Effects of high-impact jumping versus resistance exercise on bone mineral content in children and adolescents: a systematic review and meta-analysis

Protocol and registration

This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement guidelines (Page et al., 2021). The study protocol was registered in the International Prospective Register of Systematic Reviews (ID: CRD 42024625921).

Search strategy

A literature search was conducted by two researchers, Zhang and Yang, based on the inclusion and exclusion criteria, which included PubMed, Embase, The Cochrane Library and Web of Science databases, spanned from the establishment of each database to April 4, 2025, without any language restriction. In case of disagreement, the researchers will resolve the issue through discussion with the first author, and if the disagreement is still not agreed upon, the first author will serve as the final adjudicator. The PubMed database was systematically searched employing the following criteria: (((((((((“Exercise”[Mesh]) OR (“Resistance Training”[Mesh])) OR (sports)) OR (High-impact sports)) OR (Jump)) OR (physical activity)) OR (training))AND (((((((“Adolescent”[Mesh]) OR (student)) OR (Puberty)) OR (children)) OR (kids)) OR (child)) OR (pediatrics))) AND ((((bone) OR (bone health)) OR (Bone mineral density)) OR (Bone mineral content)) AND (clinical trial[Filter] OR controlled clinical trial [Filter]OR randomized controlled trial [Filter]).

Inclusion and exclusion criteria for the studies

The inclusion criteria followed the PICOS framework: (a) participants: healthy children and adolescents (<18 years) without hepatic, renal, endocrine, or other metabolic bone diseases; (b) intervention: structured high-impact exercise or resistance training programs; (c) control: routine school physical education without supplemental exercise interventions; (d) outcomes: ΔBMC (pre-to-post intervention) measured by dual-energy X-ray absorptiometry (DXA) at lumbar spine, femoral neck and whole-body (studies using dual-photon absorptiometry were excluded, e.g., Blimkie et al. (1996)); (e) study design: randomized or non-randomized controlled trials.

Exclusion criteria comprised: (a) participants ≥ 18 years; (b) concurrent pharmacotherapy; (c) non-extractable outcome data; (d) abstract-only/review articles; (e) animal studies.

Literature screening and data extraction

Two researchers independently screened the literature and extracted data according to the predefined inclusion/exclusion criteria, with any discrepancies resolved through consensus discussion involving a third reviewer. The systematic search was conducted across multiple databases using standardized search strategies. All retrieved records were imported into EndNote for duplicate removal, followed by title/abstract screening to exclude irrelevant studies. The extracted study characteristics included: (a) Basic characteristics: first author, publication year, country, study design, sample size, participants’ age and gender distribution; (b) Intervention characteristics: exercise modality (specific description), intervention duration, intervention frequency, intervention period and outcome measures.

Quality assessment

Two researchers independently evaluated the methodological quality of included studies using the Cochrane Collaboration’s Risk of Bias tool (Higgins et al., 2011), assessing seven domains: (a) random sequence generation (selection bias); (b) allocation concealment (selection bias); (c) blinding of participants and personnel (performance bias); (d) blinding of outcome assessment (detection bias); (e) incomplete outcome data (attrition bias); (f) selective reporting (reporting bias); (g) other potential biases. Each domain was judged as ‘low risk’, ‘high risk’ or ‘unclear risk’ of bias.

Data analysis

Meta-analysis was conducted using RevMan 5.4 and Stata 17 software. As all included studies measured BMC in grams (g) by Dual-energy X-ray Absorptiometry (DXA), the mean difference (MD) with 95% confidence intervals (95% CI) was selected as the effect measure. We preferentially extracted within-group change scores (follow-up minus baseline values) and their standard deviations (SDs) for both intervention and control groups. When studies did not directly report SDs of change scores, we calculated them using the following formula:

$\mathrm{S}{\mathrm{D}}_{\mathrm{change}}=\sqrt{\mathrm{S}{\mathrm{D}}_{\mathrm{baseline}}^2+\mathrm{S}{\mathrm{D}}_{\mathrm{follow}-\mathrm{up}}^2-2\times \mathrm{r}\times \mathrm{S}{\mathrm{D}}_{\mathrm{baseline}}\times \mathrm{S}{\mathrm{D}}_{\mathrm{follow}-\mathrm{up}}}$

a default correlation coefficient (r) of 0.5 was used when unreported. Heterogeneity was evaluated using Cochrane’s Q-test (P < 0.10) and I² statistics, with interpretation thresholds set as: 0%–40% (might not be significant), 30%–60% (moderate), 50%–90% (substantial), and 75%–100% (considerable) according to Cochrane Handbook guidelines (Cochrane Collaboration, 2024). A random-effects model was applied when I²  ≥  50%. To ensure result robustness, we performed subgroup analyses (by sex), sensitivity analyses. Meta-regression with robust variance estimation (RVE) was performed using the robumeta package, as described by Fisher & Tipton (2015), to evaluate the effects of moderator variables (duration, frequency, and period) on BMC at the lumbar spine, femoral neck and whole-body in jump and resistance exercises.

