Introduction

Approximately 10 million people in the USA have osteoporosis[1]. Osteoporosis-related fractures can increase disability, total health care costs, and mortality[2]. Many treatments can be used to treat osteoporosis depending on its severity. Bisphosphonates (BPs) and denosumab have been widely used to treat osteoporosis since 1995 and 2010, respectively. However, patients sometimes experience osteonecrosis of the jaw (ONJ) as a result of taking BPs and denosumab [3, 4]. ONJ is a persistent painful necrosis of bone in the maxillofacial region, which is associated with significant morbidity and reduced quality of life, although its pathogenesis remains poorly understood [5]. BPs are a class of antiresorptive medications, which were first introduced to the American market in 1995 to treat bone diseases such as bone cancer, metabolic bone diseases, and osteoporosis. In 2003, a handful of case studies reported the occurrence of ONJ following BPs use [6, 7]. In 2007, the American Association of Oral and Maxillofacial Surgeons (AAOMS) provided a definition for bisphosphonates-related osteonecrosis of the jaw (BRONJ) [8]. In 2014, the term was revised to medication-related osteonecrosis of the jaw (MRONJ) given the finding that it was also associated with the use of other antiresorptive agents (denosumab) or angiogenesis inhibitors (bevacizumab) [9]. Risk factors for MRONJ include age [10], sex, comorbidities (diabetes mellitus [DM] [5], chronic kidney disease [CKD], anemia [11], hyperthyroidism [12], hypothyroidism, rheumatoid arthritis, periodontal disease [11, 13]), dental treatment regimens (tooth extractions [11], ill-fitting dentures), and drugs (corticosteroid [14], antiangiogenic agents [bevacizumab and sunitinib] [15]). The incidence of MRONJ among patients using BPs injections has been reported as 2 to 15% [16]. In a survey study of more than 13,000 Kaiser Permanente members, the prevalence of MRONJ in osteoporotic patients receiving long-term oral BP therapy was reported as 0.1% (10 cases per 10,000), which increased to 0.21% (21 cases per 10,000) in patients who had been taking oral BPs for more than 4 years [17]. Moreover, patients undergoing dental treatment have been shown to have a higher risk of developing MRONJ. Another study that analyzed patients with osteoporosis who were exposed to yearly BPs (Zoledronate®) therapy over a 3-year period reported a MRONJ risk of 0.017% (1.7 cases per 10,000 patients) [18], while an extension of this study to 6 years showed no change in the frequency of MRONJ [19].

Denosumab, a human receptor activator of nuclear factor kappa-B ligand (RANKL) monoclonal antibody, with a different mechanism from BPs, was approved for commercial use in the USA in 2010 and in Taiwan in 2011 (Taiwan Ministry of Health and Welfare, 2011). Preliminary clinical results [20, 21] have shown that denosumab has comparable effects to BPs in treating osteoporosis and bone mass damaged by cancer [20,21,22,23,24,25,26,27]. In 2010, Aghaloo et al. [28] first reported a case of ONJ related to the use of denosumab in a patient with a sacral giant cell tumor. Although clinical cases of MRONJ have been reported, the mechanisms by which denosumab acts on the bone differ from those of BPs [29,30,31,32]. Given its 28-day half-life, it is believed that denosumab will not be retained in the body for long periods, and short-term follow-ups have demonstrated its low likelihood of inducing MRONJ [20, 21, 33]. Recent studies on osteoporotic patients exposed to denosumab have reported a risk of MRONJ of 0.04% (4 cases per 10,000 patients) [21]. Some studies have reported that the incidence of MRONJ among patients with cancer exposed to high doses of denosumab ranges from 0.7 [34] to 1.9% [35]. Therefore, the incidence of denosumab-related ONJ (DRONJ) does not differ significantly from that of BPs-related ONJ in patients with cancer [36, 37]. However, the results vary across countries and populations. Additionally, there is no consensus on whether denosumab treatment poses a lower risk of MRONJ in osteoporosis patients with long-term administration. The incidence difference of MRONJ in previous studies could be due to the difference in MRONJ risk factors and populations in these studies. Moreover, early studies focused on European and North American countries, which differ from Asian countries in terms of race, economic condition, medical status, and patient-related factors. Indeed, a recent study by Taguchi [16] revealed that the incidence of MRONJ is higher in Asian countries, highlighting the importance of conducting large-scale research on this issue in the Asian region [38,39,40,41,42,43].

