- Top of page
- Materials And Methods
Zoledronic acid reduces the risk of death by 28% after hip fracture, but the mechanisms are not known. This exploratory analysis sought to identify potential pathways for the reduction in mortality with zoledronic acid after hip fracture. This was a retrospective analysis of a randomized, controlled trial. Patients with recent hip fracture (n = 2111) were treated with zoledronic acid or placebo infusion yearly, as well as calcium and vitamin D supplementation. Causes of death were adjudicated by a blinded central review committee. Baseline comorbidities, events occurring during the study period, including subsequent fracture, change in bone density, infections, cardiovascular events, arrhythmias, and falls, were included in multivariable analyses. In a model adjusted for baseline risk factors, zoledronic acid reduced the risk of death by 25% [95% confidence interval (CI) 0.58–0.97). The effect was consistent across most subgroups. Subsequent fractures were significantly associated with death (hazard ratio 1.72, 95% CI 1.17–2.51) but explained only 8% of the zoledronic acid effect. Adjusting for acute events occurring during follow-up eliminated the death benefit, and zoledronic acid–treated subjects were less likely to die from pneumonia (interaction p = .04) and arrhythmias (interaction p = .02) than placebo-treated subjects. Only 8% of zoledronic acid's death benefit is due to a reduction in secondary fractures. Zoledronic acid may have an effect on cardiovascular events and pneumonia. Further studies of zoledronic acid in other acute illnesses may be warranted. Copyright © 2010 American Society for Bone and Mineral Research
- Top of page
- Materials And Methods
Hip fractures are associated with substantial excess mortality in older adults.1 The risk of death in the first 6 months after hip fracture, adjusted for multiple risk factors, is up to 7-fold greater for women and 40-fold greater for men compared with those without fracture.2 In many,2, 3 but not all studies,1 an increased risk of mortality persists for at least 5 years after the hip fracture. With the aging population, the risk of death owing to hip fractures is expected to increase.4 These data highlight the urgent need to identify strategies to reduce the risk of mortality following hip fractures.
The intravenous bisphosphonate zoledronic acid given as an annual 5 mg infusion after hip fracture reduced the risk of clinical fracture by 35% and the risk of death by 28% compared with placebo.5 Although osteoporotic fractures (including vertebral fractures) have been shown to be associated with increased mortality,6 this significant mortality reduction was unexpected in a trial powered to show a reduction in clinical fractures, and this was the first trial to demonstrate that an osteoporosis therapy can reduce mortality significantly. This mortality benefit was manifested after the first year of treatment and persisted after adjustment for demographic and baseline variables. However, the pathway(s) or underlying mechanism(s) producing the benefit remain unknown.
Previous research has suggested several potential physiologic mechanisms for bisphosphonates to affect mortality after hip fracture. A hip fracture is associated with substantial increases in proinflammatory cytokines such as interleukin 6 (IL-6), IL-8, IL-10, C-reactive protein, tumor necrosis factor-α (TNF-α), and vascular endothelial growth factor (VEGF).7–13 These increases can be measured in bone, in plasma, and in cerebrospinal fluid and may persist for months in some patients. Persistently elevated proinflammatory markers are associated with perioperative complications and worse long-term functional outcomes8–10 and are hypothesized to explain the excess infections and cardiovascular events observed in the 1 to 2 years after hip fracture. Intravenous bisphosphonates have immunomodulatory effects, affecting both dendritic cells and γδ T cells,14 with variable changes in proinflammatory cytokines, TNF-α, and VEGF.15 Clinically, this effect is likely responsible for the acute-phase reaction of fever and myalgia, as well as the transient lymphocytopenia and increased C-reactive protein seen in patients receiving intravenous bisphosphonates.16 With long-term administration, an anti-inflammatory effect has been observed in animal models of arthritis and colitis and in case series of patients with rheumatoid arthritis.15 In cancer patients treated with intravenous pamidronate, a significant decrease in angiogenic factors such as VEGF also has been documented.17 Therefore, by modifying the inflammatory, angiogenic, and immunologic effects of hip fracture, intravenous bisphosphonates potentially could mediate a mortality benefit through decreased cardiovascular disease, infections, or malignancies.
