1.1 BMD and BMC analyzed by photon absorptiometry.

The majority of the studies included in this review used photon absorptiometry to assess bone mass in swimmers; in fact 53 of the 64 studies included used this method to evaluate bone mass (Table 1). Dual energy X-ray absorptiometry (DXA) is the most common photon absorptiometry method used. DXA is a two dimensional measure highly influenced by body size [21]. It therefore seems necessary to adjust by covariates to minimize the differences among participants when these are compared. The decision regarding which covariates better adjust the bone mass values is taken by each researcher, taking into account participant age-range, comparison group and so on. For the purposes of this review the final results and authors conclusions presented in each published work were used, regardless of the covariates employed and whether results had been adjusted.

Nilsson et al. [22] first evaluated bone mass in swimmers, other athletes and in a non-athletic control group (CG) aged 18 to 22. They observed higher BMD in the femur of all the athletes than in the CG; however, swimmers did not differ in BMD values when compared with the CG. These findings of similar BMD values in swimmers and CG were reinforced by subsequent studies that also compared swimmers bone mass with CG who performed less than 3 hours of physical activity per week in both male and female subjects, in children [23]–[27], adolescents [28]–[34] young adults [35]–[41] or elderly populations [35]. Some of these studies included a comparison sport group that also showed no differences in BMD when compared with the swimmers [22], [31], [38], [39]. Furthermore, some of the first studies performed on swimmers such as Jacobson et al. [42] or Orwoll et al. [43] as well as other studies found higher arm BMD [16], [42]–[44] or BMC [45], [46] in swimmers than in CG. However, other measured sites in these studies such as lumbar spine [16], [42], [45], a weight bearing zone, or whole body [44], were similar in swimmers and CG. These exclusive higher arm BMD values in swimmers may be due both to the level of force applied by the forearm muscles while swimming, and to the fact that this part of the body is not overly used in daily life by the general population. In fact, Orwoll et al. [43] who included both men and women in their study, found differences only in the male group. According to the authors this was due to the greater forces applied by males, not reached by females who presented values similar to their peer CG. Out of the 50 studies included in this review using DXA, only 2 [45], [47] showed higher whole body BMD in swimmers than in CG. In both studies lean mass, which is well known to influence BMD, [48] was significantly higher in swimmers than in the CG, and was not included as a covariable in the comparisons. This fact may mask some important real differences. Nevertheless, neither of the two cross-sectional studies were graded with a 6/7 in the quality assessment.

Of the 6 studies [16], [42]–[46] that showed higher upper limb BMD values in swimmers than CG, four were performed on adults over 40 years old suggesting that when practiced in the postmenopausal period swimming might reduce the rate of normal bone mass loss accompanying age [42], [43]. However, those studies showing higher upper limb BMD in swimmers, all graded with a 4/7, and others performed with older aged populations [35], did not take into account other physical activities or sedentary behaviours during life, calcium intake [42], [45] or lean mass [42], [43], [45], all of them variables affecting bone. In fact, when Dook et al. [44] controlled by lean mass, the differences in BMD between swimmers and CG disappeared. This may imply that swimming benefits muscle mantainance but the direct effect of swimming on bone mass at these ages is not clear.

Only 2 studies [37], [45] took past physical activity into account: Andreoli et al. [45] performed a retrospective study concluding that physical activity during youth appeared to have a beneficial effect on bone mass later in life. This conclusion underlines the importance of registering past physical activity in studies evaluating bone mass in the elderly population. Greenway et al. [37] did in fact evaluate past, recent and current physical activity in addition to swim participation, and showed that swimmers, who presented fewer cases of lower bone mass than CG, had performed greater amounts of physical activity (excluding swimming) at the ages ranging from 10 to 19. These higher levels of physical activity registered in the study performed by Greenway et al. [37] may be similar physical activity patterns to those of older adult swimmers evaluated in other studies, who showed higher BMD values than CG but whose past physical activity was not registered.

