Efficacy and Safety of Single Dose Zoledronic Acid for Osteoporosis in Frail Seniors: A Randomized Clinical Trial

Abstract

Importance

85% of institutionalized elderly have osteoporosis, with fracture rates 8–9 fold higher than observed among community-dwelling elderly. Yet most are untreated and excluded from pivotal osteoporosis trials.

Objective

Determine the efficacy and safety of zoledronic acid in frail elderly women.

Design

2-year, randomized, placebo-controlled, double-blinded study conducted between December 2007 and March 2012.

Setting

Nursing home and assisted living facilities.

Participants

181 women ≥ age 65 with osteoporosis including those with cognitive impairment, immobility, and multimorbidity.

Intervention

One 5 mg dose of zoledronic acid or placebo IV and daily calcium and vitamin D.

Main Outcomes

(1) Hip and spine bone mineral density (BMD) at 12 and 24 months and (2) adverse events.

Results

There were no baseline differences in age (mean=85.4±0.6 years), BMD, or functional or cognitive status, but the treatment group included more subjects with frailty, falls history, diabetes, and anticonvulsant use. BMD was available for 87% of participants at 12 months and 73% at 24 months. BMD changes were greater in the treatment group (p< 0.01): 3.2 ± 0.7 and 3.9 ± 0.7 percentage point differences (mean ± SE) in the total hip at 12 and 24 months respectively, and 1.8 ± 0.7 and 3.6 ± 0.7 at the spine (p<0.01); adjusted analyses were similar. The treatment and placebo groups’ fracture rates were 20% and 16%, respectively (OR=1.30; 95% CI=0.61–2.78); mortality rates were 16% and 13% (OR=1.24; 95% CI=0.54–2.86). Groups did not differ in the proportion of single fallers (28% vs. 24%; OR=1.24; 95% CI=0.64–2.42; p=0.52) but more subjects in the treatment group had multiple falls (49% vs. 35%; OR=1.83; 95% CI=1.01–3.33; p=0.047); this was no longer significant when adjusted for baseline frailty.

Conclusions and Relevance

In this group of frail, osteoporotic women, one dose of zoledronic acid improved BMD over 2 years. The clinical importance of nonsignificant increases in fracture and mortality rates in the treatment group need further study. Since it is not known whether such therapy reduces the risk of fracture in this cohort, any change in nursing home practice must await results of larger trials powered to assess fracture rates.

Trial Registration

Clinical Trials. Gov Identifier NCT00558012

Introduction

Nearly 2 million frail Americans reside in long-term care (LTC) facilities, and another 4–6 million similarly-impaired seniors live in the community.¹ Eighty-five percent of such individuals have osteoporosis^2–4 and their fracture rates are 8–9 fold higher than those observed among less impaired seniors.⁵ Moreover, the impact of an associated hip fracture is dire^6,7: decreased mobility and independence, frequent hospitalizations, and a 6 month mortality rate of up to 36%.⁸ Yet osteoporosis therapy remains vastly under-prescribed to such individuals.^9,10

Few osteoporosis trials have focused on such individuals and, to our knowledge, none has evaluated the efficacy or safety of a bisphosphonate in this group. Although post-hoc analyses of the pivotal osteoporosis trials suggest that therapy is efficacious for healthy, community-dwelling participants over age 75 years,^11–13 these trials excluded functionally impaired individuals. Such exclusion may limit their applicability to this group because in such individuals the link between bone density and bone strength may be attenuated. Once sufficient bone mass is lost, structural integrity of the remaining bone may be compromised due to impaired trabecular connectivity, poor skeletal microstructure, and unhealed stress fractures.^14–16 Adding mass to the remaining non-connected “bone stubs” may add little if any bone strength. The odds of this phenomenon are likely substantial in frail LTC residents because they are 10–15 years older than those in the pivotal trials, with a longer duration of bone loss, and because bone loss over this period is more apt to be accelerated by comorbidity, frequent illness, sedentary lifestyle, and renal impairment. Support for this concern is heightened by results of the risedronate hip fracture trial which, despite having adequate power, demonstrated a reduction in hip fractures for community-based women under age 80 but not for those older than 80 years, although evaluation of the latter group did not include BMD.¹⁷ Thus, it is unclear whether anti-osteoporosis therapy is effective in frail elderly. Because of the rapid global growth of this group, multiple organizations--including NIH--have for decades reiterated the critical need for osteoporosis research and treatment of such individuals.^18–21

