Exercise Effects on Bone Mineral Density in Men

In contrast to postmenopausal women, evidence for a favorable effect of exercise on Bone Mineral Density (BMD) is still limited for men. This might be due to the paucity of studies, but also to the great variety of participants and study characteristics that may dilute study results. The aim of the present systematic review and meta-analysis was to evaluate the effect of exercise on BMD changes with rational eligibility criteria. A comprehensive search of six electronic databases up to 15 March 2021 was conducted. Briefly, controlled trials ≥6 months that determined changes in areal BMD in men >18 years old, with no apparent diseases or pharmacological therapy that relevantly affect bone metabolism, were included. BMD changes (standardized mean differences: SMD) of the lumbar spine (LS) and femoral neck (FN) were considered as outcomes. Twelve studies with 16 exercise and 12 control groups were identified. The pooled estimate of random-effect analysis was SMD = 0.38, 95%-CI: 0.14–0.61 and SMD = 0.25, 95%-CI: 0.00–0.49, for LS and FN, respectively. Heterogeneity between the trials was low–moderate. Funnel plots and rank and regression correlation tests indicate evidence for small study publication bias for LS but not FN-BMD. Subgroup analyses that focus on study length, type of exercise and methodologic quality revealed no significant difference between each of the three categories. In summary, we provided further evidence for a low but significant effect of exercise on BMD in men. However, we are currently unable to give even rough exercise recommendations for male cohorts.


Introduction
Many guidelines on osteoporosis and fracture prevention consider physical exercise as the most effective non-pharmacologic agent for increasing bone strength and reducing falls (e.g., [1][2][3]). However, in contrast to female cohorts [4,5], evidence for a favorable effect of exercise on Bone Mineral Density (BMD) in men is still limited (review in [6][7][8][9]). Recently, two systematic reviews and meta-analyses of randomized controlled trials reported data on physical activity/exercise effects on BMD in men 18 years and older [6,8]. Ashe et al. [6] reported significant exercise effects on BMD at the total hip but little or no effect of exercise on the adjacent femoral neck (FN)-BMD. In contrast, the meta-analysis of Hamilton et al. [8] indicates significant evidence for an exercise effect on FN-BMD. Of importance, both studies do not observe any relevant beneficial effect of exercise on BMD at the lumbar spine (LS). Considering this brand-new meta-analytic data, one justifiably wonders about the rationale for another meta-analysis in the area of exercise, BMD and men. However, trivially, the results of meta-analyses predominately depend on the studies included. Apart from new study results, our eligibility criteria substantially differ from both studies. This particularly relates to pre-study exercise status, prevalent diseases/conditions and pharmacologic therapy with potential impact on exercise effects (i.e., differences in BMDchanges between exercise and control) on BMD. However, applying too critical a set of criteria aggravates the paucity of exercise studies with men performed in the gynocentric field of osteoporosis. Thus, we aimed to apply a balanced compromise between eligibility and number of studies included. This embraces the approach to exclude studies that focus on cohorts with diseases (e.g., inflammatory diseases) and pharmaceutic therapy (e.g., androgen deprivation therapy) with proven negative impact on exercise effects on bone. In summary then, the aim of the present systematic review and meta-analysis was to evaluate the effect of exercise on BMD at the lumbar spine and proximal femur ROIs in men applying reasonable eligibility criteria. Additionally, we intended to identify study and exercise characteristics with impact on BMD.

Literature Search
The current systematic review and meta-analysis follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [10] and was registered in the international prospective register of systematic reviews. (PROSPERO; ID: CRD42021233194).
Although our 2018 systematic review on exercise effects in men [9] identified eligible articles before 2018, an new extensive search of electronic databases was performed through PubMed, Science direct, Scopus, Web of Science, Cochrane and ERIC for all articles published in German or English language up to 15 March 2021. A standard search protocol was developed and controlled vocabulary (Mesh term for MEDLINE) was applied. In order to include all the relevant studies, the following key words and their synonyms were used: "Exercise" or "Physical activity" or "Exercise training" or "Resistance training" or "Training") AND ("men" or "male") AND ("Bone" or "Bone mass" or "Bone density" or "Bone mineral content" or "Bone mineral density" or "BMD"). Additionally, the reference lists of the identified studies were reviewed to identify further relevant articles. Duplicate publications were identified by comparing author names, type of intervention, intervention duration and date of publication. If further information was needed, the authors were contacted via e-mail. In summary, four authors were contacted, three of them finally responded. One author submitted data for their male cohort [11], two authors were unable to provide further study information.

