Background: Osteoporosis drugs might affect fracture-healing. We therefore studied the effects of teriparatide in comparison with risedronate on recovery after pertrochanteric hip fractures.
Methods: The study was a randomized, multicenter, active-controlled, 78-week trial comparing teriparatide (20 μg/day) with risedronate (35 mg/week) initiated within 2 weeks after fixation of a low-trauma pertrochanteric hip fracture (AO/OTA 31-A1 or 31-A2). The main inclusion criteria were a bone mineral density T-score of ≤−2.0 and 25-OH-vitamin D of ≥9.2 ng/mL. During the first 26 weeks, patients received study medication with oral or injectable placebo plus calcium and vitamin D in a double-blinded fashion. Secondary (Timed Up-and-Go [TUG] test, hip pain, Short Form [SF]-36 health status, and safety) and exploratory (radiographic outcomes and ability to walk) 26-week end points are reported.
Results: Of the 224 patients who were randomized, 171 (86 teriparatide, 85 risedronate) were included in the analysis. The mean age was 77 ± 8 years, 77% were female, and 26% had a prior history of low-trauma fracture. The teriparatide group completed the TUG test in a shorter time at 6, 12, 18, and 26 weeks (differences of −5.7, −4.4, −3.1, and −3.1 seconds, respectively; p = 0.021 for the overall difference). They also reported less pain on a visual analog scale immediately after the TUG test at 12 and 18 weeks (adjusted absolute differences of 10.6 and 11.9 mm, respectively; p < 0.05). There were no significant between-group differences in the SF-36 score, Charnley hip pain score, ability to walk, or use of walking aids during follow-up. Radiographic healing at 6, 12, and 26 weeks, mechanical failure of the implant (teriparatide, 7; risedronate, 8), loss of reduction (teriparatide, 2; risedronate, 4), and nonunion (0 cases) were not significantly different. Mild hypercalcemia and hyperuricemia were more frequent with teriparatide.
Conclusions: Teriparatide was associated with less pain and a shorter time to complete the TUG test between 6 and 26 weeks compared with risedronate. Other fracture-recovery outcomes were similar. The results should be interpreted with caution as these were secondary end points.
Level of Evidence: Therapeutic Level II. See Instructions for Authors for a complete description of levels of evidence.
Approximately 1 in 6 women in North America will experience a hip fracture in their lifetime1. Patients 55 years or older with hip fractures have an overall 1-year mortality rate of up to 39%2, and the survivors have a decreased life expectancy, loss of physical function, and impaired quality of life2-4, mainly because of frailty5,6. Hip fractures are often associated with osteoporosis, and they are strong risk factors for fractures at other skeletal sites7,8.
The 2 major categories of pharmacological treatment for osteoporosis are antiresorptive and bone anabolic medications. Approved antiresorptives include bisphosphonates, selective estrogen receptor modulators, denosumab, and strontium ranelate. Teriparatide (recombinant human parathyroid hormone [PTH]-1-34) is the only currently approved anabolic medication for the treatment of osteoporosis. There is great interest in the effect of these agents on bone repair and fracture-healing in humans. To our knowledge, antiresorptive agents have not demonstrated any deleterious effects on fracture-healing9,10, and they have shown positive effect in some animal studies11,12. In animal studies, teriparatide enhanced bone-healing13-16, and post-hoc analysis of a randomized trial showed that teriparatide accelerated the time to cortical continuity of distal radial fractures at a dosage of 20 μg/day, but not at 40 μg/day17. A post-hoc subgroup analysis showed a positive dose-related effect on early radiographic callus formation18. A similar study on proximal humeral fractures failed to show any effect19. PTH-(1-84) seemed to improve fracture-healing in women with osteoporosis and pelvic fractures20, although that study had important design limitations21. Case reports and cohort studies also suggest that teriparatide accelerates bone-healing22.
We present the 26-week results of a randomized, double-blind trial comparing the effect of teriparatide with that of risedronate on functional and radiographic outcomes after a pertrochanteric hip fracture in men and postmenopausal women with low bone mass. The initial study design was to use physical function (the Timed Up-and-Go [TUG] test) as the primary outcome variable. However, in the final design, in order to more accurately calculate the sample size for the study, the TUG test became an important secondary end point and the primary outcome variable was the change from baseline to 18 months in bone mineral density (BMD) in the lumbar spine; these BMD results will be published elsewhere. Here we report a preplanned analysis of all secondary end points and exploratory variables related to recovery after fracture recovery and to drug safety.
