Wang Siyi, Shi Mingpeng, Zhou Xue, Yu Jianan, Han Mingze, Zhang Xianshuai, Li Zhenhua, Chen Xinhua
College of Acupuncture and Massage, Changchun University of Chinese Medicine, Changchun, China.
College of Traditional Chinese Medicine, Changchun University of Chinese Medicine, Changchun, China.
BMC Musculoskelet Disord. 2025 Jan 27;26(1):87. doi: 10.1186/s12891-025-08338-z.
Patients with vertebral compression fractures may experience unpredictable residual pain following vertebral augmentation. Clinical prediction models have shown potential for early prevention and intervention of such residual pain. However, studies focusing on the quality and accuracy of these prediction models are lacking. Therefore, we systematically reviewed and critically evaluated prediction models for residual pain following vertebral augmentation.
We systematically searched eight databases (PubMed, Embase, Web of Science, CNKI, WanFang, VIP, and SinoMed) for studies that developed and/or validated risk prediction models for residual pain after vertebral augmentation. The limitations of existing models were critically assessed using the PROBAST tool. We performed a descriptive analysis of the models' characteristics and predictors. Extracted C-statistics were combined using a weighted approach based on the Restricted Maximum Likelihood (REML) method to represent the models' average performance. All statistical analyses were performed using R 4.3.1 and STATA 17 software.
Fifteen models were evaluated, involving 4802 patients with vertebral compression fractures post-vertebral augmentation. The overall pooled C-statistic was 0.87, with a 95% CI of 0.83 to 0.89 and a prediction interval ranging from 0.72 to 0.94. The models included 35 different predictors, with posterior fascia injury (PFI), bone mineral density (BMD), and intravertebral vacuum cleft (IVC) being the most common. Most models were rated high risk due to concerns about population selection and modeling methodology, yet their clinical applicability remains promising.
The development and validation of current models exhibit a certain risk of bias, and our study highlights these existing flaws and limitations. Although these models demonstrate satisfactory predictive performance and clinical applicability, further external validation is needed to confirm their accuracy in clinical practice. Clinicians can utilize these models alongside relevant risk factors to predict and prevent residual pain after vertebral augmentation, or to formulate personalized treatment plans.
椎体压缩性骨折患者在椎体强化术后可能会经历不可预测的残留疼痛。临床预测模型已显示出对这种残留疼痛进行早期预防和干预的潜力。然而，缺乏关注这些预测模型质量和准确性的研究。因此，我们系统地回顾并批判性地评估了椎体强化术后残留疼痛的预测模型。
我们系统地检索了八个数据库（PubMed、Embase、Web of Science、中国知网、万方、维普和中国生物医学文献数据库），以查找开发和/或验证椎体强化术后残留疼痛风险预测模型的研究。使用PROBAST工具对现有模型的局限性进行了批判性评估。我们对模型的特征和预测因素进行了描述性分析。基于限制最大似然（REML）方法，采用加权方法合并提取的C统计量，以代表模型的平均性能。所有统计分析均使用R 4.3.1和STATA 17软件进行。
评估了15个模型，涉及4802例椎体强化术后的椎体压缩性骨折患者。总体合并C统计量为0.87，95%置信区间为0.83至0.89，预测区间为0.72至0.94。这些模型包括35个不同的预测因素，其中后纵韧带损伤（PFI）、骨密度（BMD）和椎体内真空裂隙（IVC）最为常见。由于对人群选择和建模方法的担忧，大多数模型被评为高风险，但它们的临床适用性仍然很有前景。
当前模型的开发和验证存在一定的偏倚风险，我们的研究突出了这些现有的缺陷和局限性。尽管这些模型显示出令人满意的预测性能和临床适用性，但仍需要进一步的外部验证以确认其在临床实践中的准确性。临床医生可以结合这些模型和相关风险因素来预测和预防椎体强化术后的残留疼痛，或制定个性化的治疗方案。