Yet obstacles to such studies are substantial,²² especially if the goal is to demonstrate fracture reduction. However, virtually every trial that has demonstrated fracture reduction by a potent bisphosphonate has been associated with an increase in bone density.^23,24 Although this does not prove that such an increase is required, it does suggest that fracture reduction would be unlikely in the absence of an impact on bone density. Thus, we employed a stepwise approach. Our objective in this first study was to determine the impact of a bisphosphonate on BMD and safety for 2 years in frail women residents of LTC facilities. Although frail elderly are more likely to live outside of institutions, we enrolled an institutionalized group to ensure that we could deliver therapy and closely monitor its impact and side effects. We chose zoledronic acid because it can be given as a single intravenous dose, the effect of which may last for 2 years;^25,26 this eliminates difficulties inherent in administering an oral bisphosphonate on a regular basis in this setting.

Methods

Study Design

The Zoledronic acid in frail Elders to STrengthen bone (ZEST) study was a 2-year double-blind, placebo-controlled, randomized clinical trial based in the Pittsburgh Pennsylvania area. Participants were enrolled and treated between December 2007 and March 2012 (ClinicalTrials.gov Identifier: NCT00558012).²²

Participants

We included frail women aged 65 years or older who resided in a nursing home or assisted living facility,²² were not receiving a bisphosphonate, and who either had BMD below the treatment cut off for osteoporosis (based on 2003 NOF guidelines: BMD < −2.0 SD at the spine, hip or radius)^27,28 or a prior vertebral or hip fracture. All women whose 25-hydroxyvitamin D levels were < 20 ng/dl²⁹ were supplemented with vitamin D (50,000 units/week for 2 months) and re-screened. All subjects received a daily divided dose of vitamin D (800 IU/day) and 1200 mg/d elemental calcium (supplement plus diet³⁰).

We included women who had cognitive and functional impairment, immobility, multiple medical conditions, and who were prescribed multiple medications (including glucocorticoids and anti-seizure medications). We excluded those with a projected life expectancy of < 2 years or estimated GFR < 30 ml/min.

Approval and Consent

The study was approved by the University of Pittsburgh Institutional Review Board and Pennsylvania Department of Health. Informed consent was obtained for all participants. Those with cognitive impairment provided assent and the resident’s responsible party provided written consent.

Randomization and Blinding

The study biostatistician randomized participants in a 1:1 ratio using random block sizes of 2 and 4. The research pharmacist provided identically-appearing active drug or placebo. Investigators, study personnel, providers, and participants were blind to treatment assignment.

Clinical Protocol and Intervention

Study visits were conducted at each subject’s facility. BMD was assessed in a mobile unit that included a motorized platform to facilitate access for participants with restricted mobility. Baseline blood tests and BMD were obtained, and 10 year fracture risk (FRAX³¹) was calculated.

After completing the baseline assessment, subjects were randomized to infusion with either 5 mg of intravenous zoledronic acid or placebo. Patients were assessed for immediate adverse reactions for 3 days post-infusion.³² Subsequent follow-up visits occurred at 6, 12 and 24 months.

Outcome Variables

The primary outcome was percent change in BMD of the total hip and spine at 12 months. Secondary outcomes included adverse events and bone turnover markers. Additional outcomes included change through 24 months in BMD at other skeletal sites, bone turnover markers, physical and cognitive function, comorbidity, and survival, and an exploratory assessment of fragility fractures at 12 and 24 months.

Bone Mineral Density

Measured sites included the hip (total hip, femoral neck), spine (posterior-anterior and lateral projection), and 1/3 distal radius. Dual x-ray absorptiometry (DXA) was performed using a Discovery densitometer (Hologic Inc., Bedford, MA); the precision ranged from 1.2 to1.9% at these skeletal sites.³³

Vertebral fractures

Fractures of T6–L4 were detected by Vertebral Fracture Assessment (VFA) performed by DXA at baseline, 12 and 24 months and classified by Genant’s criteria as mild, moderate, or severe.³⁴ We included new fractures and fracture grade progression. As compared to conventional x-ray, the sensitivity and specificity of VFA are 100% and 95%, respectively,³⁵ with a kappa statistic of 0.92.³⁵

Fragility fractures

In addition to chart review, we asked participants every 6 months about clinical fragility fractures (defined as a fracture following a fall from standing or sitting height); fractures were confirmed by radiology reports.

