Our objective was to assess the effects of differing modes of impact exercise on bone density at the hip and spine in premenopausal women through systematic review and meta-analysis. Electronic databases, key journals and reference lists were searched for controlled trials investigating the effects of impact exercise interventions on lumbar spine (LS), femoral neck (FN) and total hip (TH) bone mineral density (BMD) in premenopausal women. Exercise protocols were categorised according to impact loading characteristics. Weighted mean difference (WMD) meta-analyses were undertaken. Heterogeneity amongst trials was assessed. Fixed and random effects models were applied. Inspection of funnel plot symmetry was performed. Trial quality assessment was also undertaken. Combined protocols integrating odd- or high-impact exercise with high-magnitude loading (resistance exercises), were effective in increasing BMD at both LS and FN [WMD (fixed effect) 0.009 g cm(-2) 95% CI (0.002-0.015) and 0.007 g cm(-2) 95% CI (0.001-0.013); P = 0.011 and 0.017, respectively]. High-impact only protocols were effective on femoral neck BMD [WMD (fixed effect) 0.024 g cm(-2) 95% CI (0.002-0.027); P < 0.00001]. Funnel plots showed some asymmetry for positive BMD outcomes. Insufficient numbers of protocols assessing TH BMD were available for assessment. Exercise programmes that combine odd- or high-impact activity with high-magnitude resistance training appear effective in augmenting BMD in premenopausal women at the hip and spine. High-impact-alone protocols are effective only on hip BMD in this group. However, diverse methodological and reporting discrepancies are evident in published trials.