Inclusion and Exclusion Criteria
We included studies/study arms with (1) randomized and non-randomized controlled trials with at least one exercise group as an intervention versus one control group with sedentary/habitual active lifestyle without exercise or with sham exercise. (2) ≥6 months intervention duration, (3) Areal BMD of the LS, femoral neck (FN) and/or total hip (tH) region at baseline and study end as determined by (4) dual-energy X-ray absorptiometry (DXA) or dual-photon absorptiometry (DPA). We excluded studies with (1) diseases/conditions with relevant impact on bone metabolism (e.g., inflammatory diseases), (2) pharmaceutic therapy that relevantly affects bone metabolism (e.g., androgen deprivation therapy), (3) mixed-gender cohorts without separate BMD analysis for men, (4) double/multiple publications from one study and preliminary data from subsequently published trials; (5) Review articles, case reports, editorials, conference abstracts, and letters were not considered. (6) We also excluded studies with participants with pre-training exercise habits close to the volume and intensity applied by the intervention protocol.

Data Extraction
Titles and abstracts were screened by an independent reviewer (MM) carefully supervised by a second rater (WK). Full-text articles of relevant studies were evaluated by two reviewers (MM and WK) independently and they extracted data from included studies. In the case of disagreement, the third reviewer was consulted until a consensus was reached. An extraction form was utilized to consider the relevant data including publication information (i.e., name of the first author and year of publication), study details (i.e., study duration, initial sample size of the participants, dropout rate), participants' characteristics (i.e., age, BMI, health-, BMD-and exercise status, medication with impact on bone, nutrition, dietary supplementation) (Table 1), methodologic quality aspects (Table 2), exercise characteristics (i.e., type of exercise, progression of intensity, frequency, setting and supervision, duration, sets and repetition, site specificity), and adherence to exercise (including number of withdrawals) ( Table 3).

Outcome Measures
(Areal) BMD of the LS and/or the proximal femur regions "total hip" and/or the FN, as determined by dual-energy X-ray absorptiometry (DXA), are used as outcome measures. BMD assessment must be reported at least for one of the regions listed above at baseline and follow-up assessment at the end of intervention period.

Quality Assessment
Two reviewers (MM, WK) assessed the included articles independently for risk of bias using the Physiotherapy Evidence Database (PEDro) scale risk of bias tool [23]. Additionally, both reviewers used the "Tool for the Assessment of Study Quality and reporting in Exercise" (TESTEX) [24] to determine study quality and reporting. Disagreements between the two reviewers were resolved by consensus in consultation with a third independent reviewer (SvS).
Studies were screened for potential selection bias, performance bias, detection bias, attrition bias and reporting bias using the 11 criteria of the PEDro-scale ( Table 2). In total, the scale scores 10 items. Using the TESTEX criteria [24] as well allowed us to consider the following additional five exercise-relevant aspects: adverse effects, attendance report, activity monitoring in control groups, progression of relative exercise intensity and exercise volume ( Table 2).

Data Synthesis
For sub-analyses, the intervention duration was stratified as 6-9 months, 10-16 months and >16 months. We also categorized the included studies by their type of exercise into three sub-groups: (a) resistance training (RT), (b) weight bearing exercises (WB), (c) RT + WB. Finally, we classified the studies according to the PEDro-Score (low <5 vs. moderate 5-6 vs. high ≥7 score points) [25]. If the studies presented a confidence interval (CI) or standard errors (SE), these were converted to standard deviation (SD) with standardized formulas [26].

