For each study, the solid circle indicates the effect size from the random effects model; whiskers, the 95% CI of the study; and gray block, the relative size of the study compared with other studies. The diamond indicates the overall effect size of all studies combined.
eTable 1. PRISMA Checklist
eTable 2. PubMed Search Terms
eTable 3. Cochrane Risk of Bias Assessment
eFigure 1. Forest Plot of Alendronate Treatment for Osteoporosis and Overall Mortality
eFigure 2. Forest Plot of Bisphosphonate Excluding Zoledronic Acid Treatment for Osteoporosis and Overall Mortality
eFigure 3. Forest Plot of All Nitrogen Bisphosphonate Treatment for Osteoporosis and Overall Mortality
eFigure 4. Forest Plot of Nitrogen Bisphosphonate Excluding Zoledronic Acid Treatment for Osteoporosis and Overall Mortality
eFigure 5. Forest Plot of Trials of All Treatments With 3 Years or More Duration and Total Mortality
eFigure 6. Forest Plot of Bisphosphonate Treatment With 3-Year Study for Osteoporosis and Overall Mortality
eFigure 7. Forest Plot of Bisphosphonate Treatment Excluding Zoledronic Acid Treatment With 3-Year Study for Osteoporosis and Overall Mortality
eFigure 8. Meta-regression of the Association Between Risk Ratio of Mortality Associated With Treatment (Risk Ratio of Death) and Mortality Rate in the Placebo Groups of the Trials
eFigure 9. Meta-regression of the Association Between Risk Ratio of Mortality Associated With Bisphosphonate Treatment (Risk Ratio of Death) and Mortality Rate in the Placebo Groups of the Trials
eFigure 10. Funnel Plot for All Trials
eFigure 11. Funnel Plot for Bisphosphonate Trials
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Cummings SR, Lui L, Eastell R, Allen IE. Association Between Drug Treatments for Patients With Osteoporosis and Overall Mortality Rates: A Meta-analysis. JAMA Intern Med. 2019;179(11):1491–1500. doi:10.1001/jamainternmed.2019.2779
Are drug treatments, particularly treatment with bisphosphonates, for the prevention of fracture in patients with osteoporosis associated with lower overall mortality?
This meta-analysis of 38 randomized clinical trials of drug therapies, particularly bisphosphonate treatments, that included 101 642 unique patients with osteoporosis found that these therapies were not associated with reduced overall mortality rates.
This meta-analysis suggests that drug treatments, including treatment with bisphosphonates, for osteoporosis should be recommended only for the prevention of fracture and not for any additional reduction in mortality.
Previous studies have reported that drug treatments, particularly treatment with bisphosphonates, is associated with reduced overall mortality rates in addition to decreased fracture risk. If so, drug treatments should be recommended for this reason alone, regardless of a patient’s risk of fracture.
To assess whether randomized clinical trials demonstrate that treatment with bisphosphonates, particularly zoledronate, is associated with reduced mortality rates.
Science Direct, MEDLINE, Embase, and the Cochrane Library were searched for randomized placebo-controlled clinical trials of drug treatments for osteoporosis published after 2009 and published or in press before April 19, 2019. Conference abstracts from annual osteoporosis society meetings were also included in the search.
Included studies were clinical trials that (1) were randomized and placebo-controlled; (2) studied drug treatments with proven antifracture efficacy; (3) used agents at the approved dose for treatment of osteoporosis; and (4) had a duration of 1 year or more. Abstracts from the literature searches were reviewed for inclusion and exclusion criteria, and mortality rate data were abstracted from the article by 1 researcher and validated by a second. A total of 2045 records were screened; 38 (1.8%) were included in the meta-analyses.
Data Extraction and Synthesis
The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) checklist was followed for abstracting data and assessing data quality and validity. Data were pooled using random-effects models, and between-study variability was assessed using the I2 index. The risk of bias for each study was assessed, and funnel plots and Egger and Begg statistics were used to evaluate publication bias.
Main Outcomes and Measures
Associations of all drug treatments, particularly bisphosphonate and zoledronate treatments, with overall mortality.
