Diabetes, Osteoporosis, and Antiresorptive Therapy in Implant Dentistry: Biological Effects, Analytic Artifacts, and an Individual-Participant-Data Protocol to Test Whether the Bisphosphonate Association Survives Patient-Level Analysis
- Paper type
- Methodological review with an embedded individual-participant-data meta-analysis protocol
- Status
- Original Work · Working Paper
- Keywords:
- diabetes mellitus · osteoporosis · bisphosphonates · denosumab · medication-related osteonecrosis of the jaw · implant failure · unit of analysis · confounding by indication
Contents
- 1Introduction and Clinical Significance
- 2Critical Review of the Established Literature
- 3Identification of the Specific Research Gap
- 4Scientific Objective and Hypothesis
- 5Theoretical Foundation
- 6Methodology
- 7Evidence Synthesis and Explicitly Labeled Theoretical Predictions
- 8Discussion: Mechanisms and Competing Explanations
- 9Implications for Implant Surgery and Restorative Dentistry
- 10Technical and Clinical Limitations
- 11Proposed Validation
- 12Future Research Directions
- 13Evidence-Calibrated Conclusion
Abstract
Background: Diabetes, osteoporosis, and antiresorptive drugs alter bone metabolism and are common in the population seeking implants. Meta-analyses report no significant increase in implant failure with diabetes, greater marginal bone loss and peri-implantitis risk with hyperglycemia, little effect of osteoporosis on survival, and a moderately increased failure risk with bisphosphonates. A recent analysis found the bisphosphonate association at the implant level but not at the patient level.
Objective: To separate biological from analytic explanations for these associations and to design a study capable of deciding the bisphosphonate question.
Methods: Critical review of systematic reviews, cohort studies, position papers, and mechanistic literature. The inflation of variance caused by analyzing implants as independent was quantified with the design effect and applied to a published estimate as an arithmetic illustration. A protocol for an individual-participant-data meta-analysis was written to PRISMA-IPD.
Findings: Evidence that well-controlled diabetes does not materially reduce implant survival is of moderate certainty; evidence that hyperglycemia delays stabilization and increases peri-implant disease is consistent. Osteoporosis alone has not been shown to reduce survival. The bisphosphonate association rests on observational implant-level data with few events, and its confidence interval widens to include no effect under plausible intraclass correlation. The serious drug-related risk is osteonecrosis, which implant survival statistics do not capture. We hypothesize that the association between oral bisphosphonate exposure and implant failure is not distinguishable from null once patients are the unit of analysis and age, sex, smoking, diabetes, and glucocorticoid use are adjusted for. A pooled patient-level hazard ratio whose 95% confidence interval excludes one would refute the hypothesis.
Conclusions: Much of what is reported about systemic conditions and implant failure is fragile to the choice of analytic unit. Clinical decisions should weigh rare severe harms and disease control more than small, uncertain differences in failure rates. The numerical illustrations use assumed clustering parameters and are not re-analyses of patient data.
Keywords: diabetes mellitus; osteoporosis; bisphosphonates; denosumab; medication-related osteonecrosis of the jaw; implant failure; unit of analysis; confounding by indication
1. Introduction and Clinical Significance
The population receiving implants is aging, and with age come type 2 diabetes, osteoporosis, and the drugs used to treat it. A clinician planning implants for a 68-year-old woman taking alendronate, or a 60-year-old man with a glycated hemoglobin of 8 percent, wants to know two things: whether the implant is less likely to succeed, and whether the surgery exposes the patient to a harm, such as osteonecrosis or infection, that would not otherwise arise. The published answers are familiar. Diabetes is a relative contraindication when poorly controlled. Osteoporosis is not a contraindication. Oral bisphosphonates slightly increase failure and carry a small risk of osteonecrosis; intravenous antiresorptives for cancer are a contraindication to elective implant surgery [1].
These statements rest on evidence of very uneven quality, and the quantitative parts of them, relative risks of 1.07 or 1.7, are more fragile than their decimal places suggest. Nearly all of the evidence is observational. Exposure is recorded as a binary label, diabetic or not, bisphosphonate user or not, that conceals dose, duration, and control. Outcomes are counted per implant in patients who each contribute several implants. And the exposed and unexposed differ in age, sex, smoking, comorbidity, and the surgical caution with which they are treated.
This paper has two aims. The first is to set out, for each condition, what is biologically expected and what has been observed, distinguishing outcomes that are often run together: failure to integrate, late loss, marginal bone change, peri-implant disease, and drug-specific harm. The second is methodological: to show how the unit of analysis and confounding shape the estimates, and to propose an analysis that would settle the question on which the literature currently gives two answers, the effect of bisphosphonates on implant failure.
2. Critical Review of the Established Literature
Diabetes: mechanisms
Hyperglycemia affects bone healing through several routes. Reviews of diabetes and osseous healing describe reduced osteoblast proliferation and differentiation, impaired collagen synthesis and mineralization, and delayed callus maturation in experimental diabetes, partly reversed by insulin [2]. Advanced glycation end-products accumulate in bone matrix, stiffen collagen, and act on their receptor to promote inflammation and suppress osteoblast function [3]. Diabetic bone disease in humans is characterized by low turnover and impaired bone quality with fracture risk that exceeds what bone density predicts [4]. Diabetes also amplifies and prolongs inflammatory responses in periodontal tissues and impairs their resolution [5]. A review of implant placement in diabetic animals found reduced bone-to-implant contact and delayed bone formation relative to controls, improved by insulin therapy [6]. The biological expectation is therefore slower osseointegration, bone of poorer quality at the interface, and greater susceptibility to inflammatory bone loss.
Diabetes: clinical evidence
A systematic review and meta-analysis found no significant difference in implant failure between diabetic and non-diabetic patients (risk ratio 1.07, 95% confidence interval 0.80 to 1.44) and significantly greater marginal bone loss in diabetic patients, by a mean of 0.20 mm [7]. Later systematic reviews reached compatible conclusions: implant survival in patients with controlled diabetes is comparable to that in healthy patients, with indications of more peri-implant bone loss and inflammation over time, whereas poorly controlled diabetes is associated with impaired osseointegration, more peri-implantitis, and more failure [8, 9]. Meta-analyses of biological parameters and complications reported greater bleeding, probing depth, and bone loss with poorer glycemic control [10, 11], and a meta-analysis restricted to peri-implant disease found a higher risk of peri-implantitis in people with diabetes or hyperglycemia, by about half, independent of smoking [12].