The only longitudinal study performed in postmenopausal former swimmers showed lower BMD and BMC reductions during a one year follow-up in the former swimmers than the sedentary controls [49]. However, these former swimmers performed 3 times more current physical activity than the controls. Physical activity other than swimming may be the cause of higher BMD and should therefore be taken into account in further studies focused on evaluating bone in a later adulthood population.

It would therefore appear that swimming may be beneficial in later adulthood. In spite of these results, out of the 7 aboved mentioned studies [35], [37], [42]–[45], [49] which compared older swimmers with CG, four [35], [42], [44], [45] also included a sport group (SG), showing lower BMD values in swimmers than in the SG, in leg [44], [45], lumbar spine [42] and hip [35].

Lower BMD values in swimmers than other SG, were not exclusive to later adulthood. Many studies also showed lower leg BMD values in adolescent [14], [15], [17], [29], [30] and adult [36], [40], [41], [50]–[53] swimmers. More important than the leg bone mass values are the lumbar spine and hip values where osteoporotic fractures could take place later in life. Focusing on the lumbar spine, lower BMD at this site was also found in children [23], [25], [26], adolescent [17], [29], [47] and adult [36], [54] swimmers when compared to SG. Many studies also found lower values in pelvic bones such as the femoral neck, the femur intertrocanter area or the hip per se in children [24]–[26], adolescent [17], [30], [33], [55], [56] and adult [36], [50], [57]–[61] swimmers when compared to gymnasts [17], [24], [25], [36], [55], [56], [60]–[62], track runners [30], [50], [55]–[57], volleyball [36], [50], [57], [58], soccer [33], [57]–[59] or basketball [50], [58], [59] players.

These lower BMD or BMC values in swimmers compared to SG were accompanied in a small, but still relevant number of studies [32], [46], [51], [54], [56], [61] with lower values in swimmers when compared to CG.

Of the 50 studies that measured BMD or BMC only Courteix et al. [24] showed higher values in swimmers than in SG (gymnasts) although only in the skull, while swimmers showed lower BMD values in most of the studied zones. However, head BMD and BMC were higher in swimmers than in gymnasts suggesting that in prepubertal children the increased BMD induced by impact training in the stressed sites of the body could be related to a decreased skull bone mass. This is the only study that showed differences of skull mass and future studies should be performed to confirm this data. Moreover, these differences among groups could be due to bias selection.

In addition to the cross-sectional studies that revealed lower BMD or BMC in swimmers than gymnasts [17], [23]–[26], [36], [44], [53], [60], [61], Taffe et al. [60] performed a longitudinal study also showing lower BMD in swimmers than in gymnasts, but more importantly, showing that during a 12-month follow-up, adult swimmers gained less lumbar spine and femoral neck BMD than their gymnast counterparts. This occurred despite lower initial BMD values in swimmers than gymnasts and was independent of reproductive hormone status. Similar results were found in other adult cohorts [50], [59], and more recently in adolescent swimmers compared to soccer players [15].

This is extremely important because sporting participation, specially during growth, seems to be effective in reducing the prevalence of osteoporosis-related fractures [63]. However, swimmers present similar or lower BMD values compared to CG and lower than their SG and therefore, at most may present osteoporosis values equal to the general population later in life reaching in the European Union an estimated 3.79 million osteoporotic fractures in the year 2000, with an associated estimated cost of 32 billion Euros [64].

To summarize, it seems that swimming does not produce enough power to stimulate bone growth above the regular pattern, with most studies showing similar BMD or BMC values to CG [16], [17], [23], [25], [26], [30], [31], [33]–[36], [40], [43], [44], [57]–[59], [61], [65]–[67]. A deleterious effect of swimming on bone mass has even been demonstrated in some studies, due to the elevated number of hours spent training in a hypogravity environment and therefore avoiding daily impacts [32], [46], [51], [54], [56], [61], [68]. It does also seem that swimming practiced in adulthood or elderly stages of life, may reduce the rate of normal bone mass loss accompanying age [42], [43], [45], although it is unclear if this is a real direct effect of swimming or it is due to a more active lifestyle. Quality assessment sugests a medium quality level of the literature on this subject. Longitudinal studies and well controlled case-control studies including past physical activity history are needed to elucidate the independent effect of swimming in bone development and evolution from childhood to late adulthood.