Markers of bone turnover and vitamin D

Bone resorption was assessed by serum C-telopeptide crosslinks type I collagen (CTX, Crosslaps, Osteometer Biotech, Herlev, Denmark). Bone formation was assessed by serum intact N-terminal propeptide type I procollagen (P1NP, Orion Diagnostica, Espoo, Finland). Serum 25-hydroxy vitamin D was assessed by liquid chromatography/mass spectrometry.

Functional Assessments

Function was assessed using the Katz Activities of Daily Living,³⁶ Instrumental Activities of Daily Living,³⁷ gait speed,^38,39 Short Portable Mental Status Questionnaire,⁴⁰ Duke comorbidity Index,{1630} depression scale (PHQ-9),⁴² modified Fried Frailty Index (categorized as frail, pre-frail, robust),⁴³ and the Physical Performance Test.⁴⁴

Safety and Adverse Events (AEs)

We collected information on AEs from patients and charts at scheduled visits. To improve reporting of serious adverse events (SAEs) we created an electronic database alert system.²²

Sample Size

A priori assumptions, statistical power, and sample size justification have been published elsewhere.²² Briefly, assumptions were based on available data from younger, postmenopausal women in whom zoledronic acid increased BMD at 12 months by 4.5±3.6% at the spine and 3.0±2.9% at the femoral neck.^45,46 We assumed 25% less improvement in our cohort, no change in our placebo group, and 30% attrition over a year. This suggested that randomizing 180 women would yield 126 completers at 1 year and provide 95% power to detect an absolute difference in BMD between groups of 2.4 percentage points at the spine and 86% power to detect a 1.6 percentage point difference at the femoral neck (two-tailed test; α=0.05).

Statistical analysis

To compare baseline characteristics in the treatment and placebo groups, we used independent sample t-, Wilcoxon rank sum, chi square and Fisher’s exact tests. For the main analysis, we fitted a series of linear mixed models, using percent change between baseline and follow-up assessment in each of the BMD and biomarker measures as the dependent variable; treatment arm (active/placebo), follow-up assessment (6/12/24 months) and their interaction as fixed effects of interest; baseline value of the measure as a fixed effect covariate; and a subject random effect to account for multiple measurements from the same participant over time and the resulting non-independence of observations. We used appropriately constructed contrasts to compare treatment arms at each of the 6-, 12- and 24-month assessments. To assess robustness of the main results, we conducted sensitivity analyses using: raw change in measurements instead of percent change from baseline; last-value-carried-forward and a multiple imputation approach for missing data based on multivariate normality and the Markov Chain Monte Carlo method; additional covariate adjustments for participant characteristics that differed significantly at baseline (frailty, falls risk, diabetes, and anticonvulsant use); and log transformations of markers of bone turnover.^47,48 Finally, we used chi-square and Fisher’s exact tests to compare the proportions of participants in each group who experienced each and any type of adverse event. We fitted logistic regression models for events of specific interest (deaths, falls, fractures) with covariate adjustments for frailty, falls risk, and diabetes. SAS^® version 9.3 was used, with MIXED and LOGISTIC procedures for the main analyses.

Results

We contacted 733 women; 252 consented to screening and 181 received the infusion (Figure 1). At baseline (Table 1), both groups suffered from substantial impairment: 95% of subjects were categorized as frail or pre-frail, 74% were dependent in at least 1 basic ADL (and 31% in at least 3), 93% had at least one deficit in IADL (and 70% in at least 3), 85% had a gait speed characteristic of frailty (<0.8 meter/sec), and the majority of subjects were cognitively impaired. There were no group differences in age, body mass index, and calcium or vitamin D intake. However, compared with the placebo group, more women in the treatment group were categorized as frail by the modified Fried Frailty Index,⁴³ more tended to be fallers in the year prior to the study, more had a diagnosis of diabetes mellitus, and more were taking anticonvulsants. Baseline safety laboratory values were similar between the 2 groups, as were markers of bone turnover, BMD, the number of women with vertebral fractures, and the 10 year calculated risk of osteoporotic fractures. Seventy-six percent of patients completed 24 months.

Figure 1.

Enrollment and Study Design Flow Chart

Table 1.