Statistical Analysis
Briefly, a random-effects meta-analysis was conducted by using the metafor package [27] included in the statistical software R [28]. Effect size (ES) values were presented as standardized mean differences (SMDs) along with the 95% confidence interval (95%-CI). In addition to the traditional random-effects model, we applied the more robust inverse heterogeneity (IVhet) model proposed by Doi et al. [29]. A priori sensitivity analysis was applied to determine whether the overall result of the analysis is robust to the use of the imputed correlation coefficient (minimum, mean or maximum). Heterogeneity between the studies was checked using the Cochran Q test and I 2 statistics (0-40%: low, 30-60%: moderate, 50-90%: substantial heterogeneity [26]). For those studies with two exercise groups (i.e., [15,16,20,22]), the control group was split into two smaller groups for comparison against each intervention group [26]. Funnel plots with regression test and the rank correlation between effect estimates and their standard errors using the t-test and Kendall's τ statistic, respectively, were applied to explore potential small study/publication bias. To adjust the results for possible publication bias, we also conducted a trim and fill analysis using the L0 estimator proposed by Duval et al. [30]. In parallel, we used Doi plots and the Luis Furuya-Kanamori index (LFK index) [31] to check asymmetry. Finally, we applied influence analyses excluding two "critical" studies [11,14] from the analysis for BMD-LS and FN. A p-value < 0.05 was considered as the significance level for all tests. Subgroup analyses were applied for intervention length and type of exercise.

Study Characteristics and Quality Assessment
In total, our search identified twelve studies [11][12][13][14][15][16][17][18][19][20][21][22] (Figure 1) with 16 exercise and 12 control groups. The pooled number of participants was 823 (intervention groups: 461, control groups: 362). Table 1 presents the study and participant characteristics of the eligible studies that determined the effect of exercise on bone mineral density among men. In addition to the information given in Table 1, all but one study [22] focused on Caucasian men 36 ± 7 [14] to 79 ± 5 years [18] old. Although not always consistently stated, BMD was addressed as the primary outcome in 11 of 12 studies [11][12][13][14][15][16][17][18][19]21,22]. The methodological quality of the studies was rated using the PEDro and the TESTEX scale, both particularly dedicated to exercise studies. Score points vary between three and eight from a maximum of 10 points for PEDro and seven-fourteen from a maximum of 15 for the TESTEX score (Table 3). However, successful blinding of trainers and/or participants is almost impossible in conventional exercise studies; thus, maximum score-points were rather eight for PEDro and 14 for TESTEX.

Protein, Vitamin-D and Calcium Supplementation; Nutrition
Only one study [18] provided all participants with a maximum of 1000 mg/d Ca and 1400 IU/d Vit-D according to Ca-questionnaires and serum 25 OHD-levels. As per recent recommendations [32], total protein intake in this study based on dietary protocols averaged 1.5 g/kg/d in the EG and 1.2 g/kg/d in the CG. In a study arm not included in the present analysis, Kukuljan et al. [19] implemented a combined exercise and fortified milk (1000 mg Ca and 800 IU Vit-D/400 mL) that did not relevantly affect the exercise/bone The methodological quality of the studies was rated using the PEDro and the TESTEX scale, both particularly dedicated to exercise studies. Score points vary between three and eight from a maximum of 10 points for PEDro and seven-fourteen from a maximum of 15 for the TESTEX score (Table 3). However, successful blinding of trainers and/or participants is almost impossible in conventional exercise studies; thus, maximum score-points were rather eight for PEDro and 14 for TESTEX.

