Response letter to Lakin and Doe.
Response letter to Lakin and Doe.
复制标题
给莱金和多伊的回复信。
DOI：
10.1111/jgs.12108
复制
发表时间：
2013
期刊：
影响因子：
6.3
通讯作者：
Sakamoto Y et al.
中科院分区：
文献类型：
--
作者：
Ebihara S;Shannon F;Sakamoto Y et al.
编辑：Sakamoto和他的同事1进行了一项精心设计的多点纵向随机对照试验，以评估薰衣草芳香疗法贴片是否会减少养老院老年居民的跌倒。作者做了一项无可挑剔的工作，详细说明了重要的结果和变量，如功能能力、认知功能以及行为和心理方面的担忧，这些可能揭示了随机试验的失败。他们还利用对这些变量的仔细纵向监测，提出了一组有趣的次要结果，其中一些显示了薰衣草贴片的显著好处。然而，我们感到关切的是，在抽象和讨论中选择性地提出结果，让读者对审判的主要结果产生错误的印象。在摘要中，作者指出，结果显示薰衣草组比安慰剂组下降更少。在这一点上，作者没有报告摔倒次数的主要结果在统计学上没有显著差异(P=。08)，这是正确解读审判的关键。此外，他们没有报告其主要结果的未调整风险比(0.67，95%可信区间=0.40-1.10；P=。11)。在摘要中，作者报告了一个调整后的模型(模型2)产生的显著危险比，但没有说明他们这样做了。只有在调整了在基线时看起来相等的变量并通过随机分组(年龄、性别、跌倒病史、简易精神状态检查评分、科恩-曼斯菲尔德激动量表评分、转移状态、视觉状态和镇静剂使用)进行理论控制后，才显示出统计学意义。尽管有可以考虑调整的情况，但作者在方法部分没有说明为什么他们对混杂因素进行了调整。这些调整的理由是什么？此分析是预先计划的还是在数据收集后完成的？表1没有显示这些特定变量中的许多显著差异；那么为什么作者选择它们作为他们的模型？为什么他们认为这一信息不是抽象的，而是由一个调整后的模型取代，而没有说明这样做的原因？这些调整和缺乏明确和一致的标签似乎给原本设计良好的重要试验蒙上了一层阴影，该试验评估了这种低风险、有希望的干预措施的效果。
To the Editor: Sakamoto and colleagues1 generated a welldesigned, multisite, longitudinal randomized controlled trial to evaluate whether a lavender aromatherapy patch would reduce falls in elderly nursing home residents. The authors did an impeccable job of detailing important outcomes and variables such as functional ability, cognitive function, and behavioral and psychological concerns that could have revealed a failure of randomization. They also set out, using careful longitudinal monitoring of such variables, an interesting set of secondary outcomes, some of which showed significant benefit of lavender patches. Nevertheless, we are concerned about the selective presentation of results in the abstract and in the discussion that give readers a false impression of the primary outcome of the trial. In the abstract, the authors state that the results showed fewer fallers in the lavender group than in the placebo group. The authors failed to report at that point that there was no statistically significant difference in the main outcome of number of falls (P=. 08), a critical point for interpreting the trial correctly. Furthermore, they failed to report the unadjusted hazard ratio for their primary outcome (0.67, 95% confidence interval= 0.40–1.10; P=. 11). In the abstract, the authors report a significant hazard ratio yielded by an adjusted model (Model 2) without stating that they did so. Statistical significance was demonstrated only after adjustments for variables that appeared equal at baseline and theoretically controlled for by randomization (age, sex, fall history, Mini-Mental State Examination score, Cohen-Mansfield Agitation Inventory score, transfer status, visual status, and tranquilizer use). Although there are instances in which adjustment can be considered, the authors failed to describe in the Methods section why they adjusted for confounders. What was the reasoning for these adjustments? Was this analysis preplanned or completed after data collection? Table 1 did not show any significant differences in many of these particular variables; why then did the authors choose them for their models? Why did they feel that this information was appropriate to leave out of the abstract and instead be replaced by an adjusted model without describing the reasons for doing so? These adjustments and lack of clear and consistent labeling seem to cast a shadow on what was otherwise a well-designed and important trial assessing the effect of this low-risk, promising intervention.