A computerized literature search was performed on April 15, 2015 by using Koreanstudies Information Service System (KISS, http://kiss.kstudy.com/) and PubMed (http://www.ncbi.nlm.nih.gov/pubmed). Inputted keywords included “Female AND Exercise AND bone mineral density” for KISS and “Korea[Mesh] AND Female[Mesh] AND Exercise[Mesh] AND Bone Density[Mesh]” for PubMed. Reference lists of the identified articles were reviewed to identify additional relevant articles.

Studies were selected when they met all of the following inclusion criteria: 1) randomized control trials (RCT); 2) postmenopausal women and/or older women whose mean age ≥ 65 years; 3) receiving exercise training; and 4) using acceptable methods of BMD assessment such as dual energy X-ray absorptiometry (DXA). Primary outcomes for this study were changes in BMD of lumbar spine (LS), femoral neck (FN), Ward’s triangle (WT), and trochanter (Tro).

KISS and PubMed searches yielded 116 articles. One article was identified as relevant from other reference lists. Literature search was conducted independently by the primary reviewer (JEK) and the secondary reviewer (HSM). Disagreements were resolved by consensus of investigators (JEK, HMJ, and HSM). Of the 117 articles, 99 were excluded due to the following reasons: study design was not a RCT; study subjects were premenopausal women or mean age < 65 years; exercise training was combined with other treatments including hormone replacement therapy or dietary intervention. Therefore, a total of 18 articles were included and fully assessed by JEK and the HSM independently. One article was excluded because it did not report primary outcomes. Finally, 17 articles were included for this systematic review (Figure 1). One article provided multiple treatments including dietary supplement and exercise training¹⁶.The group with dietary supplement treatment in this article was excluded since dietary supplement treatment did not meet our inclusion criteria. Four articles included two exercise groups^{17, 18} or three exercise groups^{19, 20} with different type of exercise training. Therefore, they were assigned into the groups with 2 and 3 different types of exercise training, respectively.

The following data were extracted by the primary and secondary reviewers from each article using an electronic form: first author’s last name, publication year; population and sample size; details of intervention design; mean age; exercise type; intervention duration; method of outcome measures; pre- and post- intervention values; and changed value of outcome measures.

The quality of selected articles was assessed by using modified Cochrane risk of bias assessment tool²¹. Evaluated risk of bias included selection bias, performance bias, and detection bias.

Changes in BMD of LS, FN, WT, and Tro were manually calculated from raw pre- and post- intervention values reported in the articles. Percent change values of BMD were also manually calculated by dividing the change of BMD by pre-intervention value and multiplying the ratio by 100.

Participants in 17 RCTs were either postmenopausal women, older women (mean age ≥ 65 years), or women who received hysterectomy. Interventions lasted between 8 to 48 weeks. Participants conducted exercise training 3-5 days per week. Eight^22-29, five^{18, 20, 30-32}, and two^{16, 33} articles provided combined (aerobic + resistance), aerobic, or resistance exercise training, respectively. Cho, 2003¹⁷ included both aerobic and combined exercise training groups. Ahn, 2002¹⁹ included both aerobic and resistance exercise training groups (Table 2).

The qualities of selected articles are summarized in Table 3. Only one article²⁸ indicated a low risk of randomization factor using computer-generated program for study randomization. The remaining studies did not clearly report the randomization method. Moreover, all articles did not provide details regarding allocation concealment, blinding of participants, study investigator, or blinding of outcome assessment.

The effects of exercise training on BMD are summarized in Table 4. All selected articles reported BMD change in LS after exercise training. In addition, BMD of FN, WT, and Tro were reported in 12^{16-20, 22, 23, 27, 28, 30-32}, 11^{16-20, 22, 23, 28, 30-32}, and 11^{16-20, 22, 23, 28, 30-32} articles, respectively.