Of 38 clinical trials that included 101 642 unique participants, 38 were included in the meta-analyses of all drug treatments (45 594 participants randomized to placebo; 56 048 to treatment); 21 clinical trials, of bisphosphonate treatments (20 244 participants randomized to placebo; 22 623 to treatment); and 6 clinical trials, of zoledronate treatments (6944 participants randomized to placebo; 6926 to treatment). No significant association was found between all drug treatments for osteoporosis and overall mortality rate (risk ratio [RR], 0.98; 95% CI, 0.91-1.05; I2 = 0%). Clinical trials of bisphosphonate treatment (RR, 0.95; 95% CI, 0.86-1.04) showed no significant association with overall mortality. Also, clinical trials of zoledronate treatment (RR, 0.88; 95% CI, 0.68-1.13) showed no association with overall mortality rate; however, evidence existed for heterogeneity of the results (I2 = 48.2%).
Conclusions and Relevance
Results of this meta-analysis suggest that bisphosphonate treatment may not be associated with reduced overall mortality rates in addition to decreased fracture risk and should only be recommended to reduce fracture risk. Additional trials are needed to clarify whether treatment with zoledronate reduces mortality rates.
The purpose of treating patients with medications for osteoporosis is to reduce the risk of fracture and the subsequent pain and disability.1-3 Preventing fractures may also lessen the increased risk of mortality owing to fractures. Hip and vertebral fractures are often followed by increased overall mortality,4-12 and some studies have reported that most types of fractures are associated with increased mortality.12-15 However, much of the increased risk for both fracture and mortality may be owing to the patients’ poor health or other factors; thus, it is not clear whether or to what degree mortality following fractures could be reduced by preventing fractures.16-18 Studies have estimated that less than 30% of the mortality following hip and vertebral fractures may be attributed to the fracture itself and, therefore, potentially avoidable by preventing the fracture.19,20
Some studies have suggested that treatments for osteoporosis may directly reduce overall mortality rates in addition to decreasing fracture risk. Several observational studies have reported that patients with osteoporosis who have received drug treatments, particularly bisphosphonates (largely oral versions, such as alendronate sodium), experienced 25% to 60% lower overall mortality that is too large to be attributed to a reduction in fractures and may have been owing to the direct effect of the drug treatments. This substantially lower overall mortality rate has been reported among patients with several different characteristics, including those living in communities,21 living in institutional care,22 admitted to intensive care units,23 prescribed treatments after a fracture,23-25 or receiving treatments from a fracture liaison service as part of an osteoporosis management program.26,27 It seems unlikely that bisphosphonate treatments substantially reduce all-cause mortality in addition to reducing fracture risk because these drugs bind almost exclusively to bone, are cleared from the blood to extremely low or undetectable levels within 24 hours of administration, and are only detectable in extremely low concentrations in tissues other than bone.28 Although the association between drug treatments for osteoporosis and reduced mortality rates remained significant after adjusting for several potential confounders and matching propensity scores, it is possible that these associations were owing to other confounding factors that were not measured; for example, those who took drug treatments for the prevention of fracture may have had better health and nutrition, exercised more frequently, or used other preventive measures more often than those who did not.
Two randomized placebo-controlled clinical trials of zoledronate therapy suggested that patients randomized to receive zoledronate treatment experienced benefits beyond fracture prevention. One randomized clinical trial reported a statistically significant 28% lower mortality rate in patients who had a recent hip fracture.29 A recent, 6-year clinical trial of zoledronate treatment administered to older women without osteoporosis reported significant reductions in the risk of myocardial infarction and breast cancer, with a 35% lower mortality rate that was not statistically significant.30 In contrast, a larger randomized clinical trial of zoledronate treatment administered to women with osteoporosis noted no effect on mortality rate.31
A 2010 meta-analysis of clinical trials of antiresorptive drugs, including bisphosphonates, reported that those treatments reduced overall mortality by 10%.32 Another meta-analysis of 61 clinical trials of bisphosphonate treatments reported a 10% lower overall mortality rate; however, most of the studies were of patients with cancer, for whom it is known that bisphosphonate therapy reduces metastases to bone, and many of the clinical trials were not placebo-controlled.33
If drug therapies for patients with osteoporosis, particularly treatments with bisphosphonates or zoledronate, substantially reduce mortality in addition to decreasing fractures, it may be worthwhile to administer these treatments to almost all older adults, regardless of their fracture risk. Therefore, we conducted a systematic review and meta-analysis of randomized placebo-controlled clinical trials of drug treatments for osteoporosis to evaluate whether these treatments reduced overall mortality. We specifically tested the hypothesis that treatment with bisphosphonates or zoledronate reduces the overall mortality rate.
