Zoledronic acid and risedronate in the prevention and treatment of glucocorticoid-induced osteoporosis (HORIZON): a multicentre, double-blind, double-dummy, randomised controlled trial
The Lancet, Volume 373, Issue 9671, Pages 1253 - 1263, 11 April 2009
Editors' note: In this randomised, double-blind, double-dummy, non-inferiority trial, zoledronic acid was compared to risedronate in the prevention and treatment of glucocorticoid-induced osteoporosis. A single intravenous infusion of zoledronic acid provided greater increases in bone mineral density and more rapid and substantial decreases in bone turnover than a 5 mg daily dose of risedronate.
Prof David M Reid MD a Corresponding AuthorEmail Address, Prof Jean-Pierre Devogelaer MD b, Prof Kenneth Saag MD c, Prof Christian Roux MD d, Prof Chak-Sing Lau MD e, Prof Jean-Yves Reginster MD f, Philemon Papanastasiou PhD g, Alberto Ferreira PhD g, Florian Hartl MD h, Taiwo Fashola PhD g, Peter Mesenbrink PhD i, Prof Philip N Sambrook MD j, for the HORIZON investigators
Persistent use of glucocorticoid drugs is associated with bone loss and increased fracture risk. Concurrent oral bisphosphonates increase bone mineral density and reduce frequency of vertebral fractures, but are associated with poor compliance and adherence. We aimed to assess whether one intravenous infusion of zoledronic acid was non-inferior to daily oral risedronate for prevention and treatment of glucocorticoid-induced osteoporosis.
This 1-year randomised, double-blind, double-dummy, non-inferiority study of 54 centres in 12 European countries, Australia, Hong Kong, Israel, and the USA, tested the effectiveness of 5 mg intravenous infusion of zoledronic acid versus 5 mg oral risedronate for prevention and treatment of glucocorticoid-induced osteoporosis. 833 patients were randomised 1:1 to receive zoledronic acid (n=416) or risedronate (n=417). Patients were stratified by sex, and allocated to prevention or treatment subgroups dependent on duration of glucocorticoid use immediately preceding the study. The treatment subgroup consisted of those treated for more than 3 months (272 patients on zoledronic acid and 273 on risedronate), and the prevention subgroup of those treated for less than 3 months (144 patients on each drug). 62 patients did not complete the study because of adverse events, withdrawal of consent, loss to follow-up, death, misrandomisation, or protocol deviation. The primary endpoint was percentage change from baseline in lumbar spine bone mineral density. Drug efficacy was assessed on a modified intention-to-treat basis and safety was assessed on an intention-to-treat basis. This trial is registered with ClinicalTrials.gov, number NCT00100620.
Zoledronic acid was non-inferior and superior to risedronate for increase of lumbar spine bone mineral density in both the treatment (least-squares mean 4·06% [SE 0·28] vs 2·71% [SE 0·28], mean difference 1·36% [95% CI 0·67-2·05], p=0·0001) and prevention (2·60% [0·45] vs 0·64% [0·46], 1·96% [1·04-2·88], p<0·0001) subgroups at 12 months. Adverse events were more frequent in patients given zoledronic acid than in those on risedronate, largely as a result of transient symptoms during the first 3 days after infusion. Serious adverse events were worsening rheumatoid arthritis for the treatment subgroup and pyrexia for the prevention subgroup.
"reductions in both biomarkers at 12 months were significantly greater in patients on zoledronic acid than in those on risedronate in both the treatment and prevention subgroups."
A single 5 mg intravenous infusion of zoledronic acid is non-inferior, possibly more effective, and more acceptable to patients than is 5 mg of oral risedronate daily for prevention and treatment of bone loss that is associated with glucocorticoid use.
