BACKGROUND Because there are no known treatments that alter the natural course of the pathophysiology of osteoarthritis, nonoperative treatment needs to be compared with known effective treatments that seek to mitigate symptoms or with similarly invasive inert (placebo) treatments to determine effectiveness. Comparing a treatment to an uninformative control group may inappropriately legitimize and support the use of potentially ineffective treatments. We therefore investigated the prevalence of inappropriate control groups in musculoskeletal research and asked whether these are associated with reporting a positive treatment effect. QUESTIONS/PURPOSES We systematically reviewed randomized trials of nonoperative treatments of osteoarthritis and asked: (1) What proportion of randomized trials use uninformative control groups (defined as a treatment less invasive than the tested treatment, or a treatment that might possibly not outperform placebo but is not acknowledged as such)? (2) Is the use of uninformative control groups independently associated with reporting a positive treatment effect (defined as p < 0.05 in favor of the intervention, or as making a recommendation favoring the intervention over the control treatment)? METHODS In a systematic review following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we searched PubMed, Cochrane, and Embase up to September 2023 for randomized controlled trials published between 2020 to 2022 that compared one or more nonoperative treatments for the symptoms of osteoarthritis. We excluded studies that contained a surgical treatment group. We identified 103 trials that met eligibility criteria, with a total of 15,491 patients. The risk of bias was high in 60% (n = 62) of trials using the Cochrane Risk of Bias Tool, version 2. Although the high risk of bias in the included studies is concerning, it does not invalidate our design; instead, it highlights that some studies may use flawed methods to recommend treatments with unproven effectiveness beyond nonspecific effects because the kinds of bias observed would tend to increase the apparent benefit of the treatment(s) being evaluated. We used logistic regression to test the association of uninformative control groups with a positive treatment effect, accounting for potential confounders such as conflict of interest and study bias using the Cochrane Risk of Bias score. RESULTS The use of uninformative control groups (treatments less invasive than the tested treatment, or treatments that might not outperform placebo but are not acknowledged as such) was found in 46% (47 of 103) of included studies. After accounting for potential confounding, there was no association between reporting positive treatment effects and the use of an uninformative control group. Studies with a low risk of bias had a lower likelihood of reporting a positive treatment effect (OR 0.2 [95% confidence interval 0.05 to 0.9]; p = 0.04, model pseudo R2 = 0.21). CONCLUSION The finding that recent studies that mimic high-level evidence often use uninformative control groups that do not adequately account for nonspecific effects (perceived treatment benefits unrelated to a treatment's direct physiological effects) points to a high risk of legitimizing ineffective treatments. This raises the ethical imperative for patients, clinicians, journal peer reviewers, and journal editors to hold researchers to the standard of an adequate, informative control group. Awareness and risk of bias checklists might help patients and clinicians forgo new treatments based on seemingly high-level evidence that may carry only iatrogenic, financial, and psychological harm (false hope, in particular). LEVEL OF EVIDENCE Level I, therapeutic study.

中文翻译：

---

研究骨关节炎非手术治疗的随机对照试验经常使用误导性和信息量不足的对照组：系统评价。


背景 由于没有已知的治疗方法可以改变骨关节炎病理生理学的自然过程，因此需要将非手术治疗与寻求减轻症状的已知有效治疗方法或类似的侵入性惰性（安慰剂）治疗进行比较，以确定有效性。将治疗与信息不足的对照组进行比较可能会不恰当地合法化和支持使用可能无效的治疗方法。因此，我们调查了肌肉骨骼研究中不适当对照组的普遍性，并询问这些是否与报告积极的治疗效果有关。问题/目的 我们系统评价了骨关节炎非手术治疗的随机试验，并询问： （1） 有多少比例的随机试验使用无信息对照组（定义为比测试治疗侵入性更小的治疗，或可能效果可能不优于安慰剂但未被承认的治疗）？（2） 使用信息量大的对照组是否与报告积极的治疗效果独立相关（定义为 p < 0.05 支持干预，或提出支持干预优于对照治疗的建议）？方法 在遵循系统评价和荟萃分析首选报告项目 （PRISMA） 指南的系统评价中，我们检索了截至 2023 年 9 月的 PubMed、Cochrane 和 Embase，以查找 2020 年至 2022 年间发表的随机对照试验，这些试验比较了一种或多种非手术治疗对骨关节炎症状的影响。我们排除了包含手术治疗组的研究。我们确定了 103 项符合纳入标准的试验，共涉及 15,491 名患者。 在使用 Cochrane 偏倚风险工具第 2 版的 60% （n = 62） 试验中，偏倚风险较高。尽管纳入研究的高偏倚风险令人担忧，但这并不会使我们的设计无效;相反，它强调一些研究可能会使用有缺陷的方法来推荐除非特异性效果之外具有未经证实的有效性的治疗方法，因为观察到的偏倚类型往往会增加所评估治疗的明显益处。我们使用 logistic 回归来检验信息不足的对照组与积极治疗效果的关联，并使用 Cochrane 偏倚风险评分解释潜在的混杂因素，例如利益冲突和研究偏倚。结果在 46%（103 项中的 47 项）的纳入研究中发现了信息不足的对照组（比测试治疗侵入性更小的治疗，或可能不会优于安慰剂但未被承认的治疗）。在考虑了潜在的混杂因素后，报告积极的治疗效果与使用无信息量的对照组之间没有关联。低偏倚风险的研究报告积极治疗效果的可能性较低（OR 0.2 [95% 置信区间 0.05 至 0.9];p = 0.04，模型伪 R2 = 0.21）。结论 最近模仿高水平证据的研究经常使用没有信息量的对照组，这些对照组没有充分考虑非特异性效应 （感知到的治疗益处与治疗的直接生理效应无关） 表明无效治疗合法化的风险很高。这提高了患者、临床医生、期刊同行评审员和期刊编辑的道德责任，要求研究人员遵守一个充分、信息丰富的对照组的标准。 意识和偏倚风险检查表可能有助于患者和临床医生放弃基于看似高水平证据的新疗法，这些证据可能只带来医源性、经济和心理伤害（尤其是虚假的希望）。证据级别 I 级，治疗研究。