深度学习辅助胸片机会性筛查骨质疏松的临床与经济影响
本研究基于新加坡≥50岁女性,用马尔可夫微观模拟模型比较深度学习(DL)胸片机会性筛查骨质疏松(单用或联合OSTA)与不筛查的临床经济影响。结果显示DL筛查可减少骨折、增加QALY,ICER为S$46,412/QALY(17%患病率),联合OSTA效率略优。
要点
- DL胸片机会性筛查骨质疏松可减少骨折并增加QALY
- 17%患病率下ICER为S$46,412/QALY,低于支付阈值S$85,000
- 联合OSTA使ICER降低约6%并略改善临床获益
核心结论
- DL增强胸片机会性筛查在新加坡≥50岁女性中具成本效益,联合OSTA可略提效率,但单用DL已具成本效益且可扩展。
研究设计与样本
- 采用终身微观模拟马尔可夫模型,从医疗系统视角比较DL筛查、DL联合OSTA、不筛查三种策略。
- 模型输入包括筛查性能、DXA转诊率、治疗 uptake 与依从性、当地骨折发生率与成本数据。
- 基础案例假设治疗女性接受5年阿仑膦酸钠治疗。
主要发现
- 与不筛查相比,DL筛查单用减少骨折并增加QALY,17%骨质疏松患病率下ICER为S$46,412/QALY,低于S$85,000支付阈值。
- 联合OSTA使ICER降低约6%并略改善临床获益。
- 患病率9.3%-37%范围内,ICER从S$71,162至S$24,749。
- 扩大到10,000名女性时,联合策略可预防17例骨折、获得7个额外生命年和16个QALY。
关键数值与推荐
- 17%骨质疏松患病率下DL单用ICER为S$46,412/QALY,支付阈值为S$85,000/QALY。
- 联合OSTA使ICER降低约6%。
- 患病率9.3%-37%时ICER范围为S$71,162至S$24,749。
- 10,000名女性规模下联合策略预防17例骨折、获7个生命年和16个QALY。
临床怎么用
- 对≥50岁女性可考虑利用胸片进行DL机会性骨质疏松筛查,无需常规DXA即可识别高风险者。
- 联合OSTA可 marginally 提升效率,但单用DL筛查已具成本效益,适合作为可扩展的预防策略。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
UNLABELLED: Opportunistic analysis of chest radiographs with deep learning detects osteoporosis cost-effectively in women aged ≥ 50 years in Singapore. Adding OSTA marginally improves efficiency, but DL "deep learning" screening alone already reduces fractures and increases quality-adjusted life-years (QALYs) at acceptable cost. PURPOSE: Osteoporosis is underdiagnosed in older women due to limited access to dual-energy X-ray absorptiometry (DXA). Chest radiographs are commonly performed, and deep learning (DL) algorithms can opportunistically detect osteoporosis features. This study evaluated the clinical and economic impact of DL-enhanced opportunistic screening, alone or with the Osteoporosis Self-Assessment Tool for Asians (OSTA), in Singaporean women aged ≥ 50 years. METHODS: A lifetime microsimulation-based Markov model from the healthcare system perspective compared three strategies: DL screening alone, DL screening combined with OSTA, and no screening. Model inputs included screening performance, DXA referral rates, treatment uptake and adherence, and local fracture incidence and cost data. The base case assumed 5 years of alendronate therapy for treated women. Outcomes included fractures prevented, quality-adjusted life years (QALYs) gained, and incremental cost-effectiveness ratios (ICERs). RESULTS: Compared with no screening, DL screening alone reduced fractures and increased QALYs, with an ICER of S$46,412 per QALY at 17% osteoporosis prevalence, below a willingness to pay threshold of S$85,000. Adding OSTA improved efficiency, decreasing the ICER by ~ 6% while slightly improving clinical benefits. Across a prevalence range of 9.3%-37%, ICERs ranged from S$71,162 to S$24,749. When scaled to 10,000 women, the combined strategy could prevent 17 fractures, yield 7 additional life-years, and generate 16 QALYs. CONCLUSION: DL-enhanced chest radiographs offer a cost-effective approach for opportunistic osteoporosis screening in Singaporean women aged ≥ 50 years. Combining with OSTA slightly improves efficiency, but DL screening alone is already cost-effective and provides a scalable preventive strategy in real-world healthcare.
引用
- Chandran M, Reginster JY, Kim MY, et al. Clinical and economic impact of deep learning-enhanced opportunistic osteoporosis screening using chest radiographs with and without the osteoporosis self-assessment tool for Asians (OSTA). Osteoporos Int. 2026. PMID: 41563407. doi:10.1007/s00198-026-07858-1
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。