Glucocorticoid drugs are a mainstay of treatment in many inflammatory and immune-mediated disorders.1 However, persistent use is associated with side-effects, such as bone loss and increased fracture risk.2-6 This increased risk is apparent in some patients within 3 months of starting glucocorticoids.4, 7
Prevention and treatment of glucocorticoid-induced osteoporosis is best established for bisphosphonates,5 a class of drugs that increase bone mineral density and reduce vertebral fracture risk in patients beginning or continuing glucocorticoid treatment.8-10 Daily oral bisphosphonate therapy has been approved for the treatment and prevention of glucocorticoid-induced osteoporosis, but compliance and adherence with daily and weekly therapy is characteristically suboptimum.11-14 An association between poor adherence or compliance and increased fracture risk has been documented in women with postmenopausal osteoporosis who were treated with bisphosphonate.15, 16
Zoledronic acid is a potent bisphosphonate; when given every year by intravenous infusion, the drug increases bone mineral density and reduces fracture risk in women with postmenopausal osteoporosis.17 It also reduces subsequent fractures in patients who have had an osteoporosis-related fracture, and increases survival in those who have sustained a low-trauma hip fracture.18 We present the results of HORIZON (Health Outcomes and Reduced Incidence with Zoledronic acid ONce yearly): a 1-year, multicentre, double-blind, double-dummy randomised controlled trial designed to establish whether one 5 mg infusion of zoledronic acid is non-inferior to the licensed dose of risedronate (5 mg daily) for the prevention and treatment of glucocorticoid-induced osteoporosis.
Figure 1 shows the trial profile. The treatment subgroup enrolled 545 patients and the prevention subgroup enrolled 288. Overall, 93% (n=771 patients) of enrolled patients completed the study (treatment 94% [n=511], prevention 90% [n=260]).
Table 1 shows baseline characteristics of patients taking zoledronic acid or risedronate. Overall, 68% (n=568) of participants were women, of whom 66% (n=373) were menopausal. 82% (n=304) of menopausal women had reached menopause more than 5 years earlier (treatment 80% [n=188], prevention 84% [n=116]). In total, 14% (n=77) of patients in the treatment subgroup and 14% (n=41) in the prevention subgroup had one or more fractures at baseline. The proportion of patients with baseline fractures was similar between drug groups in the treatment subgroup (zoledronic acid 16% [n=65] vs risedronate 13% [n=53]), but higher for the zoledronic acid than the risedronate group (18% [n=26] vs 10% [n=15]) in the prevention subgroup. The median prednisolone-equivalent dose of glucocorticoids was the same for both drug groups and the treatment and prevention subgroups (table 1).
Both zoledronic acid and risedronate increased lumbar spine bone mineral density in the prevention and treatment subgroups adjusted for drug group, study region, and sex (figure 2). Assessment of the primary efficacy endpoint showed that the non-inferiority criterion was met. Moreover, by 12 months zoledronic acid had increased lumbar spine bone mineral density more than had risedronate in both the treatment (least-squares mean 4·06% [SE 0·28] vs 2·71% [SE 0·28], mean difference 1·36% [95% CI 0·67 to 2·05]) and prevention subgroups (2·60% [0·45] vs 0·64% [0·46], 1·96% [1·04 to 2·88]) (figure 2).
Change in mean bone mineral density of lumbar spine and femoral neck for (A) treatment and (B) prevention subgroups (modified intention-to-treat group)
Data are least-squares mean. Error bars=95% CI. BMD=bone mineral density. p values compare change in bone mineral density relative to baseline between drug groups, calculated from a three-way analysis of variance with adjustment for drug group, study region, and sex. *p=0·0005. p=0·0001. p=0·0050. p<0·0001. p=0·0156. p=0·0049.
