Introduction

Approximately four million hip fractures are projected to occur worldwide in 2025 [9] and the majority are the result of low-trauma events in the elderly, such as falls [21, 27, 34]. During recovery from hip fracture, most patients experience fracture site pain, reduced mobility, and impaired ability to perform routine daily activities, with a large proportion failing to regain their prefracture functional level after 1 year [6, 24, 25]. Slow recovery after hip fracture has been associated with negative consequences [26, 36, 39]. Thus, there is medical need to improve healing and functional recovery after hip fracture.

Most hip fractures are managed surgically with arthroplasty or with open reduction and internal fixation [10, 16, 23, 35]. Healing complications may occur after either type of surgery, with further surgeries more often required in patients receiving internal fixation [4, 23]. If a pharmacologic therapy could improve hip fracture healing, this may reduce the number of revision surgical procedures, and potentially improve other aspects important for recovery, such as early ambulation.

Locally applied pharmacologic therapies are approved in some countries to accelerate bone healing, but the use of systemic agents for this purpose is controversial [7, 17]. Teriparatide (recombinant human parathyroid hormone [PTH] [1–34]; Forteo®/ Forsteo®, Eli Lilly and Company, Indianapolis, IN, USA) is approved for treating osteoporosis and glucocorticoid-induced osteoporosis in patients at high fracture risk [12, 29]. Teriparatide enhanced bone healing in preclinical models [1, 20, 28, 32], and in a controlled clinical trial accelerated radiographic healing of wrist fractures at a dosage of 20 µg/day (post hoc analysis) but not 40 µg/day (primary endpoint) [2]. PTH 1–84 also improved fracture healing in a controlled trial of patients with pelvic fractures [30]. Published case reports suggest that teriparatide accelerates bone healing and improves functional recovery after fracture at various skeletal sites [8, 19, 31, 33, 37, 38].

To further evaluate the potential effect of teriparatide on fracture healing, two identical well-controlled trials were designed to test the hypotheses that teriparatide treatment compared with placebo would (1) decrease the frequency of revision surgery, (2) improve radiographic fracture healing, and (3) improve outcomes including pain control, gait speed, and safety in patients with low-trauma femoral neck fractures repaired with internal fixation.

Patients and Methods

Study Design

Two separate, but identically designed, prospective randomized Phase III double-blind, placebo-controlled, multicenter, multinational studies were initiated (NCT01473589 and NCT01473602) in January 2012. The program included two studies to meet FDA requirements for substantial evidence to support approval for a new indication for fracture healing. The primary endpoint was the proportion of patients with successful fracture healing at 24 months. A patient was considered to have successful fracture healing if the following four components of the composite endpoint were met: (1) no revision surgery; (2) radiographic evidence of healing; (3) functional healing measured by gait speed; and (4) pain control. The trials included features in common with Fixation Using Alternative Implants for the Treatment of Hip Fractures (FAITH) [11], which compares outcomes in patients treated with cancellous screws versus sliding hip screws, including recommendations for preoperative management and central adjudication of clinical events.

Because accrual in both trials was much slower than expected, enrollment was closed in December 2012. The decision was based on the conclusion that it would take much longer (approximately 10 years or more based on the enrollment rate during the first year) to complete the studies and the goal of the studies could not be achieved in a clinically relevant time. The primary endpoint was amended to revision surgery rate alone at 12 months as decided by the Central Adjudication Committee. Because of the reduced sample size (< 10% of the plan) and no difference in trial design and implementation, an a priori decision was made to pool the results of those studies before study completion, and pooling was specified in both protocols. All randomized patients were allowed to complete the study.

Participants

The study enrolled men and postmenopausal women 50 years or older who were ambulatory before sustaining a unilateral low-trauma, femoral neck fracture. Patients received internal fixation surgery (sliding hip screw or multiple cancellous screws) that occurred four or fewer or seven or fewer days after hospitalization for displaced and nondisplaced fractures, respectively. Surgery was not part of the protocol. Other inclusion criteria included no incapacitating disorders except hip fracture and life expectancy of 2 years or more.

