Answers are grounded in this paper and cite their sources.
Objective: To review the effect of alendronate on bone density and fractures in postmenopausal women. Data Source: We searched MEDLINE, EMBASE, Current Contents, and the Cochrane Controlled trials registry from 1980 to 1999, and we examined citations of relevant articles and proceedings of international meetings. Study Selection: We included 11 trials that randomized women to alendronate or placebo and measured bone density for at least 1 yr. Data Extraction: For each trial, three independent reviewers assessed the methodological quality and abstracted data. Data Synthesis: The pooled relative risk (RR) for vertebral fractures in patients given 5 mg or more of alendronate was 0.52 [95% confidence interval (CI), 0.43–0.65]. The RR of nonvertebral fractures in patients given 10 mg or more of alendronate was 0.51 (95% CI 0.38–0.69), an appreciably greater effect than for the 5 mg dose. We found a similar reduction in RR across nonvertebral fracture types; in particular, RR reductions for fractures traditionally thought to be “osteoporotic,” such as hip and forearm, were very similar to RR reductions for “nonosteoporotic” fractures. Individual studies showed similar results, reflected in the P values of the test of heterogeneity (P = 0.99 for vertebral and 0.88 for nonvertebral fractures). Alendronate produced positive effects on the percentage change in bone density, which increased with both dose and time. After 3 yr of treatment with 10 mg of alendronate or more, the pooled estimate of the difference in percentage change between alendronate and placebo was 7.48% (95% CI 6.12–8.85) for the lumbar spine (2–3 yr), 5.60% (95% CI 4.80–6.39) for the hip (3–4 yr), 2.08% (95% CI 1.53–2.63) for the forearm (2–4 yr), and 2.73% (95% CI 2.27–3.20) for the total body (3 yr). Heterogeneity of the treatment effect of alendronate was not consistently explained by any of our a priori hypotheses; in particular, the effect was very similar in prevention and treatment studies. The pooled RR for discontinuing medication due to adverse effects for 5 mg or greater of alendronate was 1.15 (95% CI 0.93–1.42). The pooled RR for discontinuing medication due to gastro-intestinal (GI) side effects for 5 mg or greater was 1.03 (0.81–1.30, P = 0.83), and the pooled RR for GI adverse effects with continuation of medication was 1.03 (0.98 to 1.07) P = 0.23. Conclusions: Alendronate increases bone density in both early postmenopausal women and those with established osteoporosis while reducing the rate of vertebral fracture over 2–3 yr of treatment. Reductions in nonvertebral fractures are evident among postmenopausal women without prevalent fractures and have bone mineral density (BMD) levels below the World Health Organization threshold for osteoporosis. The impact on fractures appears consistent across all fracture types, casting doubt on traditional distinctions between osteoporotic and nonosteoporotic fractures. ALENDRONATE SODIUM, a bisphosphonate and antiresorptive agent, was developed and marketed as an intervention to reduce vertebral and nonvertebral fractures in postmenopausal women (1–3). Alendronate does not impair bone mineralization at doses that maximally inhibit bone resorption (4). A previous systematic review examining the effect of alendronate on nonvertebral fractures (2) did not clearly report their search methods, address the methodological quality of the individual trials, include unpublished data, or examine the effect on vertebral fractures or BMD. Furthermore, the 95% CI on the RR bordered on no effect (0.50–1.0). We therefore undertook a systematic review and meta-analysis of the effect of alendronate on bone density and fractures. An additional motivation for the review was to explore whether the effect of alendronate was consistent across studies, and whether results differed in those with milder vs. more severe osteoporosis, those ingesting more or less calcium or vitamin D, across different doses and durations of alendronate therapy, and in different fracture sites. We followed the procedures defined by the Cochrane Collaboration for conducting a systematic review (5). Trials satisfied the following inclusion criteria: 1) randomized placebo-controlled trials (RCTs) comparing postmenopausal women receiving alendronate to those not receiving alendronate with follow-up of at least 1 yr; and 2) fracture incidence, or BMD data (including percentage change from baseline and a measure of variance) available. We made no restriction by country in which the trial occurred, nor did we limit the language to English. Eleven trials total met our inclusion