Within the same period, zoledronic acid significantly increased bone mineral density at the femoral neck compared with risedronate, in both the treatment (1·45% [0·31, n=247 patients] vs 0·39% [0·30, n=239], 1·06% [0·32 to 1·79]) and prevention (1·30% [0·45, n=126] vs -0·03% [0·46, n=135], 1·33% [0·41 to 2·25]) subgroups. Similar findings were noted at the trochanter for the treatment (1·97% [0·31, n=247] vs 0·63% [0·31, n=239], 1·34% [0·59 to 2·08], p=0·0005) and prevention subgroups (2·75% [0·55, n=126] vs 0·48% [0·56, n=135], 2·27% [1·15 to 3·39], p<0·0001), and total hip for the treatment (1·65% [0·21, n=247] vs 0·45% [0·20, n=239], 1·21% [0·71 to 1·70], p<0·0001) and prevention subgroups (1·54% [0·36, n=126] vs 0·03% [0·36, n=135], 1·51% [0·78 to 2·23], p<0·0001). However, at the distal radius zoledronic acid increased bone mineral density compared with risedronate in the treatment (0·85% [0·27, n=239] vs 0·09% [0·26, n=237], 0·76% [0·11 to 1·40], p=0.0223) but not the prevention subgroup (0·06% [0·36, n=128] vs 0·47% [0·38, n=131], -0·42% [-1·17 to 0·34], p=0·2784). Furthermore, increases in bone mineral density at 6 months were significantly higher with zoledronic acid than with risedronate for the lumbar spine (figure 2), total hip (both subgroups, data not shown), trochanter (both subgroups, data not shown), and femoral neck (prevention subgroup only, figure 2). In a per protocol analysis, zoledronic acid showed significantly greater mean increases in bone mineral density at the lumbar spine than did risedronate in both the treatment (4·03% [0·29, n=235] vs 2·70% [0·28, n=230], 1·33% [95% CI 0·64 to 2·03], p=0·0002) and prevention (2·34% [0·50, n=112] vs 0·36% [0·51, n=116], 1·98% [0·99 to 2·96], p<0·0001) subgroups.
With the treatment and prevention subgroups combined, the frequency of new vertebral fractures was very low for patients receiving both zoledronic acid (n=5) and risedronate (n=3), with no significant difference between drug groups.
Absolute and unadjusted concentrations of biomarkers for bone resorption and formation were consistently reduced by both drugs (figure 3). However, reductions in both biomarkers at 12 months were significantly greater in patients on zoledronic acid than in those on risedronate in both the treatment and prevention subgroups. With the exception of P1NP in the prevention subgroup, zoledronic acid had an enhanced inhibitory effect on serum ß-CTx and P1NP at all timepoints from day 9-11 onwards in both subgroups. In the prevention subgroup, zoledronic acid was associated with significant reductions of P1NP from 3 months onwards (figure 3).
Mean values of markers of bone turnover for (A) treatment and (B) prevention subgroups (modified intention-to-treat group)
Data are least-squares mean. Error bars=SE. ß-CTx= ß-C-terminal telopeptides of type 1 collagen. P1NP=procollagen type 1 aminoterminal propeptide. p values compare change in bone mineral density relative to baseline between drug groups, calculated from a three-way analysis of variance with adjustment for drug group, study region, and sex. *p<0·0001. p=0·0013. p=0·0210.
In this trial, almost all patients in both drug groups (treatment: 96% on zoledronic acid [n=260 patients] and risedronate [n=262]; prevention: 95% [n=137] vs 99% [n=143]) met the definition of compliance and were taking at least 80% of their oral medication.
The overall occurrence of adverse events was significantly higher in the zoledronic acid group in both the treatment and prevention subgroups (table 2), mainly caused by a higher frequency of symptoms (eg, influenza-like illness, pyrexia) that were reported within 3 days of starting the drug. After 3 days, the occurrence of adverse events was similar in the two drug groups (data not shown).