Key exclusion criteria included: diseases affecting bone metabolism other than osteoporosis; malignancies (except pretreated carcinoma of the skin or uterine cervix); abnormal baseline serum calcium (≥ 10.6 mg/dL), PTH (> 72 pg/mL), or 25-hydroxyvitamin D (< 9.2 ng/mL); concomitant other bone-active drugs; prior treatment with PTH, strontium, BMPs, or growth factors; infection or prior fracture or surgery at the current fracture site; bone grafting or osteotomies; augmentation using degradable cement, hydroxyapatite-coated implants, or noninvasive interventions; and concomitant major injuries of a lower extremity.

The study was compliant with Good Clinical Practice, the Declaration of Helsinki, and other applicable guidelines. Protocols were approved by the appropriate ethical review boards. Signed informed consent was obtained before study procedures were performed.

Randomization

Within 14 days after internal fixation, patients were randomly assigned to placebo or teriparatide at 1:1 ratio. The randomization was performed separately at each study level, by a table-based randomization scheme with a block size of two using the following factors: site country, fracture type (nondisplaced or displaced), and fixation type (sliding hip screw or multiple cancellous screws) as entered by site personnel.

Description of Experiment

The study consisted of screening (within 14 days of surgery), 6-month treatment, and 6-month observation periods. Blinded treatment was once-daily subcutaneous injection of 20 µg teriparatide or placebo (identical pen injection devices). All patients received supplemental calcium (≤ 1000 mg/day) and vitamin D (≤ 4000 International Units/day). No other osteoporosis or bone-active medications were allowed during the 6-month treatment period.

Variables, Outcome Measures, Data Sources, and Bias

The primary objective was the effect of 6 months of teriparatide treatment versus placebo on the proportion of patients with no revision surgery 12 months after internal fixation of a low-trauma femoral neck fracture. Revision surgery was defined as an additional surgical intervention at the original fracture site as recommended by an investigator or orthopaedic surgeon, except those that were planned at the time of the original procedure. The primary efficacy analysis included the revision surgeries at the original fracture site determined by the Central Adjudication Committee (comprised of four orthopaedic surgeons and one radiologist).

Secondary endpoints included radiographic evidence of fracture healing (determined by the Central Adjudication Committee), pain control (no severe pain during ambulation, during the 24 hours preceding visit, or during weightbearing), recovery of ambulation (gait speed 0.05 m/second in a 4-m portion of a 7-m walk test; canes or walkers were permitted as needed), composite successful fracture healing (including no revision surgery, radiographic fracture healing, pain control, and ambulation), and safety including adverse events as reported by patients or investigators. Efficacy endpoints were assessed at 10 weeks, 6 months, and 12 months. The Central Adjudication Committee evaluated serial radiographs for fracture healing using prespecified definitions. Postrevision efficacy data were not included in the analyses.

Study Population

Of 161 randomized patients in both studies combined from 67 study sites in 18 countries, 159 received the study drug and were included in the full analysis set (78 teriparatide; 81 placebo) (Fig. 1). Four patients (two per treatment group) were judged by the central adjudication committee as having a pertrochanteric fracture. There were no significant differences in patient disposition, including death, discontinuation owing to adverse events, withdrawal, or loss of followup. We did conduct the separated analyses for the data from either study, but because of the small sample size in each (163 patients screened from one trial and 57 patients screened from the other), the results did not change the findings, therefore data from the two trials are not presented separately.

Fig. 1
figure 1

The CONSORT diagram shows the patient disposition in our study.

Full size image

Baseline characteristic were well-balanced between treatment groups including age, sex, fracture type, and fixation type (Table 1). The majority of patients were recruited from Asia or Europe; very few were from the United States.

Table 1 Baseline characteristics
Full size table

Overall median drug exposures were 168 days (range, 4–213 days) and 167 days (range, 1–196 days) in the placebo and teriparatide groups, respectively.

