Büttner Fionn, Terry Douglas P, Iverson Grant L
School of Public Health, Physiotherapy and Sports Science, University College Dublin, Dublin, Ireland.
Department of Physical Medicine and Rehabilitation and Department of Psychiatry, Harvard Medical School, Boston, Massachusetts.
Clin J Sport Med. 2021 Nov 1;31(6):e476-e483. doi: 10.1097/JSM.0000000000000825.
To introduce a new methodology for summarizing the results from systematic reviews-a likelihood heuristic-to the field of sport-related concussion.
We applied the likelihood heuristic to the results of a systematic review published by Iverson et al (2017), containing 101 studies, on the predictors of worse clinical outcome following sport-related concussion.
We re-examined 5 individual prognostic factors that are of clinical interest and for which there is conflicting literature (female sex = 44 studies, prior concussion history = 41 studies, loss of consciousness = 31 studies, post-traumatic amnesia = 25 studies, and retrograde amnesia = 10 studies).
For each prognostic factor, likelihood ratios were generated using the (1) number of significant and nonsignificant studies, (2) study power, (3) alpha level, and (4) prior probability that the alternative hypothesis was true.
Assuming each study had 80% power and an alpha level of 5%, observing the reported number of conflicting studies for female sex, prior concussion history, and retrograde amnesia is substantially more likely if each prognostic factor is associated with worse clinical outcome following sport-related concussion. For loss of consciousness, the observed number of conflicting studies is more likely if loss of consciousness is not associated with worse clinical outcome following sport-related concussion. A secondary analysis incorporating potentially more realistic study parameters of statistical power (45%) and alpha level (25%) generates weaker likelihood evidence that the observed numbers of studies for each prognostic factor are associated with worse clinical outcome following sport-related concussion than they are not.
Using a likelihood heuristic with 80% power and 5% alpha level, there is very strong likelihood evidence that female sex, prior concussion history, and retrograde amnesia are associated with worse clinical outcome following sport-related concussion. The strength of likelihood evidence that prognostic factors are associated with worse clinical outcome reduces when statistical power is lower and alpha level is inflated.
向运动相关脑震荡领域介绍一种用于总结系统评价结果的新方法——似然启发法。
我们将似然启发法应用于艾弗森等人（2017年）发表的一项系统评价结果，该评价包含101项关于运动相关脑震荡后临床结局较差的预测因素的研究。
我们重新审视了5个具有临床意义且文献存在冲突的个体预后因素（女性 = 44项研究，既往脑震荡史 = 41项研究，意识丧失 = 31项研究，创伤后遗忘 = 25项研究，以及逆行性遗忘 = 10项研究）。
对于每个预后因素，使用以下数据生成似然比：（1）显著和非显著研究的数量，（2）研究效能，（3）α水平，以及（4）备择假设为真的先验概率。
假设每项研究的效能为80%且α水平为5%，如果每个预后因素与运动相关脑震荡后较差的临床结局相关，那么观察到的关于女性、既往脑震荡史和逆行性遗忘的冲突研究数量会更有可能。对于意识丧失，如果意识丧失与运动相关脑震荡后较差的临床结局不相关，那么观察到的冲突研究数量会更有可能。纳入统计学效能（45%）和α水平（25%）等可能更现实的研究参数的二次分析产生的似然证据较弱，表明观察到的每个预后因素的研究数量与运动相关脑震荡后较差的临床结局相关的可能性并不比不相关时更大。
使用效能为80%且α水平为5%的似然启发法，有非常强的似然证据表明女性、既往脑震荡史和逆行性遗忘与运动相关脑震荡后较差的临床结局相关。当统计学效能较低且α水平升高时，预后因素与较差临床结局相关的似然证据强度会降低。