Exercise training resulted in greater positive change in BMD of LS. The average absolute and percent changes in BMD of LS were 0.02 g/cm² (range: -0.01-0.12) and 2.3 % (range: -1.1-10.9), respectively, in the exercise group. In the control group without exercise training, the average absolute and percent changes in BMD of LS were -0.02 g/cm² (-0.10-0.01) and -1.1 % (range: -7.5-1.1), respectively. In addition, the absolute changes in BMD of FN, WT, and Tro were 0.02 (range: -0.01-0.07), 0.01 (range: -0.01-0.04), and 0.01 g/cm² (range: -0.01-0.05), respectively, in the exercise group and -0.01 (range: -0.04-0.01), -0.01 (range: -0.02-0.01), and -0.01 g/cm² (range: -0.03-0.01), respectively, in the control group.

According to exercise type, the average absolute changes in BMD of LS, FN, WT, and Tro with combined exercise training were 0.03, 0.04, 0.02, and 0.02 g/cm², respectively. The percent changes in BMD of LS, FN, WT, and Tro with combined exercise training were 3.3, 4.8, 3.2, and 3.4 %, respectively (Table 4 and Figure 2). Most of these articles indicated that both resistance and aerobic exercise trainings improved BMD. However, some articles reported that BMD was decreased after exercise trainings (Table 4 and Figure 2).

Bone is a dynamic tissue. It is very sensitive to mechanical stimuli such as exercise training that provides mechanical loading to bone^{9, 10}. It has been well-established that exercise training may increase or maintain bone mass and decrease the risk of fracture in women during the aging process⁹. The current systematic review observed an overall increase in BMD of LS, FN, WT, and Tro with exercise trainings and a decrease in BMD without exercise training in Korean postmenopausal women. These results are consistent with previous systematic reviews and meta-analyses^34-37.

Although positive impact of exercise training on BMD has been well-established, certain selected articles in this review have reported a minimal decrease or no changes in BMD after exercise training^{17, 19, 20, 31}. Such results might be partially explained by the length of the exercise training. In general, bone remodeling requires a minimum of 4-6 months to 1-2 years³⁸. The length of exercise training in these articles was ≤ 12 weeks, which is shorter than the time required for bone remodeling. Previous meta-analysis studies have reported beneficial impacts of exercise training on BMD by including articles that provided ≥ 24 weeks of exercise trainings^{36, 37}.

By contrast, another meta-analysis has reported no improvement in BMD of FN with ≥ 24 weeks of exercise training³⁹ in postmenopausal women. The discrepancy in the results between our finding and previous work in FN might be due to differences in the intensity of exercise training. Most articles selected for previous research used lower intensity activities (primarily walking) as the exercise training protocol. A recent consensus statement from the National Institutes of Health Consensus Development Panel on Osteoporosis, Prevention, Diagnosis, and Therapy has suggested that higher intensity activities may develop higher peak bone mass and reduce the risk of fall in older adults⁴⁰.

Very limited human intervention studies have investigated the effect of exercise training frequency on bone. A recent study has reported that at least two sessions per week of exercise training might be crucial to observe the beneficial impact in postmenopausal and osteopenic women⁴¹. The United States Sports Academy has also recommended that three or more sessions per week using various types of exercises may increase BMD in older adults⁴². In the current systematic review, all selected articles provided three or more sessions per week which met the general recommendation. The optimum type of exercise training to improve bone health in postmenopausal women is not fully determined yet. Therefore, more studies are needed to evaluate the adequate length, intensity, and frequency of exercise training protocol in this population.

Although exercise training has been recommended for the prevention and management of postmenopausal bone loss, the most effective exercise type has not been well examined. In general, resistance exercise training has been highlighted as the most beneficial exercise type on bone health since a variety of muscular loadings are applied to the bone during resistance exercise training which generate stimuli and promote osteogenic response of the bone⁴³. A previous systematic review has revealed that performing resistance exercise training 2-3 times per week during 1 year is able to maintain or increase the BMD of spine and hip in postmenopausal women¹⁰. A previous meta-analysis has also observed a positive effect of resistance exercise training in BMD of LS and femur in postmenopausal women⁴⁴. Our systematic review included 3 articles that performed resistance exercise training. Consistent with previous findings, Ahn, 2002¹⁹ has reported improvements in BMD of LS, FN, WT, and Tro (1.5-2.2% increases) after 1 year of resistance exercise training. Kim, 2010³³ has also reported a minimal increase (0.3%) after 12 weeks of resistance exercise training. By contrast, Sung et al., 2004¹⁶ have reported a minimal decrease or no changes in BMD of LS, FN, WT, and Tro after 8 weeks of resistance exercise training. The authors discussed that 8 weeks period might not be enough to observe the changes in BMD with resistance exercise training.