Search results

A total of 7,368 relevant studies were initially identified, including 1,724 from the PubMed database, 3,115 from Embase, 1,508 from The Cochrane Library and 1,021 from Web of Science. These studies were exported to EndNote and after removing duplicates, 3,403 articles remained. After screening titles and abstracts, 3,286 articles were excluded, leaving 117 studies for full-text review to assess eligibility for inclusion. Among them, inconsistency in the target of the intervention (n = 20); interventions did not meet the inclusion criteria (n = 46); the control group did not meet the criteria (n = 5); discrepancies in ending indicators (n = 23); unable to extract data (n = 11), and finally the remaining 12 articles were included in meta-analysis. Figure 1 shows the specific process.

Basic characteristics of the included studies

The meta-analysis included 12 studies (Arnett & Lutz, 2002; Dowthwaite et al., 2019; Fuchs, Bauer & Snow, 2001; Gómez et al., 2021; Macdonald et al., 2008; Nichols et al., 2008; Nichols, Sanborn & Love, 2001; Nogueira, Weeks & Beck, 2014b; Nogueira, Weeks & Beck, 2015; Thein-Nissenbaum et al., 2023; Weeks, Young & Beck, 2008; Witzke & Snow, 2000), published between 2000–2023, involving 940 total participants (intervention group: n = 524; control group: n = 416), aged 7.3 (Fuchs, Bauer & Snow, 2001)–16.8 (Gómez et al., 2021) years (mean age: females 12.6; males 11.1). Sample sizes in intervention groups ranged from 5 (Nichols, Sanborn & Love, 2001) to 71 (Nogueira, Weeks & Beck, 2014b) participants. Studies were conducted in the United States (n = 8) (Arnett & Lutz, 2002; Dowthwaite et al., 2019; Fuchs, Bauer & Snow, 2001; Gómez et al., 2021; Nichols et al., 2008; Nichols, Sanborn & Love, 2001; Thein-Nissenbaum et al., 2023; Witzke & Snow, 2000), Australia (n = 3) (Nogueira, Weeks & Beck, 2014b; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008) and Canada (n = 1) (Macdonald et al., 2008). Detailed baseline characteristics are presented in Table 1.

Characteristics of exercise interventions included in the studies

Eight studies (Arnett & Lutz, 2002; Fuchs, Bauer & Snow, 2001; Macdonald et al., 2008; Nichols et al., 2008; Nogueira, Weeks & Beck, 2014b; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) implemented high-impact exercise, while four studies (Dowthwaite et al., 2019; Gómez et al., 2021; Nichols, Sanborn & Love, 2001; Thein-Nissenbaum et al., 2023) involved resistance exercise. The intervention duration ranged from five (Arnett & Lutz, 2002) to 60 minutes (Gómez et al., 2021), with a frequency of two (Nichols et al., 2008; Weeks, Young & Beck, 2008) to five sessions per week (Macdonald et al., 2008) . The intervention period spanned three (Gómez et al., 2021) to 24 months (Dowthwaite et al., 2019). BMC outcomes were measured at three sites: lumbar spine (n = 11), femoral neck (n = 11) and whole-body (n = 10). Detailed intervention characteristics are presented in Table 2.

Risk of bias assessment of included studies

According to the Cochrane risk of bias assessment criteria, the methodological quality of included studies demonstrated the following characteristics: Four studies (Dowthwaite et al., 2019; Gómez et al., 2021; Thein-Nissenbaum et al., 2023; Witzke & Snow, 2000) used non-randomized grouping (by school/class), while one study (Nichols, Sanborn & Love, 2001) showed significant attrition bias (85% dropout in intervention group) and low adherence (mean compliance: 73%). Two studies (Arnett & Lutz, 2002; Gómez et al., 2021) had short intervention periods (3–4 months), potentially limiting long-term effect observation. Although no studies reported allocation concealment or blinding of participants/personnel, all utilized DXA for objective outcome measurement, resulting in low detection bias and no evidence of selective reporting bias. The risk of bias evaluation results are shown in Figs. 2 and 3.

Meta-analysis results

Effects of high-impact jumping on lumbar spine BMC

Eight studies (Arnett & Lutz, 2002; Fuchs, Bauer & Snow, 2001; Macdonald et al., 2008; Nichols et al., 2008; Nogueira, Weeks & Beck, 2014b; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) evaluated the effects of jumping on lumbar spine BMC. No significant heterogeneity was detected (P = 0.76, I² = 0%), and a fixed-effects model was applied. The meta-analysis demonstrated a statistically significant increase in lumbar spine BMC changes in the exercise group compared to controls (MD = 0.86, 95% CI [0.27–1.45], P = 0.004) (Fig. 4).

Effects of high-impact jumping on femoral neck BMC

Eight studies (Arnett & Lutz, 2002; Fuchs, Bauer & Snow, 2001; Macdonald et al., 2008; Nichols et al., 2008; Nogueira, Weeks & Beck, 2014b; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) examined the effects of jumping on femoral neck BMC. No significant heterogeneity was observed (P = 0.93, I² = 0%), warranting a fixed-effects model. Meta-analysis revealed a statistically significant increase in femoral neck BMC changes in the exercise group compared to controls (MD = 0.11, 95% CI [0.04–0.18], P = 0.001) (Fig. 5).