Protein, Vitamin-D and Calcium Supplementation; Nutrition
Only one study [18] provided all participants with a maximum of 1000 mg/d Ca and 1400 IU/d Vit-D according to Ca-questionnaires and serum 25 OHD-levels. As per recent recommendations [32], total protein intake in this study based on dietary protocols averaged 1.5 g/kg/d in the EG and 1.2 g/kg/d in the CG. In a study arm not included in the present analysis, Kukuljan et al. [19] implemented a combined exercise and fortified milk (1000 mg Ca and 800 IU Vit-D/400 mL) that did not relevantly affect the exercise/bone interaction in his cohort with high baseline Ca-(911 mg/d) but low (50 IU/d) Vitamin D-intake, however.
Unfortunately, only few studies [17][18][19]21] reported baseline dietary intake of their cohorts. Due to this limited statistical power, we do not address the effect of baseline dietary intake (e.g., Vit-D, Ca, Protein, energy intake) on the exercise bone/interaction by sub-analyses. Table 3 specifies the exercise intervention of the included studies. Most of the RCTs compared a single exercise group with a single inactive control group. The study of Allison et al. [12] applied unilateral loading ("hopping") and used the unloaded leg as control. Four other exercise trials implemented two exercise arms with different types of exercise interventions [15,16,22]. Another trial incorporated study arms with combined exercise and/or fortified milk supplementation [19] but was not included in the present analysis. Only one study established an active control group. This group was advised to undertake three × 30 min of unsupervised brisk walking/week [21]; however, adherence was not monitored in this CG. Although not always specified, pre-study exercise status or corresponding inclusion criteria were reported by most studies [11][12][13][14][15][17][18][19]21,22]. However, there is some evidence that men with exercise habits potentially relevant for subsequent intervention effects on BMD outcomes were included in some studies (e.g., [11,21]).

Exercise Characteristics
Net exercise frequency (reported exercise frequency adjusted for attendance) varied from ≈1.5 session [15,16] to ≥6 sessions/week [12]; however, most trials (Table 3) averaged between 2 and 2.5 sessions/week at least when considering attendance rate.

Results on BMD at the Lumbar Spine (LS)-ROI
Nine studies with 11 exercise groups evaluated the effect of exercise versus control on LS-BMD ( Figure 2). In summary, we observed a small-moderate exercise effect on LS in men (SMD: 0.38, 95% CI: 0.14-0.61) ( . . . translated into changes in LS-BMD: 0.012 ± 0.011 g/cm 2 or 1.2 ± 1.1%). Using the more "robust" IVhet model, the effect decreased but remained significant (SMD: 0.29, 95% CI: 0.04-0.53). There was a moderate level of heterogeneity in estimates of the exercise effect (I 2 = 46%, Q = 18). Sensitivity analysis of imputation determined that even in the worst case (i.e., imputing with minimum correlation i.e., maximum SD) there is a significant effect (SMD: 0.35, 95%-CI: 0.11-0.59, p = 0.004). However, including only studies with complete data increased SMD considerably (0.68; 0.40-0.95). Results listed in Figure 2 based on imputation with mean correlation.
interaction in his cohort with high baseline Ca-(911 mg/d) but low (50 IU/d) Vitamin Dintake, however.
Unfortunately, only few studies [17][18][19]21] reported baseline dietary intake of their cohorts. Due to this limited statistical power, we do not address the effect of baseline dietary intake (e.g., Vit-D, Ca, Protein, energy intake) on the exercise bone/interaction by subanalyses. Table 3 specifies the exercise intervention of the included studies. Most of the RCTs compared a single exercise group with a single inactive control group. The study of Allison et al. [12] applied unilateral loading ("hopping") and used the unloaded leg as control. Four other exercise trials implemented two exercise arms with different types of exercise interventions [15,16,22]. Another trial incorporated study arms with combined exercise and/or fortified milk supplementation [19] but was not included in the present analysis. Only one study established an active control group. This group was advised to undertake three × 30 min of unsupervised brisk walking/week [21]; however, adherence was not monitored in this CG. Although not always specified, pre-study exercise status or corresponding inclusion criteria were reported by most studies [11][12][13][14][15][17][18][19]21,22]. However, there is some evidence that men with exercise habits potentially relevant for subsequent intervention effects on BMD outcomes were included in some studies (e.g., [11,21]).

