判定个别病例疗效的统计方法

A Statistical Method for Testing Efficacy of Treatment in Individual Patient

作者对慢性病、罕见病单个病例的临床治疗试验提出了一种统计检验方法。资料必须满足以下假定:①患者病情可概括为5个状态;②病情可重复观察多次;③各次观察结果相互独立;④被研究的疾病在试验明间的自然过程稳定。符合以上假定的资料方可使用本法。

A statistical method for testing efficacy of treatment in individual patient with chronic or rare diseases is worked out. The basic assumptions are: 1. The condition of patients can always be shown to be one of the 5 states. The starting state, i. e. the condition of the patient just before the treatment, is always in the middle. The other four states are: better, even better, worse, even worse states. 2. Patients can be observed n + 1 times, the first observation is just before the treatment and the others are n different times after the starting of the treatment. 3. The data collected at each time are independent of each other. 4. The natural course of the disease is relatively stable when the trial is going on. Every sequence of n +1 data from one patient fornis one point in the sample space with a probability distribution that all of its points have equal probability. Under the null hypothesis that there is no effect of treatment, data observed at each time would be round the starting state, otherwise they would tend to deviate from it. A function for measuring the deviation is defined, and, by means of it, the sample space is represented as a union of its subsets (events). The more an event deviates from the starting state, the less possibility there should be for this event to happen. After calculating the probability of each event, we can then determine a rejection region. An asymptotically normally destributcd statistic is given for dealing with large n.