Statistical Analysis

Originally, planned enrollment was 1220 patients per study to detect an 8% clinically meaningful difference based on the composite endpoint between treatments with 80% statistical power. Based on published information [4, 5, 11] and advice from orthopaedic surgeons, 8% absolute difference in the assumed revision-surgery rates at 24 months (eg, 25% placebo versus 17% teriparatide) was considered clinically meaningful.

When the decision was made to suspend the enrollment early, the statistical power was recalculated in the amended protocols before locking the database. With anticipation of approximately 170 randomized patients from both studies, the statistical power was 81% for detection of a 12% absolute difference in the revision surgery rates between treatments (eg, 20% placebo versus 8% teriparatide) using a one-sided test with α = 0.10. This assumption was still in line with previous studies reporting revision surgery rates in patients with hip fracture ranging from 0% to 44% with a weighted pooled risk of 23% (95% CI, 20%–26%) [4, 11].

The proportion of patients with revision surgery at 12 months was compared between treatment groups by comparing Kaplan-Meier estimates of revision surgery at 12 months and using the Z-test statistic based on the chi-square quadratic form using the Greenwood estimator of standard errors [13, 18] of the Kaplan-Meier estimates for each treatment. The null hypothesis was that the proportion of patients with no revision surgery at 12 months was not greater in the teriparatide versus placebo group.

Comparisons between treatment groups for secondary endpoints were made using a Fisher’s exact test. To account for patient dropout in longitudinal continuous efficacy analyses, an analysis of covariance model based on the last observation carried forward method was used in some instances.

Evidence of femoral neck fracture healing also was evaluated for the composite endpoint consisting of four components. A patient had evidence of successful healing if all four components were met by the 12-month visit. Insufficient data for assessment of any component were deemed a fracture-healing failure.

The primary endpoint was tested at a one-sided α level of 0.10. Other endpoints were tested at a one-sided α level of 0.05. No adjustments for multiplicity were applied to comparisons of secondary measures, baseline characteristics, or safety parameters.

The data from patients who were randomized and had received the study drug (modified intent-to-treat population) were included in the analyses.

Results

Revision Surgery

There was no between-group difference in the proportion of patients undergoing revision surgery at 12 months with the numbers available; 14% (11 of 81) of patients treated with placebo and 17% (13 of 78) treated with teriparatide had actual or planned revision surgery as reported by the investigators. Two of 11 patients treated with placebo had revision surgery after their femoral neck fractures healed and therefore were excluded from Central Adjudication Committee analyses. Thus, per the Central Adjudication Committee, 11% (nine of 81) of patients treated with placebo and 17% (13 of 78) treated with teriparatide had actual or planned revision surgery. The proportion of patients with no revision surgery at 12 months (primary endpoint) was 88% (90% CI, 0.79–0.93) in the placebo group and 84% (90% CI, 0.75–0.90) in the teriparatide group per the Central Adjudication Committee (p = 0.743; Table 2). No treatment group differences were noted in assessments at any time for either investigator or Central Adjudication Committee revision surgery.

Table 2 Revision surgery as assessed by the Central Adjudication Committee and reported by the investigator
Full size table

The reasons for investigators to recommend revision surgery and the type of surgeries were variable, and no major difference were noted between the groups (Table 3). The most common mode of implant failure as reviewed by the Central Adjudication Committee was varus collapse with screw backout in four of 11 patients treated with placebo and seven of 13 treated with teriparatide.

Table 3 Revision surgeries reported by investigators
Full size table

Radiographic Assessment of Fracture Healing

There were no differences in radiographic healing between the teriparatide and placebo groups at 10 weeks, 6 months, or 12 months (Table 4), with the numbers available. Based on the available radiographs, 75% (61 of 81) of patients treated with placebo and 73% (53 of 78) treated with teriparatide had radiographic evidence of fracture healing at 12 months (odds ratio [OR], 0.89; 90% CI, 0.46–1.72; p = 0.692). Twelve patients in each group had no radiographic evidence of healing at study completion or early discontinuation, and the remaining patients had no radiograph data (Table 4). In nearly all patients, there was no callus formation at the fracture site during or after fracture healing.