Healing kinetics have been measured directly. In a prospective study of patients with type 2 diabetes across a range of glycated hemoglobin values, including poorly controlled patients, implant stability during the first weeks was lower and its recovery slower with higher glycated hemoglobin, yet one-year survival was not reduced [13]. A critical review by the same group argued that glycemic control should be regarded as a modifier of healing time and not as a threshold for exclusion, and noted how few studies had enrolled patients with poor control [14].
The clinical picture thus matches the biological expectation in kind: integration is delayed and peri-implant tissues are more vulnerable, without a demonstrable reduction in survival over the periods studied. Two cautions apply. Patients with poorly controlled diabetes are under-represented, since clinicians avoid treating them or defer treatment until control improves; the estimate for "diabetes" is largely an estimate for controlled diabetes. And follow-up is short relative to the time over which peri-implantitis develops.
Osteoporosis
Osteoporosis reduces bone mass and degrades trabecular architecture, which would be expected to lower primary stability in cancellous bone. Whether it impairs the biological process of osseointegration is less clear; the jaws differ from the axial skeleton in loading and turnover, and the relation between systemic bone density and jaw bone quality is loose. A systematic review with meta-analysis found no difference in implant survival between patients with and without osteoporosis at either the implant or the patient level, with slightly greater marginal bone loss in the osteoporosis group [15]; an earlier systematic review concluded that the evidence was weak and did not support osteoporosis as a contraindication [16]. A prospective controlled multicenter study of women with osteoporosis or osteopenia reported five-year cumulative survival of about 91 percent in the osteoporosis group and 100 percent in controls, a difference that was statistically significant when implants were the unit and not when patients were, with similar marginal bone changes, and concluded that implant therapy was reliable in these patients [17], and a retrospective study found that osteoporotic status did not affect survival of titanium implants [18]. Vitamin D deficiency has been examined as a related exposure; a retrospective series suggested more early failures with very low serum levels without statistical significance, and a systematic review found the evidence insufficient [19, 20].
Osteoporosis seldom appears in isolation in these datasets. It is treated, and the treatment is the more consequential exposure.
Antiresorptive drugs: pharmacology
Nitrogen-containing bisphosphonates bind to bone mineral, are taken up by osteoclasts during resorption, and inhibit an enzyme of the mevalonate pathway, impairing osteoclast function and survival; they persist in bone for years [21]. Denosumab, a monoclonal antibody against the ligand that drives osteoclast formation, suppresses resorption profoundly and reversibly [22], with rapid rebound of turnover and reports of multiple vertebral fractures after discontinuation [23]. Both drug classes suppress remodeling. For an implant, suppressed remodeling could cut either way: resorption of necrotic bone at the osteotomy margin and its replacement might be delayed, while net bone mass at the interface might be preserved or increased.
Antiresorptive drugs: implant failure
A meta-analysis of comparative studies found a higher risk of implant failure in patients taking bisphosphonates (risk ratio 1.73, 95% confidence interval 1.21 to 2.48), and its authors described the result as not well established because of the limited number and quality of studies and uncontrolled confounding [24]. An updated meta-analysis reported an odds ratio of about 1.65 with a confidence interval whose lower bound was close to 1 [25]. A further systematic review of antiresorptive drugs and implant therapy found that low-dose oral bisphosphonate intake for osteoporosis did not, in general, compromise implant therapy, while emphasizing that information on high-dose therapy was lacking and that osteonecrosis was the principal concern [26]; other reviews came to similarly qualified conclusions [27, 28]. In a case review of patients taking oral bisphosphonates, all but 2 of 468 implants integrated and no osteonecrosis was found [29]. The most recent synthesis, including denosumab, reported an association of bisphosphonates with failure when implants were the unit of analysis (risk ratio 1.74, 95% confidence interval 1.10 to 2.75) and none when patients were (risk ratio 1.01, 0.35 to 2.91), rating the certainty of evidence as very low [30].
The discrepancy between implant-level and patient-level results is the central observation for this paper. It is what would be expected if a few patients with multiple failures drive the implant-level estimate.
Antiresorptive drugs: osteonecrosis
Medication-related osteonecrosis of the jaw is defined and staged in the position paper of the American Association of Oral and Maxillofacial Surgeons, which summarizes risk by drug, dose, indication, and duration and regards dentoalveolar surgery as the principal local precipitant [1]; aspects of its definition and staging have been challenged in correspondence [31]. Osteonecrosis associated with dental implants has two presentations: soon after placement, in the manner of extraction-triggered disease, and years later around implants that had integrated before or during drug therapy [32, 33]. In one series most affected patients were taking oral bisphosphonates for osteoporosis, and only about one case in six could be attributed to the implant surgery itself [32]. In a series of peri-implant osteonecrosis, a relation to peri-implantitis was considered, the inflammatory lesion acting as the trigger in drug-exposed bone [34]. Because this complication can arise at long-standing implants, it is poorly captured by studies that count implant failures in the first years after placement, and an implant lost to osteonecrosis is a qualitatively different event from one that fails to integrate.
The statistical issue
That implants within a patient are not independent, and that analyses ignoring this misstate precision, has been demonstrated for implant data repeatedly [35–37]. The general result is that when m observations per cluster are correlated with intraclass correlation coefficient ICC, the variance of an estimate is inflated by the design effect [38]. Meta-analyses of implant risk factors typically take event counts per implant from primary studies and pool risk ratios as though implants were independent. Their confidence intervals are therefore too narrow by an unknown factor, and their point estimates are weighted toward patients with many implants.