Baseline Clinical Characteristics, Functional Status, Markers of Bone Turnover, and Bone Mineral Density

Characteristic	Zoledronic Acid (n=89)	Placebo (n=92)	P-value
Age (years)	85.4±0.6	85.5±0.5	0.85
Body Mass Index (kg/m2)	28.2±0.6	26.9±0.5	0.29
Total calcium intake (mg/d)	807±48	763±44	0.50
Vitamin D intake (IU/day)	163±15	168±15	0.83
Anticonvulsant use (%)	10	1	0.007
Calcium antacids use (%)	8	0	0.006
ADL (0–14)†*	11.5±0.3	11.7±0.3	0.45
IADL (0–14)‡*	7.7±0.4	8.2±0.4	0.42
Gait speed (m/sec)	0.5±0.0	0.6±0.3	0.16
Physical Performance Test (0–24)*	19.2±0.5	20.1±0.5	0.20
Fried Frailty Index (% frail)	72	58	0.04
Comorbidity Index (0–8 domains)Δ	3.5±0.2	3.3±0.1	0.39
Diabetes (%)	26	13	0.03
SPMSQ score (0–10 errors)Δ	1.8±0.3	1.8±0.3	0.99
PHQ-9 score (0–27)Δ	4.2±0.5	3.6±0.4	0.33
Fall once previous year (%)	50	36	0.06
Recurrent falls previous year (%)	26	18	0.16
Albumin (g/dl)	3.8±0.03	3.8±0.04	0.64
Albumin corrected calcium (mg/dL)	9.5±0.03	9.6±0.04	0.28
25 hydroxy vitamin D (ng/mL)	29.6±1.4	29.8±1.4	0.89
eGFR (ml/min/1.73 m2)	64.3±2.3	68.4±2.0	0.18
CTX (nmol/L BCE)	0.425±0.024	0.405±0.024	0.56
P1NP (ug/L)	49.6±2.9	48.2±2.5	0.72
Bone Mineral Density (gm/cm2)
Spine	0.932±0.020	0.967±0.021	0.22
Total Hip	0.681±0.015	0.698±0.015	0.42
Femoral Neck	0.607±0.128	0.620±0.013	0.50
Bone Mineral Density T-Scores (SD)
Spine	−0.8±0.2	−0.6±0.2	0.48
Total Hip	−2.1±0.1	−2.0±0.1	0.69
Femoral Neck	−2.3±0.1	−2.1±0.1	0.49
BMD Osteoporosis Classification (%)	48	45	0.92
Vertebral Fractures (%)	52	41	0.16
10 Year Hip Fracture Risk (%, FRAX)	9.7±1.5	7.7±0.8	0.25
10 Year Major Osteoporosis Fracture Risk (%, FRAX)	22.7±1.5	20.3±1.1	0.21

Results as mean ± standard error or N (%)

^*Low score worse

^ΔHigh score worse

^†ADL= Activities of Daily Living (eat, dress, groom hair, walk, transfer, bathe, get to bathroom).

^‡IADL= Instrumental Activities of Daily Living (use phone, travel, shop, prepare meals, housework, take medication, handle money).

Bone Mineral Density

Total hip BMD increased more in the treatment group than in the placebo group, both at 12 months (2.8±0.5 vs. −0.5±0.4%, mean ± SE, p<0.0001) and 24 months (2.6±0.6 vs. −1.5±0.7%, p<0.0001); the adjusted difference was 3.9±0.7 percentage points at 24 months (p<0.0001, Figure 2). Similar differences were observed in the femoral neck with an adjusted difference of 2.7±1.0 percentage points at 24 months; p<0.01.

Figure 2.

Mean ± SE percent change in bone mineral density (BMD) from baseline to 24 months (unadjusted).

* p<0.05, ** p<0.01 change from baseline using paired t-test. # p<0.05, ## p<0.01 for comparison between zoledronic acid (solid line) and placebo (dashed line) groups using linear mixed models.

Spine BMD also increased more in the treatment group than in the placebo group at 12 months (3.0±0.5 vs. 1.1±0.5%, p< 0.01) and 24 months (4.5±0.8 vs. 0.7±0.5%, p<0.0001), with an adjusted difference of 3.6±0.7 percentage points at 24 months (p<0.0001, Figure 2). At 24 months, the improvement from zoledronic acid was 4.8 ±1.1 percentage points greater (p<0.0001) at the lateral spine and 1.2±0.6 (p<0.05) at the 1/3 distal radius. When BMD was stratified by diabetes, trends were similar, as were trends when group differences were assessed for missing data by the last value carried forward and multiple imputation methods.

Results were nearly identical when adjusted for the baseline imbalance in frailty, diabetes, and anticonvulsant use: the adjusted difference for spine BMD was 2.0 ± 0.7 percentage points at 12 months and 3.8 ± 0.7 at 24 months (p< 0.001); for total hip BMD, it was 3.2 ± 0.7 percentage points at 12 months and 3.8 ± 0.7 at 24 months (both p< 0.0001); and for femoral neck BMD, it was 3.7 ± 0.9 percentage points at 12 months and 2.9 ± 1.0 at 24 months (both p < 0.01).

