Osteoporotic fractures are a major cause of morbidity and mortality whose burden is expected to increase in the future( 1 , Reference Lidgren, Smolen and Bentley 2 ). Research has been focused on understanding the mechanisms of age-related bone loss, but growing evidence emphasizes the influence of early-life factors in the attainment of adult bone mass( Reference Heaney, Abrams and Dawson-Hughes 3 , Reference Javaid and Cooper 4 ). It has been suggested that high peak bone mass protects against fragility fractures later in life( Reference Heaney, Abrams and Dawson-Hughes 3 – Reference Bachrach 5 ). Diet is known to modulate achievement of the full genetic potential for skeletal mass during adolescence( Reference Javaid and Cooper 4 , Reference Ilich and Kerstetter 6 ) but research has typically examined the effects on bone mineral density (BMD) of single nutrients or foods (e.g. Ca, P, vitamin D, vitamin K and protein) or food groups such as dairy products, fruit and vegetables( Reference Cashman 7 – Reference Weaver 9 ). However, meals consist of a variety of foods with complex combinations of nutrients that are likely to be interactive and many foods and nutrients are highly correlated, making it difficult to separate their effects. Effects of single nutrients may be too subtle to detect with the available measurement instruments, but combined effects as expressed in a dietary pattern may be sufficiently large.
Dietary patterns analysis is a more comprehensive approach to dietary exposure assessment and to examine its relationship with bone health( Reference Rizzoli, Bonjour and Chevalley 10 , Reference Kontogianni and Yiannakouris 11 ). Dietary choices in epidemiological studies may be assessed through compliance with dietary patterns defined a priori based on presumed health effects( Reference Kant 12 ). Indices to measure adherence to the Mediterranean diet (MD) have been used to explore the risk of obesity, CVD and cancer( Reference Bach, Serra-Majem and Carrasco 13 ). Although some components of this pattern have been associated with bone quality in children( Reference Prentice, Schoenmakers and Laskey 14 ), the overall effect of adherence to this type of diet on bone accrual remains unknown.
The Dietary Approaches to Stop Hypertension (DASH) diet is a Ca-rich diet that emphasizes fruits, vegetables and low-fat dairy products( Reference Sacks, Svetkey and Vollmer 15 – Reference Appel, Moore and Obarzanek 17 ). Although originally designed for blood pressure reduction, several aspects of this pattern may benefit bone such as the rich Ca, K and Mg contents( Reference Lin, Ginty and Appel 18 ).
MD and DASH do not specifically assess foods with a possible negative effect on bone quality. Such information is of major interest and can be obtained with the Oslo Health Study (OHS) dietary index built on prevailing knowledge about foods that potentially influence bone health. The OHS is computed as the ratio between soft drinks and fruit and vegetables consumption and has been directly associated with metabolic syndrome and inversely associated with BMD( Reference Hostmark 19 , Reference Hostmark, Sogaard and Alvaer 20 ).
In addition to being a critical period in the acquisition of bone mass( Reference Bachrach 5 ), adolescence is also a period of social and psychological influences on food choices( Reference Shepherd and Dennison 21 ).
To the best of our knowledge, there are no studies examining the association between a priori-defined dietary patterns and BMD in adolescents. By using prospective data from a population-based cohort, our objective was to quantify the associations between forearm BMD in early and late adolescence and adherence to a priori-defined dietary patterns in early adolescence, using the Mediterranean Diet Quality Index for children and adolescents (KIDMED), the DASH diet index and the OHS dietary index.
Experimental methods
The EPITeen cohort was assembled during the 2003/2004 school year, when all public and private schools in Porto that provided teaching to adolescents born in 1990 were approached. Nineteen private and all public schools (79·2 % v. 100 %, P = 0·04) agreed to participate in the study. Executive boards were asked to provide contact information for each student's family. Parents received written explanation of the purpose and the study and meetings were arranged in each school in order to explain research aims and procedures. We identified 2787 eligible adolescent students, of whom 2160 (1044 boys) agreed to participate and provided information for at least part of the protocol (77·5 % participation at the individual level). Similar participation proportions were obtained in public and private schools. The follow-up evaluation of the cohort took place in the 2007/2008 school year and 1716 participants were successfully re-evaluated (79·4 %). Sampling procedures and detailed methods have been described elsewhere( Reference Ramos and Barros 22 ). The study was approved by the Ethics Committee of the University Hospital of São João, Porto and policies and procedures were developed to guarantee data confidentiality and protection. Written informed consent was obtained from both parents and adolescents.