Recently, mounting evidence has indicated that a combination of several types of exercise training may be more effective than any single type of exercise training for improving BMD in postmenopausal women^{34, 45}. Particularly, a combination of resistance exercise training and weight-bearing aerobic exercise training including walking, running, or jumping is recommended since resistance exercise training provides muscular loading while weight-bearing aerobic exercise training provides additional mechanical loading to the bone above gravity. A recent systematic review has summarized that the majority of research studies with combined exercise training have shown improved BMD among postmenopausal women³⁴. A previous meta-analysis to assess the effects of the types of exercise training has also revealed that a mixture of resistance and aerobic exercise trainings is effective in reducing postmenopausal bone loss at the hip and spine whereas other forms of exercise training are less effective in preserving BMD in this population⁴⁶. Similar to previous findings, the articles that provided combined exercise training in the current review all reported improvements in BMD of LS, FN, WT, and Tro (0.7-8.5% increases). However, caution is warranted in the interpretation of positive impact of combined exercise training on BMD because methodology differences and reporting discrepancies still exist^{34, 46}.

There is a paucity of data regarding the impact of aerobic exercise training alone on bone health. Studies performed to determine the impact of aerobic exercise training on BMD in postmenopausal women have reported disparate results. One possible explanation for such disparate results is the application of weight-bearing versus non-weight-bearing aerobic exercise training in the exercise program. Many studies have observed positive effects of moderate to high impact weight-bearing aerobic exercise trainings on changes in BMD, including walking, jogging, stair climbing or mountain climbing, and jumping^{34, 45}. However, particularly high-impact weight-bearing aerobic exercise trainings are not always suitable for older adults due to the presence of musculoskeletal impairment in this population⁴⁷. Therefore, non-weight-bearing aerobic exercise training such as swimming has been recommended for this population. However, non-weight-bearing aerobic exercise training applied no or very low impact on bone. They had fewer osteogenic responses than weight-bearing aerobic exercise trainings in older adults⁴⁸. The current systematic review also observed inconsistent results in BMD after aerobic exercise training among articles. Walking, jogging, aerobic dance, and mountain climbing (-1.0-10.9% increases) had greater improvements in BMD of LS, FN, WT, and Tro compared to swimming (-1.0-1.5% increases). Less favorable changes in bone status after aerobic exercise training have been reported compared to resistance^{12, 49} or combined³⁴ exercise trainings. Consistent with previous studies, selected articles in the current review also reported greater increases of BMD with resistance and combined exercise trainings than aerobic exercise training alone. Ahn, 2002¹⁹ conducted aerobic or resistance exercise training and found that resistance exercise training (1.5-2.2% increases) provided more favorable changes in BMD of LS, FN, WT, and Tro compared to aerobic exercise training (-1.0-1.5% increases). Cho, 2003¹⁷ provided both aerobic and combined exercise training and found that combined exercise training (2.0-28.5% increases) had more improvements in BMD of LS, FN, WT, and Tro compared to aerobic exercise training alone (-1.0-1.4% increases). Although resistance and combined exercise trainings are more likely to improve BMD than aerobic exercise training in postmenopausal and older women, further studies are needed to evaluate the optimum type of exercise training in this population without causing difficulties to this population during exercise training.

The findings from the current systematic review were limited by various intervention design, intervention duration, and exercise training protocols. In addition, according to results from quality assessments, except Park, 2008²⁸, all articles did not provide details regarding randomization, allocation concealment, blinding of participants and study investigator, and blinding of outcome assessment. Nonetheless, the current review systematically evaluated the impact of exercise training and the type of exercise training on BMD responses in Korean postmenopausal women who are at risk for osteoporosis. Moreover, LS, FN, WT, and Tro are the regions strongly attributable to osteoporosis⁵⁰. Thus, the findings from this systematic review might be clinically useful for healthcare providers and Korean postmenopausal women.