Exercise Characteristics
Net exercise frequency (reported exercise frequency adjusted for attendance) varied from ≈1.5 session [15,16] to ≥6 sessions/week [12]; however, most trials (Table 3) averaged between 2 and 2.5 sessions/week at least when considering attendance rate.

Results on BMD at the Lumbar Spine (LS)-ROI
Nine studies with 11 exercise groups evaluated the effect of exercise versus control on LS-BMD ( Figure 2). In summary, we observed a small-moderate exercise effect on LS in men (SMD: 0.38, 95% CI: 0.14-0.61) (…translated into changes in LS-BMD: 0.012 ± 0.011 g/cm 2 or 1.2 ± 1.1%). Using the more "robust" IVhet model, the effect decreased but remained significant (SMD: 0.29, 95% CI: 0.04-0.53). There was a moderate level of heterogeneity in estimates of the exercise effect (I 2 = 46%, Q = 18). Sensitivity analysis of imputation determined that even in the worst case (i.e., imputing with minimum correlation i.e., maximum SD) there is a significant effect (SMD: 0.35, 95%-CI: 0.11-0.59, p = 0.004). However, including only studies with complete data increased SMD considerably (0.68; 0.40-0.95). Results listed in Figure 2 based on imputation with mean correlation. Inspecting the funnel plot suggests evidence for a publication bias (Figure 3). The analysis indicates missing studies on the lower right-hand side (i.e., small studies with negative outcomes). A trim and fill analysis resulted in non-significant effects sizes (SMD: 0.12, 95%-CI: −0.16 to 0.40) after adjusting for missing studies. Further, both tests for funnel plot asymmetry, the regression (p = 0.02) and rank correlation test (p = 0.09) indicate relevant asymmetry. Applying the LFK Index (2.35), the result of significant asymmetry was confirmed. mean difference (SMD) with 95% CI for changes in exercise and control groups. CG: control group, EG: exercise group Inspecting the funnel plot suggests evidence for a publication bias (Figure 3). The analysis indicates missing studies on the lower right-hand side (i.e., small studies with negative outcomes). A trim and fill analysis resulted in non-significant effects sizes (SMD: 0.12, 95%-CI: −0.16 to 0.40) after adjusting for missing studies. Further, both tests for funnel plot asymmetry, the regression (p = 0.02) and rank correlation test (p = 0.09) indicate relevant asymmetry. Applying the LFK Index (2.35), the result of significant asymmetry was confirmed.

Results on BMD at the Proximal Femur
Ten studies with 13 exercise groups determined the effect of exercise on femoral neck ( Figure 4) and eight studies [11][12][13]15,18,19,21,22] with 10 exercise arms addressed the total hip-BMD. Due to the higher statistical power, we decided to focus our analysis on the femoral neck ROI. In summary, we observed a small effect size for FN-BMD (SMD: 0.25, 95% CI: 0.00-0.49, p = 0.048) ( . . . or 0.009 ± 0.016 g/cm 2 (1.0 ± 1.9%)). Using the more "robust" IVhet model, the effect was slightly lower (SMD: 0.21, 95% CI: −0.04-0.47). In parallel to LS-BMD, we observed a moderate level of heterogeneity between the trials (I 2 = 46%, Q= 21). Sensitivity analysis of imputation shows that when imputing with minimum correlation (i.e., maximum SD) differences between exercise and controls were non-significant (SMD: 0.16; 95% CI: −0.01 to 0.34, p =0.06). The same was true when including only studies with complete data (0.49; −0.53 to 1.50). Results listed in Figure 4 based on imputation with mean correlation.  In contrast to LS, we observed no evidence for a small study bias for BMD at the FN ( Figure 5). Trim and fill analysis did not impute missing studies on the left side; neither regression (p = 0.061) nor rank correlation test indicate evidence for funnel plot asymmetry. Applying the LFK index, the result (0.78) also indicates no relevant asymmetry. In contrast to LS, we observed no evidence for a small study bias for BMD at the FN ( Figure 5). Trim and fill analysis did not impute missing studies on the left side; neither regression (p = 0.061) nor rank correlation test indicate evidence for funnel plot asymmetry. Applying the LFK index, the result (0.78) also indicates no relevant asymmetry. In contrast to LS, we observed no evidence for a small study bias for BMD at the FN ( Figure 5). Trim and fill analysis did not impute missing studies on the left side; neither regression (p = 0.061) nor rank correlation test indicate evidence for funnel plot asymmetry. Applying the LFK index, the result (0.78) also indicates no relevant asymmetry.