1.2 Bone geometry and structure.

Bone strength is determined by BMD and bone geometric properties [69]–[71]. However, despite the fact that the use of pQCT allows volumetric (vBMD) to be measured, distinguishes between different bone sections and their respective BMD, and is independent of physical size [72], to our knowledge only 4 studies have used this technique in swimmers [73]–[76] (Table 2). To ascertain whether the mechanical properties of bone in response to long-term physical exercise are related to geometric adaptation and not to vBMD in swimmers as they are in jumpers and tennis players [74], [77], [78] would seem to be an important area of study. In order to determine this, bone could be evaluated with pQCT or using DXA combined with other techniques such as magnetic resonance imaging (MRI) [31], [79] or hip structural analysis (HSA) [14]–[17], [67].

A basic means of studying bone structure is by performing a radiography as Xia Qu Ma [80] did in 1992. He performed a radiograph of the anterior side of both humeri and measured length, cortical thickness and diameters at the proximal, middle and distal thirds finding that swimmers had the lowest medial and lateral cortex at both humeri of all the sport groups compared, and that they presented similar values to the CG. Further studies also measured swimmers humeri [73], [74], tibia [74]–[76] and femur [79] with different devices revealing similar [16], [76], [79] or higher cortical cross-sectional areas (CSA) and bone strength indexes [14], [73], [75] in swimmers than in CG. The 3 studies using pQCT that found an improved structure in swimmers than CG were all performed in the same sample age groups, young adults (20 years old). When HSA was used to evaluate bone structure, adolescent [16], [17] or adult [67] (18–35 years old) swimmers presented values similar to CG. These different results could be due to different age samples or to differences between imaging techniques. However, when compared with other sports, swimmers had lower cortical thickness [17], [74], [75], cortical CSA [17], [31], [75], [76] and lower strength indexes [31], [67], [75], [76]. Results were similar independently of the technique used; pQCT [73]–[76], MRI [31] or HSA [14], [17], [67]. The lower cortical mass described in some studies allows a larger medullary cavity CSA in swimmers, resulting in lower trabecular vBMD as found by Nikander et al. [74] and Duncan et al. [31], and a bone with its mass distributed relatively distally from the centroid.

Similar results were found when bone geometry was assesed by Liang et al. [81], using a totally different technique: the mechanical response tissue analyzer (MRTA), invented for the National Aeronautics and Space Administration (NASA) to evaluate bone strength in astronauts after space flight, using the response of a long bone to a low-frequency vibratory stimulus. Results showed that although synchronized swimmers exhibited lower BMD (calculated with a single photon densitometer) wrist values than GC and SG, the bone bending stiffness in the tibia and ulna were greater in synchronized swimmers than in the CG, reinforcing the fact that structure has a great influence on bone strength independently of BMD.

Ferry et al. [15] compared bone geometry during an 8-month period in swimmers and soccer players finding that swimmers did not increase sub-periosteal width, while soccer players did, as well as bone strength indexes such as the cross-sectional moment of inertia and section modulus Z-scores. Buckling ratio (BR) which is an index of bone instability (BR =  maximum distance from the center of mass to the medial or lateral surface divided by the cortical thickness), was the only value improved in swimmers and maintained in soccer players. BR still remained higher in swimmers after the longitudinal period suggesting that swimmers had weaker bone.

In summary, although swimmers may show weaker bone when compared to other weight-bearing sports as a result of smaller cross-sectional areas, the characteristics of the sport may adapt the bone to a higher trabecular CSA [31] and similar [16], [17], [67], [76], [79] or higher bone strength indexes [73]–[75], [81] when compared to a sedentary group, making it more resistant to bending and torsion than a sedentary bone, even though it may have lower BMD. However, studies including both methods of assessment, DXA with HSA and pQCT or MRI, and different sample age groups should be performed in order to test this hypothesis.