Physical examination
In both evaluations, forearm BMD (g/cm2) was measured at the distal radius of the non-dominant forearm by dual-energy X-ray absorptiometry (DXA) using a Lunar® PIXI device (GE Medical Systems, Madison, WI, USA). In cases of reported previous fracture of the non-dominant arm, the dominant arm was assessed. Anthropometry was obtained while the student stood barefoot in light indoor clothing. Weight was measured using a digital scale to the nearest tenth of a kilogram (Tanita Corporation, Tokyo, Japan) and height was measured in centimetres, to the nearest tenth, using a portable stadiometer (Seca Deutschland, Hamburg, Germany); BMI was calculated.
Home questionnaires variables
At 13 years of age, dietary intake was evaluated using an FFQ designed according to Willett et al.( Reference Willett, Sampson and Stampfer 23 ) and adapted for the Portuguese population and previously validated in Porto adults( Reference Lopes, Aro and Azevedo 24 , Reference Ramos 25 ). The FFQ was adapted for adolescents through the addition of foods more frequently eaten by this age group( Reference Silva, Rego and Guerra 26 , Reference Araujo, Severo and Lopes 27 ). The FFQ comprised ninety-one food items and an open-ended section to add items not listed in the questionnaire but eaten at least once per week. Participants were asked to report, over the prior 12-month period, the frequency of consumption of each food in one of nine categories, ranging from ‘never or less than once per month’ to ‘six or more times per day’. The FFQ did not include specific questions on portion size and the frequency reported was multiplied by a standard portion size to estimate the average daily intake. Seasonal variation of food consumption was considered by multiplying the frequency and portion of seasonal items by a seasonal factor of 0·25 (i.e. equivalent to consumption during a 3-month period). Food consumption was converted into energy and nutrient intakes with the software program Food Processor Plus (1997; ESHA Research, Salem, OR, USA) based on values from the US Department of Agriculture and which was updated with values for typical Portuguese foods( Reference Lopes, Aro and Azevedo 24 ).
Information on regular practice of physical activity at 13 years was quantified through the frequency spent in sport activities of at least twenty consecutive minutes beyond compulsory school activities. Adolescents were classified according to parental educational level measured as the number of successfully completed years of formal schooling of the parent with higher education. Since examination at school limited conditions to evaluate adolescents’ pubertal development according to Tanner stages( Reference Tanner 28 ), age at menarche referred at 13 years of age was recorded as a pubertal development indicator for girls.
Data analysis
The degree of adherence to the MD pattern was measured using the Mediterranean Diet Quality Index for children and adolescents (KIDMED)( Reference Serra-Majem, Ribas and Ngo 29 ); adherence to the DASH diet was evaluated through a DASH diet index score proposed by Fung and colleagues( Reference Fung, Chiuve and McCullough 30 ); and the total score of the OHS dietary index( Reference Hostmark, Sogaard and Alvaer 20 ) was calculated.
The original KIDMED includes sixteen components. Items denoting a negative connotation with respect to the MD are assigned a value of −1, and those with a positive aspect are assigned +1. This index assesses the daily consumption of at least one serving of fruit and vegetables, rating higher when consumption is greater. Weekly consumption of at least two to three servings of nuts and fish and of more than one serving of pulses is evaluated. It also assesses the consumption of pasta or rice at least five times weekly; grains or cereals, daily for breakfast; dairy products, three servings daily; and olive oil for culinary use. Items denoting a negative connotation with respect to the MD include frequent intake of sweets, candy, commercially baked goods, pastries, fast foods, and non-consumption of breakfast( Reference Serra-Majem, Ribas and Ngo 29 ). The higher the KIDMED score, the more Mediterranean and favourable is the dietary pattern. We eliminated the items regarding breakfast since our dietary evaluation was based on an FFQ. Excellent agreement (κ = 0·86) between the original KIDMED and the adapted version was found in a different sample of 117 adolescents aged 13 years.
We used a DASH score awarding points for high intake of fruit, vegetables, nuts, legumes, dairy products and whole grains according to quintile rankings (i.e. participants in the lowest quintile were assigned 1 point and those in the highest quintile were assigned 5 points). Regarding the intake of Na, red and processed meat, and sweetened beverages, participants in lower quintiles of intake scored higher( Reference Fung, Chiuve and McCullough 30 ). Higher scores reflect higher compliance with DASH characteristics.
