Physical activity during life course and bone mass: a systematic review of methods and findings from cohort studies with young adults

The literature search was conducted in the databases Pubmed, LILACS, Scielo and Web of Science. The search was performed by a single author and occurred up to May 2012 without date limits or language restrictions. Three command groups were employed to find articles. In the first group, we included the terms related to bone mineral density or content (bone density; bone mineral density; bone mass; bone mineral content; bone content). In the second one the terms related to physical activity were entered (physical activity; motor activity; inactivity; sedentarism; sedentary; sports; exercise). In the third group, we added the terms to restrict the study design (longitudinal; cohort; prospective; follow-up). Within each group, we used the Boolean operator ‘OR’ and between the groups we used the Boolean operator ‘AND’. In the Pubmed database we restricted the search for studies performed with adults (19–44 years), whereas in the other ones we added a fourth group of commands related to age group (adults, young adults, adulthood).

A database with the search results was generated, excluding duplicate references, totaling 750 articles. The selection of articles included in the final review was performed independently by two reviewers (RMB and JMM), based on inclusion and exclusion criteria previously defined. In the case of disagreement, the selection was evaluated by a third reviewer (DPG). Initially, each reviewer selected the titles for articles of interest. The second step consisted of the examination of abstracts from those papers previously selected. Then, we proceeded to search the full text. The references of all selected full text articles were also reviewed.

Criteria used to identify the manuscripts were regarding subjects, study design and measurement of outcomes. Concerning the subjects, studies should be conducted in healthy adult subjects with age from 20 to 40 years (or average in this interval) and not specifically athletes. The age was limited up to 40 years though the maintenance phase of bone mass occurs during young to middle adulthood [1], decreases on bone mass occur at earlier ages, mainly in women, due to premenopausal or menopausal periods. Another criterion was that the studies should evaluate bone mass using the method of dual energy x-ray absorptiometry (DXA) in at least one out of these three sites: total body, lumbar spine and femoral neck. The choice of this method was due to the evidence shown by the literature that DXA is the main method for evaluation of bone mineral density. Besides, it is the gold standard to diagnose osteoporosis [1, 7‐9]. Regarding the study design, we included only cohort studies, which performed at least one longitudinal analysis between physical activity and outcomes.

We excluded studies in which the sample was made up by subjects with diseases which are known to affect the bone metabolism (i.e. lupus erythematosus, cerebral palsy, cancer, etc.) and those that used other methods to measure bone mass. Cohort studies which performed only cross-sectional analysis between main exposure and outcomes were also excluded.

The evaluation of the quality of evidence was also performed independently by two authors (RMB and JMM). The disagreements were discussed between the two authors and the final decision was made by consensus between the two examiners. The instrument proposed by Downs and Black [10] was used to assess the quality of studies. These authors devised an instrument consisting of 27 questions that evaluate reporting, external validity, internal validity (bias and confounding), and statistical power. In items 4, 14, and 15, “intervention” was interpreted as “exposure,” and in no. 19 “compliance with the intervention” was replaced by “avoidance of misclassification error of the exposure”. Since the instrument was originally conceived for the evaluation of clinical trials, items applicable specifically to this study design (8, 13, 23, and 24) were not considered. All questions received scores 0 or 1, with the exception of question 5, which ranged from 0 to 2, depending on whether the statistical power of the survey was explicitly stated in the article as being at least 80%. Thus, the maximum score achievable by an article was 24 points. This manuscript was written according to current recommendations of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [11].

Figure 1 shows the study selection flowchart. Out of the 750 references initially located, 576 of the potential articles were excluded in the first step as the focus was not on PA; were conducted in unhealthy subjects, children, adolescents, postmenopausal women, elderly or athletes; or were cross-sectional studies. From the 174 papers with abstracts assessed, 49 were selected for reading the full text, based on the inclusion criteria. Out of these, 33 were excluded. The main reasons were the study design and age of subjects included in the sample [12‐30]. Other reasons for exclusion were the method for evaluation of bone mass [31‐35], as studies did not evaluate the effect of physical activity on bone mass [36‐41] or were a review of findings showed in other articles conducted with the same sample [42]. The reference lists of all selected papers were examined to detect other publications eligible for this review. In this process we identified one article which was not found before [43]. In the end, two other studies were found by search using the all author’s names of included manuscripts followed of terms related to bone mineral density or content previously described. In total, 19 articles were selected for this review.

The description of these studies is shown in Table 1. The most part of studies have been published in the last 12 years. We only found studies carried out in high income country. Out of these, the majority (n = 14) was performed in Europe. Concerning sample size, few studies (n = 6) had more than 200 subjects. Seventeen studies were carried out with females, whereas 11 were performed with males. Only three of these cohort studies did not evaluate the effect of physical activity during childhood or adolescence on bone mass. Three studies performed analysis between physical activity during adulthood and bone mass. Few studies used bone mineral content as outcome (n = 6), whereas bone mineral density was not evaluated only in three out of these 19 studies. Lumbar spine was the skeletal site most studied (n = 15). Femoral neck was evaluated by 13 studies, whereas association between physical activity and total body bone mineral density or content was showed in only five articles.

Table 2 shows other characteristics of the studies included in this present review. Although 19 articles were found in this present search, only 11 different samples were studied. For example, 5 of the manuscripts included in this review were written using data from Amsterdam Growth and Health Longitudinal Study (AGAHLS) [43‐47]. Twelve of the studies starts when subjects were up to 15 years-old [43, 45‐55] and the mean of time between first measurement of physical activity and measurement of considered outcome was 14.1 years (sd = 6.2 years). Different questionnaires to assess physical activity were used in these studies. Although the most part of these questionnaires were created by researchers themselves, other known questionnaires such as Baecke, Physical Activity Questionnaire (PAQ) and Kaiser Physical Activity Survey (KPAS) were also used [48, 49, 51, 52, 56]. About temporality of the information, two studies estimated physical activity during adolescence using retrospective questionnaire [57, 58].