1.3 Stiffness, speed of sound and broadband ultrasound attenuation.

The use of quantitative ultrasound (QUS) is an option that avoids results biased by differences in body size among subjects. The values of SOS, BUA, and the Stiffness index (SI) derived from both of the former, provided by the QUS, are related to bone density and structure [82] but not to cortical thickness [83]. SI is the default parameter used by the manufacturer for demographic comparison of patient data because SOS and BUA are given as absolute values with no normative values for them. Thus T-scores for these two parameters cannot be calculated [84], [85].

Taaffe et al. [41] compared swimmers with jumpers and a CG observing no differences among the 3 groups in BUA. Similar values between swimmers and other SG were found in only one other study [86]. However, a lack of differences between swimmers and CG was also found in two further studies [35], [87]. In contrast, Falk et al. [86], [88], [89] observed higher tibial SOS values in both male and female swimmers than in the CG in their 3 studies; as others [90], [91] did. Despite the above findings, some of these studies included a SG and with further studies showed that either children [89], adults [90], [91] or elderly [35] swimmers presented lower values than their high impact comparison groups. Showing in 2 of the previous studies lower SI values in swimmers than SG [35], [90].

To our knowledge only one longitudinal 12 month follow-up study using QUS has been performed in swimmers [49]. The study sample was composed of former postmenopausal swimmers that were not swimming when the study took place. Swimmers showed a lower decrease in SI values than CG, although they performed more current physical activity (other than swimming). Therefore it is not possible to confirm that the lower SI decrease was due to swimming (in earlier stages of life) or was a consequence of current physical activity.

We can briefly conclude that swimmers have higher QUS values than CG although these values seem to be higher in high impact sports than in swimmers. It is worth noting that the higher QUS values in swimmers when compared to CG were generally present in the lower limbs. However, the major forces applied in some swimming styles such as crawl or backstroke are applied with the upper limbs. This would suggest that the higher QUS values presented by the swimmers but not by the synchronized swimmers [87], might be due to the push that swimmers perform against the wall, although future studies should be performed comparing swimmers with similar training and history habits who train in a 25 and 50 meter pool to compare their lower limb QUS values.

1.4 Bone markers.

Bone is a metabolically active tissue that is constantly changing, with BMD being the result of bone formation and resorption which are closely linked in time and space within the bone multicellular unit [92]. To assess these changes we can measure bone turnover markers which are usually able to provide an early indication of an effect on bone and can be quite sensitive, their main limitations being poor specificity of response and lack of validated connection with the functional outcome [93].

Regarding bone formation markers, several studies showed higher bone-specific alkaline phosphatase (B-ALP) values in swimmers than in CG [47] and SG [40], [47], [94]. However, Morgan et al. [59] found lower values of B-ALP in swimmers than in high impact sports. These diverse results may be due to the fact that the sample ages of the studies were adolescents [47], 20 year old adults [59], [94] and 30 year old adults [40]; all stages of life with different bone dynamics. Another bone formation marker used is osteocalcin (OC), that was lower [57], similar [16], [17], [28], [59], [87] or higher [40] in swimmers than in CG and lower [57] or similar [17], [40], [59] when compared to SG. These differences between studies are a clear example of the difficulties inherent in the use of biochemical markers owing to the wide range of interactions possible depending on age, gender, nutritional status, season and time following intense training [95], [96]. Added to these difficulties and as described by PASSCLAIM [93], the OC molecule exhibits considerable immunological heterogeneity. This, combined with the fact that no internationally recognized assay standard exists, make OC measurements difficult to interpret meaningfully. This may be one of the explanations of the heterogeneity of this bone marker in swimmers.

The use of two bone markers to evaluate bone formation [17], [40], [59] permits differing results to be compared and is therefore a recommended methodology to follow for future studies.