criteria (1, 3, 6–12, 15, 14). To identify relevant studies of alendronate therapy, we used the search strategy outlined in Section I and used the following key and text words: bisphosphonates, diphosphonates, osteoporosis, postmenopausal, and alendronate (15). We reviewed citations of relevant articles and conference proceedings. Two reviewers (A.C., V.R.) examined each title generated from the search and identified potentially eligible articles for which we obtained the abstracts. For abstracts consistent with study eligibility, we obtained the full article text. Three reviewers judged study eligibility, assessed the methodological quality, and abstracted the data. We achieved a high level of chance corrected agreement in our rating of methodological quality for blinding (κ 1.0) and moderate agreement for loss to follow-up (κ 0.6). We examined the effect of alendronate on vertebral and nonvertebral fractures, bone density of the total body, lumbar spine, hip, and forearm, as well as adverse effects of the drug as manifested in patients unable to continue study medication. We developed, for fractures and bone density, a priori hypotheses that might explain heterogeneity of study results as outlined in Section I. Specifically, we compared groups according to 1) prevention vs. treatment; 2) concurrent treatments including total calcium intake (1250 mg of total calcium vs. <1250 mg); 3) vitamin D supplementation; 4) for nonvertebral fractures, sites in which fractures are strongly associated with low bone density and those not as strongly associated; and 5) individual components of the quality assessment, including concealed randomization, blinding, loss to follow-up, and intention-to-treat analysis. Section I includes a full description of the methodology used to separate prevention from treatment studies. To further evaluate heterogeneity of the severity of osteoporosis in women with BMD 2 sd values below the mean, we grouped trials with a prevalent vertebral fracture rate of greater than 10% at study inclusion (1, 3, 8, 12, 13), in comparison to those without fractures or prevalence below 10% (6, 7, 9, 10, 14). We chose the 10% cut because there was a cluster of studies just less than 10%, and a number of studies with fracture rates much greater than 10%. We also evaluated whether the treatment effect varied depending on baseline total calcium intake (cut point of 1250 mg) and presence of vitamin D supplementation. Reviewers abstracted data regarding study design, patient characteristics, treatment duration, dosage, mean change, and sd values for bone density, and number of fractures, both vertebral and nonvertebral. We did not include data from the calcitonin treatment arm of Adami et al. (9) or the hormone replacement arm of the Hosking et al. (7) trial. We sought data missing from the original reports through correspondence with the primary investigators and with the company, Merck, that sponsored the trials. Merck provided us with the clinical study reports for all of the published trials included in this meta-analysis (1, 3, 6–10). For the trials not yet published at the time we collected data, we were able to obtain the clinical studies reports from Merck for the Clinical Fracture Arm of the Fracture Intervention Trial (FIT) (14), the draft manuscript of two trials (Refs. 11 and 12 ; both have since been published), and the data from the most recently completed RCT (13). When we found duplicate reports of the same study in preliminary abstracts and articles (16–18), we analyzed data from the most complete data set (1, 9). A random effects model provided the strategy for final estimates of all treatment effects, whether for fractures, bone density, or toxicity (19). We only used data on one fracture per person in the analysis. We conducted separate analyses using regression models for each bone density site (lumbar spine, combined hip, forearm, and total body) using the difference between the change in bone density for each dose group and the change in the placebo arm. Across both fractures and bone density, the dominant parsimonious models generally allowed combining doses of 1 and 2.5 mg, almost invariably allowed combining the results from 10-, 20-, and 40-mg doses, and sometimes allowed combining the 5-mg dose with doses of 10–40 mg. Because the effect was smallest with doses of 1–2.5 mg, and because clinicians are not using doses lower than 5 mg, we present data only from arms using doses of 5–40 mg. Although clinicians are not using doses of 20–40 mg, in all but one instance we found their effect is similar to 10 mg, and including these trials allows a more precise estimate of the treatment