The effect of teriparatide on patients with atypical femur fractures: a systematic review and meta-analysis

This SR and MA included studies based on the PICOS criteria: patients, intervention, control, outcomes, and study design. The patients of interest were patients with AFFs. The control group consisted of patients who received everything in the intervention group except for TPTD. The studies must report the results of the outcomes of interest to be included. Given the scarcity of controlled studies on the effect of TPTD on AFFs, we searched for studies whose designs were randomized controlled trials as well as comparative observational, prospective, and retrospective studies. There were no restrictions on race, country, publication date, or follow-up duration. To increase our sample size, we included groups with complete or incomplete AFFs, unilateral or bilateral AFFs, and regardless of the site of the femur fracture, such as subtrochanteric or diaphyseal. We included only studies that assessed TPTD after the occurrence of AFFs and were either treated with surgery or conservatively, such as with BP cessation, dietary calcium, and vitamin D.

We excluded single-arm studies, animal studies, conference abstracts, non-English papers, and studies that did not report our outcomes of interest separately for the TPID group.

Relevant articles were identified through a comprehensive search of the PubMed, Web of Science, Cochrane, and Scopus databases from inception to August 2, 2023. Other relevant studies were found by searching the reference lists of the eligible papers.

A search was conducted in PubMed, Web of Science, Scopus, and Cochrane for comparative studies published using a combination of the following terms: “teriparatide”, “atypical”, “femur”, and “fracture”. There were no filters applied. Supplementary Table 1 contains the complete search strategy for each database.

All records were pooled using Endnote. The data were exported to an Excel sheet, which was subsequently submitted in two stages to find the eligible studies. The title and abstract screening step was carried out first, and records that passed this stage were moved to the full-text screening stage. It is worth noting that the eligibility of each article in each phase was independently examined by two authors. Any disagreements were settled by a third senior author.

The lead author prepared formatted Excel sheets in which the review authors extracted baseline data and study characteristics, as well as quality assessment and outcomes of interest. Two authors extracted data from each study independently and then discussed it. Any disagreements were settled by a third senior author. Any incomplete or incompatible data were handled using the Cochrane Handbook’s suggested methods [15]

The primary outcomes were the incidence of complete bone healing and the time to bone union. Complete bone healing was defined as bridging across three or four cortices and/or disappearance of a visible fracture line on standard antero-posterior and lateral femoral radiographs, and/or a clinical lack of pain at the fracture site on palpation and weight-bearing [16, 17]. It included the incidence of bone healing as reported at the end of the study, regardless of how long it took for the healing to occur. Secondary outcomes included the incidence of early bone union, the incidence of delayed union, the incidence of progression to a complete fracture, and the incidence of non-union. Early bone union was defined if the fracture healed within 6 months. Delayed union was defined as the lack of evidence of bone union within 6 months. Non-union was defined as a fracture that did not achieve union at the end of the study.

Two authors independently extracted study characteristics and baseline data. Study characteristics included: study ID, study design, follow-up duration, AFF diagnostic criteria applied by each study, laterality, degree, and site of the AFF, AFF treatment employed by the studies, such as surgery or conservative treatment, and description of the intervention group and control group. Baseline data included sample size, age, gender, duration of BP use, number of patients who used BP, number of patients who stopped BP after AFFs, and BP agent used.

Two authors independently assessed the quality of the included studies using Newcastle–Ottawa Scale [18]. The quality of the studies was determined by the overall score they received, which was as follows: very good (9–10 points), good (7–8 points), satisfactory (5–6 points), and unsatisfactory (0–4 points).

Review Manager (RevMan) version 5.4 [28] was used to conduct all the analyses. All the data was collected as means ± standard deviation (SD), or event and total for continuous and dichotomous outcomes, respectively. The continuous outcome data of time to union was measured using the inverse variance statistic method and reported as mean differences with a 95% confidence interval (CI), and the Mantel–Haensze equation to calculate the pooled RR and 95% CI was used for the remaining dichotomous outcomes. Cochrane’s Q test and the I² statistic were used to assess heterogeneity. Significant heterogeneity was considered if the P value was less than 0.1 and the I² was greater than 60%. We used the random effects model regardless of heterogeneity due to differences in studies and patient characteristics, as well as limited data, which did not justify assuming the presence of a true effect size among the included studies and using the fixed effects model. To solve and identify the source of heterogeneity, the leave-one-out strategy was employed. Subgroup analyses were carried out to determine the impact of TPTD on surgically treated cases and cases with complete fractures.

The database search yielded a total of 333 records. After duplicates were removed, 206 records were entered into the selection process and evaluated for eligibility criteria. Finally, our study included eight eligible studies [16, 17, 19‐24]. Figure 1 shows the detailed process of search strategy results and study selection.