Regarding bone resorption, Cross-linked N-telopeptides of type I collagen (NTx) was the most used biomarker [57], [59], [87], showing no differences between swimmers, CG and other SG. Other resorption markers used were pyridinoline and deoxypiridinoline both lower in male swimmers than judo athletes [94] and higher than CG [47]. Type I collagen C-telopeptide (CTx), another resorption marker that has been recommended by The International Osteoporosis Foundation and the International Federation of Clinical Chemistry [1] was similar [16], [17] or higher in swimmers when compared to CG [40]. This latter point, added to the previously named higher B-ALP values, suggests a higher bone turnover in swimmers than in controls which reflects an intense remodelling process without producing any differences in BMD. Perhaps the use of other techniques to assess bone quality or structure such as pQCT might have shown differences among bones and therefore explained the differences revealed among bone markers.

The Osteoprotegerin(OPG)/Rank-ligand (RANKL) system which is known to have considerable influence on bone formation and degradation, has only been evaluated in one study [17], that showed lower levels of RANKL in swimmers than in gymnasts and controls. Further studies evaluating OPG and RANKL in swimmers are needed in order to confirm these results and elucidate the effect that swimming might have on this system.

Summarizing all the previous studies that include swimmers and bone metabolic markers we can conclude that swimming seems to involve a high bone turnover [40], [47], [94] although in most cases this is not translated into higher BMD. Out of the 14 studies included in this review including bone metabolic markers, only one [87] was performed with a technique other than DXA; QUS, which showed no differences between QUS parameters or bone metabolic markers between swimmers and CG, suggesting that bone metabolic markers might have a high relation to bone structure. Moreover, Maïmoun et al. have performed HSA in two recent studies [16], [17] finding no differences in bone markers nor bone structure between swimmers and CG, thus reinforcing the previous hypothesis. However, further studies including bone metabolic markers and techniques which allow the evaluation of bone structure are needed in order to confirm this hypothesis.

2.1 Hormonal profile.

It is well known that hormone concentrations change during growth and vary during or after the practice of exercise[97], [98]. In addition, some parameters of the hormonal profile affect bone metabolism; for example estrogens have been demonstrated to stimulate the proliferation of osteoblasts [99] and insulin-like growth factor-I (IGF-I) has been shown to activate bone turnover [100]. It is therefore important to describe whether hormone concentrations were similar in swimmers and CG or SG.

Lima et al. [47] first found lower testosterone (TT) values in swimmers than in high impact athletes. However, when compared to other non-impact athletes such as cyclists, swimmers showed higher values of TT. Comparison with CG exhibit inconclusive results showing lower [27], and similar [16], [39] values in swimmers. These differences could be due to different training loads as previously suggested by others [101], or due to the different age-range in the samples of the studies [27], [39].

Estradiol levels were also studied, with no differences found between young swimmers, CG and other sports [39]. A similar case is the Luteinizing hormone that was also measured in three studies and also revealed no differences in values between swimmers, CG and other sports in adolescents [16], [47] and adults [39].

Jürimae et al. focused on the influence of Ghrelin (GR) on BMD, showing that GR appeared to be an important hormonal predictor for BMD in swimmers [27]. However, in a further study they showed that GR was not related to measured BMD values in swimmers, suggesting that GR concentration did not have a direct influence on bone mineralization in female swimmers [102].

Jürimae et al. [27], [102], also studied leptin which is directly related to fat mass as well as to BMD in children and adolescents [103], [104] finding similar values in swimmers, GC and sport reference values. Further studies [56] also found similar values in swimmers suggesting that leptin concentration in swimmers is not related to bone mineral parameters.

IGF-I which stimulates endochondral bone formation and activates bone turnover [100] was lower in swimmers than in CG [87] and lower increases during pubertal development were found in a further longitudinal study [27]. However in the latest published study regarding IGF-I, swimmers and CG presented similar values [16]

We can therefore conclude that the values of the majority of the hormones studied were similar in swimmers and CG. However, the concentrations of some of these hormones may vary according to training loads, making the measuring period critical regarding hormone concentrations and associated effects.