Biochemical Markers of Bone Turnover

As assessed by CTX, bone resorption decreased in the treatment group (at 12 and 24 months, p<0.01) and increased in the placebo group (at 12 and 24 months, p<0.05); the adjusted between-group difference was 0.135 ±0.035 nmol/L Bone Collage Equivalents (BCE) at 24 months (p<0.001, Figure 3). As expected, P1NP, the marker of bone formation, decreased in the treatment group at 12 and 24 months (p<0.01); the adjusted between-group difference was 16.95±3.15 µg/L (p<0.001) at 24 months.

Figure 3.

Mean ± SE change in biochemical markers of bone turnover including CTX (nmol/L BCE) and P1NP (µg/L) from baseline to 24 months (unadjusted). * p<0.05, ** p<0.01 change from baseline using paired t-test. #p<0.05, ##p<0.01 for comparison between zoledronic acid (solid line) and placebo (dashed line) groups using linear mixed models.

Function and mental status

Both cognitive and physical function declined significantly in both groups over time, but differences between the two groups were not significant.

Adverse Events

97% of participants had an adverse event and 64% had a serious adverse event, but there were no group differences (Table 2). As expected, in the 3 days following the infusion the treatment group experienced more acute symptoms, but there were no group differences in the laboratory safety parameters other than a small drop in serum calcium on day 2 in the zoledronic acid group (0.52 mg/dL less than control, p<0.001) that was not significantly different at 6 months. Overall, there were no significant differences in number of deaths, fractures, or cardiac disorders, including atrial fibrillation (Tables 2 and 3). Patients who fractured were neither unmasked nor treated. Although there were no differences in serious falls, there were more fallers in the treatment group than in the placebo group over the 2 year study. There was no significant difference between groups in the number of single fallers (28% vs. 24%; OR=1.24; 95% CI=0.64–2.42; p=0.52) but more subjects in the treatment group had multiple falls (49% vs. 35%; OR=1.83; 95% CI=1.01–3.33; p=0.047). When adjusted for baseline frailty, however, the difference in the proportion of multiple fallers was no longer significant (OR=1.60; 95% CI=0.85–2.99; p=0.142). When incident falls were stratified by diabetes, there were more falls in the treatment group than in the placebo group (87% [n=20] vs 33% [n=4], OR=13.3 (2.42–73.5, p<0.001) but numbers were small and confidence intervals wide. Results did not differ appreciably in any of the sensitivity analyses.

Table 2.

Adverse Events and Serious Adverse Events Through 24 months and Specific Adverse Events Up To 3 Days Post Infusion

Event N (%)	Zoledronic Acid	Placebo	P-value
Any adverse event	87 (98)	88 (96)	0.68
Serious adverse event	60 (67)	55 (60)	0.29
Cardiac disorders	28 (32)	25 (27)	0.53
Atrial fibrillation	5 (6)	5 (5)	0.96
All fallers	69 (78)	54 (59)	0.01
Single fallers	25 (28)	22 (24)	0.52
Multiple fallers	44 (49)	32 (35)	0.046
Serious fallers	4 (4)	5 (4)	1.00
Within three days post-infusion
Headache	14 (16)	6 (7)	0.048
Pyrexia	7 (8)	0 (0)	0.01
Fatigue	21 (24)	14 (15)	0.15
Arthralgias	10 (11)	6 (7)	0.26
Myalgias	7 (8)	3 (3)	0.21
Flu-like illness	6 (7)	2 (2)	0.16
Falls	4 (5)	3 (3)	0.72
Albumin adjusted calcium day 2 < 8.4 mg/dL	2 (2)	0 (0)	0.24
eGFR<30 at any follow-up	3 (3)	3 (3)	1.00
Albumin adjusted calcium 6 months < 8.4 mg/dL	0 (0)	0 (0)	1.00

Results as number of distinct women (%); p-Value from chi-square or Fisher’s exact test

Table 3.