This study followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline (eTable 1 in the Supplement).34 We systematically searched MEDLINE (via PubMed), Science Direct, Embase, and the Cochrane Library databases for all randomized placebo-controlled clinical trials of drug treatments for osteoporosis with data on mortality rates published after 2009 and published or in press before April 19, 2019. Studies published before 2009 were added based on the previous meta-analysis by Bolland et al.32 Additional published studies were identified by reviewing the reference lists of meta-analyses and the conference abstracts from annual meetings of the American Society for Bone and Mineral Research and the European Calcified Tissue Society. The search strategy is described in eTable 2 in the Supplement. The protocol for the meta-analysis was registered on Prospero, the international prospective register of systematic reviews (CRD42018106037).
Studies included in this meta-analysis were clinical trials that (1) were randomized, double-blind, and placebo-controlled; (2) studied drug treatments with proven antifracture efficacy; (3) used agents at the approved dosage for treatment of osteoporosis; and (4) had a duration of 1 year or more.
Studies excluded were clinical trials that (1) studied estrogen, tibolone, and strontium therapies; (2) were duplicate publications in which only the final, complete trial results were reported; (3) studied patients who were treated with corticosteroids, generally owing to underlying inflammatory diseases; (4) studied patients with cancer; and (5) evaluated open-label drug treatments. If a clinical trial studied more than 1 type of drug treatment, each treatment arm was compared with the placebo group.
Citations and abstracts identified from the literature searches were reviewed for inclusion and exclusion criteria by a single reviewer (I.E.A.), and full text review and data extraction from published articles were reviewed by 2 researchers (L.-Y.L. and I.E.A.). Mortality rate data were abstracted from the text of the article by 1 researcher (I.E.A.) and validated by a second researcher (L.-Y.L.).
Data were pooled using random-effects models to control for any between-study variability, which was assessed using the I2 index and the Cochran Q statistic. The risk of bias for each study was assessed using the Cochrane risk-of-bias chart (eTable 3 in the Supplement). Funnel plots and Egger and Begg statistics were used to examine publication bias. Risk ratios (RRs) for mortality comparing treatment with placebo were calculated with a 95% CI, and all analyses were performed using Stata statistical software, version 15.1 (StataCorp LLC).
The prespecified analyses included the effects of treatment with all drugs, treatment with bisphosphonates, and treatment with zoledronate; we designated these analyses as primary because they had reported an association between drug treatment and reduced mortality in observational studies and, for zoledronate treatment, in randomized clinical trials. We also analyzed the association between overall mortality rates and patients receiving drug treatment for 3 years or more. In post hoc analyses, we assessed the association between overall mortality rates and specific drug treatments, including alendronate only, bisphosphonates without zoledronate, and nitrogen-containing bisphosphonates with and without clinical trials of zoledronate as well as treatments without zoledronate lasting 3 years or more.