2.2 Calcium intake.

Most of the studies that registered calcium intake showed no differences among groups. Eigth studies showed significantly higher calcium intake values in swimmers than in CG [14], [15], [32], [34], [35], [40], [49], [105]. Calcium is closely related with bone mass and its intake through diet is crucial. According to the recommended dietary allowance guidelines [106] the studies performed on children [23], [25], [66], [89], adolescents [14], [15], [31], [33], [34], [47], [86], [88], young adults [86], [88], [90] and older adults [37], [44], [49] indicate that swimmers did not reach the recommendations established by these guidelines for calcium intake. However, some of these studies showed higher BMD than reference values independently of calcium intake.

Research has demonstrated that calcium supplementation increases bone mass significantly during childhood and adolescence [107], [108]. A bone-exercise-nutrition interaction exists as shown in previous studies [109]. It would therefore be interesting to perform specific studies evaluating the effect of calcium intake in swimmers in order to ascertain the possible interactions between physical activity, calcium and bone. Other nutritional aspects like magnesium, phosphorus or vitamin D that have not been registered and also affect bone may also be important regarding results. We would therefore suggest that future studies evaluating bone take these variables into account.

2.3 Gender.

Gender differences in peak bone mass acquisition are well documented in humans [110], it is therefore important to describe whether swimming might affect these differences.

It seems that results in studies with either male or female participants are similar. Some studies included both genders showing no differences between groups [32], [94]. However, Orwoll et al. [43], who first included both male and female swimmers, and a further study [28] observed higher BMD values in male swimmers than in CG, whereas female swimmers showed similar bone values to the CG. As previously commented, these exclusive differences for the male group were thought to be due to the greater forces generated by males resulting in a greater effect on bone remodelling. On the contrary, Grimston et al. [55] found lower bone mass values only in male swimmers when compared to CG. Different values between genders were also found by Liu et al [75] that showed higher periosteal area, polar moment of inertia and strain strength index only in female swimmers when compared to CG. These higher values only in females could be due to delayed puberty that results in larger periosteal and endocortical area in girls but not in boys [75], female swimmers having the latest menarche in the study and therefore a later puberty. Similar results were found in another study [46] where female swimmers showed higher arm BMD; however, these differences were inexistent between males.

2.4 Age.

We have observed that swimming does not seem to negatively affect bone mass and might provide a stronger bone structure than that of the CG, but weaker than that of other impact sports independently of the age group studied: children, adolescents or adults. However, as previously stated, it seems that swimming might be more beneficial later in life. Of the 5 studies performed with DXA that found benefits of swimming compared to CG, 4 were performed in adults over 40 years old, suggesting that swimming may be a beneficial activity to practice in later adulthood in order to maintain bone. It is possible that these higher bone mass values in this older population were due to a more active life than their CG, as found in some studies [49]. Therefore, it would be interesting to take into account in future studies other sports practiced, in addition to swimming, to explain whether the differences within groups might be due to other activities or whether they are exclusive to swimming.

2.5 Training influence.

The number of years practiced and weekly hours trained could influence bone in different ways. Unfortunately, this information is not available in all the articles included in this review. Weekly training hours were therefore evaluated when specified by the authors and no evidence of its influence on bone mass was found, in fact, the data presented in each article regarding weekly training hours and years of practice were not clearly specified and therefore cannot be taken into consideration. Training intensity may also differ between groups that perform the same sport modality and train for a similar number of hours. Different intensities can result in different sport performances and also in different body composition and bone adaptations. Magkos et al. [51] were the only researchers that divided swimmers into sprint and endurance swimmers, finding lower BMD values in the latter. The higher BMD values of the sprinters could be due to the higher muscle force and as a result bone stimulation that sprinters perform during training and competition. However, future studies comparing groups that have different training routines should be perfomed to throw more light on this question.

In addition to swimming hours, many swimmers also perform strength training that could also affect bone. This type of training was generally not reported in most of the studies included in this review. However, due to the osteogenic effect that this type of training has, it is important that future studies evaluating bone in swimmers take this parameter into account