The frequency of serious adverse events recorded by the investigators was similar between drug groups (table 2). In the treatment subgroup, the primary system organ classes that affected more than 2% of participants given zoledronic acid versus risedronate were musculoskeletal and connective tissues disorders (13 [5%] patients vs 17 [6%]); infections and infestations (12 [4%] vs 15 [5%]), respiratory, thoracic, and mediastinal disorders (nine [3%] vs eight [3%]); cardiac disorders (seven [3%] vs three [1%]); general disorders and those related to drug delivery site (six [2%] in each drug group); and gastrointestinal disorders (three [1%] vs seven [3%]). The most often reported serious adverse event for patients given zoledronic acid and risedronate was worsening rheumatoid arthritis, which was judged to be severe in six (2%) patients in each drug group. In the prevention subgroup, the four primary system organ classes that affected more than 2% of participants given zoledronic acid and risedronate were infections and infestations (five [3%] vs two [1%]), musculoskeletal and connective tissues disorders (four [3%] vs five [3%]), gastrointestinal disorders (three [2%] in each drug group), and cardiac disorders (two [1%] vs four [3%]). The most frequently reported serious adverse event was pyrexia, which was judged to be severe in two (1%) patients given zoledronic acid and one (1%) given risedronate. No meaningful differences were recorded between the drug groups in either the treatment or prevention subgroups within the cardiac disorders primary system organ class.
The adjudication committee independently assessed the case data and judged that of seven deaths during the trial (four on zoledronic acid and three on risedronate), none was related to the study drug. No cases of atrial fibrillation were judged to be serious adverse events. On the basis of the criteria defined by the adjudication committee, three patients given zoledronic acid had serious adverse events related to cardiac arrhythmia, with no confirmed events in the risedronate group. In two of the patients given zoledronic acid, an underlying medical disorder was deemed to be responsible for the arrhythmia. The third event, which took place 3 days after zoledronic acid infusion, was adjudicated as supraventricular tachycardia. No cases of osteonecrosis of the jaw or of delayed or non-union of fractures were noted. One case of osteonecrosis of long bones occurred in a patient given zoledronic acid, but osteonecrosis was suspected before enrolment and the patient was on high-dose corticosteroids of more than 7·5 mg daily for rheumatoid arthritis.
Most patients (treatment 93% [n=506 patients], prevention 88% [n=254]) had normal renal function (calculated creatinine clearance 60 mL/min or greater) at baseline. Confirmed adjudicated clinically significant renal events occurred in nine patients given zoledronic acid and six given risedronate, all but one of which were reversible. One event in a patient given zoledronic acid was adjudicated to be possibly associated with comorbid disorders (obstructive uropathy due to benign prostatic hyperplasia) and unlikely to be related to study medication. Three cases of acute renal failure (one on zoledronic acid and two on risedronate) were reported but all were adjudicated to be related to underlying diseases. Six patients in each drug group had confirmed adjudicated ocular events. In the zoledronic acid group, five patients had conjunctivitis and one had blepharitis. In the risedronate group, two patients had blurred vision, and one had each of episcleritis, conjunctivitis, diplopia, and increase of lacrimation. No clinically relevant changes were recorded for any haematological, biochemical, or urinalysis indices in either drug group. One patient given zoledronic acid had asymptomatic hypocalcaemia 11 days after infusion.
EQ-5D health-related quality-of-life data showed no significant differences between drug groups for either the visual analogue or utility score at 6 months or 12 months (table 3). The only significant difference was recorded for the utility score at 3 months in the prevention subgroup.
At the end of the study patients were asked which drug delivery method they preferred with respect to convenience, satisfaction, and willingness to use for a long period (many years). Most patients (n=785) stated a preference: 81% (n=636) preferred the intravenous preparation and 9% (n=68) the oral preparation for convenience, 78% (n=613) preferred the intravenous preparation and 8% (n=64) the oral preparation for satisfaction, and 84% (n=659) were willing to take the intravenous preparation long term and 9% (n=73) the oral preparation.
We have shown that one 5 mg infusion of zoledronic acid increases bone mineral density of the lumbar spine and femoral neck, trochanter, and total hip more than does 5 mg oral risedronate daily in patients already taking glucocorticoids. The enhanced effectiveness of zoledronic acid was evident within 6 months of starting treatment. The increases in lumbar spine bone mineral density induced by risedronate are of similar magnitude to those reported previously for glucocorticoid-induced osteoporosis prevention and treatment (table 4). Moreover, risedronate prevents loss of, but does not substantially increase, bone mineral density.8 By contrast, zoledronic acid induced substantial increases in bone mineral density of the lumbar spine and hip.