Materials and Methods
From April 2009 to February 2014, hospital-based physicians from 17 countries in North America, Mexico, and Europe who were experienced in treating patients with hip fractures screened over 2,400 patients for eligibility. The majority could not enter the trial, mainly because of dementia, patient decision, or major comorbidities. A total of 389 patients with a recent pertrochanteric hip fracture were enrolled. Key eligibility criteria are summarized in Table I.
This study was a Phase-IV, randomized, multicenter, active-controlled trial with 3 periods: (1) a screening phase lasting a maximum of 14 days from surgery to randomization; (2) a double-blind, double-dummy treatment period from randomization to the 26-week visit; and (3) an open-label treatment period during which patients continued treatment up to 78 weeks with the same study drug to which they had been randomized. Here we present results from the double-blind phase of the trial.
The trial was registered at ClinicalTrials.gov (NCT00887354). It was approved by the responsible institutional review boards at each center and was conducted in accordance with the Declaration of Helsinki, good clinical practices, and applicable laws and regulations. Written informed consent was obtained from all patients before any screening procedures.
At the screening visit, all patients started oral supplements of calcium (500-1,000 mg/day) and vitamin D (800 IU/day) and discontinued any ongoing osteoporosis drug. If the baseline serum 25-OH-vitamin D level was between 9.2 and 16 ng/mL, patients received a single oral loading dose of 100,000 IU of vitamin D. At the baseline visit, patients were randomized in a 1:1 ratio to teriparatide (20 μg/day subcutaneous teriparatide injection plus weekly oral placebo) or risedronate (daily subcutaneous placebo injection plus 35 mg/week oral risedronate). Treatment assignment was stratified by the type of fracture (AO/OTA 31-A1 and 31-A2)23 and determined by a computer-generated random sequence. All patients and investigators were blinded to the study treatments.
Secondary Outcomes (Efficacy)
Self-reported hip pain was assessed immediately after completion of the TUG test using a linear visual analog scale (VAS) on which 0 mm represented no pain and 100 mm, the most severe pain possible26. In addition, a modified Charnley pain score was used to estimate the worst hip pain in the preceding 24 hours (see Appendix)27.
Patient-rated health status was estimated with the Short Form (SF)-36 questionnaire27-30, self-administered by the patients at the randomization visit and the 6, 12, 18, and 26-week visits. Patients also completed the questionnaire during the screening phase to assess their status during the 4 weeks before the fracture (recall SF-36 value).
The sequence for the patient-reported outcomes was (1) SF-36 survey, (2) TUG test, (3) VAS pain assessment, (4) modified Charnley hip pain score, and (5) ability to walk.
Secondary Outcomes (Safety)
Safety analyses were conducted on all patients who received ≥1 dose of medication. These included treatment-emergent adverse events (TEAEs); incident clinical fractures; analgesic use for hip pain; serum levels of 25-OH-vitamin D, calcium, and uric acid; clinical chemistry and hematology; and vital signs.
Hip radiographs were centrally adjudicated by 2 independent radiologists who were blinded to treatment assignment (Synarc). Adequate fracture reduction was defined as a femoral neck-shaft angle of 15° valgus to 10° varus relative to the contralateral, unfractured hip on an anteroposterior radiograph, posterior angulation of <20° on a lateral radiograph, and proximal fragment alignment with or superior to the distal fragment on an anteroposterior radiograph. The lag screw position was considered adequate on the basis of a tip-apex distance (TAD) of <20 mm31 together with the criteria defined by Parker32. Radiographic healing was assessed by conventional anteroposterior and lateral radiographs at 6, 12, and 26 weeks. The variables for evaluating healing were predefined and included a combination of cortical bridging or softened cortical continuity, disappearance of the fracture line, stable fracture alignment compared with the preceding visit, and/or progressive sclerosis at the fracture site.
