It has been suggested that early postmenopausal women and patients treated with steroids should receive preventive therapy (calcium, vitamin D, vitamin D analogs, estrogens, or bisphosphonates) to preserve their bone mineral density (BMD) and to avoid fragility fractures. We designed the present study to compare the effects of native vitamin D to its hydroxylated analogs alfacalcidol 1-alpha(OH)D and calcitriol 1,25(OH)(2)D. All randomized, controlled, double-blinded trials comparing oral native vitamin D and its analogs, alfacalcidol or calcitriol, to placebo or head-to-head trials in primary or corticosteroids-induced osteoporosis were included in the meta-analysis. Sources included the Cochrane Controlled Trials Register, EMBASE, MEDLINE, and a hand search of abstracts and references lists. The study period January 1985 to January 2003. Data were abstracted by two investigators, and methodological quality was assessed in a similar manner. Heterogeneity was extensively investigated. Results were expressed as effect-size (ES) for bone loss and as rate difference (RD) for fracture while allocated to active treatment or control. Publication bias was investigated. Fourteen studies of native vitamin D, nine of alfacalcidol, and ten of calcitriol fit the inclusion criteria. The two vitamin D analogs appeared to exert a higher preventive effect on bone loss and fracture rates in patients not exposed to glucocorticoids. With respect to BMD, vitamin D analogs versus placebo studies had an ES of 0.36 (P < 0.0001), whereas native vitamin D versus placebo had an ES of 0.17 (P = 0.0005), the interclass difference being highly significant (ANOVA-1, P < 0.05). When restricted to the lumbar spine, this intertreatment difference remained significant: ES = 0.43 (P = 0.0002) for vitamin D analogs and ES = 0.21 (P = 0.001) for native vitamin D (analysis of variance [ANOVA-1], P = 0.047). There were no significant differences regarding their efficacies on other measurement sites (ANOVA-1, P = 0.36). When comparing the adjusted global relative risks for fracture when allocated to vitamin D analogs or native vitamin D, alfacalcidol and calcitriol provided a more marked preventive efficacy against fractures: RD = 10% (95% Confidence interval [CI-2] to 17) compared to RD = 2% (95% CI, 1 to 2), respectively. The analysis of the spinal and nonspinal showed that fracture rates differed between the two classes, thereby confirming the benefits of vitamin D analogs, with significant 13.4% (95% CI 7.7 to 19.8) and. 6% (95% CI 1 to 12) lower fracture rates for vitamin D analogs, respectively. In patients receiving corticosteroid therapy, both treatments provided similar global ESs for BMD: ES = 0.38 for vitamin D analogs and ES = 0.41 for native vitamin D (ANOVA-1, P = 0.88). When restriced to spinal BMD, D analogs provided significant effects, whereas native vitamin D did not: ES = 0.43 (P < 0.0001) and ES = 0.33 (P = 0.21), respectively. The intertreatment difference was nonsignificant (ANOVA-1, P = 0.52). Neither D analogs for native vitamin D significantly prevented fractures in this subcategory of patients: RD = 2.6 (95%CI, -9.5 to 4.3) and RD = 6.4 (95%CI, -2.3 to 10), respectively. In head-to-head studies comparing D analogs and native vitamin D in patients receiving corticosteroids, significant effects favoring D analogs were found for femoral neck BMD: ES = 0.31 at P = 0.02 and spinal fractures: RD = 15% (95%CI, 6.5 to 25). Publication bias was not significant. Our analysis demonstrates a superiority of the D analogs atfacalcidol and calcitriol in preventing bone loss and spinal fractures in primary osteoporosis, including postmenopausal women. In corticosteroid-induced osteoporosis, the efficacy of D analogs differed depending on the comparative approach: indirect comparisons led to nonsignificant differences, whereas direct comparison did provide significant differences. In this setting, D analogs seem to prevent spinal fractures to a greater extent than do native vitamin D, but this assumption should be confirmed on a comprehensive basis in multiarm studies including an inactive comparator.