Zoledronic acid showed a faster, more substantial inhibitory effect on bone turnover biomarkers than did risedronate. This effect might be important for the prevention of glucocorticoid-induced osteoporosis because the reduction of bone mineral density and associated increased risk of fracture can occur within a few months of starting glucocorticoid treatment.7, 25 Notably, the lowest values of bone resorption were recorded in the first 9-11 days in the group given zoledronic acid and the values increased towards the final month of the study. Therefore, one 5 mg infusion of zoledronic acid provided sufficient reduction of bone turnover for 12 months, without complete inhibition of bone turnover.
The most important adverse effect on bone caused by glucocorticoid use is vertebral facture, which might happen at higher bone mineral density values than in postmenopausal osteoporosis.26 However, in our study the frequency of vertebral fractures was very low for the groups receiving zoledronic acid and risedronate, which is possibly an indication of the antifracture efficacy of both drugs. Wallach and co-workers'27 post-hoc analysis of 111 patients suggested that risedronate reduced vertebral fracture risk by 70% in patients receiving moderate-to-high doses of glucocorticoids. However, the recorded frequency of vertebral fractures was much higher (placebo 16% [n=18 patients], risedronate 5% [n=6])27 than in our study (risedronate <1%, zoledronic acid 1%). Our lower frequency of vertebral fractures might be attributable to differences in the method of fracture assessment, and the reduced age of the study cohort, which is associated with an increased baseline bone mineral density and therefore a reduced probability of fracture in glucocorticoid-induced osteoporosis.26 Furthermore, we used semi-quantitative assessment,21 whereas risedronate trials have used quantitative morphometry,27, 28 which might have identified further vertebral fractures.
The percentage difference of lumbar spine bone mineral density between zoledronic acid and risedronate after 12 months (prevention 1·96%, treatment 1·36%) is about half that recorded in studies comparing risedronate with placebo (prevention 3·4%,8 treatment 2·5%9). Although glucocorticoids might affect bone indices other than bone mineral density, such as microarchitecture, no randomised controlled trials in glucocorticoid-induced osteoporosis have assessed such markers. The effect of zoledronic acid on percentage change in bone mineral density, and the overall low vertebral fracture rate suggests that a much larger trial than this study is needed to establish whether the changes in bone mineral density and bone turnover as surrogate endpoints translate to improved fracture risk reduction.
5 mg oral risedronate daily is associated with a significant treatment effect on vertebral fractures, according to a systematic review.29 The main comparator trials to our study are two risedronate trials: one prevention study8 and one treatment study.9 Few active comparator studies of patients with glucocorticoid-induced osteoporosis have been reported.30-32 However, the results of our study suggest that the increase in bone mineral density that was induced by zoledronic acid, is generally greater than that induced by risedronate, and possibly other bisphosphonates..8-10 Teriparatide might be considered for the treatment of established glucocorticoid-induced osteoporosis because of its beneficial effect on bone formation, which is often suppressed in such a setting. Teriparatide reduced the frequency of vertebral fractures in the post-hoc analysis of one clinical trial.32 Similarly bisphosphonates prevent fracture according to a meta-analysis of clinical trials.29 Cost considerations and licensing laws might restrict the use of teriparatide, however, to short-term treatment of patients with pre-existing vertebral fractures or very low bone mineral density, who are taking glucocorticoids.
The primary endpoint of all glucocorticoid-induced osteoporosis studies until now has been lumbar spine bone mineral density, not vertebral fractures. The association remains unclear between change in surrogate endpoints, such as bone mineral density or biomarkers, and vertebral fracture. Despite the fact that the long-term efficacy of these surrogate endpoints remains unstudied, particularly with zoledronic acid, such endpoints remain widely used in clinical practice to assess fracture risk and treatment options. Furthermore, clinical studies of drugs for prevention of glucocorticoid-induced osteoporosis-related fractures have been quite short (eg, 1-2 years), and therefore physicians should be cautious when making assumptions about the long-term effectiveness of these drugs with respect to fractures.