Table 4 Radiographic evidence of healing as assessed by the Central Adjudication Committee
Full size table

At 12 months or the time of the last available radiographs, the Central Adjudication Committee determined that 14% (10 of 71) of patients treated with placebo and 23% (16 of 69) treated with teriparatide had radiographic healing complications, including femoral neck shortening (six with placebo, 14 with teriparatide), nonunions (three with placebo), malunions (one with placebo), implant breakage or failure (one with teriparatide), and screw migration (two in each group). Avascular necrosis was noted in three patients treated with placebo and five treated with teriparatide.

Other Outcomes

Pain Control

With the numbers available, there were no differences between treatment groups in any of the measured pain scores at any time (Table 5) (data not shown). Among the patients who reported analgesic medication use at 6 months, 48% (33 of 69) of patients treated with placebo and 34% (23 of 67) treated with teriparatide took analgesic medications (p = 0.077), and the rest of the patients reported not taking analgesic medications. At 12 months, 43% (29 of 67) of patients treated with placebo and 31% (19 of 61) treated with teriparatide were still using analgesic medications (p = 0.108).

Table 5 Fracture site pain and gait speed at 12 months (last observation carried forward)
Full size table

Gait Speed

Among the patients who were able to take the gait speed test, 73% (47 of 64) treated with placebo and 89% (51 of 57) treated with teriparatide could ambulate (gait speed > 0.05 m/second) without decline at 12 months (OR, 3.07; 90% CI, 1.20–8.49; p = 0.021) (Table 5). The proportions of patients treated with placebo and teriparatide who regained their prefracture ambulatory status were 58% (43 of 74) and 67% (49 of 73), respectively (OR, 1.47; 90% CI, 0.79–2.73; p = 0.169) at 12 months.

Composite Measure of Successful Fracture Healing

There were no differences between groups in the proportion of patients who met all four components of successful fracture healing (no revision surgery, radiographic fracture healing, pain control, and ambulation) at 12 months by either Central Adjudication Committee assessment (58% [46 of 79] with placebo versus 59% [46 of 78] with teriparatide; OR, 1.03; 90% CI, 0.58–1.85; p = 0.527) or investigator reporting (57% [46 of 81] with placebo versus 59% [46 of 78] with teriparatide; OR, 1.09; 90% CI, 0.61–1.95; p = 0.453).

Safety

There were no differences between groups in the proportion of patients reporting adverse events (Table 6), with the numbers available. During the treatment period, three patients in the placebo group and four in the teriparatide group discontinued participation in the study owing to adverse events (gastrointestinal disorder, lethargy, and erythema multiforme in the placebo group; asthenia, pneumonia, subdural hematoma, and blood glucose fluctuation in the teriparatide group, but none was drug-related per investigators).

Table 6 Adverse events (full-analysis set) reported by investigators
Full size table

Two deaths occurred in the teriparatide group during the treatment period (pneumonia, subdural hematoma), and one death occurred in the placebo group during the observation period (myocardial infarction). Per Central Adjudication Committee assessment, these three deaths were not related to the study drug or the original hip fracture.

Discussion

In current practice, more patients with hip fractures are treated with arthroplasty, which is largely attributable to the lack of viable options to repair hip fractures in a manner that maintains the anatomy (and favorable biomechanics) of a normal hip. We hypothesized that adjunctive therapy to internal fixation of hip fractures with teriparatide could provide a biologic advantage compared with surgical internal fixation alone.

We initiated two large Phase III trials to evaluate whether teriparatide could improve clinical outcomes in patients with a low-trauma femoral neck fracture repaired with internal fixation. However, during the first 12 months of recruitment, despite substantial effort, only 161 patients had been randomized in both trials, which were approximately 7% of the planned study cohort, and few of these patients were from the study sites in the United States. The study sponsor decided to close enrollment early, based on the conclusion that the goal of the studies could not be achieved in a clinically relevant time. Patients who already had enrolled continued in the blinded studies until they completed their 12-month visits.