Therefore, the objective of this study was to compare the risk of MRONJ between osteoporosis patients using BPs vs. denosumab based on data collected from the private database of multiple large-scale medical institutions, called the Chang Gung Research Database (CGRD) [44, 45]. This database accounts for 1/10 of Taiwan’s National Health Insurance Database (NHIRD). The NHIRD was considered as a reference because it covers all medical claims from 99.9% of the entire Taiwanese population. We also validated whether the long-term use of denosumab poses a lower risk of MRONJ than the long-term use of BPs. Finally, the severity of MRONJ was classified based on the treatment regimens to better understand whether differences exist in the severity of MRONJ following the use of BPs and denosumab. These results are expected to serve as a reference for clinicians to administer suitable medications for patients with osteoporosis.

Methods

Study design and population

We conducted a multi-institutional retrospective cohort study using the Chang Gung Research Database (CGRD) [44, 45]. The study population comprised patients who were diagnosed with osteoporosis between January 2010 and December 2018. The ICD-9-CM and ICD-10-CM diagnosis codes for osteoporosis are 733.0 and 733.1, respectively. The exclusion criteria were as follows: (1) < 40 years old; (2) previously received radiotherapy; (3) diagnosed with cancer; (4) diagnosed with ONJ before being diagnosed with osteoporosis; (5) diagnosed with ONJ before starting anti-osteoporotic drugs; (6) < 1 year of continuous anti-osteoporotic drug use; (7) no drug use within a year prior to ONJ diagnosis; (8) failure to attend follow-ups after seeking treatment; and (9) used drugs for osteoporosis other than BPs or denosumab or did not seek treatment at all. The eligible patients were divided into BPs and denosumab groups based on their choice of drug (Fig. 1).

Fig. 1
figure 1

Study patient enrollment

Full size image

Co-variables and propensity score matching

The patients’ age, sex, comorbidities, use of other drugs, and dental treatment regimens were recorded and analyzed in both groups given that these are relevant factors on ONJ. Propensity score matching was used for group matching to diminish the influence of relevant factors on ONJ in group comparison. The variables included in the matching process were age, sex, comorbidities (DM, CKD, anemia, hyperthyroidism, hypothyroidism, rheumatoid arthritis), dental treatment regimen, and drugs (steroids). After matching both groups at a 1:1 ratio, the incidence of ONJ and the correlation between the duration of drug use and the occurrence of ONJ were derived. The severity of ONJ was assessed based on the dental treatment regimen that was used.

Outcome measures: ONJ complication

All of the patients in both groups had continuously and exclusively used the respective drug for more than 1 year. We evaluated the occurrence of ONJ in each group after the use of anti-osteoporotic drugs from January 2010 to December 2018. The occurrence of ONJ was confirmed by the ICD-9-CM and ICD-10-CM diagnosis codes and the related dental procedure (Appendix).

Statistical analysis

Descriptive statistics are expressed as the mean ± standard deviation (SD), median (range), or frequency (percentage), as appropriate. All numerical variables were tested with normality by P-P plots or Kolmogorov–Smirnov test. Data were log-transformed to the approximate normal distribution with the right skewness. Student’s t-test was used to compare continuous variables between groups, and the chi-squared test or Fisher’s exact test was used to compare categorical variables between groups. We estimated the propensity score by modeling the probability of being in the BPs group vs. the denosumab group. The cumulative incidences of ONJ were described by the Kaplan–Meier method. The differences in the time to ONJ development were compared with the log-rank test. The Cox proportional hazards regression model was employed to analyze the hazard ratio of the associated factors for ONJ development. A two-tailed p value < 0.05 indicated statistical significance. All statistical analyses were performed with SPSS version 20.0 (SPSS Inc, Chicago).

Results

Based on the private database of the large-scale medical institution (CGRD), 84,102 patients were diagnosed with osteoporosis between January 2010 and December 2018. We excluded 2463 patients who were < 40 years; 1,515 patients who had received head and neck radiotherapy before enrollment; 2764 patients who were diagnosed with cancer before enrollment; 612 patients with ONJ occurrence before osteoporosis; 138 patients with ONJ occurrence before the use of osteoporosis drugs; 43,693 patients who were using drugs other than BPs or denosumab or did not seek treatment at all; 20,334 patients who did not continuously use anti-osteoporotic drugs more than 1 year; and 3621 patients who failed to attend follow-ups after seeking treatment. After excluding 75,140 patients, the 8962 eligible patients were divided into two groups based on their drug use; the BPs group included 5139 patients, and the denosumab group included 3823 patients (Fig. 1). Following propensity score matching at a 1:1 ratio, each matching group included 3665 patients. The two groups showed no significant differences in related systemic diseases and drug use, but certain factors, such as tooth extraction, were significantly higher percentages in the denosumab group (p < 0.04) (Table 1). After matching, the incidence of ONJ in both groups was calculated, along with the correlation between the duration of drug use and MRONJ occurrence, and the severity of MRONJ was assessed based on the dental treatment regimen used (Table 2). The incidence density of ONJ in the BPs and denosumab groups was 2.49 and 1.47 (per 1000 person-years), respectively. The hazard ratio (95% confidence interval) of denosumab vs. BPs was estimated as 0.581 (0.33–1.04) (p = 0.07). Regarding the correlation between the duration of drug use and ONJ occurrence, the cumulative incidence rates of ONJ in both groups were similar for the first and second years after drug use, but significant differences were observed from the third year onwards (Figs. 2, 3). Moreover, the likelihood of ONJ occurrence stabilized over time in the denosumab group but increased in the BPs group. As shown in Fig. 2, the cumulative incidence rate of ONJ in the BPs group was approximately twofold more than that of the denosumab group from the fourth year onwards. The severity of MRONJ comprises three stages, each with its own treatment approach: Stage 1 involves drug treatments; stage 2 involves local debridement; and stage 3 involves surgical debridement and resection [46]. Even though there was little difference between the BPs and denosumab groups in terms of the severity of ONJ, in the denosumab group, 13 of 18 patients with ONJ received drug treatment, while the remaining five patients received surgical debridement and resection to remove large areas of damaged tissue; in contrast, most patients with ONJ in the BPs group received drug treatment and local debridement. Those differences were not statistically significant in the severity of ONJ (Table 2).