Subgroup Analysis
Our subgroup analyses focused on study length (classified in 6-< 10 vs. 10-16 vs. >16 months; Table 1) and type of exercise (RT vs. WB vs. RT-WB, Table 3). In summary, no significant differences for exercise effects on LS or FN for study length (

Subgroup Analysis
Our subgroup analyses focused on study length (classified in 6-<10 vs. 10-16 vs. >16 months; Table 1) and type of exercise (RT vs. WB vs. RT-WB, Table 3). In summary, no significant differences for exercise effects on LS or FN for study length (

Discussion
In the present systematic review and meta-analysis, we provided further evidence for a favorable effect of exercise on BMD in predominately middle-aged to older men. In contrast to two recent studies [6,8] that reported positive effects for the proximal femur ROI only (Hamilton et al. [8]: SMD: 0.21, 95% CI: 0.03-0.040 for femoral neck BMD; Ashe et al. [6]: MD: 0.03 g/cm 2 , 95%-CI 0.01-0.05 g/cm 2 for the total hip ROI.), we determined a low but significant positive effect of exercise on BMD at the FN and LS-ROI. Despite statistical differences between the studies, the main reason for the diverging results was the varying eligibility criteria. This concerns in particular pre-study health and exercise status and also pharmaceutic therapy with impact on bone metabolism. Hamilton et al. [8] rigorously excluded men with every type and volume of regular exercise prior to the intervention and, in parallel to Ashe et al. [6], did not focus on pharmaceutic therapy. In the present study, on the other hand, two reviewers carefully checked conditions (i.e., health and pre-study exercise status) and diseases/pharmaceutic therapy with relevant impact on BMD and, even more important on their potential influence on exercise effects on BMD. Although debatable, the latter resulted in the inclusion of HIV-infected men with stable antiretroviral therapy [14] or quiescent/mildly-active Crohn's disease [11] ( Table 2). However, our results indicate (Figures 2 and 4) that, in diametral contrast to our expectation, positive effect on exercise induced-BMD-changes were reported for both cohorts [11,14] which relate in particular to the HIV study of Ghayomzadeh et al. [14]. In this context, we conducted two separate a priori (sensitivity) analyses (not given in the result section) without the studies of Jones et al. [11] or Ghayomzadeh et al. [14]. While the exclusion of the study of Jones et al. did not relevantly affect our results, the exclusion of the HIV-infected men with stable antiretroviral therapy [14] led to a pronounced decrease in the effect size for BMD-LS and a shift to a non-significant effect for BMD-FN.
We also included studies with men who reported pre-study exercise habits that should not or not relevantly impact the subsequent intervention effect on BMD (Table 3) [11,15,18]. Nevertheless, some studies do not report pre-study exercise status (Table 3) or pharmacologic therapy (Table 2). Consequently, there is some evidence that failure of an intervention effect on BMD might be related to (a) diseases/pharmaceutic therapy [33][34][35]), (e.g., Androgen Deprivation Therapy that induces adverse metabolic effects, including reduced muscle mass, increased fat mass and loss of bone mineral density (BMD) [36]) with striking impact on bone metabolism or/and (b) low difference between pre-intervention and intervention exercise with a corresponding lack of effective stimuli for bone. This might also relate to studies with an active control group (e.g., [21,37]) and potentially effective exercise characteristics that dilute differences between EG and CG.
Another aspect that decreases the effect sizes of exercise on any given outcome, but in particular on BMD with its complex mechanisms of action, is the "try and error approach" in phase III exercise studies [38]-an incompatible procedure in pharmacological research. Indeed, the approach of testing the effect of an intervention without (a) properly respecting basic principles of exercise application on bone ( [39], e.g., [40,41]) or/and (b) determining the isolated effect of selected exercise characteristics (e.g., type of exercise, [4,42,43]) will provide at least suboptimal study results. An example for the latter aspect might be the included FrOST study [18] that focuses solely on the isolated effect of machine-based DRT on BMD in older men, whilst fully aware that the effect of a combined high impact/DRT would be potentially superior.