effect associated with doses of 10–40 mg. With regard to the duration of therapy, the dominant parsimonious model for bone density kept all years separate. In the one exception, we found we could combine yr 2 and 3 for total body bone density. Because clinicians should appropriately offer alendronate therapy for at least 2 yr, we present only data from two or more years of follow-up. For the bone density analysis, if there was statistically significant heterogeneity between studies (P < 0.05), we divided the studies into two groups according to a priori hypotheses and tested whether the effect sizes differed in the two groups of studies (20). For each fracture analysis, we calculated a RR and tested for heterogeneity using a χ2 procedure (19). The same analytic strategy was used to deal with the proportion of patients who discontinued medication because of adverse effects. We reviewed a total of 358 articles and abstracts. Of these, 46 warranted closer examination, and 19 proved to be possible RCTs (Fig. 1). Of these 19, we excluded 8 for the following reasons: duplicate report or earlier report of another study (16–18), lack of a control group (21), duration less than 1 yr (22–24), and outcomes limited to histomorphometric data (25). Search results for alendronate review. Table 1 presents the characteristics of the 11 trials that met our eligibility criteria. The trials included a total of 12,855 women, of whom 5,561 received placebo. Two trials dealt with prevention (6, 7), and the other nine trials involved women whose densitometry showed low bone density (1, 3, 8, 9–14). Of the nine trials in women with low bone density, six included women with a prevalence of vertebral fractures at the beginning of the study of over 10% (Table 1). The Fracture Intervention Trial, which was the largest treatment trial, used a dose of 5 mg for the initial 2 yr and then switched to a 10-mg dose for the third year. Study characteristics for Alendronate trials Refer to a priori hypotheses defining prevention and treatment. Mean BMD of femoral neck site. BMD g/cm2 corrected to Hologic measurements with sdin parentheses. Study characteristics for Alendronate trials Refer to a priori hypotheses defining prevention and treatment. Mean BMD of femoral neck site. BMD g/cm2 corrected to Hologic measurements with sdin parentheses. All of the trials effectively concealed randomization, and blinded patients, clinicians, and those assessing outcome. Heterogeneity in study methodological quality was restricted to loss to follow-up. Two trials achieved loss to follow-up of less then 5%, losses to follow-up varied between 5 and 20% in six trials, and three trials had losses to follow-up over 20% (Table 1). For vertebral fractures, our analysis failed to reject pooling across the entire range of doses from 5 to 40 mg. The pooled estimate of the RR of vertebral fractures with alendronate from 8 trials randomizing patients to a dose of 5 mg of alendronate or greater (1, 3, 6–10, 14) showed a RR of 0.52 (95% CI 0.43–0.65), which was consistent across trials (heterogeneity P value 0.99; Fig. 2). Thus, it is not surprising that, of our five a priori hypotheses, none showed statistically significant differences between subgroups. In particular, the point estimates for the reduction in fracture risk was similar for the prevention trials (RR of 0.45, 95% CI 0.06–3.15) and treatment trials (RR of 0.53, 95% CI 0.43–0.65. P value on difference between estimates 0.87). RR for vertebral fractures with alendronate (5 mg and greater). Table 2 presents the RR for fracture with alendronate in the groupings of the nonvertebral fractures we examined. We found statistically significantly smaller effect sizes for 5 mg than for 10–40 mg of alendronate in every fracture category. At doses in the 10- to 40-mg range, the effect on each fracture category is similar, with the RRs varying from 0.45 to 0.57. The results across studies are quite similar, reflected in both the heterogeneity P values in Table 2 and Fig. 3, which presents the results of doses of 10 mg or more for individual studies for total nonvertebral fractures. Once again, the nonsignificant reduction in risk of all nonvertebral fractures in the one prevention trial (RR of 0.79, 95% CI 0.28–2.24) was not clearly different from the five treatment trials (RR 0.49, 95% CI 0.36–0.67, P value on difference between estimates 0.40). These results did not change for the 10- to 40-mg dose if we used a femoral neck T-score cut point of −2.5, instead of −2.0, to classify the trials into prevention and treatment. When we pooled across studies using 5 mg and greater for outcome of hip fracture, the pooled