Mechanical failure of the implant included cutting-out of the screw from the femoral head (defined as projection of the screw from the femoral head by >1 mm32); excessive migration of the tip of the screw within the femoral head (change in TAD of >6 mm); varus collapse with a decrease in femoral neck-shaft angle of >10°; excessive progression of offset; a sliding plate that was bent, broken, or pulled off the shaft; or loosened or broken cortical screws. Fracture nonunion was defined as the absence of radiographic healing combined with hip pain (Charnley categories 2 to 5; see Appendix) and inability to walk without assistance (nonfunctional ambulatory or nonambulatory) at 26 weeks.
Ability to Walk
Ambulatory functioning was assessed at all post-baseline visits using 4 categories: community, household, nonfunctional, and nonambulatory33. In addition, use of walking aids was assessed using the following categories: no aid, 1 cane, 1 crutch, 2 canes, 2 crutches, a walker, a person as support, and not walking.
Sample size estimation was based on the primary outcome of lumbar spine BMD after 18 months (results to be presented elsewhere). A difference of 0.023 g/cm2 in BMD between the treatment groups was considered to be clinically important34. Assuming a common standard deviation of 0.047 g/cm2 in each group, 76 patients per treatment arm were planned to yield 85% power at a significance level of 0.05 (2-sided test). Allowing for a 30% loss, we planned to enroll 109 patients per treatment arm. Efficacy analyses for effects on functional recovery were conducted on the full analysis set (FAS), which follows the intent-to-treat principle and includes data from all patients who received ≥1 dose of the trial drug (active or placebo) and had ≥1 follow-up visit that included any efficacy assessment.
Baseline characteristics were summarized descriptively. The analysis of the TUG test, patient-rated health status, and hip pain VAS was performed with a mixed-effects model for repeated measures (MMRM), which accounts for data missing at random by using the correlation of observations within each patient and without the need of any explicit imputations35. Treatment, visit, treatment-visit interaction, and type of fracture were always included as fixed effects in the MMRM. Other variables were included as prespecified, depending on the end point. TUG test results were derived from log-transformed data. All non-missing data from the FAS were analyzed. At each follow-up visit, covariate-adjusted (least squares) mean changes from baseline with standard errors were derived for the 2 treatments and p values were reported for their differences. Hip pain using the Charnley score and radiographic healing were analyzed with logistic regression with repeated measures to model the probability of a positive outcome. Frequencies of patients with surgical complications and TEAEs were compared using the Fisher exact test. All data were analyzed using SAS software (version 9.4).
Patient Disposition and Baseline Characteristics
Overall, 224 patients were randomized, and 171 (86 teriparatide, 85 risedronate) contributed to the efficacy analysis. Ninety-four patients (42%) did not complete the 26-week visit (Fig. 1). Patient characteristics and characteristics of the index hip fracture and surgery are summarized in Tables II and III. Details regarding the implants are reported in the Appendix.
Overall, the time required to complete the TUG test was shorter with teriparatide than with risedronate at 6, 12, 18, and 26 weeks (differences of −5.7, −4.4, −3.1, and −3.1 seconds, respectively; p = 0.021 for the overall between-treatment difference). The least-squares mean times with teriparatide and risedronate were 26.4 and 32.1 seconds (p = 0.029) at 6 weeks, 20.2 and 24.7 seconds (p = 0.026) at 12 weeks, 18.0 and 21.2 seconds (p = 0.074) at 18 weeks, and 16.8 and 20.0 seconds (p = 0.058) at 26 weeks, respectively (Fig. 2; see Appendix). Post-hoc analysis of unadjusted data showed similar results (data not shown).
Self-Reported Hip Pain
VAS-assessed hip pain during the TUG test was reduced with teriparatide compared with risedronate (p = 0.032 for the between-treatment difference). The adjusted absolute difference was 10.6 mm at 12 weeks (p = 0.041), 11.9 mm at 18 weeks (p = 0.023), and 10.1 mm at 26 weeks (p = 0.054) (Fig. 3; see Appendix). Logistic regression analysis found no significant difference in the number of patients experiencing a satisfactory Charnley hip pain score between treatment arms at any time point (see Appendix).
Patient-Rated Health Status
There was no significant between-treatment difference in the change from baseline for any of the 8 domains of the SF-36 at any time point (Table IV). Depending on the domain, values returned to the pre-fracture status by 12 to 18 weeks. Age and sex were associated with the physical function domain results. In contrast, the type of fracture and adequacy of the reduction were not associated with any of the domain results (data not shown).