The OHS dietary index was constructed based upon the literature about foods that potentially influence Ca balance or acid load, and consequently bone health. It is a ratio between the intake of soft drinks (negative for bone health owing to its role in acid load increase and promotion of urinary loss of Ca) and the intake of fruit/berries, fruit juice and cooked and raw vegetables (positive for bone owing its role in promotion of Ca absorption and reduction of Ca loss)( Reference Hostmark, Sogaard and Alvaer 20 ). Thus, it is expected that higher scores on this index, reflecting unhealthier patterns, are negatively associated with BMD. Our index was based on the sum of intake frequency categories of three items (range 3–27): ‘colas’, ‘ice tea soft drinks’ and ‘other soft drinks’, divided by the sum of intake frequency categories of three items (range 3–27): ‘vegetables’ (fresh or cooked), ‘vegetable soup’ and ‘fruit’ (fresh, canned or juice). This adaptation was due to differences in the FFQ used.
Categorical and continuous variables were summarized as percentages and as means and standard deviations, respectively, by tertile of the final KIDMED, DASH diet index and OHS dietary index scores, separately for boys and girls. Differences in proportions were tested with the χ 2 test. One-way ANOVA and Kruskal–Wallis one-way ANOVA were used to compare continuous variables between independent samples.
Using adherence to the MD dietary pattern, to the DASH diet and to the OHS dietary index as the main exposures and forearm BMD as the outcome, associations were estimated cross-sectionally (KIDMED score13 v. BMD13, DASH diet score13 v. BMD13 and OHS dietary index score13 v. BMD13) and prospectively (KIDMED score13 v. ΔBMD, DASH diet score13 v. ΔBMD, OHS dietary index score13 v. ΔBMD, KIDMED score13 v. BMD17, DASH diet score13 v. BMD17 and OHS dietary index score13 v. BMD17), using linear regression coefficients and 95 % confidence intervals. BMD was used in mg/cm2 to improve readability. In addition to crude estimates, linear regression coefficients were also adjusted for height and weight at 13 years and at 17 years old, total energy intake at 13 years old, parental educational level, regular practice of physical activity and mean Ca intake. In girls, coefficients were further adjusted to menarche age to address confounding by pubertal status.
Sample description
From a total of 2160 adolescents who participated in the first evaluation, 1264 (591 boys) were included in the cross-sectional analysis (complete information on forearm BMD, anthropometric variables, parental educational level, age at menarche and FFQ). Of those 1264 adolescents, 1075 (85·0 %) participated in the follow-up evaluation. The final sample in the longitudinal analyses included 1023 adolescents (474 boys) after exclusion of participants with missing information in the follow-up evaluation (Fig. 1).
Fig. 1 Flowchart of the included and excluded participants in the cross-sectional and longitudinal analyses (BMD, bone mineral density)
When participants with and without missing information for any of the above-mentioned variables in the baseline evaluation were compared (896 v. 1264 participants, respectively), we found no differences in female proportion, in BMD and in BMI means at 13 years. However, adolescents with missing information had lower median (25th percentile, 75th percentile (P25, P75)) parental educational level (9 (6, 12) years v. 11 (7, 16) years, respectively; P < 0·001).
No significant differences in female proportion, in BMD and in BMI means at 13 years were found between participants with complete information in both baseline and follow-up evaluations and participants with missing information (1023 with complete information v. 896 with missing information in the baseline evaluation + 189 lost to follow-up + fifty-two with missing information in the follow-up evaluation). However, adolescents with complete information had higher median (P25, P75) parental educational level (12 (8, 16) years v. 9 (6, 12) years, P < 0·001).
Results
Mean forearm BMD increased from 0·360 (sd 0·057) g/cm2 at 13 years to 0·435 (sd 0·052) g/cm2 at 17 years in girls and from 0·342 (sd 0·050) to 0·452 (sd 0·075) g/cm2 in boys.
There were no statistically significant differences in mean KIDMED and DASH diet scores by gender (KIDMED score: 5·2 (sd 2·0) for boys, 5·1 (sd 2·1) for girls, P = 0·410; DASH diet score: 23·7 (sd 4·5) for boys, 23·7 (sd 4·5) for girls, P = 0·908). The mean OHS dietary index score was significantly higher in boys than in girls (1·1 (sd 0·6) v. 1·0 (sd 0·5), respectively, P = 0·005).
Average values of anthropometric, sociodemographic and behavioural characteristics in the baseline evaluation, stratified according to tertile of each a priori dietary pattern, are presented in Table 1 for girls and boys. Proportional distributions are presented for categorical variables.
