Department of Biostatistics, School of Public Health, University of Ghana, Legon, Accra, Ghana.
Malar J. 2013 Jul 19;12:253. doi: 10.1186/1475-2875-12-253.
Sample size has increasingly become a prerequisite for grant approval. Study size calculations for multicentre trials are more complicated because these sites present different assumptions on incidence of disease expected in the control group; this then changes the mechanism of sample size determination. This paper suggested an alternative approach to estimating study size in multicentre vaccine efficacy trials.
The approach suggested in this paper was to determine the expected number of events for a given sample size under set of different assumptions. The power was then calculated given the expected number of events under the set of assumptions so as to assess the sensitivity of the sample size. The approach was then illustrated assuming a malaria vaccine efficacy trial planned in four centres.
The approach showed that by assuming 30% cumulative incidence of malaria in three of the centres and 10% cumulative incidence in the other centre, a sample size of 460 children in each centre (total 1,840) corresponding to a total of 339 events gives 90% power to detect vaccine efficacy of 30% at 5% level of significance, allowing for 15% loss to follow-up. However, if the incidence is lower than anticipated or a centre drops out altogether the power will be low. But this would not have much effect if it were a low incidence centre. Rather, it might have major effect if it were a high incidence centre.
Decision on recruitment depends on whether separate estimates of efficacy in each transmission level are reasonable. If not, equal numbers can be recruited, which then gives safety data for each site and overall efficacy. Recruiting all or most subjects in the highest transmission site can minimize sample size but may be better to spread the risk due to uncertainty about incidence due to year to year variation and also the possibility of a site dropping due to political or other unforeseen problems.
The approach demonstrated the potential of estimating the expected number of events required to give a specified power for multicentre efficacy trails of blood stage malaria antigens.
样本量已日益成为获得资助的前提条件。由于这些研究中心对预期对照组疾病发生率有不同的假设，因此多中心试验的研究规模计算更加复杂；这会改变样本量确定的机制。本文提出了一种用于估计多中心疫苗效力试验研究规模的替代方法。
本文建议的方法是根据不同假设确定给定样本量下的预期事件数。然后根据假设下的预期事件数计算功效，以评估样本量的灵敏度。然后，以计划在四个中心进行的疟疾疫苗效力试验为例说明了该方法。
该方法表明，假设三个中心的疟疾累积发病率为 30%，而另一个中心的发病率为 10%，则每个中心招募 460 名儿童（总计 1840 名），总共 339 例事件，可在 5%的显著性水平下检测到 30%的疫苗效力，允许 15%的失访率。然而，如果发病率低于预期，或者某个中心完全退出，那么功效就会很低。但如果是低发病率中心，影响不会太大；而如果是高发病率中心，影响可能会很大。
招募决策取决于是否可以对每个传播水平的疗效进行单独估计。如果不可行，则可以招募相同数量的患者，这样每个中心和总体疗效都会有安全性数据。在发病率最高的中心招募所有或大部分患者可以最小化样本量，但由于发病率每年的变化以及由于政治或其他不可预见的问题而导致某个中心退出的可能性存在不确定性，分散风险可能会更好。
该方法展示了估计多中心血期疟原虫抗原效力试验所需的预期事件数以达到特定功效的潜力。