A total of 2765 records were identified in the initial search. After duplicates were removed and additional records from other sources were added, 2045 records were screened (Figure 1). Of those screened, 108 clinical trials were evaluated for inclusion and exclusion criteria, and 38 randomized placebo-controlled clinical trials were included in the meta-analyses and comprised 101 642 unique patients: 45 594 participants were randomized to placebo and 56 048 to treatment. These studies are listed in the Table.29-31,35-68
There was no significant association between receiving a drug treatment for osteoporosis and overall mortality rate (RR, 0.98; 95% CI, 0.91-1.05; P = .56; Figure 2).29-31,35-68 Although the analysis included diverse types of drugs—bisphosphonates, denosumab (a receptor-activated nuclear factor–κB ligand), selective estrogen receptor modulators, parathyroid hormone analogues, odanacatib (a cathepsin K inhibitor), and romosozumab (an antisclerostin antibody)—no evidence existed for heterogeneity of the associations with mortality (I2 = 0%). An analysis limited to 21 clinical trials of bisphosphonate treatments (20 244 participants randomized to placebo and 22 623 to treatment) also found no significant association of treatment with overall mortality (RR, 0.95; 95% CI, 0.86-1.04; P = .17; Figure 3).29-31,35-52 In the 6 randomized placebo-controlled clinical trials of zoledronate (6944 participants randomized to placebo and 6926 to treatment), in which patients received 5 mg of zoledronate every 12 to 18 months, the overall mortality rate was also not statistically significant (RR, 0.88; 95% CI, 0.68-1.13; P = .31); however, there was evidence for heterogeneity of the results (I2 = 48.2%; Figure 4).29-31,50-52
We conducted 7 exploratory analyses of subgroups in the clinical trials. Clinical trials of alendronate, the most common treatment evaluated in observational studies, reported no association with a reduction in mortality (RR, 1.00; 95% CI, 0.71-1.40; P = .98; eFigure 1 in the Supplement). Clinical trials of treatment with bisphosphonates without zoledronate also reported no association with a reduction in mortality (RR, 0.98; 95% CI, 0.88-1.10; P = .42; eFigure 2 in the Supplement). Analysis of only clinical trials of nitrogen-containing bisphosphonate treatments (alendronate, risedronate, ibandronate, and zoledronate) suggested an association with lower overall mortality that was not statistically significant (RR, 0.90; 95% CI, 0.81-1.00; P = .06; eFigure 3 in the Supplement); this association was nonsignificant when clinical trials of zoledronate treatment were excluded (RR, 0.92; 95% CI, 0.79-1.07; P = .29; eFigure 4 in the Supplement). Clinical trials of drug treatment lasting 3 years or more also reported no significant association with mortality (RR = 0.97; 95% CI, 0.88-1.08; P = .59; eFigure 5 in the Supplement), nor did clinical trials of bisphosphonate treatments lasting 3 years or more that included zoledronate (RR,0.94; 95% CI, 0.83-1.06; P = .28; eFigure 6 in the Supplement) or excluded zoledronate (RR, 1.00; 95% CI, 0.89-1.12; P = .94; eFigure 7 in the Supplement).
A previous review of a small number of clinical trials of drug treatments for osteoporosis suggested that an association may exist between the increased mortality of participants in the clinical trial, reflecting the comorbidity and age of the participants and the difference in mortality in the clinical trial.69 However, in this larger analysis, we noted no significant association between the mortality rates in the placebo groups of the clinical trials and the association of treatment with mortality for all drug treatments (slope, -0.003; SE, 0.002; P = .52; eFigure 8 in the Supplement) or for bisphosphonate treatments (slope, -0.0066; SE, 0.0008; P = .35; eFigure 9 in the Supplement).
Funnel plots of the risk ratio by standard error for all clinical trials showed no pattern of potential bias for large clinical trials, with a preponderance of published clinical trials showing no reduction in risk ratio. In addition, results from Egger and Begg tests for small sample bias were not significant (bias, -0.288; SE, 0.214; P = .19 and P = .27, respectively; eFigure 10 in the Supplement). The Egger and Begg results remained nonsignificant (bias, -0.424; SE, 0.299; P = .17 and P = .27, respectively) for clinical trials of bisphosphonate treatments (eFigure 11 in the Supplement).
This meta-analysis of randomized clinical trials suggested that drug treatments, particularly treatment with bisphosphonates, for patients with osteoporosis were not associated with reduced overall mortality rates. We found less certainty regarding the association of intravenous zoledronate treatment with overall mortality rates because heterogeneity was noted between the results of the clinical trials. In particular, 2 large clinical trials of zoledronate treatment observed 28% and 35% reductions in mortality that were not observed in other clinical trials.29,30 More data from placebo-controlled clinical trials of zoledronate therapy and mortality rates are needed to resolve whether treatment with zoledronate is associated with reduced mortality in addition to decreased fracture risk.