Table 1 Description of average BMI, regular physical activity, parental educational level and total energy intake at 13 years old (baseline evaluation, 2003/2004) according to tertile of adherence to the Mediterranean diet (MD) pattern, to the Dietary Approaches to Stop Hypertension (DASH) diet and to the Oslo Health Study (OHS) dietary index among Portuguese girls and boys, EPITeen cohort
P25, 25th percentile; P75, 75th percentile.
We observed that adolescent girls and boys with healthier dietary patterns, i.e. whose dietary patterns were more similar to the MD pattern, to the DASH diet and who had lower scores in the OHS dietary index, were significantly more likely to have parents with higher education (all P < 0·001).
For both sexes, adolescents with higher adherence to the MD pattern, with higher adherence to the DASH diet and with lower OHS dietary index scores tended to practise physical activity more frequently. Significantly higher frequency of physical activity was observed in girls when adherence to the MD pattern (P = 0·008) and the DASH diet (P = 0·032) was higher and in boys when OHS dietary index score was lower (P = 0·038).
Although adolescents in the highest tertile of adherence to the MD pattern, as compared with those in the first and second tertiles of adherence, had significantly higher energy intake (P < 0·001), these adolescents had no statistically significant differences in BMI in the baseline evaluation. We also found no meaningful differences between tertiles of adherence to the MD pattern with regard to mean BMI at 17 years in the follow-up evaluation (data not shown).
As in the baseline evaluation (Table 1), we also found no significant differences in mean follow-up BMI by tertile of adherence to the DASH diet and of OHS dietary index score among males and by tertile of OHS dietary index score among females (data not shown). Although significant differences in girls’ mean follow-up BMI by tertile of adherence to the DASH diet have been observed (first tertile: 21·7 (sd 3·4); second tertile: 22·6 (sd 3·5); third tertile: 21·9 (sd 3·2); P = 0·027), a dose–response relationship was not present.
Table 2 shows the average baseline and follow-up BMD for each of the a priori dietary patterns in tertiles. We observed no statistically significant differences in mean BMD, measured in the baseline and in the follow-up evaluations, or in BMD change by tertile of adherence to the a priori-defined dietary patterns.
Table 2 Bone mineral density (BMD) at 13 years old (baseline evaluation, 2003/2004) and at 17 years old (first follow-up evaluation, 2007/2008) according to tertile of adherence to the Mediterranean diet (MD) pattern, to the Dietary Approaches to Stop Hypertension (DASH) diet and to the Oslo Health Study (OHS) dietary index in Portuguese girls and boys, EPITeen cohort
Table 3 summarizes the cross-sectional and prospective associations between adherence to the MD pattern, to the DASH diet and to the OHS dietary index and BMD in girls and boys, by presenting crude and adjusted linear regression coefficients and 95 % confidence intervals. Both crude and adjusted linear regressions coefficients confirm the lack of clear associations between the a priori dietary patterns studied and BMD at 13 years and longitudinal changes in BMD at 17 years in this sample of adolescents. Among males, higher adherence to the MD pattern was significantly associated with higher BMD at 13 years (P = 0·013) but not to its annual variation. Further adjustment of the linear regression coefficients for adolescents’ regular practice of physical activity and mean Ca intake did not change the results appreciably (data not shown).
Table 3 Linear regression coefficients (95 % confidence intervals) for the cross-sectional and prospective associations between adherence to the Mediterranean diet (MD) pattern, to the Dietary Approaches to Stop Hypertension (DASH) diet and to the Oslo Health Study (OHS) dietary index and bone mineral density (BMD) in mg/cm2 among Portuguese girls and boys, EPITeen cohort
BMD13, bone mineral density (mg/cm2) at 13 years old; BMD17, bone mineral density (mg/cm2) at 17 years old.
Further adjustment of the linear regression coefficients for regular practice of physical activity and mean Ca intake did not change the results appreciably (data not shown).
*Adjusted for height and weight at 13 years of age, total energy intake, parental educational level and, in girls, age at menarche.
†Adjusted for height and weight at 17 years of age, total energy intake, parental educational level and, in girls, age at menarche.
Discussion
In the present study we observed no significant difference in annual variation of BMD, between 13 and 17 years of age, by adherence to the different a priori dietary patterns. However, among boys, a significant linear trend towards increased BMD at 13 years with increasing adherence to the MD pattern was observed.