Effects of high-impact jumping on whole-body BMC

Six studies (Macdonald et al., 2008; Nichols et al., 2008; Nogueira, Weeks & Beck, 2014b; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) evaluated the effects of jumping on whole-body BMC. No significant heterogeneity was observed (P = 0.89, I² = 0%) and a fixed-effects model was employed. The meta-analysis found no statistically significant difference in whole-body BMC changes between exercise and control groups (MD = 5.11, 95% CI [−42.18–52.40], P = 0.83) (Fig. 6).

Effects of resistance exercise on lumbar spine BMC

Three studies (Dowthwaite et al., 2019; Nichols, Sanborn & Love, 2001; Thein-Nissenbaum et al., 2023) examined the effect of resistance exercise on lumbar spine BMC. No significant heterogeneity was detected (P = 0.99, I² = 0%) and a fixed-effects model was applied. The meta-analysis showed no statistically significant difference in lumbar spine BMC changes between exercise and control groups (MD = 0.70, 95% CI [−0.35–1.75], P = 0.19) (Fig. 7).

Effects of resistance exercise on femoral neck BMC

Three studies (Dowthwaite et al., 2019; Nichols, Sanborn & Love, 2001; Thein-Nissenbaum et al., 2023) evaluated the effect of resistance exercise on femoral neck BMC. No significant heterogeneity was observed (P = 0.92, I² = 0%) and a fixed-effects model was used. The meta-analysis revealed no statistically significant difference in femoral neck BMC changes between exercise and control groups (MD = 0.06, 95% CI [−0.04–0.15], P = 0.25) (Fig. 8).

Effects of resistance exercise on whole-body BMC

Four studies (Dowthwaite et al., 2019; Gómez et al., 2021; Nichols, Sanborn & Love, 2001; Thein-Nissenbaum et al., 2023) investigated the effect of resistance exercise on whole-body BMC. The analysis showed no significant heterogeneity (P = 0.92, I² = 0%), and a fixed-effects model was employed. Meta-analysis results indicated no statistically significant difference in whole-body BMC changes between exercise and control groups (MD = 56.22, 95% CI [−3.98–116.43], P = 0.07) (Fig. 9).

Subgroup analysis results

The subgroup analysis by sex revealed differential effects of high-impact jumping on BMC. Three studies involving boys (Macdonald et al., 2008; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008) and five studies involving girls (Arnett & Lutz, 2002; Macdonald et al., 2008; Nogueira, Weeks & Beck, 2014b; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) were analyzed for lumbar spine BMC. Subgroup analysis demonstrated that jumping significantly increased lumbar spine BMC in girls compared to controls (MD = 1.40, 95% CI [0.16–2.63], P = 0.03), while no significant effect was observed in boys (MD = 0.55, 95% CI [−1.19–2.29], P = 0.54) (Fig. 10).

Three studies (Macdonald et al., 2008; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008) examined the effects on femoral neck BMC in boys, while five studies (Arnett & Lutz, 2002; Macdonald et al., 2008; Nogueira, Weeks & Beck, 2014b; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) evaluated girls. Subgroup analysis demonstrated that jumping significantly increased femoral neck BMC in girls compared to controls (MD = 0.11, 95% CI [0.00–0.21], P = 0.04), but showed no significant effect in boys (MD = 0.02, 95% CI [−0.13–0.17], P = 0.79) (Fig. 11).

Three studies (Macdonald et al., 2008; Nogueira, Weeks & Beck, 2015; Weeks, Young & Beck, 2008) evaluated the effects on whole-body BMC in boys, while four studies (Macdonald et al., 2008; Nogueira, Weeks & Beck, 2014b; Weeks, Young & Beck, 2008; Witzke & Snow, 2000) assessed girls. The subgroup analysis revealed no statistically significant effects of jumping on whole-body BMC changes in either girls (MD = 13.62, 95% CI [−57.10–84.3], P = 0.71) or boys (MD = 20.06, 95% CI [−50.25–90.38], P = 0.58) compared to control groups (Fig. 12).

All resistance exercise studies exclusively involved female participants. As previously reported in the meta-analysis, resistance exercise demonstrated no statistically significant effects on lumbar spine, femoral neck or whole-body BMC in this population.

Sensitivity analysis

The sensitivity analysis confirmed the stability of effect sizes for both high-impact jumping and resistance exercise on femoral neck, lumbar spine and whole-body BMC. The pooled effect sizes (95% CI) remained robust without significant alterations upon sequential exclusion of individual studies Fig. S1.

Meta-regression analysis

The robust meta-regression analysis revealed distinct patterns of association between exercise parameters and BMC across anatomical sites. For jumping, longer period showed a significant negative association with whole-body BMC (β = −12.74, P = 0.028), while no significant effects were observed at lumbar spine or femoral neck sites (all P > 0.05). In resistance exercise, both longer duration (β = −1.68, P = 0.032) and period length (β = −3.83, P = 0.014) were negatively associated with whole-body BMC, with consistent null findings at other anatomical sites Table S1.