We have shown that zoledronic acid has an acceptable safety and tolerability profile. The overall occurrence of adverse events was similar in patients on zoledronic acid and on risedronate in the treatment and prevention subgroups, except for the first 3 days after the infusion of zoledronic acid. During this period, the difference between drug groups was mainly caused by transient symptoms that have been associated with first intravenous infusion of bisphosphonates.19, 33 A lower frequency of pyrexia (8·7%) than in our study (12-15%), was noted when ibuprofen or paracetamol was co-prescribed with zoledronic acid.18
Ethical considerations precluded inclusion of a placebo group in our study, which meant that the overall effects of zoledronic acid and risedronate on adverse events could not be assessed. However, events that have previously been reported in patients given bisphosphonates (eg, osteonecrosis of the jaw, delayed or non-union of fractures, renal problems, hypocalcaemia, and ocular events) were uncommon. Moreover, zoledronic acid was not associated with a higher risk of these disorders than was risedronate.
Our study has several advantages for identification of the effects of bisphosphonates on glucocorticoid-induced osteoporosis-ie, large sample size, generalisability of the results by inclusion of both prevention and treatment subgroups for glucocorticoid-induced osteoporosis, and excellent retention rate. Although, the trial was limited by its short duration of 12 months, many diseases for which glucocorticoids are prescribed need only medium-term treatment of 12-18 months.
Guidelines recommend prescription of risedronate or alendronate in many patients given glucocorticoids, who are at increased risk of fracture.5, 6 However, our study has shown that one intravenous infusion of zoledronic acid provides greater increases in bone mineral density and more rapid and substantial decreases in bone turnover than does daily risedronate.
Men and women aged 18-85 years were eligible for inclusion in the study if they were receiving at least 7·5 mg oral prednisolone daily (or equivalent) and were expected to receive glucocorticoids for at least another 12 months. Patients were enrolled from 54 centres in 16 countries (Australia, Belgium, Czech Republic, Estonia, Finland, France, Hong Kong, Hungary, Israel, Lithuania, Poland, Romania, Spain, Switzerland, UK, and USA), and the study took place in teaching and community hospitals and clinics. Patients were selected from two cohorts: those who started taking glucocorticoids within the last 3 months, and those who had been taking glucocorticoids for longer than 3 months. Patients were screened for eligibility by radiography of the lumbar spine to ensure that at least three vertebrae in the region L1-L4 could be tested for bone density. Radiographs were assessed at Bioimaging Technologies (Leiden, Netherlands).
Major exclusion criteria were previous treatment with bisphosphonates or other drugs that affect the skeleton (except in accordance with a predefined washout schedule), serum 25-hydroxyvitamin D concentration of less than 30 nmol/L, recent history of cancer or parathyroid disease, and renal impairment (calculated creatinine clearance19 of less than 30 mL/min or proteinuria). Serum 25-hydroxyvitamin D concentration was measured with the Nichols20 assay except when the measured value was less than 29·9 nmol/L, in which case the DiaSorin20 assay was used for higher precision; from August, 2005, the DiaSorin assay was used exclusively by the central laboratory (Covance Central Labs, Geneva, Switzerland). Since the precision of the assays is different, the results were not combined. Pregnant women and those of childbearing potential who were not using adequate contraception were also excluded. The protocol was approved by the responsible centre-specific ethics committees. All patients gave written informed consent.