Little research has been done examining the relationship between a priori-defined dietary patterns and bone quality( Reference Hostmark, Sogaard and Alvaer 20 , Reference Kontogianni, Melistas and Yannakoulia 31 – Reference Zagarins, Ronnenberg and Gehlbach 34 ). To our knowledge, the present study is the first one to explore both cross-sectional and prospective associations between a priori-defined dietary patterns and BMD during the adolescent growth spurt.
We identified two studies examining the association between adherence to the MD pattern and BMD( Reference Kontogianni, Melistas and Yannakoulia 31 , Reference Whittle, Woodside and Cardwell 33 ). Adherence to the MD was not significantly related to total body bone mineral content or to BMD at the lumbar spine in adult women( Reference Kontogianni, Melistas and Yannakoulia 31 ). Also no evidence of an association between BMD or bone mineral content and adherence to this pattern was found in a sample of young adults( Reference Whittle, Woodside and Cardwell 33 ). The association between adherence to the DASH diet and bone metabolism was studied as a 3-month DASH intervention study in adults and this pattern significantly reduced bone turnover( Reference Lin, Ginty and Appel 18 ).
The MD and DASH diets are characterized by high intakes of fruit and vegetables, wholegrain bread and cereals, pulses, nuts, dairy products and fish( Reference Serra-Majem, Ribas and Ngo 29 ). Although these patterns value the consumption of alkali-forming food groups that have been associated with higher BMD (i.e. fruit and vegetables)( Reference Prynne, Mishra and O'Connell 35 – Reference McGartland, Robson and Murray 37 ), they also include increased quantities of acid-forming foods (i.e. cereals, pulses, nuts and dairy products). The latter can change the acid–base balance and may preclude expression of the beneficial effects of fruit and vegetables( Reference Remer and Manz 38 , Reference Bushinsky 39 ).
One of the main limitations of the a priori dietary patterns approach is related to the use of dietary guidelines that generally are not disease specific and the adherence to them may reduce the risk of some diseases but not others( Reference Michels and Schulze 40 ). Therefore, we additionally studied the association between BMD and the OHS dietary index. Although this pattern was associated with BMD in a sample of adults( Reference Hostmark, Sogaard and Alvaer 20 ), our study found no indication of such an association among adolescents.
In our study, the ascertainment of usual diet was done using an FFQ and, as with any method to assess dietary intake, under-reporting of usual intake is possible because of social desirability bias( Reference Togo, Osler and Sorensen 41 , Reference Livingstone and Black 42 ), particularly in overweight and obese individuals( Reference Heitmann and Lissner 43 , Reference Newby 44 ). Differential under-reporting of diet components is a relevant limitation to pattern analysis( Reference Togo, Osler and Sorensen 41 ). For example, diet under-reporting of foods high in fat or sugar, such as fast foods, pastries or sweetened beverages, will be reflected in higher scores in the MD and DASH indices and in lower scores in the OHS dietary index. Additionally, there could be inherent imprecision associated with the option of not quantifying food portion sizes. However, research has shown that the majority of variation in food intakes is captured by frequency of consumption( Reference Noethlings, Hoffmann and Bergmann 45 ). The lack of validation of the FFQ in this adolescent population may also be a limitation. However, the FFQ had been previously validated in the adult population of the same city( Reference Lopes 46 ).
Moreover, in our study dietary intake was evaluated at baseline and we do not know whether adherence to dietary patterns changed throughout the follow-up. Although some studies have been providing evidence that variations of dietary patterns occur during adolescence( Reference Madruga, Araujo and Bertoldi 47 , Reference Patterson, Warnberg and Kearney 48 ), it is also likely that dietary patterns remain relatively constant over time( Reference Cutler, Flood and Hannan 49 – Reference Oellingrath, Svendsen and Brantsaeter 52 ).
Taking into account the weak and few significant associations found in the present study, we hypothesize that dietary patterns may not be the primary factor in determining BMD in this age group. It is possible that dietary patterns may not have the impact on BMD in adolescents as they do in adult populations. One possible explanation may be the fact that, contrary to what happens in heterogeneous samples of adults, our sample of adolescents is very homogeneous, since they were born in the same year and share the urban environment of Porto city, making it much more difficult to find consistent associations between diet and BMD.
In our study we observed gender differences in the association between adherence to the MD pattern and BMD, possibly attributable to sex-specific difference in BMD response to biological and environmental determinants at this age( Reference Macdonald, Kontulainen and Petit 53 ). Indeed, it is possible that the environmental determinants of bone development during adolescence differ according to sex as was previously described for the relationship between physical activity and BMD( Reference Weeks and Beck 54 ).
