density and high consumption frequencies of these food groups in the studied population (data not shown) [44]. So, despite the observed relationship between hair minerals and food consumption, insufficient evidence is provided to consider hair minerals as direct reflection of dietary habits. We expect other mechanisms and processes to be involved in the mineral incorporation and accumulation in scalp hair (with inter-individually varying rates and extents) which may complicate the direct relationship between diet and hair minerals.

Concerning this complex relationship between nutrient intake and biochemical indicators in general (e.g. serum, urine, hair), it should be noted that nutrient intake is just one of the determinants affecting the nutrient status. Other determinants are, e.g., the homeostatic control mechanisms and a wide array of genetic, environmental and life-style factors [45]. For these reasons, associations between bio-indicators (in e.g. serum or urine) and nutrient intake have not always been evident. Although recovery bio-indicators such as 24-h urine samples have sometimes offered valid measures of nutrient intake, the selection and validation of an appropriate method (i.e. matrix) for estimating a particular nutrient intake has been recommended to be made on a nutrient-by-nutrient basis [45].

Even though hair mineral determination should not (yet) be considered a clear-cut diagnostic means for assessing the child’s dietary habits and/or mineral status, or should not replace serum or urinary nutritional biomarkers, they could be used as complementary means to investigate the minerals status with the possibility to detect deviations from homeostasis or changes between groups of individuals with a different health status, nutrient intake or geographical location over a long period retrospectively [1, 45].

Strengths, Limitations and Further Research

To our knowledge, this study was the first to present age-specific hair mineral reference values for Flemish girls and to specifically investigate the relationship between food consumption frequencies on hair mineral variations by the use of RRR analysis, a technique that has gained increasing importance in nutritional epidemiology [30]. Other strong methodological features of this study are the use of the validated and state of the art ICP-MS technique to measure the hair mineral concentrations; application of the LMS method to establish age-specific reference values and its large sample size. However, some limitations should be considered when interpreting results. The CEHQ-FFQ did not include portion sizes or school and canteen meals and was not designed or evaluated for assessing nutrient intakes. Therefore, no mineral intakes could be calculated from the FFQ and a direct comparison between mineral intakes and hair status was not possible. Also, the CEHQ-FFQ assesses children’s dietary habits during the last month, while our hair minerals represent a period of approximately 5–6 months in the past (i.e. 6 cm of hair sample). Possibly, results could be stronger or more accurate if a time-matched hair sample would have been analysed (1 cm hair sample), although both measurements represent ‘chronic’ assessments. As our population only consisted of females in a small age range, findings should be confirmed in a more heterogeneous population sample (i.e. boys and girls, childhood to adolescence), as hair mineral concentrations and dietary habits vary between sexes and age groups which thus influences the relationship between hair minerals and diet. Nevertheless, this study may initiate further hair mineral research in relation to diet in children, as hair mineral analysis offers considerable advantages for large-scale epidemiological research in children: hair sampling is easy, non-invasive, inexpensive and the samples are easily stored. Moreover, hair minerals are less influenced by short-term (hourly or daily) concentration fluctuations in, e.g., food intake or other environmental factors compared to blood or urinary samples. Further research (1) examining intra-individual mineral statuses in different types of biological matrices in parallel (e.g. hair, 24-h urine, blood), (2) investigating detailed mechanisms and processes involved in hair mineral accumulation and (3) further validating hair minerals as bio-indicators for dietary habit and/or mineral intake, may enlarge their use in nutritional epidemiology.

Conclusion

This study strengthened previous indications of a relationship between diet and hair minerals. More specifically, we observed that a significant proportion of variation in hair minerals may be explained by the consumption frequency of specific food groups. However, up to now, there is insufficient evidence to consider hair minerals as a direct reflection of dietary habits. Our findings should be confirmed in a more heterogeneous population and future research should investigate the mechanisms and processes involved in the mineral incorporation and accumulation in scalp hair and study the relationship with the total body burden of these minerals or the mineral status in other biomatrices, in order to fully understand the importance and influence of diet on hair minerals.