Treatment compliance (defined as not having missed >25% of the study medication at 2 consecutive visits) was 98.6% for teriparatide and 100% for risedronate. This information was not fully available in 37 patients. Fewer patients receiving teriparatide than risedronate died or reported serious TEAEs or clinical fractures, but the differences were not significant (Table V). There were 3 new fractures in the teriparatide group and 8 in the risedronate group (p = 0.14); no clinical vertebral fractures were diagnosed. At 26 weeks, the overall rate of analgesic use was 21% and was similar in both treatment arms (p = 0.85). Hyperuricemia and hypercalcemia were more frequent with teriparatide at the 6 and 26-week follow-up visits, respectively (Table V). Serum alkaline phosphatase was higher (p = 0.01) and serum cholesterol was lower (p = 0.04) with teriparatide (Table V). Other laboratory parameters and vital signs were not significantly different between the 2 treatment arms.
At 26 weeks, 100% of the 62 patients with data who received teriparatide and 98% of the 64 with data who received risedronate had achieved radiographic healing. By 12 weeks, the rate had been 89% to 90% in both groups. There were no significant differences in the frequency of implant failure or loss of reduction between the 2 treatment arms (Table VI). No patient showed fracture nonunion (Table VI).
Ability to Walk
There were no significant differences between treatment groups at any follow-up period with respect to the ability to walk or the use of walking aids. By 26 weeks, all patients were ambulatory except for 1 in the teriparatide arm. No walking aid was needed by 36 patients (58%) receiving teriparatide and 36 (55%) receiving risedronate (p = 0.8).
The overall trial was designed to evaluate a primary variable not related to recovery from fracture. However, judging by secondary outcomes (TUG test and pain during the test), teriparatide appears to be associated with a better early functional outcome after pertrochanteric hip fractures compared with risedronate. This earlier recovery of function might possibly reflect an earlier mature fracture union due to teriparatide, even though the study was not designed to show that per se. The alternative that teriparatide had other non-fracture-related effects appears unlikely, lacking biological plausibility. A third possibility would be that risedronate had a negative effect, which appears unlikely, as bisphosphonates have been shown not to impair healing in comparison with placebo36-38, with the potential exception of distal radial fractures39, and likely improve implant fixation40. A positive effect of teriparatide on human fracture recovery would not be surprising, considering the positive effect on bone-healing in various animal models13-16,21,41.
Other secondary efficacy outcomes (SF-36, Charnley score, and ability to walk) showed no significant differences. However, the TUG test might be considered more clinically relevant than the SF-36 in these patients, as it better predicts future walking activity and activity level25, and a rapid return to function is likely to diminish the rate of late complications and falls24,42. Moreover, patient-reported health-related quality of life is a consequence of both hip function and other unrelated factors, which add sources of variation. We therefore consider the TUG test not only more important, but also more sensitive, than the other secondary outcome variables. The absolute differences in the time to complete the TUG test between the 2 treatments were larger than the 1.4 to 2.4 seconds considered to be the minimal clinically important difference for this test43,44. Still, these results are only hypothesis-generating.
The absence of significant differences in radiographic healing is notable. However, the radiographic variables were coarse and appeared to miss the relevant time points, as no fractures were radiographically “healed” at the first examination (at 6 weeks) and the vast majority were at the second time point (at 12 weeks). Moreover, development of callus or disappearance of the fracture line is unreliable for assessing clinical healing of hip fractures45,46. Given the small number of radiographs that can be made to monitor fracture-healing—especially in frail subjects with hip fractures—this end point was exploratory, and not among the predetermined secondary outcomes.
