Statistical aspects
Statistical aspects
复制标题
统计方面
DOI：
10.1136/bmj.328.7438.506
复制
发表时间：
2004
期刊：
影响因子：
--
通讯作者：
M. Campbell
中科院分区：
文献类型：
--
作者：
M. Campbell
本研究的主要问题是使用一项n/1试验来评价妊娠期恶心和呕吐的治疗。这些试验用于决定哪些治疗对个体患者有益，并且是比全科医生使用的通常方法更科学的决策方法：如果第一次治疗不起作用，要求病人返回，并尝试第二次治疗。正如我之前指出的，R A Fisher，最伟大的统计学家之一，详细讨论了一项1对1的试验，调查一名妇女是否能分辨出牛奶是在茶之前还是之后加入的，2这甚至成为一本书的标题。3它们通常用于慢性疾病，如关节炎，来评估病人的最佳长期治疗方案。最近的一个很好的例子表明，在一些骨关节炎患者中，扑热息痛可以代替非甾体抗炎药。4
然而，值得一提的是与1项试验中的n相关的一些问题。5在设计方面，研究者需要决定：
多少个治疗期
每次治疗需要多长时间。
由于妊娠期恶心的持续时间明显有限，因此可用的治疗周期很少，并且必须对周期数与每个周期的长度进行折衷。该试验只有五对治疗期，因此没有太大的力量来检测效果，除非效果很大。然而，这正是我们在这种情况下所希望的。为了在时间跨度内容纳甚至五对治疗期，仅给予两天的治疗，一些答复者评论说，两天的积极治疗不太可能足以纠正维生素缺乏症。
N/1试验是交叉试验，与交叉试验一样，它们也容易产生残留效应，其中治疗效应延伸至下一个治疗期。在这种情况下，这可能是一个问题，因为如果治疗确实治愈了缺陷，那么两天后它不太可能再次出现。
一个有趣的问题（统计学家进行了详细讨论，因为它影响P值）是雷诺兹女士对分配的了解。她是否只是被告知，她将接受随机分配的维生素B-6或安慰剂治疗25天，或者她是否知道有五个匹配的时期？尽管有这些警告，但很高兴看到更多的全科医生进行这类试验。
The main issue in this study is the use of an n of 1 trial for the evaluation of treatments for nausea and vomiting in pregnancy. These trials are used to decide what treatment benefits an individual patient and are a more scientific method of decision making than the usual method used by general practitioners: asking the patient to return if the first treatment doesn't work and trying a second one.1 As I have pointed out before, R A Fisher, one of the greatest statisticians, discussed at length an n of 1 trial investigating whether a woman could tell whether milk was added before or after the tea,2 and this has even become the title of a book.3 They are usually used in chronic illness, such as arthritis, to evaluate the best long term treatment for a patient. A recent good example showed that paracetemol may be used instead of non-steroidal anti-inflammatory drugs in some patients with osteoarthritis.4
However it is worth mentioning a few problems associated with n of 1 trials.5 In terms of design an investigator needs to decide:
How many treatment periods
How long to treat within each period.
Since nausea in pregnancy obviously has a limited duration, few treatment periods will be available, and a compromise will have to be made about the number of periods compared with the length of each one. The trial had only five pairs of treatment periods, and so did not have much power to detect an effect unless it was large. However, that is what we would be hoping for in this case. To accommodate even five pairs of treatment periods within the time span, the treatment was given for only two days, and several respondents commented that two days of active treatment is unlikely to be enough to correct a vitamin deficiency.
N of 1 trials are crossover trials, and like crossover trials they are also vulnerable to carry over effects, where the effect of treatment extends to the next treatment period. This may be a problem in this case, because if the treatment did in fact cure the deficiency, it is unlikely to reappear again after two days.
An interesting problem (discussed at length by statisticians since it affects the P value) is what Ms Reynolds knew of the allocation. Was she simply told that she was going to be treated for 25 days with either vitamin B-6 or placebo allocated at random or did she know that there were five matched periods? Despite these caveats, it would be nice to see more general practitioners conducting these type of trials.
副岛林三，日野二郎：现代医学 21(1) (1989)。
DOI：
--
发表时间：
--
期刊：
影响因子：
--
作者：
通讯作者：
--