Participants were randomised in a 1:1 ratio (double-blind) to receive either 5 mg zoledronic acid as one 100 mL intravenous infusion over 15-20 min on day 1 and daily oral placebo, or 5 mg oral risedronate daily and one intravenous infusion of placebo on day 1; both regimens lasted 12 months. Randomisation to a drug group was done with an interactive voice response system (IVRS). At every patient's first visit to the clinic or hospital, IVRS was contacted and the patient was assigned a unique two-part number. Patients were screened for eligibility at the second visit, and the IVRS system was contacted again at the third visit to allocate a randomisation number to the prevention or treatment subgroup. Assignment of randomisation numbers was done with a minimisation algorithm to ensure balance across drug groups and within strata, according to the duration of glucocorticoid treatment before randomisation (≦3 months, >3 months). This algorithm ensured that across the treatment and prevention subgroups, at least 33% of the randomised patients were men and at least 60% were women. An automated validated system was used to randomly assign packs of the study drug to patients in the appropriate drug groups. Masking was achieved by the use of study drugs with identical packaging, labels, appearance, and odour. All patients, investigators, and analysts were blinded, with the exception of members of the data and safety monitoring board and the Novartis independent team (programmer, statistician, and physician), but the team did not assess patients or participate in writing of the report. The first patient was screened on June 16, 2004, and the last patient visit was on April 4, 2007. All patients were prescribed 400-1200 IU per day of supplemental vitamin D and 1 g per day calcium starting up to 28 days before the infusion and continuing throughout the trial.
Bone density was measured by dual-energy X-ray absorptiometry with Hologic (Waltham, MA, USA) or GE Lunar (Madison, WI, USA) axial bone densitometers at 6 and 12 months. We calculated the percentage change from baseline of bone mineral density at these timepoints. All bone density scans were read at BioImaging Technologies. The primary endpoint for drug efficacy was the percentage change from baseline in bone mineral density of the lumbar spine (L1-L4) at 12 months. As a secondary endpoint, we measured the percentage change from baseline in bone mineral density at other appendicular sites (total hip, femoral neck, trochanter, and distal radius) and established the occurrence of thoracic and lumbar vertebral fractures at 12 months, defined according to the Genant and colleagues'21 semi-quantitative method.
Changes in bone turnover biomarker concentrations were measured as secondary endpoints. We assessed relative change from baseline in fasting serum concentrations of both ß-C-terminal telopeptides of type 1 collagen (ß-CTx) with Elecsys 2010 Immunoassay System (Roche, Basel, Switzerland) (coefficient of variation within assay <7%, and between assay <10%), and procollagen type 1 aminoterminal propeptide (P1NP) with UniQ P1NP RIA (Orion Diagnostica Oy, Espoo, Finland) (coefficient of variation within and between assay <8%), at 9-11 days, 3 months, 6 months, and 12 months. ß-CTx and P1NP measured bone resorption and formation, respectively. Assays were done in duplicate. Each mean result was validated if the coefficient of variation after duplicate measurement was lower than 15%. Only two retests were possible if the coefficient of variation was higher than 15%. These analyses took place in a laboratory at the University of Liege (Liege, Belgium).
Safety was assessed from adverse events and serious adverse events recorded by the investigators. Independently, masked expert committees of between three and five members then adjudicated all laboratory criteria and targeted adverse events: ocular events (review by an external expert), osteonecrosis of the jaw, avascular necrosis at other skeletal sites, cardiac arrhythmias (reported as serious adverse events), deteriorating renal function, hypocalcaemia, delayed fracture healing, and primary cause of death. Each expert committee created a set of predefined search terms based on codes from the Medical Dictionary for Regulatory Activities22 and the WHO drug reference list,23 as described previously.17 The adverse event and concomitant medication databases were then searched for these terms. Investigators at each clinical centre gathered medical documentation for the cases. This documentation was forwarded to expert panels, whose members were unaware of drug group or subgroup assignments and adjudicated the events.
Renal function was assessed mainly by monitoring serum creatinine and calculated creatinine clearance19 before treatment, 9-11 days after treatment had started (along with assessment for hypocalcaemia), and at 3 months, 6 months, and 12 months. A significant increase was defined as a rise of more than 44 ƒÊmol/L in serum creatinine compared with the baseline value before the first infusion.