Unadjusted Comparison of Mortality and Fracture Outcomes across Groups

Outcome	Zoledronic Acid, N (%)	Placebo, N (%)	Zoledronic Acid vs Placebo, Odds Ratio (95% CI)	Odds Ratio p-Value
Mortality	14 (16%)	12 (13%)	1.24 (0.54–2.86)	0.61
All fractures	18 (20%)	15 (16%)	1.30 (0.61–2.78)	0.50
Vertebral fractures	6 (7%)	8 (9%)	0.76 (0.25–2.28)	0.62

Discussion

This is the first randomized trial of a potent antiresorptive therapy administered to a group of frail elderly women. We found that, as compared with calcium and vitamin D alone, adding a single dose of intravenous zoledronic acid significantly improved BMD of the hip and spine over 2 years. Improvements were similar to those found in healthy, less aged women in the pivotal zoledronic acid trial.³² Changes in bone turnover suggest that zoledronic acid had a continued impact for 2 years after one dose. There were no significant differences in serious adverse events through 24 months although fracture and mortality rates were higher in the zoledronic acid group.

Our cohort was more impaired than the one we previously investigated from assisted living communities.⁴⁹ Those subjects were 8 years younger than these and all were cognitively intact, mobile, able to self-administer a daily medication, and able to attend a centralized study site for assessments. Nonetheless, bone mass response to a single dose of zoledronic acid in the current group was similar to that seen in the less impaired group who took an oral bisphosphonate (alendronate) daily for two years.

Despite the robust increase in BMD at the hip and spine, we did not observe a reduction in total or vertebral fractures. Our study was neither designed nor powered to examine absolute fracture reduction, but our point estimates contrast with findings from the pivotal zoledronic acid study, which reported a 60% reduction in vertebral fractures after 1 year and a 30% reduction in clinical fractures after 2 years.³² It is possible that the significantly greater frailty of our treatment group masked beneficial differences in fracture rates. It is also possible that, despite the improvement in bone density, such frail individuals may not benefit from therapy because skeletal integrity may be so compromised that poor connectivity, impaired microstructure, altered microgeometry, unhealed stress fractures, restricted mobility (less weight bearing), reduced life expectancy, or other factors may prevent fracture reduction.^15,16 However, neither the limited statistical power of this study nor the point estimates observed can be used to infer that fracture reduction would or would not be seen in a larger study;⁵⁰ strong computational evidence suggests that only a fracture reduction study will suffice. Based on our data demonstrating a positive impact on both bone mass and bone turnover, we believe that such a study is now justified and essential. The need for such a trial is also timely since many payers use bisphosphonate treatment as a quality assessment measure in this population despite the lack of safety and fracture reduction data.

Approximately 30% of community dwelling seniors fall annually, as do at least half of LTC residents.^51,52 Our treatment group experienced more non-injurious falls than the control group. However, the treatment group also included more subjects at baseline who met criteria for frailty, had a history of falls and diabetes, and took anticonvulsants. Moreover, there were no differences in serious falls or fractures and, after adjusting for baseline frailty, the difference in falls was no longer significant. We do not know of other studies that reported an increase in falls as a potential side effect of zoledronic acid or other bisphosphonates; this finding is likely due to poorer baseline status of subjects in the treatment arm in our study or simply due to chance.

In addition to its randomized design, a strength of our study was the use of a single infusion of zoledronic acid to benefit skeletal health for at least 2 years. Extended benefit has also been reported in younger patients.^25,26 Another strength was inclusion of residents with immobility and cognitive impairment, as well as those taking glucocorticoids and anti-seizure medications; such patients are generally excluded from pivotal trials. Third, we were able to capture serious adverse events in a timely matter through use of an electronic record system rather than relying on patient recall. Finally, as in all LTC facilities, falls were well-captured owing to regulation.

Our study also had limitations. First, despite randomization, the treatment group contained more subjects with frailty, falls, diabetes, and anticonvulsant use. Despite the baseline differences, however, the adjusted analyses were similar, confirming the robustness of the findings. Second, our study was neither designed nor powered to examine fracture reduction, although we did gather fracture data from adverse events and VFA exams. Instead, this was a proof of concept study. It was designed to assess the impact of therapy on the surrogate markers of BMD and bone turnover and thereby to determine whether a larger trial of fracture reduction would be justified in frail elders. If we had found that these patients were too frail or too sick to achieve differences in BMD or bone turnover, a fracture study would not be worthwhile. Finally, and not surprisingly in such a debilitated population, there were a number of dropouts by 24 months. However, the use of two different approaches to account for their missing data did not change the findings.

In summary, we found that a single infusion of zoledronic acid in frail, cognitively challenged, less mobile elderly women improves bone density and reduces bone turnover for 2 years. This suggests that even a very frail cohort may benefit. However, prior to changing practice, larger trials are needed to determine whether improvement in these surrogate measures will translate into fracture reduction for vulnerable seniors.