Despite a considerable number of patients with osteoporosis in China, most of them remain undiagnosed and untreated. The China Osteoporosis Prevalence Study reported the incidence of osteoporosis, vertebral fracture, and clinical fracture in the past 5 years among Chinese women was 20.6%, 9.7%, and 4.2%, respectively. But the percentage of osteoporosis patients receiving anti-osteoporotic drugs was only 1.4%[6]. One potential reason is the lack of BMD testing instruments, especially in rural areas.
FRAX is a tool to evaluate fracture risk using independent CRFs that are easy to obtain, with or without BMD information. Although FRAX has been widely used and studied the research is still at the early stage in China. Our study used FRAX Chinese version with BMD to estimate MOF and HF possibilities for 1018 Beijing participants. After the 10-year observation period, we collected data for prevalence of new osteoporotic fractures and compared it with FRAX-based fracture possibilities to find the discriminative ability of FRAX. Our study is the first report to evaluate its discrimination using 10-year data in Mainland China.
In our study, the observed 10-year incidence of MOF and HF were 10.4% and 1.1%, respectively. The prevalence of fractures was higher than the study conducted by Cheung et al[15] and the study conducted by Tamaki et al[16] but similar to the study conducted by Su et al[17]. Cheung found the observed fracture rate in 4.5 years was 4.7% for MOF and 0.9% for HF [15]. Tamaki reported the incidence of MOF and HF were 5.3% and 0.5%, respectively[16]. According to the Mr. OS and Ms. OS cohort study, 6.6% of men and 11.0% of women experienced one or more MOFs in a period of 9.94 and 8.82 years, respectively [17]. With the aging of the Chinese population, incidence of osteoporosis-related fractures will continue to increase, which causes a heavy burden on health care system. The previous studies and our current study highlighted the urgent need for finding patients at high risk to receive proper treatment in order to reduce fractures.
Of interest, except for differences in FRAX-based fracture possibilities and some other CRFs which were shown in Table 3, we also found participants who sustained a MOF also had a lower frequency of osteoporosis but a higher frequency of a diagnosis of osteoporosis or osteopenia. Previous research has shown that most of the fractured patients had their BMD T-score higher than − 2.5 [18]. And similar findings were also reported by Schuit et al [19]. However, we did not find similar results between individuals with or without HF. This may be because our study had a very small number of hip fractures. Our results suggested using BMD alone as an indicator for fracture had a low sensitivity. Thus, finding a reliable method to identify individuals at high fracture risk is of great importance in order to relieve the burden caused by osteoporotic fractures.
Our study suggested FRAX had the ability, although suboptimal, to predict osteoporotic fracture. The AUCs for MOF and HF were 0.683 and 0.776, respectively. According to a survey conducted by Crandall et al, the AUC value of FRAX for predicting MOF was 0.698. For predicting HF, the AUC value was 0.776 [20]. These results are similar to our study. However, researchers also found that compared to simple models (e.g., age + BMI model or age alone), the performance of FRAX was not superior[20]. Thus, there is still room for improvement of FRAX in fracture discrimination.
Two main approaches are taken in setting intervention thresholds with FRAX [10]. The first one is a fixed threshold that could be used without consideration of age. The NOF set the fixed threshold of MOF probability of 20%, or HF probability of 3% [11]. In Japan, a 15% FRAX-based MOF probability was considered to be the threshold [21]. The second approach is age-dependent thresholds. The threshold probability is equivalent to the fracture probability of women in the absence of other CRFs except for a prior fragility fracture [12]. However, there is no Chinese-specific intervention threshold available at present. The osteoporosis clinical guidelines recommended intervention thresholds of FRAX-based MOF possibility ≥ 20% or HF possibility ≥ 3%. Nevertheless, these thresholds had shown a low sensitivity for prediction of fractures [15] and discrimination of osteoporosis [22]. Our study has also proved that guideline-recommended thresholds could not recognize high-risk patients. When using the greatest Youden index, cut-offs of MOF possibility of 2.95% and HF possibility of 0.45% were identified. However, we should also take multiple factors (e.g., economic burden, cost-effectiveness, and awareness of osteoporosis in our society) into consideration.
Our study has limitations. First, the sample size was relatively small. Therefore, there were only a few participants reported MOF events, and the number of HFs was very low. Second, participants were not enrolled by random selection from geographically different regions. Third, the original participants had a low follow-up rate. On the other hand, we also compared the baseline characteristics between these two groups. In comparison, the results showed that no more differences were found except for age and FRAX-based fracture possibilities. Thus, the representativeness of analyzed participants was acceptable. Fourth, we identified a fracture based on self-reported data. It may have a recall bias. Additionally, some asymptomatic vertebral fractures may not be diagnosed. Finally, because we did not follow up with participants yearly, it was unknown when these fractures occurred.