In order to optimize feasibility and to minimize radiation exposure we used forearm BMD measured by DXA. The DXA-derived BMD is based on the two-dimensional projected area of a three-dimensional structure which does not capture true volumetric density or bone geometry( Reference Crabtree and Ward 55 ). However, BMD remains a valid index of bone quality and DXA-derived measures have been shown to predict fracture risk accurately, which is ultimately the goal of bone quality assessment( Reference Flynn, Foley and Jones 56 ). Since we studied bone quality only by examining forearm BMD measured with DXA, we cannot rule out the possibility that other measures of bone quality are associated with the dietary patterns considered.
One possible confounder of the association between dietary patterns and BMD is body size. Although we did not find substantial variation in mean BMI, weight and height between different classes of adherence to the dietary patterns studied, which may be related to the aforementioned limitations of the dietary assessment method, there is a well-documented important weight-dependent positive association between adiposity and bone strength( Reference Zhao, Jiang and Papasian 57 ) that has been found in a previous work evaluating the girls of this cohort( Reference Lucas, Ramos and Severo 58 ). For this reason we found it essential to account for confounding by body size in the present analysis.
Owing to the difficulty in ensuring privacy at school, adolescents’ pubertal development by the Tanner criteria( Reference Tanner 28 ) was not assessed. This may be a limitation since it is possible that the higher height observed in boys of the third tertile of the MD pattern may reflect greater maturity. However, for a sub-sample of 121 boys, we identified no significant difference in mean level of serum collagen type 1 cross-linked C-telopeptide, a marker of bone resorption that has been associated with skeletal maturation( Reference Silva, Goldberg and Nga 59 ), according to tertile of MD pattern adherence. In girls, age at menarche was recorded as an indicator of pubertal development in girls.
Physical activity is also a possible confounder of the association studied, but in the present study it was not consistently associated with either dietary patterns or BMD (data not shown). One possible explanation is the low validity of self-reported information to assess physical activity. However, the most probable reason is the previously described high levels of sedentary behaviour in this population( Reference Ramos and Barros 22 ), which probably situated the vast majority of the sample below the threshold level for exercise-induced bone formation. In fact, there is evidence that light or moderate physical activity is not associated with bone properties( Reference Sayers, Mattocks and Deere 60 ). Although we cannot exclude that physical activity might modify the effect between diet and BMD, the further adjustment to this confounder did not change the results.
As with any observational analysis, it is possible that other confounders that were not measured in the study or error in the measurement of the existing confounders could lead to residual confounding that could not be accounted for. However, we controlled for other potential confounders, which did not substantially affect our estimates.
Another limitation is the large number of comparisons made in the present study, which makes it difficult to rule out the role of chance in any one result. However, adjustment for multiple comparisons by the Bonferroni method did not substantially change the results( Reference Rothman 61 ).
Although the dietary patterns approach has certain advantages over traditional methods of examining the relationship between diet and health outcomes, results can be more challenging to interpret. In fact, more than the food combinations consumed, dietary patterns probably reflect individual food preferences modulated by a mix of genetic, cultural, social, health, environmental, lifestyle and economic determinants( Reference Kant 12 ). As a result of this complexity, mechanisms to explain the observed associations are not clarified by the present results. In fact, more healthy dietary patterns are often reported with a constellation of other desirable health behaviours, thus confounding the pattern and health association( Reference Kant 12 ).
Conclusions
We did not find consistent or strong associations between dietary patterns defined a priori and forearm BMD in early or late adolescence. The selected dietary patterns may not capture the elements of diet that are truly important in determining adolescent bone quality or, given the overall adequacy of nutrient intake in high-income populations, dietary patterns may not add substantially to other determinants of BMD at this age.
Acknowledgements
Sources of funding: This work was supported by the Portuguese Foundation for Science and Technology (PTDC/SAU-ESA/108407/2008, PTDC/SAU-EPI/115254/2009 and FCOMP-01-FEDER-0124-015750). Conflicts of interest: The authors have no conflicts of interest to disclose. Authors’ contributions: T.M. was responsible for data analysis and interpretation and for drafting the first version of the paper. R.L. contributed substantially to the interpretation of results and critically revised the manuscript. E.R. supervised the design of the study as well as the data collection protocols, contributed to the interpretation of data and critically revised the manuscript for intellectual content. H.B. supervised the conception and the design of the study as well as the interpretation of results and critically revised the paper for intellectual content. Acknowledgements: The authors gratefully acknowledge Associação Portuguesa de Osteoporose (APO) for making the bone densitometry equipment available for both evaluations.
