Results of literature screening

After developing search strategies based on the study topic, a total of 6966 related publications were retrieved from the databases up to December 31, 2023. Through preliminary screening after title or abstract review, non-English studies, studies not related to the subject of our study, review and meta-analysis, responses and comments, basic experiments, and case reports were excluded. Upon further scrutiny of the methodology and data analysis of the remaining 34 studies, non-RCTs and studies with irrelevant interventions and controls or without extractable data were excluded. Ultimately, 10 studies [18,19,20,21,22,23,24,25,26,27] were included in the analysis (Fig. 1).

Fig. 1

Flowchart of the systematic search and selection process

Basic study characteristics

The subjects in the 10 included studies were postmenopausal women, the trial regions were from Europe, the Americas and Asia, and all studies were RCTs. The interventions included in the study were various regimens of bone formation promoters followed by bone resorption inhibitors: (i) Parathyroid hormone followed by Alendronate; (ii) Teriparatide followed by Alendronate; (iii) Teriparatide followed by Denosumab; (iv) Teriparatide followed by Raloxifene; (v) Abaloparatide followed by Alendronate; (vi) Romosozumab followed by Alendronate; (vii) Romosozumab followed by Denosumab. The duration of therapy ranged from 24 to 48 months. The study design and baseline characteristics of the study population of the 10 included studies are provided in Table 1. No statistically significant differences in age and body mass index (BMI) were found between the intervention and comparator groups, allowing for comparative analysis, as shown in Fig. 2.

Table 1 Basic characteristics of the included studiesFig. 2

Meta-analysis of age and body mass index of the included studies population

Methodological quality assessment of the studies included

As per the Cochrane Handbook (v5.1.0), the studies included were evaluated for quality across seven domains, encompassing random sequence generation, allocation concealment, the blinding of participants and personnel, the blinding of outcome assessment, completeness of outcome data, selective reporting, and other bias sources. All (10) studies employed randomized allocation methods, with no selective bias or other sources of bias, indicating low risk in these aspects; the use of blinding in assessing outcome measures was of moderate risk; risks associated with allocation concealment, patient and experimenter blinding, and outcome data integrity are relatively high, which may be due to the study’s reliance on long-term drug therapy studies, where open-label and loss to follow-up rates were unavoidable. The quality assessment of these studies is described in Fig. 3: The risk of bias graph is shown in Fig. 3a, and the risk of bias summary is shown in Fig. 3b.

Fig. 3

Quality assessment of the included studies

Meta-analysis resultsBMDSpine BMD

Data on changes in Spine BMD were synthesized from 8 studies, including 6196 patients (3078 in the intervention group versus 3118 in the comparator group). Pooled analysis revealed significantly higher changes of the spine BMD in the intervention group (SMD: 1.64; 95% CI: 0.97, 2.31; P < 0.00001) with a significant heterogeneity (I2 = 99%, P < 0.00001)(Fig. 4a). The funnel plot (Fig. 5) and Egger’s test (P = 0.224) revealed no publication bias.

Fig. 4

Meta-analysis of bone mineral density change of spine, bone mineral density change of femoral neck, bone mineral density change of total hip, bone mineral density change of 1/3 distal radius, fracture risk, and incidence of adverse events

Fig. 5

Funnel plot of bone mineral density change of spine, bone mineral density change of femoral neck, bone mineral density change of total hip, bone mineral density change of 1/3 distal radius, fracture risk, and incidence of adverse events

Femoral neck BMD

Analysis of changes in femoral neck BMD was conducted in 7 studies with 5919 patients (2943 in the intervention group versus 2976 in the comparator group). Pooled analysis detected significantly higher changes of the femoral neck BMD in the intervention group (SMD: 0.57; 95% CI: 0.16, 0.99; P = 0.007) with a significant heterogeneity (I2 = 96%, P < 0.00001) (Fig. 4b). The funnel plot (Fig. 5) and Egger’s test (P = 0.578) revealed no publication bias.

Total hip BMD

Five studies involving 5503 patients (2740 in the intervention group versus 2763 in the comparator group) were included in the analysis. Pooled analysis revealed significantly higher changes in the total hip BMD in the intervention group (SMD: 0.82; 95% CI: 0.16, 1.48; P = 0.002) with a significant heterogeneity (I2 = 97%, P < 0.00001) (Fig. 4c). The funnel plot (Fig. 5) and Egger’s test (P = 0.701) indicated no publication bias.