In addition to the traditional random-effects model approach, we applied the more robust inverse heterogeneity (IVhet) model proposed by Doi et al. [29]. The IVhet model might be the better choice to consider heterogeneity that is frequently prevalent in metaanalyses on exercise [44]. As expected, the effect sizes for LS-and FN-BMD does decrease slightly, but the result still remains significant for LS-BMD. Thus, differences in the results between the present study and the study of Hamilton et al. [8] cannot be attributed to differences in statistical procedures. We think that diverging eligibility criteria and thus the different studies included predominately account for differences in results. This may be confirmed when excluding the HIV study of Ghayomzadeh et al. [14] (see above).
Another potential limitation, but one which applies to most meta-analyses in the area of exercise and BMD, is the limited number of studies that reported (baseline) dietary parameters and/or corresponding changes of dietary habits with impact on the exercise bone interaction (e.g., Ca, Vitamin D, Protein, energy intake). This at least aggravates the decision of the author of systematic reviews and meta-analysis on excluding studies with relevant dietary effects on bone or to run sub-analyses that addresses the effect of nutritional parameters on the exercise effect on BMD at LS and FN. We recommend that future exercise trials focus on this important aspect.
Apart from generating further general evidence, another aim of the present work was to determine the relevance of selected exercise parameters on BMD effects. Considering that important exercise characteristics were either hard to categorize across the different types of exercise and with respect to given cohort (e.g., strain magnitude) or unequally distributed (e.g., BMD-status), quite homogeneous (e.g., training frequency) or simply not stated (e.g., strain rate), we used intervention length and type of exercise as moderators in our sub-analyses. In summary, we observed no significant differences between the categories; however, the statistical power for our approach was rather limited. In detail, we observed tendentially more favorable effects of RT and combined exercise protocols compared with weight bearing protocols particularly for the FN-ROI. However, three out of five WB-protocols scheduled low intensity exercise such as brisk walking or Tai Chi [17,20,22]. Nevertheless, this result confirmed the results of a recent meta-analysis with 84 included trials that did not observe relevant differences on BMD at LS and FN after resistance vs. weight bearing vs. combined exercise protocols in postmenopausal women [45]. We thus conclude that the ability of comprehensive meta-analyses to derive even raw exercise recommendations is rather limited [46,47], independently of the outcome. The close interaction between exercise parameters (e.g., strain magnitude, rate, cycle number, rest periods, training frequency [39,46]) along with the aspect that even slight differences in exercise composition might significantly modify the exercise effect on BMD [48] underscore the complexity of exercise effects on BMD that obviously collide with simple meta-analytic/meta regression approaches. A possible solution to nevertheless derive (more) dedicated exercise recommendations might be to focus on trials with comparative study arms for a given exercise parameter (e.g., exercise frequency, intensity) but otherwise identical exercise and participant characteristics (e.g., [49]).
In summary, we provided further evidence for a low, but significant effect of exercise on BMD at LS and proximal femur in men. However, we are unable to provide even rough recommendations for an exercise program dedicated to bone. Due the limited number of exercise trials with men, it might be a successful approach to determine gender differences in bone adaptation to exercise. This would legitimize the transfer of the much more extensively evaluated recommendations of exercise studies in women to men. Funding: This research received no external funding; however, the project S3-Guideline "körperliches Training zur Frakturprophylaxe" was supported by the Elsbeth Bonhoff Foundation.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author (WK) upon reasonable request.