A better understanding of the patient and disease factors that influence the reduction in mortality of zoledronic acid may inform both clinical practice and future research. Therefore, the purpose of this retrospective analysis was to determine how much of the mortality effect could be explained by the reduction in fracture incidence and how much was due to other bisphosphonate effects or other medical conditions. Specifically, we hypothesized that the reduction in mortality risk could be related to secondary fracture reduction, reduction in cardiovascular disease, or reduction in infections.
- Top of page
- Materials And Methods
Baseline characteristics, including risk factors for mortality, were similar in the zoledronic acid and placebo groups (Table 1). The cumulative mortality rates were 9.6% in the zoledronic acid group and 13.3% in the placebo group [hazard ratio (HR) 0.72; 95% confidence interval (CI) 0.56–0.93; P = .01]. The mortality curves did not separate until approximately 16 months after the first infusion (Fig. 1), which was approximately 12 months after the first separation in clinical fracture incidence rates (approximately month 4).
Table 1. Baseline Characteristics for Zoledronic Acid and Placebo-Treated Subjects
| ||Zoledronic acid, N = 1054||Placebo, N = 1057||P valuea|
|Female (%)||810 (76.8)||796 (75.3)||0.41|
|Age (%)|| || ||0.25|
| <65||172 (16.3)||191 (18.1)|| |
| 65–74||305 (28.9)||268 (25.3)|| |
| 75–84||440 (41.7)||447 (42.3)|| |
| ≥85||137 (13.0)||151 (14.3)|| |
|Race (%)|| || ||0.76|
| Caucasian||962 (91.3)||960 (90.8)|| |
| Other||92 (8.7)||97 (9.2)|| |
|Geographic region (%)|| || ||0.90|
| North America||297 (28.2)||313 (29.6)|| |
| Latin America||132 (12.5)||131 (12.4)|| |
| Western Europe||356 (33.8)||353 (33.4)|| |
| Eastern Europe||269 (25.5)||260 (24.6)|| |
|BMI (kg/m2)|| || ||0.90|
| <19||79 (7.5)||72 (6.8)|| |
| 19–25||502 (47.6)||506 (47.9)|| |
| >25||441 (41.8)||450 (42.6)|| |
|SPMSQ Score (%)|| || ||0.78|
| 0||524 (49.7)||523 (49.5)|| |
| >0–2||283 (26.9)||285 (27.0)|| |
| >2||168 (15.9)||180 (17.0)|| |
| Missing||79 (7.5)||69 (6.5)|| |
|Femoral neck T-score|| || ||0.94|
| −2.5 or less||448 (42.5)||437 (41.3)|| |
| Greater than −2.5 to −1.5||360 (34.2)||374 (35.4)|| |
| Greater than −1.5||122 (11.6)||121 (11.4)|| |
| Missing||124 (11.8)||125 (11.8)|| |
|Hip fracture type (%)|| || ||0.78|
| Intertrochanteric||335 (31.8)||342 (32.4)|| |
| Subtrochanteric||48 (4.5)||57 (5.4)|| |
| Subcapital||221 (21.0)||211 (20.0)|| |
| Femoral neck||354 (33.6)||373 (35.3)|| |
| Other||76 (7.2)||74 (7.0)|| |
|Prior residence (%)|| || ||0.35|
| Private home||945 (89.7)||928 (87.8)|| |
| Assisted living||60 (5.7)||61 (5.8)|| |
| Skilled nursing facility||31 (2.9)||44 (4.2)|| |
| Other||18 (1.7)||24 (2.3)|| |
|Hypertension (%)||543 (51.5)||567 (53.6)||0.34|
|Diabetes (%)||173 (16.4)||160 (15.1)||0.44|
|Stroke (%)||181 (17.2)||190 (18.0)||0.65|
|Coronary artery disease (%)||209 (19.8)||224 (21.2)||0.45|
|Tachyarrhythmia (%)||61 (5.8)||79 (7.5)||0.14|
The investigator-reported primary causes of death are listed in Table 2. There was a trend toward more cardiac and respiratory failure–related deaths in the placebo group (Fig. 2, top panel). The blinded central adjudication committees further classified arrhythmias that led to hospitalization and cardiac death based on available source documentation. However, 58 of 88 (65.9%) deaths coded by the site investigator as cardiac could not be classified by the adjudication committee owing to inadequate source documents; e.g., deaths occurring at home frequently lacked evaluable medical records. For the cardiac events that could be adjudicated (n = 29), 11 for zoledronic acid and 18 for placebo), 3 of 29 (10.3%) were due to sudden cardiac death, 10 of 29 (34.5)% were due to pump failure, 11 of 29 (37.9%) were due to fatal myocardial infarctions, 2 of 29 (6.9%) were due to fatal arrhythmias, 1 of 29 (3.5%) was due to other cardiac deaths, and 7 of 29 (24.1%) were due unspecified cardiac causes. Note that it was possible for deaths to be attributed to more than one cardiac factor. Other illnesses, such as pneumonia, were confirmed but not further classified by the blinded adjudication committee.