Table 1 Demographic information of the study population before/after propensity score matching
Full size table
Table 2 Risk of osteonecrosis of the jaw (ONJ) (denosumab vs. bisphosphonates)
Full size table
Fig. 2
figure 2

Likelihood of osteonecrosis of the jaw (ONJ) occurrence between the denosumab and bisphosphonates (BPs) groups

Full size image
Fig. 3
figure 3

ONJ occurrence is less likely with denosumab compared to BPs use beyond 2 years

Full size image

The clinical characteristics of patients with ONJ receiving different antiresorptive agents are shown in Table 3. In the denosumab group, patients with ONJ had a significantly higher incidence of hypothyroidism (p = 0.025), tooth extraction (p = 0.018), and steroid use (p = 0.041), and were younger (p = 0.048) than those patients without ONJ. In the BPs group, patients with ONJ had a significantly higher incidence of DM (p = 0.035), tooth extraction (p < 0.001), deep periodontal curettage (p < 0.001), and steroid use (p = 0.044).

Table 3 Clinical characteristics of patients with and without ONJ in both propensity score matching groups
Full size table

Discussion

The results of this study demonstrated that after a 2-year use, the likelihood of ONJ being induced by denosumab use was lower than that of BPs use for osteoporosis patients, and the difference increased with time. However, there was no significant difference in the severity of MRONJ between the two treatment groups.

In our study, the incidence density of BRONJ was 2.49 (per 1000 person-years). Among previous studies in osteoporosis patients in Asia, the closest value reported was 2.83 (per 1000 person-years), which was reported in the National Taiwan University study by Chiu [43]. However, remarkable differences existed in other Asian studies, with reported values ranging from 0.08 to 2.8 (per 1000 person-years) [16]. Therefore, the value obtained in our study was within a reasonable range in Asia. The range of incidence rate or density reported in previous studies may be due to the inclusion of different countries, evaluated years, ONJ case definitions, and risk factors. Moreover, in previous studies, there seems to be a tendency that the more years evaluated, the higher the incidence density of BRONJ.

In this study, the likelihood of developing ONJ following the long-term use of denosumab is lower than that of BPs (1.47 vs. 2.49 per 1000 person-years, respectively). However, the likelihood of ONJ occurrence as a result of the exclusive use of low-dose denosumab for treating osteoporosis is lower than that reported for cancer treatment [39]. Moreover, a report by Khan [5] revealed that the incidence density of DRONJ was 0 to 30.2/100,000 patient-year in osteoporotic patients [20, 21, 47]. Previous studies were mostly conducted on a small scale with small sample sizes, which contributed to the low incidence of ONJ reported in those studies. Short-term studies are also less able to reflect the cumulative effects of drugs on ONJ incidence. To the best of our knowledge, no previous study has conducted a long-term head-to-head comparison of BPs vs. denosumab in patients with osteoporosis. Moreover, we used propensity score matching with a 1:1 ratio to increase the comparability between groups using different antiresorptive medications (BPs vs. denosumab). In consideration of the time effect, survival analysis was used with a Cox proportional hazards regression model and the Kaplan–Meier method. As our study was based on 8 years of data from multiple large-scale medical institutions, the likelihood of developing BRONJ or DRONJ is likely to be higher than that reported in previous studies. Despite this, the likelihood of MRONJ being induced by low doses of denosumab is still lower than that of long-term BPs. Additionally, differences were observed between the duration of drug use and ONJ occurrence. The cumulative incidence rate of ONJ in the first and second years of drug use were very similar (p = 0.062, log-rank test), but remarkable differences began to surface from the third year onwards (p = 0.022, log-rank test). The patients in the BPs group were twice as likely to develop ONJ compared to those in the denosumab group after the third year of drug use (Fig. 2). The likelihood of DRONJ occurrence in this study was five-fold higher than that reported in previous studies [5]. Therefore, when time is considered, the likelihood of developing ONJ following long-term use of denosumab is lower than that following BPs use, which could be due to the relatively low cumulative effect caused by the mechanisms of action of denosumab. Additionally, in Taiwan, the National Health Insurance scheme and the high accessibility of healthcare resources may contribute to the higher incidence density due to earlier detection and a higher treatment rate, in contrast to patients in other countries, who, when presenting with mild symptoms of ONJ, tend to not seek treatment due to the difficult health insurance environment.