The 25% to 60% reductions in total mortality rates reported by observational studies were too large to be owing to a decrease in fracture risk; therefore, the reduced mortality rates would likely be owing to the direct biological effects of these treatments rather than the reduced fracture risk.21,22,24,25 However, the results of this meta-analysis of randomized placebo-controlled clinical trials suggest that observational studies reporting that patients receiving bisphosphonate therapy had lower mortality may not have measured confounding factors that may have contributed to lower mortality.18 The apparent reduction in mortality may be an example of the “healthy adherer effect,” which has been documented in studies reporting that participants who adhered to placebo treatment in clinical trials had lower mortality.70,71 For example, in a Women’s Health Initiative study, women in the placebo group who had at least 80% adherence to placebo treatment experienced a 36% lower overall mortality than those with less than 80% adherence, despite adjustments for numerous potential confounding factors; this result suggests that patients who adhere to treatments may be healthier than those who do not.71 This effect is particularly applicable to observational studies of treatments for osteoporosis because only an estimated half of women taking oral drugs for the treatment of osteoporosis continued the regimen for 1 year, and even fewer continued longer.72-74 Patients who adhere to treatment are more likely to be identified and classified as “taking a bisphosphonate” in an observational study than those who discontinue treatment soon after the first prescription. Similar discrepancies have been noted between observational studies reporting lower mortality in women taking estrogen therapy75,76 and randomized clinical trials reporting no effect on mortality.77,78 A meta-analysis of observational studies noted a 32% lower risk of invasive breast cancer in women taking oral bisphosphonates79; however, this finding was not confirmed by randomized clinical trials,80 and the association is plausibly attributable to the confounding influence of a lower level of estrogen, which may decrease bone density and increase fracture risk while reducing the risk of breast cancer.81-84
It is not clear whether there are biological mechanisms that could lead to an association between bisphosphonate treatments and overall mortality rates, particularly because the concentration of bisphosphonates in blood and tissue other than bone is undetectable or very low within days after bisphosphonate administration.28 Treatment with zoledronate is associated with a 28% reduction in mortality in patients after hip fracture, but only 8% of this reduction may be attributed to the prevention of fractures.85 Bisphosphonate treatment was not associated with a reduced risk for diseases but was associated with reduced mortality rates after diagnoses, most notably in patients with pneumonia and arrhythmia. This reduction in mortality rates may result from the periodic inflammation stimulated by intravenous zoledronate, which may increase resistance to a subsequent inflammatory disease in a phenomenon called hormesis.86 Bisphosphonates bind to arterial calcium deposits and may decrease the rate of further calcification, although a meta-analysis of clinical trials reported no evidence of a reduction in cardiovascular events.33 In vitro studies suggest that zoledronate may reduce the accumulation of DNA damage in mesenchymal stem cells.87 Nitrogen-containing bisphosphonates, such as zoledronate, inhibit the farnesyl pyrophosphate synthase enzyme in the mevalonate pathway, and this inhibition reduces the lipid prenylation of regulatory proteins, such as ras and rho.88 In a mouse study of premature aging, administration a statin and bisphosphonate combination was associated with a reduction in the abnormal accumulation of pathogenic proteins and an increased life span.89 However, whether these laboratory effects in mice are relevant to the effects of bisphosphonate treatment in humans is uncertain.