Confounding compounds the problem. Bisphosphonate users are older and predominantly postmenopausal women; many have comorbidities and take glucocorticoids, which impair bone healing independently. Diabetic patients differ from non-diabetic ones in obesity, smoking, periodontal status, and vascular disease. The direction of confounding is not uniform: clinicians may also select exposed patients carefully, defer surgery, or use more conservative protocols, which would bias toward the null.
3. Identification of the Specific Research Gap
Established: hyperglycemia delays early implant stabilization and is associated with more peri-implant inflammation and bone loss; controlled diabetes has not been shown to reduce implant survival; osteoporosis has not been shown to reduce survival; antiresorptive drugs cause osteonecrosis at rates that depend strongly on dose and indication, and implants can be a site of it.
Uncertain: whether oral bisphosphonates for osteoporosis increase the risk of ordinary implant failure, and by how much; whether denosumab does; and whether poor glycemic control, as opposed to a diagnosis of diabetes, reduces long-term survival.
Existing methods are inadequate in four respects. Aggregate-data meta-analysis cannot correct the unit of analysis, because primary reports do not supply the within-patient correlation or patient-level counts. It cannot adjust consistently for confounders, because studies report different covariates. It cannot examine dose, duration, or drug holiday, because these are not tabulated. And it merges early with late failure, and failure with osteonecrosis.
The specific gap is an analysis in which the patient is the unit, exposure is characterized by drug, route, duration, and indication, confounders are adjusted uniformly, and outcomes are separated by type. Individual-participant data make this possible; the method is established and has a reporting standard [39, 40].
4. Scientific Objective and Hypothesis
The objective is to estimate the association between oral bisphosphonate therapy for osteoporosis and implant failure with the patient as the unit of analysis and with uniform adjustment for confounders.
Hypothesis H12: in an individual-participant-data meta-analysis of cohort studies, the hazard ratio for first implant failure in patients exposed to oral bisphosphonates for osteoporosis at the time of implant placement, relative to unexposed patients, adjusted for age, sex, smoking, diabetes, systemic glucocorticoid use, jaw, and number of implants, has a 95% confidence interval that includes 1; and the implant-level association reported in aggregate meta-analyses is attributable to clustering of failures in a small number of patients.
H12 is falsified if the pooled, adjusted patient-level hazard ratio has a 95% confidence interval that excludes 1. It is not confirmed merely by a non-significant result: the analysis is also required to be informative, defined in advance as a confidence interval whose upper limit lies below 2.0, so that an absence of evidence produced by too few events is reported as inconclusive. A secondary hypothesis, H12b, is that any excess risk is concentrated in early failure and in patients with exposure longer than four years, the duration threshold highlighted in the position paper's discussion of risk [1]; H12b is falsified by the absence of a duration gradient.
The hypothesis is confined to oral bisphosphonates for osteoporosis and to implant failure. It makes no claim about osteonecrosis, which is a separate outcome with established causality, or about high-dose antiresorptive therapy in cancer.
5. Theoretical Foundation
What altered turnover should do to an implant
Osseointegration requires bone formation on and toward the implant surface and, over months, remodeling of the damaged bone at the osteotomy wall into load-bearing lamellar bone. Antiresorptive drugs act on the second process more than the first. A plausible expectation is intact or enhanced early bone apposition, delayed replacement of marginal necrotic bone, and a long-term interface of older, more mineralized, less frequently renewed bone whose capacity to repair microdamage and to respond to infection is reduced. This expectation fits the observed pattern of little effect on integration with a specific late vulnerability, osteonecrosis, triggered by infection or surgery. In diabetes, the expectation is the reverse in timing: early formation is impaired and the tissue's response to biofilm is exaggerated, so that delayed stabilization and later peri-implantitis are the predicted manifestations, with survival affected only when control is poor or follow-up long.
These expectations imply that "implant failure" is the least informative outcome for both conditions, because the mechanisms act chiefly on other endpoints.
Design effect and effective sample size
Suppose n implants are placed in patients with a mean of m implants each, and failures within a patient have intraclass correlation ICC. The design effect is
DE = 1 + (m - 1) · ICC (1)
and the effective number of independent observations is n/DE [38]. The standard error of a log risk ratio computed as if implants were independent, SE0, should be multiplied by the square root of DE:
SE_(adj) = SE_0 √(DE) (2)
The corrected confidence interval for a risk ratio RR is
exp(ln RR ± 1.96 · SE_(adj)) (3)
Application to a published estimate
The following is an arithmetic illustration using a published summary estimate and assumed clustering parameters; it is not a re-analysis of data. A risk ratio of 1.73 with a 95% confidence interval of 1.21 to 2.48 [24] implies SE0 of about 0.183 on the log scale. If patients received three implants on average and the intraclass correlation were 0.3, the design effect would be 1.6, the adjusted standard error 0.232, and the interval 1.10 to 2.72. With four implants per patient and an intraclass correlation of 0.5, the design effect would be 2.5 and the interval 0.98 to 3.05, which includes 1. The intraclass correlation for implant failure is not reported in the bisphosphonate literature; cluster analyses in large cohorts, in which a small fraction of patients account for the majority of failures [41], indicate that it is not small. The illustration shows that the statistical significance of the pooled implant-level estimate depends on an unreported quantity. It does not show that the true effect is null.
Confounding and the strength needed to explain an association
For an observed risk ratio RR greater than 1, the E-value is the minimum strength of association, on the risk-ratio scale, that an unmeasured confounder would need to have with both exposure and outcome to explain away the association [42]:
E = RR + √(RR (RR - 1)) (4)
For RR of 1.73 the E-value is about 2.85, and for the lower confidence limit of 1.21 it is about 1.71. A confounder associated with both bisphosphonate use and implant failure by a risk ratio of 1.7, after adjustment for measured covariates, would suffice to move the interval to include 1. Systemic glucocorticoid therapy, rheumatoid disease, and advanced age are candidates of at least that strength for exposure; their association with implant failure is less well quantified. The calculation indicates that the published association is not robust to confounding of moderate strength.
Competing events and outcome separation
Patients taking antiresorptive drugs are older and have higher mortality; death removes them from risk of implant failure. Treating death as non-informative censoring overestimates the cumulative incidence of failure in the group with higher mortality, while comparing crude proportions underestimates it. Cause-specific hazards and cumulative incidence functions that treat death as a competing event are the appropriate tools, and early failure, late failure, and osteonecrosis involving the implant site should be analyzed as distinct outcomes.
6. Methodology
Review methods
The review is narrative and critical. Meta-analytic estimates are quoted from the cited sources. The numerical illustration in Section 5.3 uses only the published point estimate and interval and stated assumptions.
Design of the proposed study
The study is a systematic review with individual-participant-data meta-analysis, registered prospectively and reported according to PRISMA-IPD [40]. Its framing follows the logic of emulating a target trial: the hypothetical trial would randomize patients with osteoporosis who are candidates for implants to continue oral bisphosphonate therapy or to have never started it, and would follow them for implant failure [43]. The observational analysis is specified to resemble that trial as closely as the data permit, with eligibility, exposure definition, time zero (implant placement), and outcome fixed in advance.
Eligibility and data sources
Eligible studies are cohort studies, prospective or retrospective, of adults receiving dental implants, in which antiresorptive exposure at placement was recorded and at least 20 exposed patients were followed for a minimum of one year. Studies are identified by systematic search and by contacting authors of the studies included in existing meta-analyses [24, 25, 30]. Institutional registries with medication data are approached directly. Investigators are asked for de-identified patient-level and implant-level records under data-sharing agreements; studies that cannot provide them are retained in an aggregate-data sensitivity analysis so that availability bias can be assessed.
Variables
The requested variables are listed in Table 1. Exposure is classified by drug, route, indication, cumulative duration at placement, and whether therapy was interrupted around surgery. The primary exposure contrast is oral bisphosphonate for osteoporosis versus no antiresorptive therapy; denosumab for osteoporosis is a secondary contrast; high-dose therapy for malignancy is described and not pooled with the others.
| Domain | Variables | Level |
|---|---|---|
| Identification | Study, center, patient identifier, implant identifier | All |
| Exposure | Antiresorptive drug, route, indication, start date, cumulative duration at placement, drug holiday, concurrent denosumab or anabolic therapy | Patient |
| Confounders | Age, sex, smoking, diabetes and glycated hemoglobin where available, systemic glucocorticoids, rheumatoid disease, other immunosuppression, history of periodontitis, radiotherapy | Patient |
| Surgical and site | Jaw, region, grafting, immediate placement, loading protocol, implant dimensions and surface, number of implants, operator experience category | Implant |
| Outcomes | Early failure (before loading or within 6 months), late failure, date of failure, osteonecrosis involving the implant site with stage, death, date last seen | Implant and patient |
Outcomes
The primary outcome is time from placement to first implant failure in a patient, excluding implants lost to osteonecrosis, which are analyzed separately. Secondary outcomes are early failure, late failure, osteonecrosis involving an implant site, and, where recorded, marginal bone change. Death is a competing event.
Statistical analysis
The primary analysis is two-stage. Within each study, a Cox model for time to first failure per patient is fitted with the prespecified confounders; the number of implants placed enters as a covariate, since patients with more implants have more opportunities to fail. Study-specific log hazard ratios are pooled by random-effects meta-analysis with a Hartung-Knapp adjustment. A one-stage analysis, a Cox model stratified by study with a random treatment effect, is performed for comparison. To make the contrast with the existing literature explicit, three implant-level analyses are run on the same data: a naive model treating implants as independent; a marginal model with robust standard errors clustered on patient; and a shared-frailty model. The intraclass correlation of failure within patients is estimated and reported, which no prior synthesis has done. Missing covariate data are multiply imputed within study. Heterogeneity is described by the between-study variance and prediction interval. Duration of exposure is examined as a continuous variable with restricted cubic splines. The E-value is reported for the primary estimate and its confidence limit [42]. Competing-risk regression for the subdistribution hazard is a sensitivity analysis.
Information size
The precision of a hazard ratio depends on the number of events. With D patients experiencing a first failure, distributed between exposed and unexposed in proportions p and 1 − p, the standard error of the log hazard ratio is approximately the square root of 1 divided by D p (1 − p). For the analysis to be informative by the criterion in Section 4, an upper confidence limit below 2.0 when the true hazard ratio is 1, the standard error must be at most about 0.35, which with 10 percent of events in exposed patients requires roughly 90 patients with a first failure. If 6 percent of patients experience a failure during follow-up, a figure adopted for planning from population-based data on implant loss and subject to revision, about 1,500 patients, 150 or more of them exposed, would be needed. The exposed group is the limiting quantity; published comparative studies together include several hundred exposed patients, so the target is attainable only if most investigators share data.
7. Evidence Synthesis and Explicitly Labeled Theoretical Predictions
Synthesis
Table 2 summarizes the evidence by condition and outcome. Certainty labels are the authors' qualitative judgments and do not result from a formal grading exercise.
| Exposure and outcome | Finding | Certainty | Main limitation |
|---|---|---|---|
| Diabetes and implant failure [7–9] | No significant increase overall | Moderate for controlled diabetes | Poorly controlled patients under-represented |
| Hyperglycemia and early stability [13] | Slower stabilization with higher glycated hemoglobin; one-year survival unaffected | Low to moderate | Single prospective cohort |
| Diabetes and marginal bone or peri-implantitis [7, 11, 12] | Greater bone loss; higher peri-implantitis risk | Moderate | Observational; variable definitions |
| Osteoporosis and failure [15, 17] | No significant difference in pooled survival; one prospective study significant at implant level only | Low to moderate | Confounded with treatment; few studies; unit of analysis |
| Oral bisphosphonates and failure, implant level [24, 25] | Risk ratio about 1.7 | Low | Clustering ignored; confounding |
| Oral bisphosphonates and failure, patient level [30] | No significant association | Low | Few events; wide interval |
| Antiresorptives and osteonecrosis at implants [1, 32, 33] | Established, dose- and indication-dependent; early and late presentations | Moderate for causality; low for incidence | Case series; uncertain denominators |
Theoretical predictions
The following are predictions; the proposed analysis has not been performed.
Prediction 1. The intraclass correlation of implant failure within patients will be substantial, and the naive implant-level confidence interval for the bisphosphonate effect will be markedly narrower than the patient-level or frailty-based interval from the same data.
Prediction 2. The adjusted patient-level hazard ratio for oral bisphosphonates will be closer to 1 than the unadjusted implant-level risk ratio, and its interval will include 1 (H12).
Prediction 3. Adjustment for systemic glucocorticoid use and age will account for a larger share of the attenuation than adjustment for other covariates.
Prediction 4. Any residual association will be concentrated in early failures and in long-duration exposure (H12b), compatible with delayed remodeling at the osteotomy.
Prediction 5. Osteonecrosis involving implant sites will be rare in the osteoporosis-dose group, will occur both early and years after placement, and will not be predicted by the factors that predict ordinary failure.
Prediction 6. Results from studies providing individual data will not differ systematically from aggregate results of those that do not, an assumption the availability analysis will test.
8. Discussion: Mechanisms and Competing Explanations
The hypothesis that the bisphosphonate association is an artifact competes with a biological explanation that should be taken seriously. Suppressed osteoclast activity could delay the clearance of surgically damaged bone and so prolong the period of reduced stability, and bone that has been under antiresorptive therapy for years may be less vascular and less able to resist contamination at surgery. Either would produce a real excess of early failures. The observation that osteonecrosis can begin at an implant site soon after placement [32] shows that the drug-exposed jaw does not always heal normally. If such effects are real and of moderate size, the patient-level analysis will detect them given sufficient events, and H12 will be refuted. The requirement that the analysis be informative is meant to prevent an underpowered null from being read as exoneration.
The artifact explanation has its own support. The disappearance of the association at the patient level in the most recent synthesis [30] is the expected signature of clustering. Large cohort data show that implant failures concentrate in few patients [41]. Authors of the principal meta-analysis themselves cautioned that the association was not well established [24]. The same reversal between implant-level and patient-level significance appears in the prospective osteoporosis study [17]. And a case review of several hundred implants in bisphosphonate users found almost no failures [29].
Confounding could act in either direction. Indication is the obvious positive confounder: osteoporosis, older age, female sex, and associated diseases and drugs. A negative confounder is clinical selection. Surgeons aware of antiresorptive therapy may decline to treat patients they consider at risk, may avoid grafting and immediate loading, and may operate more carefully, so that exposed patients who receive implants are a favorable subset treated conservatively. If this effect is strong, an observed null could conceal a real hazard under equal treatment. The target-trial framing makes the estimand explicit, the effect among patients judged suitable for implants, and the recorded surgical variables allow part of the selection to be examined, though not all.
Misclassification of exposure is a further concern. Medication histories in dental records are incomplete, cumulative duration is often unknown, and some patients classified as unexposed will have taken bisphosphonates in the past, with the drug still in their bone [21]. Non-differential misclassification biases toward the null and would favor H12 spuriously. Linking dental cohorts to pharmacy dispensing records is the remedy where it is possible, and a sensitivity analysis restricted to studies with verified medication data is specified.
For diabetes, the corresponding methodological point is that the exposure of biological interest is glycemia over time, and the recorded exposure is a diagnosis. A null association between a diagnosis and failure says little about the effect of sustained hyperglycemia. The stability data that used glycated hemoglobin as a continuous exposure [13] are more informative than many larger studies using a label. The consistent signal for peri-implantitis [12] suggests that the important effect of diabetes lies in long-term tissue health, a domain where five-year survival is an insensitive measure.
For osteoporosis, the absence of an effect on survival [15, 17] is credible on biological grounds, since the disease reduces the quantity of bone more than the capacity to form it, and since implant sites are selected and prepared with local bone quality in view. The relevant practical consequences are lower primary stability in cancellous bone and the therapeutic choices that accompany the diagnosis.
A general lesson concerns how risk factors are communicated. A risk ratio from an implant-level meta-analysis is routinely transcribed into textbooks and consent discussions as a property of the drug or the disease. The illustration in Section 5.3 shows that such a figure can owe its statistical significance to an analytic convention. The remedy is not skepticism about all associations; it is routine reporting of patient-level results, of the within-patient correlation, and of how many patients contributed the failures.
9. Implications for Implant Surgery and Restorative Dentistry
The clinical implications follow the certainty of the evidence. For patients with diabetes, implant therapy is supported when the disease is controlled; glycated hemoglobin should be known before surgery; longer healing before loading is a reasonable accommodation to slower stabilization [13, 14]; and maintenance should be more frequent because the risk that is best documented is peri-implant disease [12]. For patients with osteoporosis who are not taking antiresorptive drugs, there is no evidence-based reason to withhold implants; primary stability in soft bone should be managed by site preparation and loading protocol.
For patients taking oral bisphosphonates or denosumab for osteoporosis, the decision should turn on osteonecrosis, the severe if uncommon harm, and on the alternatives to implants, more than on a possible modest increase in ordinary failure that current evidence cannot confirm. Risk stratification by drug, dose, duration, and comorbidity follows the position paper [1]. Patients should be told that osteonecrosis has been reported at implant sites years after placement and that long-term control of peri-implant inflammation is therefore part of risk reduction [34]. Decisions about interrupting therapy belong to the prescribing physician; for denosumab in particular, discontinuation carries a recognized risk of rebound bone loss and fracture [23], and dental convenience does not justify it. For patients receiving high-dose antiresorptive therapy for malignancy, elective implant placement is generally avoided [1].
For restorative dentists, medication histories should be updated at every recall, since a patient may begin antiresorptive therapy years after implants were placed and thereby acquire a risk that did not exist at surgery.
10. Technical and Clinical Limitations
The review is narrative. The clustering illustration assumes values for implants per patient and intraclass correlation that are not known for the pooled studies, and applies a simple variance inflation to a pooled estimate, which is an approximation. The E-value quantifies sensitivity to a single unmeasured confounder under simplifying assumptions. The proposed meta-analysis depends on investigators' willingness to share data and on the quality of medication and covariate recording in retrospective cohorts. Residual confounding and selection cannot be eliminated. The primary outcome, first failure per patient, discards information on multiple failures, which the frailty analysis partly recovers. Follow-up in most cohorts is too short for late events. Denosumab-exposed patients are likely to be too few for a precise estimate. The proposal addresses failure and osteonecrosis and does not cover marginal bone change or peri-implantitis under antiresorptive therapy, for which data are sparser still.
11. Proposed Validation
The individual-participant-data meta-analysis is the primary test. Two complementary studies would strengthen or challenge its result. First, a registry-linkage cohort: national or insurer dental claims linked to pharmacy dispensing and medical records would provide verified exposure, large numbers of exposed patients, and adjustment for comorbidity, at the cost of less clinical detail and reliance on procedure codes for failure. Agreement between that design and the clinical-cohort meta-analysis would be persuasive; disagreement would localize the problem to exposure measurement or outcome ascertainment. Second, a prospective cohort of patients starting implants while on oral bisphosphonates or denosumab, with serial stability measurement during healing, would test the mechanistic prediction of delayed stabilization directly, in the manner of the diabetes stability study [13], and would require far fewer patients than a failure endpoint. Reporting of both would follow STROBE [44].
A randomized trial of drug interruption before implant surgery is sometimes proposed. For bisphosphonates, whose skeletal half-life is measured in years, a short interruption is unlikely to change bone exposure; for denosumab, interruption has skeletal risks. Such a trial would need the participation of the prescribing specialties and a safety monitoring plan, and it is not proposed here. Refutation criteria for H12 and H12b are given in Section 4.
12. Future Research Directions
Three directions are suggested. Exposure measurement should move from labels to quantities: cumulative dose and duration for antiresorptives, longitudinal glycated hemoglobin for diabetes. Outcome measurement should extend beyond survival to stabilization kinetics, marginal bone trajectory, peri-implantitis incidence, and osteonecrosis with adequate follow-up, with patient-level denominators. And the newer osteoporosis drugs, anabolic agents and sclerostin inhibition, whose effects on implant healing are largely unstudied in humans, deserve prospective observation as their use grows. A standing collaborative of implant cohorts with harmonized variables would allow these questions to be addressed as the pharmacological landscape changes, without assembling a new dataset each time.
13. Evidence-Calibrated Conclusion
Controlled diabetes has not been shown to reduce implant survival, with moderate certainty; hyperglycemia slows early stabilization and is associated with more peri-implant bone loss and peri-implantitis. Osteoporosis without antiresorptive treatment has not been shown to reduce survival. Oral bisphosphonates are associated with a higher failure risk in implant-level meta-analyses and not in a patient-level analysis; the certainty of either statement is low. Antiresorptive drugs cause osteonecrosis of the jaw, which can involve implant sites early or late and is the clinically decisive risk. We have shown by arithmetic illustration that the significance of the implant-level bisphosphonate estimate depends on unreported clustering, proposed that the association will not survive patient-level, confounder-adjusted analysis, and specified the individual-participant-data study and the result that would refute that proposal. No patient data were analyzed for this manuscript.
Cite this paper
Jain S, Gupta A. Diabetes, Osteoporosis, and Antiresorptive Therapy in Implant Dentistry: Biological Effects, Analytic Artifacts, and an Individual-Participant-Data Protocol to Test Whether the Bisphosphonate Association Survives Patient-Level Analysis. Working paper. Center for Implant Dentistry; 2026. Available from: https://bayareaimplantdentistry.com/clinician-research-and-publications/bone-metabolic-conditions-and-implant-outcomes
References
- Ruggiero SL, Dodson TB, Aghaloo T, Carlson ER, Ward BB, Kademani D. American Association of Oral and Maxillofacial Surgeons' Position Paper on Medication-Related Osteonecrosis of the Jaws-2022 Update. J Oral Maxillofac Surg. 2022;80(5):920-943. doi:10.1016/j.joms.2022.02.008 (opens in a new tab) PMID: 35300956 (opens in a new tab).
- Retzepi M, Donos N. The effect of diabetes mellitus on osseous healing. Clin Oral Implants Res. 2010;21(7):673-81. doi:10.1111/j.1600-0501.2010.01923.x (opens in a new tab) PMID: 20465554 (opens in a new tab).
- Yamagishi S. Role of advanced glycation end products (AGEs) in osteoporosis in diabetes. Curr Drug Targets. 2011;12(14):2096-102. doi:10.2174/138945011798829456 (opens in a new tab) PMID: 22023404 (opens in a new tab).
- Napoli N, Chandran M, Pierroz DD, Abrahamsen B, Schwartz AV, Ferrari SL, et al. Mechanisms of diabetes mellitus-induced bone fragility. Nat Rev Endocrinol. 2017;13(4):208-219. doi:10.1038/nrendo.2016.153 (opens in a new tab) PMID: 27658727 (opens in a new tab).
- Graves DT, Ding Z, Yang Y. The impact of diabetes on periodontal diseases. Periodontol 2000. 2020;82(1):214-224. doi:10.1111/prd.12318 (opens in a new tab) PMID: 31850631 (opens in a new tab).
- Kotsovilis S, Karoussis IK, Fourmousis I. A comprehensive and critical review of dental implant placement in diabetic animals and patients. Clin Oral Implants Res. 2006;17(5):587-99. doi:10.1111/j.1600-0501.2005.01245.x (opens in a new tab) PMID: 16958701 (opens in a new tab).
- Chrcanovic BR, Albrektsson T, Wennerberg A. Diabetes and oral implant failure: a systematic review. J Dent Res. 2014;93(9):859-67. doi:10.1177/0022034514538820 (opens in a new tab) PMID: 24928096 (opens in a new tab).
- Naujokat H, Kunzendorf B, Wiltfang J. Dental implants and diabetes mellitus-a systematic review. Int J Implant Dent. 2016;2(1):5. doi:10.1186/s40729-016-0038-2 (opens in a new tab) PMID: 27747697 (opens in a new tab).
- Wagner J, Spille JH, Wiltfang J, Naujokat H. Systematic review on diabetes mellitus and dental implants: an update. Int J Implant Dent. 2022;8(1):1. doi:10.1186/s40729-021-00399-8 (opens in a new tab) PMID: 34978649 (opens in a new tab).
- Lagunov VL, Sun J, George R. Evaluation of biologic implant success parameters in type 2 diabetic glycemic control patients versus health patients: A meta-analysis. J Investig Clin Dent. 2019;10(4):e12478. doi:10.1111/jicd.12478 (opens in a new tab) PMID: 31638347 (opens in a new tab).
- Jiang X, Zhu Y, Liu Z, Tian Z, Zhu S. Association between diabetes and dental implant complications: a systematic review and meta-analysis. Acta Odontol Scand. 2021;79(1):9-18. doi:10.1080/00016357.2020.1761031 (opens in a new tab) PMID: 32401121 (opens in a new tab).
- Monje A, Catena A, Borgnakke WS. Association between diabetes mellitus/hyperglycaemia and peri-implant diseases: Systematic review and meta-analysis. J Clin Periodontol. 2017;44(6):636-648. doi:10.1111/jcpe.12724 (opens in a new tab) PMID: 28346753 (opens in a new tab).
- Oates TW Jr, Galloway P, Alexander P, Vargas Green A, Huynh-Ba G, Feine J, et al. The effects of elevated hemoglobin A(1c) in patients with type 2 diabetes mellitus on dental implants: Survival and stability at one year. J Am Dent Assoc. 2014;145(12):1218-26. doi:10.14219/jada.2014.93 (opens in a new tab) PMID: 25429035 (opens in a new tab).
- Oates TW, Huynh-Ba G, Vargas A, Alexander P, Feine J. A critical review of diabetes, glycemic control, and dental implant therapy. Clin Oral Implants Res. 2013;24(2):117-27. doi:10.1111/j.1600-0501.2011.02374.x (opens in a new tab) PMID: 22111901 (opens in a new tab).
- de Medeiros FCFL, Kudo GAH, Leme BG, Saraiva PP, Verri FR, Honório HM, et al. Dental implants in patients with osteoporosis: a systematic review with meta-analysis. Int J Oral Maxillofac Surg. 2018;47(4):480-491. doi:10.1016/j.ijom.2017.05.021 (opens in a new tab) PMID: 28651805 (opens in a new tab).
- Giro G, Chambrone L, Goldstein A, Rodrigues JA, Zenóbio E, Feres M, et al. Impact of osteoporosis in dental implants: A systematic review. World J Orthop. 2015;6(2):311-5. doi:10.5312/wjo.v6.i2.311 (opens in a new tab) PMID: 25793172 (opens in a new tab).
- Temmerman A, Rasmusson L, Kübler A, Thor A, Merheb J, Quirynen M. A Prospective, Controlled, Multicenter Study to Evaluate the Clinical Outcome of Implant Treatment in Women with Osteoporosis/Osteopenia: 5-Year Results. J Dent Res. 2019;98(1):84-90. doi:10.1177/0022034518798804 (opens in a new tab) PMID: 30205020 (opens in a new tab).
- Holahan CM, Koka S, Kennel KA, Weaver AL, Assad DA, Regennitter FJ, et al. Effect of osteoporotic status on the survival of titanium dental implants. Int J Oral Maxillofac Implants. 2008;23(5):905-10. PMID: 19014161 (opens in a new tab).
- Guido Mangano F, Ghertasi Oskouei S, Paz A, Mangano N, Mangano C. Low serum vitamin D and early dental implant failure: Is there a connection? A retrospective clinical study on 1740 implants placed in 885 patients. J Dent Res Dent Clin Dent Prospects. 2018;12(3):174-182. doi:10.15171/joddd.2018.027 (opens in a new tab) PMID: 30443302 (opens in a new tab).
- Bazal-Bonelli S, Sánchez-Labrador L, Cortés-Bretón Brinkmann J, Cobo-Vázquez C, Martínez-Rodríguez N, Beca-Campoy T, et al. Influence of Serum Vitamin D Levels on Survival Rate and Marginal Bone Loss in Dental Implants: A Systematic Review. Int J Environ Res Public Health. 2022;19(16). doi:10.3390/ijerph191610120 (opens in a new tab) PMID: 36011753 (opens in a new tab).
- Russell RG, Watts NB, Ebetino FH, Rogers MJ. Mechanisms of action of bisphosphonates: similarities and differences and their potential influence on clinical efficacy. Osteoporos Int. 2008;19(6):733-59. doi:10.1007/s00198-007-0540-8 (opens in a new tab) PMID: 18214569 (opens in a new tab).
- Cummings SR, San Martin J, McClung MR, Siris ES, Eastell R, Reid IR, et al. Denosumab for prevention of fractures in postmenopausal women with osteoporosis. N Engl J Med. 2009;361(8):756-65. doi:10.1056/NEJMoa0809493 (opens in a new tab) PMID: 19671655 (opens in a new tab).
- Anastasilakis AD, Polyzos SA, Makras P, Aubry-Rozier B, Kaouri S, Lamy O. Clinical Features of 24 Patients With Rebound-Associated Vertebral Fractures After Denosumab Discontinuation: Systematic Review and Additional Cases. J Bone Miner Res. 2017;32(6):1291-1296. doi:10.1002/jbmr.3110 (opens in a new tab) PMID: 28240371 (opens in a new tab).
- Chrcanovic BR, Albrektsson T, Wennerberg A. Bisphosphonates and dental implants: A meta-analysis. Quintessence Int. 2016;47(4):329-42. doi:10.3290/j.qi.a35523 (opens in a new tab) PMID: 26824084 (opens in a new tab).
- Sulaiman N, Fadhul F, Chrcanovic BR. Bisphosphonates and Dental Implants: A Systematic Review and Meta-Analysis. Materials (Basel). 2023;16(18). doi:10.3390/ma16186078 (opens in a new tab) PMID: 37763356 (opens in a new tab).
- Stavropoulos A, Bertl K, Pietschmann P, Pandis N, Schiødt M, Klinge B. The effect of antiresorptive drugs on implant therapy: Systematic review and meta-analysis. Clin Oral Implants Res. 2018;29 Suppl 18:54-92. doi:10.1111/clr.13282 (opens in a new tab) PMID: 30306695 (opens in a new tab).
- Gelazius R, Poskevicius L, Sakavicius D, Grimuta V, Juodzbalys G. Dental Implant Placement in Patients on Bisphosphonate Therapy: a Systematic Review. J Oral Maxillofac Res. 2018;9(3):e2. doi:10.5037/jomr.2018.9302 (opens in a new tab) PMID: 30429962 (opens in a new tab).
- Walter C, Al-Nawas B, Wolff T, Schiegnitz E, Grötz KA. Dental implants in patients treated with antiresorptive medication - a systematic literature review. Int J Implant Dent. 2016;2(1):9. doi:10.1186/s40729-016-0041-7 (opens in a new tab) PMID: 27747701 (opens in a new tab).
- Grant BT, Amenedo C, Freeman K, Kraut RA. Outcomes of placing dental implants in patients taking oral bisphosphonates: a review of 115 cases. J Oral Maxillofac Surg. 2008;66(2):223-30. doi:10.1016/j.joms.2007.09.019 (opens in a new tab) PMID: 18201600 (opens in a new tab).
- Lin L, Ren Y, Wang X, Yao Q. Effects of Bisphosphonates and Denosumab on Dental Implants: A Systematic Review With Meta-Analysis. Oral Dis. 2025;31(10):2835-2847. doi:10.1111/odi.15373 (opens in a new tab) PMID: 40326505 (opens in a new tab).
- Campisi G, Mauceri R, Bedogni A, Fusco V. Re: AAOMS Position Paper on Medication-Related Osteonecrosis of the Jaw-2022 Update. J Oral Maxillofac Surg. 2022;80(11):1723-1724. doi:10.1016/j.joms.2022.07.149 (opens in a new tab) PMID: 36332990 (opens in a new tab).
- Kwon TG, Lee CO, Park JW, Choi SY, Rijal G, Shin HI. Osteonecrosis associated with dental implants in patients undergoing bisphosphonate treatment. Clin Oral Implants Res. 2014;25(5):632-40. doi:10.1111/clr.12088 (opens in a new tab) PMID: 23278625 (opens in a new tab).
- Jacobsen C, Metzler P, Rössle M, Obwegeser J, Zemann W, Grätz KW. Osteopathology induced by bisphosphonates and dental implants: clinical observations. Clin Oral Investig. 2013;17(1):167-75. doi:10.1007/s00784-012-0708-2 (opens in a new tab) PMID: 22415216 (opens in a new tab).
- Troeltzsch M, Cagna D, Stähler P, Probst F, Kaeppler G, Troeltzsch M, et al. Clinical features of peri-implant medication-related osteonecrosis of the jaw: Is there an association to peri-implantitis? J Craniomaxillofac Surg. 2016;44(12):1945-1951. doi:10.1016/j.jcms.2016.09.018 (opens in a new tab) PMID: 27836551 (opens in a new tab).
- Weyant RJ, Burt BA. An assessment of survival rates and within-patient clustering of failures for endosseous oral implants. J Dent Res. 1993;72(1):2-8. doi:10.1177/00220345930720010201 (opens in a new tab) PMID: 8380286 (opens in a new tab).
- Herrmann I, Lekholm U, Holm S, Karlsson S. Impact of implant interdependency when evaluating success rates: a statistical analysis of multicenter results. Int J Prosthodont. 1999;12(2):160-6. PMID: 10371919 (opens in a new tab).
- Chuang SK, Tian L, Wei LJ, Dodson TB. Kaplan-Meier analysis of dental implant survival: a strategy for estimating survival with clustered observations. J Dent Res. 2001;80(11):2016-20. doi:10.1177/00220345010800111301 (opens in a new tab) PMID: 11759013 (opens in a new tab).
- Killip S, Mahfoud Z, Pearce K. What is an intracluster correlation coefficient? Crucial concepts for primary care researchers. Ann Fam Med. 2004;2(3):204-8. doi:10.1370/afm.141 (opens in a new tab) PMID: 15209195 (opens in a new tab).
- Riley RD, Lambert PC, Abo-Zaid G. Meta-analysis of individual participant data: rationale, conduct, and reporting. BMJ. 2010;340:c221. doi:10.1136/bmj.c221 (opens in a new tab) PMID: 20139215 (opens in a new tab).
- Stewart LA, Clarke M, Rovers M, Riley RD, Simmonds M, Stewart G, et al. Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data: the PRISMA-IPD Statement. JAMA. 2015;313(16):1657-65. doi:10.1001/jama.2015.3656 (opens in a new tab) PMID: 25919529 (opens in a new tab).
- Chrcanovic BR, Kisch J, Albrektsson T, Wennerberg A. Analysis of risk factors for cluster behavior of dental implant failures. Clin Implant Dent Relat Res. 2017;19(4):632-642. doi:10.1111/cid.12485 (opens in a new tab) PMID: 28332286 (opens in a new tab).
- VanderWeele TJ, Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value. Ann Intern Med. 2017;167(4):268-274. doi:10.7326/M16-2607 (opens in a new tab) PMID: 28693043 (opens in a new tab).
- Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. Am J Epidemiol. 2016;183(8):758-64. doi:10.1093/aje/kwv254 (opens in a new tab) PMID: 26994063 (opens in a new tab).
- von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370(9596):1453-7. doi:10.1016/S0140-6736(07)61602-X (opens in a new tab) PMID: 18064739 (opens in a new tab).
- Reference verification note. 44 of 44 references are PubMed-indexed; their bibliographic records (authors, title, journal, year, volume, pages, DOI, PMID) were retrieved programmatically from the PubMed database on November 17, 2026. Database retrieval confirms that each record exists as cited; it does not replace full-text reading by the responsible authors of every statement attributed to a source.