Three studies considered only weight-bearing physical activities in the analysis [47, 49, 57], while the rest used general physical activity. The nineteen included studies showed twelve different ways to classify general physical activity by questionnaires. They used the following: a standard value for groups of activities according to intensity, times the resting metabolic rate (RMR) x minutes per week [44]; physical activity at least once per week (yes/no) [57]; membership of a sports club (yes/no) [57]; scores using different ranges of values [48‐53, 56]; categories of outdoor walking [58]; participation in sports at school for at least 2 hours (yes/no) [58]; hours of sports activity per week [49, 54]; number of metabolic equivalents (METs) per week [46]; MET score in levels determined according to intensity of each activity multiplied by the measured duration in minutes [45, 47]; four categories of physical activity, the first category was the active group and the third and fourth categories were the inactive group [59, 60]; individuals who performed two or more sessions of physical activity exceeding 30 minutes per performance were considered active (1) and inactive (0) for the others – subjects had the sum of the three years' answers ranging from 0 to 3 for physical activity from adolescence to adulthood [61] and; MET-times per week – annual average of metabolic equivalent for each activity multiplied by weekly frequency [55].

Physical activity was also analyzed using peak strain scores created by Groothausen [43]. Five manuscripts used this score [43‐46, 54], whereas four out of these manuscripts were conducted with AGAHLS sample. Peak strain score consists of evaluation of physical activity based on ground reaction forces of different physical activities. Activities with ground reaction force less than 1 time the body weight such as cycling and swimming have the peak score 0, activities with peak score between 1 and 2 times the body weight – weight bearing activities such as jogging, walking and ballroom dancing – have the peak score 1, activities with ground reaction force between 2 and 4 times the body weight – activities including sprinting and turning actions such as tennis, aerobics and soccer have the peak score 2, activities including jumping actions with ground reaction force greater than 4 times the body weight such as basketball and gymnastics have the peak score 3. Peak strain score may be used in two ways. Firstly, the peak scores of each activity are added up to others. Second option consists in selecting only the highest peak scores [43]. The evaluation in these studies was performed independent of frequency and duration of activities.

Concerning quality assessment, results of evaluation criteria adapted from Downs & Black [10] are shown in Table 3. Studies could reach the maximum of 24 points, divided into 5 different aspects – reporting, external validity, bias, confounding and power. No study reached this limit. Scores were on average 16.6 points (SD = 3.0). The lowest score was 14 points [43, 48, 49, 58], whereas only one study reached the highest score of 20 points [52]. Concerning questions about reporting, only 2 manuscripts had maximum score of 10 points [51, 57]. The main problem in this sub-scale was the lack of studies reporting the characteristics of patients lost to follow-up. Only 7 studies reported no difference between followed-up subjects and those who dropped out [45, 46, 51‐53, 56, 57]. Regarding sub-scale of external validity, around half of the manuscripts did not report at least one out of the two questions about representativity of the recruited sample at the baseline and about representativity of the followed-up subjects. More frequent fragility of all studies in sub-scale of bias was no attempting to blind the subjects and those who were measuring the outcomes to the exposures. No study reported these questions in the methods section. Concerning sub-scale of confounding, few studies took into account the losses of subjects to follow-up. On the other hand, all studies recruited the subjects of different grades of physical activity from the same population. No study reported sample size calculation, sufficient power to detect an important difference or minimum detectable difference on values of bone mass between grades of physical activity.

Findings in this section were summarized by analyses results. More details are presented in the Additional file 1.

Figure 2 shows the number of studies with at least one positive association between general physical activity and bone mineral content or density according to the period of physical activity measurement and sex. In this figure were included only first published manuscripts using each studied sample and each period of assessment of physical activity, to avoid possible biased conclusions caused by inclusion of more than one study that used the same sample. Out of 18 manuscripts included in this review, thirteen manuscripts are shown in Figure 2. Concerning manuscripts that evaluated these associations in males, all studies that performed association between general physical activity during adolescence and bone mineral density or content on young adulthood found at least one positive association with at least one anatomical site [47‐49, 52]. There seems to be no consensus on literature about existence of positive or absent association between general physical activity on young adulthood and bone mass at same period of life [46, 49, 52]. In contrast, regarding general physical activity from adolescence to adulthood, the only two existing studies showed positive association with bone mineral density or content in young adulthood [47, 61].

In females, there seems to be no consensus or lack of association between physical activity during adolescence and bone mass in young adulthood, since more studies reported absence than positive associations [47, 48, 52, 53, 55, 57, 58]. Furthermore, the findings of studies carried out with females showed that there was no association between general physical activity during adulthood and bone mass measurements [26, 46, 57, 58, 60]. However, the majority (n = 4) of the studies that evaluated association between cumulative general physical activity from adolescence to adulthood and bone mineral density or content in young adulthood showed positive associations [47, 50, 56, 61].

Concerning the only two studied samples (AGAHLS and LLSLFH) [43‐46, 54] in which peak strain scores were used to evaluate physical activity in addition to general physical activity, it seems that physical activity evaluated by peak score showed more positive associations with bone mass than general physical activity. Moreover, analyses performed with peak score in adulthood were more positively associated with bone mass than analyses using the adolescence period. Since one study [54] was carried out only with males, it is impossible to make pooled conclusions concerning differences in effect of peak score by gender.