ONJ case definitions were inconsistent before 2007 due to the lack of a specific diagnosis code. In 2007, the AAOMS proposed a case definition of BRONJ based on existing literature and clinical observations [8]. However, the adoption rate of these criteria in clinical practice is unknown. Although the ONJ diagnosis code (ICD-9: 733.45) was used after 2007, Solomon et al. [48] found that most patients with this diagnosis code did not actually have ONJ and that many cases reported an osteoporosis-related fracture instead of ONJ until 2010. Hence, we confirmed the occurrence of ONJ by both the potential ICD diagnosis codes and the related dental procedure (Appendix).

At present, there is no consensus on the treatment of ONJ, with some physicians preferring a conservative approach, and others preferring aggressive methods. According to the guidance from the AAOMS [49], the management strategies of MRONJ are based on the severity of MRONJ (MRONJ staging 0–3) and its progression. The choice of management strategies, either nonoperative or operative therapies, are based on patient factors and surgical judgement in a shared decision-making model with the patient, their family, and medical and dental providers, which should be patient-specific and tailored to individual needs. Nonoperative therapies were documented to be useful in all stages and led to cures in earlier stages or disease stabilization, which included patient education, pain control, secondary infection control, and sequestration of the exposed necrotic bone [49, 50]. In stage 1, patients could use chlorhexidine to improve oral hygiene by chemically removing biofilm from the necrotic bone surface [50]. Surgery is not indicated in stable cases with an adequate quality of life. In stage 2, patients may experience difficulty with local wound care and may require antibiotics to control symptoms. In stage 2 or 3, patients who are poor surgical candidates may be indicated for nonoperative therapies [50]. Nonoperative therapies do not always result in the sequestration of the exposed necrotic bone with disease resolution [51]. Thus, operative interventions, including marginal or segmental resection or partial maxillectomy, should be presented as a treatment option to reduce the progression of disease and can be applied to patients with all stages of MRONJ [52]. In patients with more advanced disease at presentation or progressive clinical or radiographic aspects, surgical resections should be performed; these resections require margins beyond the borders of the necrotic bone to an area of vital bone [53]. Based on the data in this study, there are no significant differences in the severity of ONJ induced by BPs and denosumab. Even though surgery is more effective than conservative treatment [54, 55], the patient’s psychological and physiological conditions must be taken into account whenever necessary, which is why different patients were subjected to different treatment approaches.

The histopathological and radiologic features varied between patients receiving denosumab and those receiving BPs [56]. For instance, denosumab-related ONJ showed significantly lower numbers of osteocytes per area. Additionally, BPs-related ONJ showed numerous bone resorption lacunae on the necrotic bone surface [57], which were limited in denosumab-related ONJ. Radiologic features of BPs-related ONJ are clearly described in the literature [58] and include thickened lamina dura, bone sclerosis, prominence of the inferior alveolar canal, and pathological fractures, in addition to the features of sequestrum, subperiosteal bone formation, and lysis of the cortical border of the jaw [59].

Regarding the limitations of our research, as the CGRD database was used in this study, data on the drugs taken by the patients that were not administered by Chang-Gung medical institutions were not accounted for. Moreover, stage I ONJ cases may not be detected due to mouthwash use alone without oral antibiotic use and dental procedure [8]. Furthermore, as the incidence density of DRONJ was calculated in the propensity-score matching group, the matching process may produce selection bias, leading to a misestimated incidence density of DRONJ. However, the study by Lin [42] showed that the incidence densities were similar between groups before and after propensity-score matching.

In conclusion, this is the first Asian study to provide a head-to-head comparison of the risk of ONJ with BPs vs. denosumab in osteoporotic patients using a multi-institutional retrospective cohort study in Taiwan. Our results demonstrated that after 2-year use, the likelihood of ONJ being induced by long-term denosumab is lower than that of BPs, and the difference increases with time. These findings have the potential to directly impact the therapeutic strategy employed in patients with osteoporosis.