EQ-5D health-related quality-of-life scores were obtained at baseline, 3 months, 6 months, and 12 months, with both a visual analogue and utility score.24
Background and demographic characteristics (obtained during or before randomisation) were summarised for participants analysed on an intention-to-treat (ITT) and modified ITT (mITT) basis. The mITT group included all patients in the ITT population who received the study drug and who had measurements for the endpoint of interest at baseline and on at least one occasion post-baseline. For continuous variables, the number of patients, mean (SD), and median (IQR) were presented by drug group and treatment or prevention subgroup. For categorical variables, the number and percentage of patients in each category were presented for each drug group and subgroup. Comparisons between selected background and demographic characteristics were done with statistical tests for descriptive purposes and to assess comparability across drug groups and subgroups. Categorical variables were analysed by ƒÔ2 test, and continuous variables by one-way ANOVA. Note that these comparability tests were undertaken for descriptive reasons only.
The magnitude of the effect of risedronate on change in bone mineral density differs for glucocorticoid-induced osteoporosis treatment and prevention,8, 9 and therefore the statistical analysis was done separately for the treatment and prevention subgroups. To show that the percentage change from baseline in lumbar spine bone mineral density at 12 months with zoledronic acid was non-inferior to risedronate, different non-inferiority margins were used for the treatment and prevention subgroups. We calculated the sample size that was necessary to achieve a statistical power of 85% after adjustment for a 10% dropout rate-504 participants for the treatment subgroup and 256 for the prevention subgroup. If the one-sided 97·5% CIs for the absolute difference in percentage change between the groups on zoledronic acid and risedronate exceeded -0·70% (treatment subgroup) and -1·12% (prevention subgroup), then non-inferiority was established. In this case, superiority was investigated with an ANOVA model that included drug group, sex, and study region as covariates. A closed testing procedure was used to first assess the treatment subgroup, and then the prevention subgroup for non-inferiority.
Drug efficacy from bone mineral density and biomarkers was analysed in the mITT group. Lumbar spine bone mineral density was measured in at least two assessable vertebrae in the region L1-L4 at both baseline and at month 12; these vertebrae had to be available for assessment at baseline and at least one visit after randomisation for the patient to be included in the mITT group. Differences between groups were investigated with ANOVA or ANCOVA. Analyses of bone mineral density data were based on three-way ANOVA models adjusted for drug group, sex, and study region. Bone biomarker analyses were done with log(e) ratios of drug group to baseline (ie, geometric mean) at baseline, day 10, and months 3, 6, and 12 with an ANCOVA model adjusted for drug group, study region, and log(e) of baseline. For all statistical comparisons, two-sided p values of less than 0·05 showed statistical significance, and a closed testing procedure was used to control for the type 1 error rate of secondary efficacy variables. No interim analyses were undertaken.
We summarised adverse events that were not present at baseline and became evident after the first dose, and events that were present at baseline and increased in severity after the first dose. Adverse events were categorised according to codes used in the Medical Dictionary for Regulatory Activities,22 and analysed in the ITT group. The number and percentage of patients reporting adverse events were summarised by drug group according to primary system organ class, preferred term, and maximum severity. If a patient reported more than one adverse event with the same preferred term, the event with the greatest severity was recorded. If a patient reported more than one adverse event within the same primary system organ class, the patient was counted only once with the class of greatest severity recorded. Analyses were done with Fisher's exact test.
Visual analogue and utility scores were calculated from descriptive summary statistics (number of patients, mean [SD], and median [minimum-maximum]) for baseline, each visit (3, 6, and 12 months), and change from baseline at each visit, by drug group. Number and percentage of patients who answered the five parts of the questionnaire were assimilated. Differences in change from baseline scores between drug groups were assessed at each visit with an ANCOVA model (two-sided, ƒ¿=0·05) in the ITT group, adjusted for drug group and study region, and with baseline score as a covariate.
This trial is registered with ClinicalTrials.gov, number NCT00100620.
Role of the funding source
The sponsor of the study participated in study design, data analyses, data interpretation, and writing of the report via the 13-member trial steering committee, which included six representatives of the sponsor. The sponsor had responsibility for data collection and quality control. An independent data monitoring committee met twice a year to oversee the conduct of the study, and to monitor patient safety. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.