Table 2. Causes of Death in Placebo- and Zoledronic Acid–Treated Patients
| ||Zoledronic acid, N = 1054||Placebo, N = 1057||P valuea|
|All causes (%)||101 (9.6)||141 (13.3)||0.01|
|Cardiac||36 (3.4)||52 (4.9)||0.10|
|Infectious||18 (1.7)||14 (1.3)||0.48|
|Respiratory||13 (1.2)||22 (2.1)||0.17|
|Neoplasms||7 (0.7)||13 (1.2)||0.26|
|Cerebrovascular||10 (1.0)||11 (1.0)||1.00|
|Gastrointestinal||2 (0.2)||4 (0.4)||0.69|
|Cachexia/failure to thrive||1 (0.1)||1 (0.1)||1.00|
|Renal||1 (0.1)||3 (0.3)||0.62|
|Other vascular||3 (0.3)||3 (0.3)||1.00|
Figure 2. Incidence of common conditions in the zoledronic acid (ZOL) and placebo groups (top panel) and the risk of dying from a given condition among subjects who experienced that condition (bottom panel).
Download figure to PowerPoint
The magnitude and direction of the mortality benefit were consistent across most subgroups (Table 3). Males experienced a greater absolute mortality benefit than females (6.4% versus 2.8%) but had a similar reduction in the risk of death. Subjects living in a nursing facility prior to their hip fracture and subjects with baseline SPMSQ scores greater than 2 (indicating greater cognitive impairment) had a similar risk of death regardless of treatment arm.
Table 3. Relative Risk of Death in Zoledronic Acid–Treated Subjects Compared With Placebo-Treated Subjects by Subgroup
|Subgroup||Cumulative death rates, zoledronic acid, N = 1054 (%)||Cumulative death rates, placebo, N = 1057 (%)||Hazard ratio (95% CI)|
|Gender|| || || |
| Male||32/244 (22.0)||51/261 (25.5)||0.71 (0.46, 1.31)|
| Female||69/810 (11.9)||90/796 (15.6)||0.74 (0.54, 1.02)|
|Race|| || || |
| Caucasian||93/962 (14.6)||128/960 (18.2)||0.73 (0.56, 0.95)|
| Other||8/92 (10.2)||13/97 (16.7)||0.63 (0.26, 1.51)|
|Age|| || || |
| <75 years||23/477 (5.1)||31/460 (9.8)||0.75 (0.44, 1.28)|
| 75–84 years||44/440 (16.7)||64/447 (20.4)||0.71 (0.49, 1.05)|
| ≥85 years||34/137 (31.7)||46/150 (36.7)||0.72 (0.46, 1.12)|
|Region|| || || |
| North America||54/297 (22.0)||74/313 (25.4)||0.76 (0.53, 1.07)|
| Latin America||9/132 (7.8)||17/131 (20.2)||0.54 (0.24, 1.21)|
| Western Europe||23/356 (9.9)||32/353 (13.4)||0.71 (0.42, 1.21)|
| Eastern Europe||15/269 (6.3)||18/260 (8.4)||0.81 (0.41, 1.61)|
|BMI (kg/m2)|| || || |
| <19||11/79 (19.0)||23/72 (42.8)||0.43 (0.21, 0.89)|
| 19–25||54/502 (17.2)||66/506 (15.7)||0.82 (0.58, 1.18)|
| >25||28/441 (6.6)||47/450 (17.4)||0.59 (0.37, 0.95)|
|SPMSQ score|| || || |
| 0||24/524 (5.7)||37/523 (10.8)||0.65 (0.39, 1.09)|
| >0–2||25/283 (13.4)||49/285 (24.8)||0.47 (0.29, 0.77)|
| >2||40/168 (35.1)||47/180 (31.6)||0.96 (0.63, 1.46)|
|Femoral neck T-score−2.5 or less|| || || |
| Greater than −2.5 to −1.5||52/448 (16.3)||76/437 (22.6)||0.70 (0.49, 0.99)|
| or less||19/360 (9.4)||28/374 (9.9)||0.71 (0.40, 1.27)|
| Greater than −1.5||3/122 (4.1)||12/121 (13.5)||0.22 (0.06, 0.77)|
|Residence prior to hip fracture|| || || |
| Private home||73/945 (11.9)||103/928 (14.9)||0.70 (0.52, 0.95)|
| Assisted living||11/60 (25.6)||17/61 (41.0)||0.71 (0.33, 1.51)|
| Skilled nursing facility||14/31 (55.6)||18/44 (51.0)||1.02 (0.51, 2.05)|