To our knowledge, there is no trial that has unequivocally demonstrated improved fracture-healing with teriparatide in humans to date. A proof-of-concept, placebo-controlled trial has shown accelerated healing of wrist fractures, but the primary outcome of that study, using the higher teriparatide dose, was not significant17,18. A similar trial on proximal humeral fractures could not show any effect19. A trial on pelvic fractures in women with osteoporosis treated using human PTH-(1-84) showed an extraordinary acceleration of healing20. However, there is a risk of bias in that study, as the investigator was aware of the planned treatment when deciding whether or not to include a patient. Moreover, the TUG test time appeared to be remarkably long in that investigation21. A retrospective study of unstable pertrochanteric fractures suggested a reduction of approximately 3 weeks in the time to radiographic union with teriparatide compared with unblinded controls47. Apart from these studies, there are only case series available, which often suggest a beneficial effect of osteoporosis treatment on fracture-healing. The adjustment for confounders in those studies was often insufficient because of limited data or sample size21,48.
To our knowledge, this study is the first prospective, randomized, blinded study to analyze osteoporosis drug effects on functional recovery after a hip fracture, which is a difficult clinical model to assess49. Methodological strengths are the predefined analysis plan, using validated instruments to measure functional outcomes and pain, and the central, blinded reading of the radiographs using prespecified diagnostic criteria. Biases were avoided through a double-dummy design and complete blinding. We allowed patients who were already using osteoporosis drugs to participate.
Limitations include the fact that the study was primarily designed to measure the effects on spinal BMD and that fracture recovery was a secondary outcome. Thus, all variables either were testing secondary hypotheses or were exploratory. We had not formally ranked the SF-36, TUG test, and pain for importance, so our emphasis on the TUG test might have been biased by the outcome. The strict eligibility criteria for the trial were aimed at reducing the variability in the patient cohort, but it created a selection bias toward a healthier population and thus limited external validity for more impaired subjects with comorbidities such as dementia or a history of malignancy. Finally, we had a very high dropout rate and low compliance with visits and study procedures. This is not unexpected in frail, elderly subjects with a severe fracture, and it stresses the difficulties of performing randomized clinical trials involving frequent, cumbersome postoperative assessments in elderly patients with fractures in whom transportation issues are substantial.
In conclusion, the improved functional performance and reduced hip pain with teriparatide concur with previous studies suggesting, but not unequivocally showing, that teriparatide may improve acute fracture-healing.
Tables showing the modified Charnley pain score and measured pain categories, details of the surgical implants, and measured TUG test times and hip pain levels are available with the online version of this article as a data supplement at jbjs.org.
Note: Medical writing support was provided by Dr. Pradnya Kulkarni from Trilogy Writing and Consulting GmbH, Frankfurt, Germany. The study had the following principal investigators (with ≥1 patient enrolled). Austria: C.M. Blauth; Canada: A. Cheung; Croatia: V. Bozikov; Czech Republic: T. Philipp, J. Pesek, J. Stepan; Denmark: L. Borris; J.-E. Jensen, N.K. Jensen, S. Overgaard; Finland: H. Aro, J. Leppilahti; France: C. Benhamou, G. Cormier, M. Laroche, E. Lespessailles, H.O. Ollagnon, G. Rajzbaum, A. Talha; Germany: A. Berner, K. Dresing, S. Ruchholtz; Greece: G. Kapetanos, V. Lykomitros, K. Malizos, N. Papaioannou; Ireland: B.J. Walsh; Italy: C. Corradini, M. D’Arienzo, M. Innocenti, V. Patella, U. Tarantino; Mexico: F. Cons-Molina, P.A. García-Hernández; Norway: F. Frihagen; Spain: P. Caba, J.R. Caeiro, P. Cano, M.A. Froufe, E. Guerado, R. Larrainzar, J. Malouf; Sweden: E. Waem, S. Ponzer; United Kingdom: T. Chesser, J. Cobb; and United States: J. Bibiloni, P. Candelora, D. Cole, E. Kurland, M. Lillestol, R. Recker, C. Recknor, K. Shrock, R. Zura.
Disclosure: The trial was funded by Eli Lilly, and two of the authors are employees of Eli Lilly. The sponsor designed the protocol with advice from external advisors and was responsible for the quality control of data collected and the statistical analysis. On the Disclosure of Potential Conflicts of Interest forms, which are provided with the online version of the article, one or more of the authors checked “yes” to indicate that the author had a relevant financial relationship in the biomedical arena outside the submitted work and “yes” to indicate that the author had other relationships or activities that could be perceived to influence, or have the potential to influence, what was written in this work.
- Copyright © 2016 by The Journal of Bone and Joint Surgery, Incorporated