RR is 0.63 (0.43,0.92). The point estimate lies between the 5 mg and 10 to 40 mg estimates, and the 95% CI excludes no effect. Risk ratios for nonvertebral fractures with alendronate (10 mg and greater). Weighted mean difference of bone density after treatment with alendronate We interpreted the heterogeneity P ≤ 0.05 as indicating important between-study differences in results. Osteoporotic fractures defined by all fractures in which calcaneal bone density was associated with a RR of fracture of 1.5 or greater in a prior study (forearm, hip, rib, leg, femur, tibia and fibula, patella, pelvis, and hands) (29). Nonosteoporotic fractures defined by all fractures in which the RR of fracture was less than 1.5 in a prior study (29). Weighted mean difference of bone density after treatment with alendronate We interpreted the heterogeneity P ≤ 0.05 as indicating important between-study differences in results. Osteoporotic fractures defined by all fractures in which calcaneal bone density was associated with a RR of fracture of 1.5 or greater in a prior study (forearm, hip, rib, leg, femur, tibia and fibula, patella, pelvis, and hands) (29). Nonosteoporotic fractures defined by all fractures in which the RR of fracture was less than 1.5 in a prior study (29). The methods of densitometry included dual x-ray absorptiometry—Hologic, Lunar, and Norland. Table 3 presents the results of the pooled estimates of effect of alendronate on bone density across the four sites. All sites had a significant positive response to treatment with alendronate (P < 0.01), but we found consistently larger effects with the higher doses of alendronate than with the 5-mg dose, and larger effects as duration of follow-up increased in the lumbar spine (Fig. 4) and forearm and hip sites. For combined forearm, we found no statistically significant heterogeneity, whereas for total body, lumbar spine, and hip there were differences in the treatment effect between trials in some subgroups (Table 3). Even when statistically significant heterogeneity existed, all studies still demonstrated appreciable differences in favor of alendronate. Weighted mean difference for lumbar spine with alendronate (10–40 mg). Weighted mean difference of bone density after treatment with alendronate We interpreted the heterogeneity P ≤ 0.05 as indicating important between-study differences in results. If a trial reported more than one forearm site, our order of preference was 1/3 distal radius and ulna and then 1/3 distal radius. If a trial reported more than one hip site, our order of preference was total hip, femoral neck, and then trochanter. Weighted mean difference of bone density after treatment with alendronate We interpreted the heterogeneity P ≤ 0.05 as indicating important between-study differences in results. If a trial reported more than one forearm site, our order of preference was 1/3 distal radius and ulna and then 1/3 distal radius. If a trial reported more than one hip site, our order of preference was total hip, femoral neck, and then trochanter. None of the possible explanations satisfactorily explained significant heterogeneity. In three comparisons (lumbar spine 10–40 mg, 2–3 yr; combined hip 5 mg and 10–40 mg, 2 yr), we found larger effects in trials in which neither treatment nor control received vitamin D. The magnitude of the effect was very similar in prevention and treatment studies. In the one instance (total body bone density) in which it differed, the effect was greater in the prevention trials. Funnel plots provided no suggestion of publication bias for any of bone density or vertebral or nonvertebral fractures. pooled estimate of the RR of discontinuing medication as a of adverse effects from trials using 5 mg of alendronate or more of 1.15 (95% CI was very consistent across trials (heterogeneity P value The pooled RR for discontinuing medication to GI side effects for 5 mg or greater from trials was 1.03 (0.81–1.30, heterogeneity P = The pooled RR for GI adverse effects from 10 trials with continuation of medication was 1.03 heterogeneity P = meta-analysis that alendronate results in a reduction in the RR of vertebral and nonvertebral fractures. For nonvertebral fractures, the effect was larger at doses of 10 mg or greater than for lower Although we found a similar larger risk reductions for vertebral fractures with doses of 10 mg or we could not chance as the for this to heterogeneity was limited for some due to the low number of fractures in some Of in this regard is our of differences in effect between treatment and prevention trials. We found very similar effects on bone density in prevention and treatment trials and a