In a model adjusting for multiple baseline risk factors (e.g., time to dosing following hip fracture repair, age, sex, private residence before hip fracture, private residence after hip fracture, cognitive impairment, BMI, diabetes, coronary artery disease, congestive heart failure, hypercholesterolemia, and baseline femoral neck T-score), zoledronic acid significantly reduced the risk of death relative to placebo (HR 0.75, 95% CI 0.58–0.97, p = .03; Table 4).
Table 4. Comparison of Models for Mortality
| ||Hazard ratio||96% CI|
|Adjusting for baseline risk||0.75||0.58–0.97|
|Adjusting for baseline risk and new clinical fractures||0.77||0.59–0.99|
|Adjusting for baseline and competing risk including acute phase reaction||0.97||0.74–1.28|
Subsequent postrandomization fractures were significantly associated with higher risk of death (HR 1.62, 95% CI 1.09–2.40) independent of treatment. Including this factor in the model reduced the risk reduction of zoledronic acid slightly (HR = 0.77, 95% CI 0.59–0.99), suggesting that 0.02 of the 0.25 risk reduction or approximately 8% of zoledronic acid's death benefit is mediated through its reduction of fracture rates. A treatment-by-fracture interaction term was not significant, indicating that subjects who fractured had a similar risk of mortality regardless of treatment group.
To explore whether zoledronic acid might affect death through an immunomodulatory effect, subjects with acute-phase reactions occurring during the first infusion were compared with subjects without such a reaction. Although subjects with acute-phase reaction were only half as likely to die as subjects without acute-phase reactions (5.2% versus 11.9%, p = .02), this variable was not significant when adjusting for the other risk factors in the multivariate models.
Including other competing risks for death that occurred after first-study drug infusion (change in bone mineral density, falls, pneumonia, other infection, acute-phase reaction, cardiovascular disease, and cardiac arrhythmia) and further performing stepwise model selection on baseline risks to the model eliminated the mortality benefit (see Table 4). This suggests that either these events occurred more often in placebo-treated participants by chance or that zoledronic acid's death benefit is mediated through one or more of those pathways. Most notably, cardiac arrhythmias (HR 14.3, 95% CI 10.4–19.8) and pneumonia (HR 3.46, 95% CI 2.40–4.98) were highly associated with mortality. Significant interactions for zoledronic acid treatment by pneumonia (p = .04) and zoledronic acid by arrhythmia (p = .02) were observed. The parameter estimates indicate a greater reduction in death for patients who received zoledronic acid and experienced these events than those who received placebo and experienced these events (see Fig. 2, bottom panel) despite a similar incidence of pneumonia and arrhythmia in the groups (see Fig. 2, top panel). No other notable treatment-by-factor interactions were observed, although there was a nonsignificant trend toward decreased risk of dying from other cardiac deaths, respiratory failure, and neoplasia despite overall similar incidence of these conditions in the study arms (see Fig. 2).
