To the Editor: We read with great interest the article published in a recent issue of Critical Care by Brouwer et al. [1]. Authors showed the beneficial effect of addition of CytoSorb filter to continuous renal replacement therapy (CRRT) for hemoadsorption of cytokines in management of septic shock. We would like to raise few points which may be important for authors and readers of the article. This was an investigator-initiated retrospective study of non-consecutive patients, treated by Cytosorb plus CRRT versus CRRT alone. This study was non-randomized, and analysis was not “intention to treat”. We believe both group populations were significantly different (disproportionately higher number of patients in control group with diabetes, hypertension, heart failure, and renal disease and nonsurgical patients) and difference in mortality may just be the result from this difference rather than intervention for the following reasons. Authors describe inclusion of all factors in regression model for determination for its association with treatment allocation; they should acknowledge that investigator initiation itself is a significant factor for bias which cannot be adjusted. Investigator-led management group will likely to have better attention and management. Relatively imperfect attention of physician taking care of patients in control group is reflected by the inappropriate assignment of renal failure as sepsis in 60 out of 109 patients. Authors used propensity score, though treatment effect made using propensity-score matching are valid only if a matched sample of treated and untreated subjects has similar distributions of measured baseline covariates [2]. Figure 2 suggests the difference in mortality only appears after application of “stabilized inverse probability of

中文翻译：

---

CytoSorb 过滤器对细胞因子的血液吸附：一项没有人为因素的模拟研究——试验是不同的

致编辑：我们饶有兴趣地阅读了 Brouwer 等人在最近一期《重症监护》上发表的文章。[1]。作者展示了在连续性肾脏替代疗法 (CRRT) 中添加 CytoSorb 过滤器对细胞因子的血液吸附在感染性休克管理中的有益效果。我们想提出一些对文章作者和读者可能很重要的观点。这是一项由研究者发起的回顾性研究，对非连续患者进行了 Cytosorb 加 CRRT 与单独 CRRT 治疗。这项研究是非随机的，分析不是“意向治疗”。我们认为两组人群有显着差异（对照组中患有糖尿病、高血压、心力衰竭、和肾脏疾病和非手术患者）和死亡率的差异可能只是这种差异的结果，而不是干预，原因如下。作者描述了将所有因素纳入回归模型以确定其与治疗分配的关联；他们应该承认，调查员的启动本身是无法调整的偏见的重要因素。以研究者为主导的管理小组可能会得到更好的关注和管理。109 名患者中有 60 名将肾功能衰竭归类为败血症，这反映了医生对对照组患者的照顾相对不完善。作者使用倾向评分，尽管使用倾向评分匹配得出的治疗效果只有在治疗和未治疗受试者的匹配样本具有相似的测量基线协变量分布时才有效[2]。图 2 表明死亡率的差异仅在应用“稳定的逆概率