similar RR of fracture in trials with varying baseline levels of bone density. Although our bone density results are the number of fractures in the prevention trials was very and the very a our analyses have very to differences in relative risk reduction of fracture with alendronate and control across prevention and treatment lack of to individual patient data limited our to explore this An analysis from one study that effect in reducing nonvertebral fractures is greater in patients who treatment with lower bone density. have been in clinical trials with other bisphosphonates, including and In the data have not the of whether important differences in risk reduction across groups of patients with varying of osteoporosis The impact of alendronate on the relative risk of nonvertebral fracture in without in bone density further to the of whether our estimates of alendronate effect to all nonvertebral fractures, most investigators have that reduce fracture rates that are associated with low bone density and In with this of we made a priori hypotheses that we might larger reduction in the RR of fractures of the hip, spine, forearm, or other fractures associated with bone density. we found that the RR reduction with alendronate doses of 10 mg or greater was hip, and for fractures traditionally To this that our pooled estimate of RR and the associated CI to all nonvertebral fractures. to an appreciable across osteoporotic and nonosteoporotic fractures prior in this With to bone density, we found that doses of alendronate 10 mg or greater had larger effects on bone density than smaller Alendronate at doses of 10–40 mg increased bone density with larger in sites with a high proportion of Although we larger effects as time the of treatment over control in bone density with each year. In analyses of differences in percentage change in bone mineral density, there was with statistically significant heterogeneity, alendronate increased bone density in all trials. When significant heterogeneity was we failed to a the impact of alendronate on fractures, we not the duration of follow-up of the trials in our review. Data from trials that are whether the effect on vertebral fractures is or our complete data from all published and unpublished studies. a we were in relevant data. in data a of this systematic review. In the trials were methodological was the loss to follow-up (Table 1). two trials achieved loss to follow-up of less than 14). is that the proportion of patients to follow-up did not to the magnitude of the treatment effect in any is the of follow-up, yr or less in all studies to The RR for discontinuing alendronate due to adverse effects was 1.15 (95% CI 0.93–1.42). of the of data on adverse effects from is that in RCTs to be with and the results not be to clinical For in a number of the alendronate trials, women were excluded if had a of or a of study In RCTs are for effects, and of RCTs generally drug Table presents the reductions in fracture risk we could over a in women with varying fracture risk receiving a dose of alendronate of 5 mg or greater for vertebral fractures. To these we have baseline risk from pooled estimates from the prevention trials of patients with bone density and from the treatment trials that women with low BMD Table also presents the reduction in fracture risk for high risk women receiving 10 mg or greater for nonvertebral fractures. We have not included the reduction for the low risk as we have the magnitude of the RR reduction in nonvertebral fractures with alendronate in to RR reduction and number to for all estimates of vertebral and nonvertebral fractures pooled RR of 0.52 for vertebral fractures and 0.51 for nonvertebral and hip fractures. RR reduction and number to for all estimates of vertebral and nonvertebral fractures pooled RR of 0.52 for vertebral fractures and 0.51 for nonvertebral and hip fractures. With the pooling of data in this analysis, we have provided a much more precise estimate of the magnitude in reduction of not only vertebral but nonvertebral fracture with alendronate than was available. Furthermore, our model in the of the CI of still a reduction in RR of nonvertebral studies should on such as whether alendronate nonvertebral fractures in women with and whether calcium effect on fractures. The to and from Merck for their additional data, and for the of this review and on earlier of the The also to and for their in additional data and for the search was in through from Merck, mineral confidence randomized placebo-controlled relative
No takes yet. Share an insight, caveat, or question.
Cranney et al. (2002) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: