Class- and dose-dependent heterogeneity in the risk of medication-related osteonecrosis of the jaw among antiresorptive agents: a nationwide new-user cohort study using Korean National Health Insurance claims data

Abstract

Antiresorptive agents are widely used to treat osteoporosis and skeletal complications of malignancy, but long-term use can lead to medication-related osteonecrosis of the jaw (MRONJ). No nationwide Korean study has directly compared oral and intravenous bisphosphonates with low- and high-dose denosumab in a single new-user cohort or examined risk heterogeneity by indication. Using Health Insurance Review and Assessment Service (HIRA) claims data, we constructed a retrospective new-user cohort of adults aged 19 years or older who initiated an antiresorptive agent (oral or intravenous bisphosphonate, denosumab 60 mg or 120 mg) in 2020, after a 2015–2019 washout period. MRONJ was defined by a dental claim with diagnosis code M87. Incidence (per 1,000 person-years) was calculated by drug group, and independent risk factors were assessed using multivariable Cox regression. Because of the sparse event count, a planned cancer-stratified Cox model was not statistically feasible; we instead fitted an expanded model that also adjusted for history of malignancy. The effects of preventive dental care and physician-dentist collaboration were evaluated using propensity-score-matched Cox models. The cohort comprised 357,889 patients (oral bisphosphonate, 159,601; intravenous bisphosphonate, 108,667; denosumab 60 mg, 87,363; denosumab 120 mg, 1,935; combination, 323). MRONJ occurred in 136 patients (0.038%; incidence 0.07 per 1,000 person-years, 95% CI 0.06–0.08). Incidence was similar across oral bisphosphonate, intravenous bisphosphonate, and denosumab 60 mg (0.05–0.06 per 1,000 person-years) but higher with denosumab 120 mg (2.85 per 1,000 person-years; adjusted hazard ratio [HR] 90.84, 95% CI 54.32–151.92, core model). When malignancy history was adjusted for along with other measured covariates, the hazard ratio decreased to 44.69 (95% CI 24.04–83.07); malignancy itself was a strong, independent risk factor (HR 2.67, 95% CI 1.71–4.18), consistent with a substantial, though not fully separable, contribution of confounding by indication. Significant risk factors included tooth extraction (HR 5.56), periodontal procedures (HR 2.04), alveolar bone surgery (HR 1.77), and diabetes (HR 1.82) (none of which included the null value). Quasi-complete separation precluded stable estimates in the physician-dentist collaboration analysis. Among patients receiving antiresorptive agents, MRONJ risk differed significantly by drug class and indication. Invasive dental procedures were the strongest risk factor. To our knowledge, this is the first nationwide Korean new-user cohort study to compare the full spectrum of antiresorptive agents directly. Because this was an observational study with residual confounding by indication and without the originally planned cancer-stratified analysis, these findings should be regarded as hypothesis-generating rather than a direct basis for clinical guidelines or reimbursement decisions.