The meta-analysis included six retrospective studies [17, 19‐24] and one prospective study [16], representing a total of 238 patients, 86 of whom received teriparatide and 152 of whom did not. The majority of the included studies were conducted on Asian populations, with the exception of Shin et al. [21] and Chiang et al. [16], which were conducted in the USA and Australia, respectively. One study [22] included patients with incomplete fractures; three studies [17, 19, 24] included patients with complete AFFs; and one study [21] included patients with complete AFFs; however, six patients suffered from contralateral incomplete fractures. In two studies [17, 23], the site of the fracture was diaphyseal, while the majority of the rest of the studies included both diaphyseal and subtrochanteric fractures. Except for Cho et al. [17], which addressed only unilateral fractures, all other studies included both unilateral and bilateral fractures. All patients were treated surgically in five studies [17, 19‐21, 24], conservatively in one [22], and surgically or conservatively in the remaining two [16, 23], one of which reported the results separately for patients treated conservatively or surgically [23]. The summary of the study characteristics are shown in Table 1. Almost all of the patients were old females who had been taking BPs prior to the onset of AFFs but had stopped taking them after the onset of AFFs. The most commonly used BP drug was alendronate. The patients characteristics are summarised in Table 2.

All of the studies included were of good quality (7–8 points). Seven studies had a total of eight points [16, 17, 19‐21, 23, 24], while only one study received seven points [22]. The quality assessment of the studies included is shown in Supplementary Table 2.

The analysis included eight studies [16, 17, 19‐24], with a total of 91 patients in the teriparatide arm and 154 patients in the non-teriparatide arm. The analysis showed insignificant increase in the incidence of complete bone healing in patients who received teriparatide compared to those who didn’t (RR = 1.09, 95% CI [0.99, 1.13], P = 0.12). The pooled analysis was homogenous (P = 0.56, I² = 0%) (Fig. 2).

The analysis included six studies [17, 19‐21, 23, 24], with a total of 83 patients in the teriparatide arm and 106 patients in the non-teriparatide arm. The time required for bone union to occur was significantly reduced by TPTD (MD = −1.56, 95% CI [−2.86, −0.26], P = 0.02) (Fig. 3a). The pooled analysis was heterogeneous (P = 0.04, I² = 57%), which was resolved after the exclusion of Cho et al. [17] (P = 0.17, I² = 38%) without significant effect on the overall estimate (MD = −2.36, 95% CI [−4.08, −0.63], P = 0.007) (Fig. 3b).

The analysis included seven studies [16, 19‐24], with a total of 85 patients in the teriparatide arm and 142 patients in the non-teriparatide arm. The analysis showed that TPTD insignificantly reduced the incidence of bone non-union (RR = 0.48, 95% CI [0.22, 1.04], P = 0.06). The pooled analysis was homogenous (P = 0.78, I² = 0%) (Fig. 4).

The analysis included three studies [19, 23, 24] with a total of 43 patients in the teriparatide arm and 63 patients in the non-teriparatide arm. The analysis found that the incidence of bone healing within 6 months of TPTD treatment in patients who received TPTD was significantly higher than that in the control group (RR = 1.45, 95% CI [1.13, 1.87], P = 0.004). The pooled analysis was homogenous (P = 0.97, I² = 0%) (Fig. 5). The incidence of delayed bone healing, on the other hand, was significantly lower in the TPTD group compared to the control group (RR = 0.47, 95% CI [0.22, 0.99], P = 0.05). The pooled analysis was homogenous (P = 0.55, I² = 0%) (Fig. 5).

The analysis included three studies [22‐24] with a total of 12 patients in the teriparatide arm and 80 patients in the non-teriparatide arm. TPTD lowered the incidence of progression to complete fracture but insignificantly (RR = 0.27, 95% CI [0.04, 1.97], P = 0.19). The pooled analysis was homogenous (P = 0.84, I² = 0%) (Fig. 6).

Subgroup analysis showed that in patients who received surgical treatment, TPTD demonstrated a significantly shorter time to bone union (MD = −1.72, 95% CI [−3.09, −0.35], P = 0.01). Furthermore, it significantly increased the incidence of early bone union in patients undergoing surgical treatment and those with complete fractures (RR = 1.52 95% CI [1.16, 1.98], P = 0.002) and (RR = 1.54 95% CI [1.18, 2.01], P = 0.001). Table 3 shows the results of subgroup analysis.

This SR and MA benefit from a thorough search that includes the most recent relevant trials as well as all accessible data, which was either examined as primary, secondary, or exploratory outcomes. Furthermore, all of the pooled analyses were homogeneous, and we performed sensitivity analysis to test the results’ robustness in the presence of heterogeneity. However, our meta-analysis is not without limitations. The small sample size is the main concern. Furthermore, there were variations in the characteristics of the included studies and patients. However, we employed a random effect model to provide more robust results and performed a subgroup account for some of these variations. However, due to limited data, we couldn’t perform further subgroup analyses.