- Top of page
- Materials And Methods
These analyses attempt to disentangle the mechanism(s) for zoledronic acid's (Aclasta/Reclast, Novartis Pharma AG) impact on mortality. While a definitive explanation cannot be obtained from a retrospective analysis of clinical trial data, the insights from this study will help to guide further basic research and epidemiologic inquiries.
Zoledronic acid's benefit was similar across a wide range of subgroups and disease states. Even the oldest age group (>85 years) derived a death benefit, although there is a suggestion that subjects with characteristics associated with greater frailty, such as prior nursing home residence or high degrees of cognitive impairment, do not receive as large a death benefit. While making treatment decisions based on post hoc subgroup analyses is inadvisable, clinicians should give consideration to zoledronic acid's death benefit, which manifests in the second year of treatment, to their patients' remaining life expectancy and competing risks of mortality.
In contrast, the statistically significant fracture benefit of zoledronic acid is seen as early as 12 months, and previous studies have shown a strong association between fractures and mortality. Acute and subacute fracture complications such as myocardial ischemia, venous thromboembolism, and fat embolism and the hazards associated hospitalization and delirium may explain some of this excess risk.18 However, after adjusting for secondary fractures and other baseline risk factors, the risk of mortality was still 23% lower in zoledronic acid–treated participants, suggesting that other pathways also may be involved.
One of our a priori hypotheses was that zoledronic acid might affect cardiovascular mortality. There is a clear epidemiologic association between osteoporosis and cardiovascular disease,19 with suppression of monocyte-macrophages,20 alteration in cytokines, and vascular calcium deposition suggested as common pathophysiologic mechanisms. A meta-analysis of subjects enrolled in clinical trials testing the oral bisphosphonate risedronate showed a trend toward lower cardiovascular mortality driven primarily by a reduction in strokes.21 Given the 29% prevalence of elevated markers of myocardial injury in hip fracture patients and its association with subsequent mortality,22 any impact of bisphosphonates on cardiovascular disease may be particularly important in hip fracture patients. We found a trend toward decreased arrhythmias and cardiovascular deaths in zoledronic acid–treated patients and a significant treatment by postrandomization arrhythmia interaction. This contrasts with the previous report of increased atrial fibrillation serious adverse events with zoledronic acid in postmenopausal women.23 Bisphosphonates affect ion channels in several cell lines, including cardiac myocytes24, 25; thus a true bisphosphonate impact on arrhythmia is physiologically plausible. Alternatively, since many arrhythmias in older adults are precipitated by myocardial ischemia, the reduction in arrhythmia incidence and death rates may reflect an impact on cardiovascular disease rather than on the conduction system. Further studies are needed to disentangle this controversy.26
Another possible pathway is through a bisphosphonate effect on the immune system. In addition to the changes in cytokines and the monocyte-macrophage system noted earlier, there is epidemiologic evidence to support this hypothesis. A greater than 10-fold increased relative risk of infectious death, particularly septicemia and pneumonia, was seen in a cohort of hip fracture patients relative to the general population over 2 years.27 We found a treatment by postrandomization pneumonia interaction that suggested that zoledronic acid–treated patients were less likely to die from a pneumonia event than placebo-treated patients even though the overall incidence of pneumonia was similar between treatment groups (5.5% versus 5.6%).