BMD at the 1/3 distal radius

Two articles reported the data of BMD at the 1/3 distal radius, including 169 patients (86 in the intervention group versus 83 in the comparator group). Pooled results demonstrated that the changes in BMD at the 1/3 distal radius were similar between the two groups (SMD: -0.25; 95% CI: -1.49, 0.99; P = 0.069) with a significant heterogeneity (I2 = 92%, P < 0.0003) (Fig. 4d). The funnel plot (Fig. 5) and Egger’s test (P = 0.242) revealed no publication bias.

Fracture risk

Data on the fracture risk were available in 5 studies with 14,163 patients (6961 in the intervention group versus 7202 in the comparator group). Evidence synthesis showed that the intervention group had lower fracture risk (RR: 0.60; 95% CI: 0.43, 0.82; P = 0.001) (Fig. 4e) with a significant heterogeneity (I2 = 75%, P = 0.003). The funnel plot (Fig. 5) and Egger’s test (P = 0.708) revealed no publication bias.

Incidence of adverse events

Five studies with 13,219 patients (6541 in the intervention group versus 6678 in the comparator group) were included in the analysis for incidence of adverse events. Pooled analysis indicated that the intervention group had a lower incidence of adverse events (RR: 0.85; 95% CI: 0.76, 0.95; P = 0.004) with a significant heterogeneity(I2 = 97%, P < 0.00001)(Fig. 4f). The funnel plot (Fig. 5) and Egger’s test result (P = 0.282) revealed no publication bias.

Fig. 6Sensitivity analysis

We conducted one-way sensitivity analyses to compare changes in spinal BMD, femoral neck BMD, total hip BMD, fracture risk, and the incidence of adverse events to evaluate the influence of each study on the combined outcomes by removing the individual study one. The sensitivity analysis revealed that the new combined outcomes remained constant after the exclusion of any individual study for the spine BMD (Fig. 6a) and fracture risk(Fig. 6d). However, when we excluded the data reported by Eastell et al. [24] or Saag et al. [26], the statistical results in femoral neck BMD between the two groups have changed (Fig. 6b). Analogously, changes in the statistical results for total hip BMD were observed when the data reported by Bone et al. [25], Eastell et al. [24], McClung et al. [19], and Saag et al. [26] were excluded one by one (Fig. 6c). Additionally, excluding the data reported by Hagino et al. [22] led to changes in the statistical results for the incidence of adverse events (Fig. 6e). Statistical changes obtained may be related to differences in study design, intervention duration, intervention measures, and sample sizes, as shown in Fig. 6.

Subgroup analysis

Subgroup analysis was performed on the 10 included studies, which were divided into different subgroups according to therapy time, different experimental areas, and various regimens. Five outcome indicators including spine BMD, femoral neck BMD, total hip BMD, distal third of radius BMD, fracture risk, and adverse events were analyzed. The research areas cover Asia, Europe, and the Americas, and the therapy time covers groups greater than or equal to 36 months and less than 36 months. The various regimens cover the Parathyroid hormone followed by the bone resorption inhibitors subgroup, Teriparatide/Abaloparatide followed by the bone resorption inhibitors subgroup, and Romosozumab followed by the bone resorption inhibitors subgroup.

Statistically differences in changes of spine BMD were observed across the American subgroup, various therapy duration subgroups, the parathyroid hormone followed by bone resorption inhibitors subgroup, and the teriparatide/abaloparatide followed by bone resorption inhibitors subgroup. Notably, heterogeneity decreased in the parathyroid hormone followed by the bone resorption inhibitors subgroup, while it remained consistent with the overall population in the other subgroups. Variations in changes of femoral neck BMD were found in the Parathyroid hormone followed by the bone resorption inhibitors subgroup and the romosozumab followed by the bone resorption inhibitors subgroup, with a marked reduction in heterogeneity in the former. Differences in changes in total hip BMD were statistically significant in the European subgroup and the subgroup with a therapy duration of less than 36 months, maintaining consistent heterogeneity with the overall analysis. Regarding fracture risk, differences were statistically significant in the Asian and American subgroups, various therapy duration subgroups, and the romosozumab followed by the bone resorption inhibitors subgroup. Heterogeneity was notably reduced in the subgroup with therapy duration less than 36 months, while it remained consistent with the overall population in the other subgroups. As for adverse events, a statistical difference was observed in the Asian subgroup, based on a single study. No statistical differences in adverse event rates were found in other subgroups, suggesting low reliability of the initial results and insufficient evidence to conclude a difference in adverse event rates between the two groups. The above subgroup analyses are summarized in Table 2.

Table 2 Subgroup analysis