Aino Hiroshi, Yanagisawa Shinichiro, Kamae Isao
Division of Health Informatics and Sciences, Research Center for Urban Safety and Security, and Graduate School of Medicine, Kobe University, 7-5-1 Kusunoki-cho, Chuou-ku, Kobe, Japan 650-0017.
J Public Health (Oxf). 2004 Mar;26(1):84-7. doi: 10.1093/pubmed/fdh101.
The number needed to treat (NNT) is a practically useful indicator that represents how many patients must be treated to prevent one adverse event when provided with a new intervention instead of the standard one. The NNT associates the net-benefit of an experimental treatment with the number of patients, or the size of trials, expecting one outcome of success. The NNT, however, also suggests that we assume an implicit execution of independent Bernouilli trials--as it were, the hypothetical NNT trials--independently repeated the same number of times as the value of NNT and with the same occurrence-probability of success as the value of absolute risk reduction. These independent Bernouilli trials, of course, have some probabilities of failure. Most decision-makers in practice would be more interested in how much the hypothetical NNT trials can achieve 'success/failure' with 'how many' patients, or the 'odds' of success versus failure, rather than 'one' outcome of success as the mean value.
We investigated the properties of hypothetical NNT trials. A binomial distribution was employed to develop formulae for estimating the odds of success versus failure to gain net-benefit in the NNT-associated trials.
Most of the estimates of odds expected by the new intervention are between three and 1.72, converging to e-1 as the NNT increases.
When basing decisions on an NNT, clinicians and public health specialists should take account of the odds of achieving the theoretical NNT.
需治疗人数（NNT）是一个实用的指标，它表示在采用新干预措施而非标准干预措施时，必须治疗多少患者才能预防一例不良事件。NNT将实验性治疗的净效益与预期获得一次成功结果的患者数量或试验规模联系起来。然而，NNT还意味着我们假设进行了独立的伯努利试验——可以说，是假设的NNT试验——独立重复的次数与NNT的值相同，且成功发生概率与绝对风险降低的值相同。当然，这些独立的伯努利试验存在一定的失败概率。实际上，大多数决策者更感兴趣的是假设的NNT试验用“多少”患者能够实现“成功/失败”，即成功与失败的“几率”，而不是将一次成功结果作为均值。
我们研究了假设的NNT试验的特性。采用二项分布来推导公式，以估计在与NNT相关的试验中获得净效益的成功与失败几率。
新干预措施预期的大多数几率估计值在3到1.72之间，随着NNT的增加收敛于e - 1。
在基于NNT做出决策时，临床医生和公共卫生专家应考虑实现理论NNT的几率。