A third hypothesis is that rather than influencing one particular organ system, zoledronic acid influences physiologic reserve more generally and thus the ability to “recover” from acute illnesses. Hormetic effects, where low-level or intermittent exposure to a substance produces enhanced ability to withstand stressors, while high-level or chronic exposure is toxic, have been noted in osteoblasts.28 A hormetic effect could explain the mortality pattern we observed: a lower risk of death from a variety of illnesses despite a similar overall incidence of these illnesses. Although direct evidence is lacking, zoledronic acid might produce a hormetic response directly or through impact on inflammatory cytokines or other mediators. The lower death rate in participants who experienced an acute-phase reaction is intriguing and consistent with a hormetic response. However, this finding also could be due to preexisting inflammatory or malignant conditions that decrease the risk of experiencing an acute-phase reaction and are also associated with higher subsequent mortality. Persistently elevated inflammatory cytokines such as IL-6 after hip fracture have been associated with adverse clinical outcomes in older patients, such as reduced lower extremity function29 and cognitive decline.30, 31 Little is known about the long-term effects of zoledronic acid on the immune system, and further studies exploring the link between changes in inflammatory markers with zoledronic acid and mortality may be warranted.
This analysis has several limitations. First, the subjects enrolled in this clinical trial were likely healthier than the general population of hip fracture patients; this could affect the rate and types of deaths we observed. However, the distribution of the causes of mortality in the HORIZON Recurrent Fracture Trial was similar to that reported in previous cohort studies. Second, the primary cause of death could not be adjudicated in a high proportion of subjects owing to unwitnessed events, missing source documentation, or multiorgan system disease with an unclear precipitant. Since a common cause of unwitnessed sudden death is cardiovascular events, the distribution of the missing data is likely not random. Finally, we relied on investigator-reported adverse events and baseline comorbidities for our models, and the accuracy and completeness of their reporting are not known.
In summary, our analysis suggests that zoledronic acid's impact on mortality is mediated only to a small extent through its fracture-reduction benefit. Subjects treated with zoledronic acid were less likely to die from cardiac arrhythmias and pneumonia than placebo-treated subjects. The mechanism for this finding is currently unknown and warrants further prospective studies.
- Top of page
- Materials And Methods
Cathleen Colón-Emeric, Kenneth Lyles, Jay Magaziner, Carl Pieper, and Steven Boonen have received research funding and are consultants for Novartis. Peter Mesenbrink is an employee of Novartis Pharmaceuticals Corporation and a shareholder in the company. Erik Eriksen was an employee of Novartis Pharma AG at the time of the development of this article and is now a consultant for the company. Dr. Colón-Emeric is supported by Paul A. Beeson Award K23 AG024787. Dr. Boonen is senior clinical investigator of the Fund for Scientific Research-Flanders, Belgium (F.W.O.-Vlaanderen) and holder of the Leuven University Chair in Metabolic Bone Diseases. Dr. Lyles holds patent “Methods for preventing or reducing secondary fractures after hip fracture” (US Patent Application 20050272707) and is Provisional Patent Application Inventor of “Medication Kits and Formulations for Preventing, Treating or Reducing Secondary Fractures after Previous Fracture.” Dr. Magaziner's effort was supported, in part, by grants from the National Institute on Aging (NIA) and the Claude D. Pepper Older Americans Independence Center P30-AG028747, R37 AG009901, R01 AG029315, and R01 AG018668.
- Top of page
- Materials And Methods
The HORIZON Recurrent Fracture Trial was funded by Novartis Pharmaceuticals. The authors wish to thank BioScience Communications for editorial support in preparation of this manuscript.
Cathleen S. Colon-Emeric participated in the planning of the study, the collecting of data, the supervision of analyses, the interpretation of results, authoring of the first draft of this article, and final responsibility for the decision to submit it for publication. Peter Mesenbrink participated in planning the study, performing the statistical analyses, interpreting the results, and authoring of the first draft of this article. Kenneth W. Lyles participated in the planning of the study, the collection of data, the supervision of analyses, and the interpretation of results. Carl F. Pieper participated in the planning of the study, the collection of data, and the interpretation of results. Steven Boonen participated in the planning of the study, the collection of data, and the interpretation of results. Pierre Delmas participated in the planning of the study, the collection of data, and the interpretation of results. Erik F. Eriksen participated in the planning of the study, the collection of data, and the interpretation of results. Jay Magaziner participated in the planning of the study, the collection of data, and the interpretation of results. All authors were involved in planning the article, critical review of the first draft, and subsequent revisions to the paper. All have seen and approved the final version, excepting Pierre Delmas, who passed away before it was completed.