The inability to enroll patients was multifactorial, including the methodologic aspect requiring daily injections of a study drug, initiation of treatment shortly after the hip fracture in an elderly population, early mobilization of patients in different healthcare systems, and the increasing preference of arthroplasty for displaced femoral neck fractures. The outcome of our program highlights the need to further improve methods for conducting studies in this population.

Because the cohort size was smaller than planned, the studies were substantially underpowered to detect the treatment effect included in the original study protocols. Since the treatment effect size was not known with certainty, the study was continued to explore whether a larger treatment effect could have been present. Therefore, the results presented here are exploratory and should be interpreted with caution. However, the results from the pooled studies are unlikely to overestimate a treatment effect as occurred in other trials that were terminated early [3, 14].

Pooled results from the 159 patients who received study drugs showed that teriparatide, given for 6 months, did not reduce the frequency of revision surgery or improve radiographic fracture healing at 12 months compared with a placebo. Although there were no between-group differences in measures of pain control, a greater proportion of patients treated with teriparatide was able to ambulate without worsening or regained prefracture ambulatory status at 12 months. However, major differences between groups regarding the functional endpoints were not observed at all times.

The overall proportion of patients undergoing revision surgery in this study was 15% (24 revision procedures among 159 study patients), which is at the lower end of revision surgery frequencies reported after internal fixation [4, 5]. Causes for revision surgery in the teriparatide and placebo groups in our study were similar to those reported for internal fixation of hip fractures [22]. The relatively low revision surgery rate in our study may be the result of most fractures being nondisplaced (73%) and the relatively healthy patient population (eg, community-dwelling, ambulatory, no or mild systemic disease, median age 70 years prefracture). Three deaths were reported in our study, but none was related to study treatment or hip fracture. This is consistent with this being a healthier, more-ambulatory patient population than that observed in general clinical practice [15], as the study entry criteria required that patients have no other incapacitating conditions that are common in the elderly and also have the ability to consent for the study.

Systemic therapy with teriparatide has been shown to improve healing in preclinical fracture models [1, 20, 28, 32], in clinical studies of wrist fractures [2], and in case reports [8, 19, 31, 33, 37, 38]. However, the results from these pooled trials revealed no treatment effect of teriparatide versus placebo in the rates of revision surgery or radiographic fracture healing in femoral neck fractures. Reasons for the lack of treatment effect are speculative. Possibly, teriparatide does not have a treatment effect in this context or the sample size was too small to detect a difference in revision surgery. Additionally, plain radiographs obtained at the selected intervals may be too crude for evaluation of femoral neck fracture healing, but it was a tradeoff with consideration of limiting radiation exposure during the study.

Findings from a negative trial are reported here, which improves transparency on the design and performance in pharmacologic trials with patients with hip fracture and potentially may limit publication bias. Additionally several study design elements are reported including methodologic rigor in design and several safeguards that ensured patient randomization and blinding, independent blinded central clinical outcomes adjudication, and use of a patient-important primary outcome measure. The study also is one of only a few well-controlled studies of clinical outcomes with a systemically active biologic drug in patients with fracture. Although preclinical studies using PTH in animal models are promising [1, 20, 28, 32] and sporadic case reports of humans [8, 19, 31, 33, 37, 38] showed an anecdotal benefit of teriparatide in various fractures, there were no well-controlled clinical trials to support the hypothesis that teriparatide improve fracture healing and our study with a much-reduced sample size showed no such benefit. Thus, the effect of teriparatide on fracture healing remains uncertain. The safety results from this study did not change the benefit-risk assessment of teriparatide, an approved bone anabolic agent for treatment of osteoporosis in men and in women who are postmenopausal at high fracture risk. Therefore teriparatide remains a treatment option for elderly patients with low-trauma hip fractures who are at high risk for subsequent fractures.