Our meta-analysis was much larger and included many more recent clinical trials than a 2010 meta-analysis reporting a 10% reduction in mortality, and it explored subgroups of drug therapies, specifically, zoledronate treatments, as well as clinical trials lasting 3 years or more. A more recent meta-analysis of 61 clinical trials of bisphosphonate treatments also reported a 10% overall reduction in mortality; however, the analysis included 32 clinical trials for cancer and only 22 for osteoporosis as well as clinical trials without placebo controls.33 A meta-analysis of clinical trials of fracture liaison services, which focus on increasing drug treatment after fracture among patients with osteoporosis, also reported a reduction in mortality; however, this reduced mortality may have been owing to the increase in comprehensive health care provided by some of the services rather than to drug therapy.26
The results of randomized clinical trials are sometimes criticized as not representative of real world patients, and it has been suggested that the findings of observational studies may be more applicable to patients who are not eligible for clinical trials. Real world studies using large databases to examine the effects of drug treatments can detect rare adverse effects or differences in the effects of drugs administered for the same indication in similar patients. However, real world studies of the association between drug treatments and overall mortality are observational and therefore may be unable to control for confounding factors that are not measured, which can produce the misleading impression that drug treatments reduce mortality rates. It is difficult to imagine a biological basis for the difference between the biological effects of bisphosphonate treatments on people who participated in clinical trials and those who did not. For instance, bisphosphonates have similar effects on human and rodent cells; thus, it is unlikely that the biological effects of bisphosphonates on cells from those eligible for clinical trials compared with those ineligible would differ sufficiently to explain large reductions in mortality.
Our meta-analysis had several limitations. We did not include the results of a large clinical trial of odanacatib, an antiresorptive with a different mechanism of action than bisphosphonates, because the study was published only in abstract form.90 The clinical trial reported no effect on mortality (327 deaths among 8028 patients randomized to placebo and 378 deaths among 8043 patients randomized to treatment; RR, 1.15; 95% CI, 1.00-1.34); therefore, inclusion of the clinical trial in our meta-analysis of all drug treatments would not have changed our conclusion that drug treatments for osteoporosis were not associated with a reduction in mortality rates.90 Although an association between treatment with bisphosphonates and mortality rates may take years to emerge,33 we found no significant association among clinical trials of all treatments or clinical trials of bisphosphonate treatments lasting 3 years or more. A nonsignificant association between mortality rates and nitrogen-containing bisphosphonate treatments may have been attributed to the inclusion of zoledronate, and the association should be regarded with caution because of the many post hoc exploratory analyses that were performed.
Although this meta-analysis did not support the claim that drug treatments for osteoporosis have an association with reduced overall mortality rates owing to causes other than decreased fracture risk, it did not exclude the possibility that decreasing the risk of fractures may be associated with reducing the mortality caused by those fractures. For example, if hip and vertebral fractures were associated with a preventable 20% mortality rate, and drug treatment reduced the absolute risk of these fractures by 10%, then this treatment would have reduced the mortality rate by only 2%, a small effect that would have been detectable only with more data than was available from current clinical trials. In this case, however, prevention of fracture would have remained the only rationale for prescribing drug treatments to patients with osteoporosis.
This meta-analysis of randomized placebo-controlled clinical trials suggests that drug treatments for osteoporosis, and treatments with bisphosphonates in particular, are not associated with reduced overall mortality rates in addition to decreased risk of fracture. Drug treatments for patients with osteoporosis should only be recommended for reducing fracture risk in accordance with clinical guidelines.1-3
Accepted for Publication: May 23, 2019.
Corresponding Author: Steven R. Cummings, MD, San Francisco Coordinating Center, 550 16th St, Second Floor, Mission Hall, Box 0560, San Francisco, CA 94143 (firstname.lastname@example.org).
Published Online: August 19, 2019. doi:10.1001/jamainternmed.2019.2779
Author Contributions: Dr Cummings and Dr Allen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Cummings, Eastell, Allen.
Acquisition, analysis, or interpretation of data: Lui, Eastell, Allen.
Drafting of the manuscript: Cummings, Allen.
Critical revision of the manuscript for important intellectual content: Cummings, Lui, Eastell.
Statistical analysis: Lui, Allen.
Administrative, technical, or material support: Lui, Allen.
Conflict of Interest Disclosures: Dr Cummings reported receiving grants and personal fees from Amgen during the conduct of the study. Dr Eastell reported receiving grants from Amgen and Alexion and personal fees from Amgen, Alexion, Lilly, Mereo, Sandoz, and AbbVie outside the submitted work. No other disclosures were reported.