1 Division of Anaesthesia, Addenbrooke's Hospital, University of Cambridge , Cambridge, United Kingdom .
2 Department of Surgery, Rady Faculty of Health Sciences, University of Manitoba , Winnipeg, Manitoba, Canada .
J Neurotrauma. 2018 Jul 15;35(14):1559-1568. doi: 10.1089/neu.2017.5596. Epub 2018 Apr 5.
The study objective was to derive models that estimate the pressure reactivity index (PRx) using the noninvasive transcranial Doppler (TCD) based systolic flow index (Sx_a) and mean flow index (Mx_a), both based on mean arterial pressure, in traumatic brain injury (TBI). Using a retrospective database of 347 patients with TBI with intracranial pressure and TCD time series recordings, we derived PRx, Sx_a, and Mx_a. We first derived the autocorrelative structure of PRx based on: (A) autoregressive integrative moving average (ARIMA) modeling in representative patients, and (B) within sequential linear mixed effects (LME) models with various embedded ARIMA error structures for PRx for the entire population. Finally, we performed sequential LME models with embedded PRx ARIMA modeling to find the best model for estimating PRx using Sx_a and Mx_a. Model adequacy was assessed via normally distributed residual density. Model superiority was assessed via Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), log likelihood (LL), and analysis of variance testing between models. The most appropriate ARIMA structure for PRx in this population was (2,0,2). This was applied in sequential LME modeling. Two models were superior (employing random effects in the independent variables and intercept): (A) PRx ∼ Sx_a, and (B) PRx ∼ Sx_a + Mx_a. Correlation between observed and estimated PRx with these two models was: (A) 0.794 (p < 0.0001, 95% confidence interval (CI) = 0.788-0.799), and (B) 0.814 (p < 0.0001, 95% CI = 0.809-0.819), with acceptable agreement on Bland-Altman analysis. Through using linear mixed effects modeling and accounting for the ARIMA structure of PRx, one can estimate PRx using noninvasive TCD-based indices. We have described our first attempts at such modeling and PRx estimation, establishing the strong link between two aspects of cerebral autoregulation: measures of cerebral blood flow and those of pulsatile cerebral blood volume. Further work is required to validate.
研究目的是利用基于无创经颅多普勒（TCD）的收缩期血流指数（Sx_a）和平均血流指数（Mx_a）推导模型，这两个指数均基于平均动脉压，以估算创伤性脑损伤（TBI）患者的压力反应指数（PRx）。我们使用 TBI 患者颅内压和 TCD 时间序列记录的回顾性数据库，推导了 PRx、Sx_a 和 Mx_a。我们首先基于以下方法推导了 PRx 的自相关结构：（A）代表性患者的自回归综合移动平均（ARIMA）建模，以及（B）整个人群中具有各种嵌入式 PRx ARIMA 误差结构的序贯线性混合效应（LME）模型。最后，我们使用嵌入 PRx ARIMA 建模的序贯 LME 模型，以找到使用 Sx_a 和 Mx_a 估算 PRx 的最佳模型。通过正态分布残差密度评估模型的充分性。通过赤池信息量准则（AIC）、贝叶斯信息量准则（BIC）、对数似然（LL）和模型间方差分析评估模型的优越性。该人群中 PRx 最合适的 ARIMA 结构为（2,0,2）。这在序贯 LME 建模中得到了应用。两个模型（在自变量和截距中采用随机效应）更优越：（A）PRx∼Sx_a，以及（B）PRx∼Sx_a+Mx_a。这两个模型中观察到的和估计的 PRx 之间的相关性为：（A）0.794（p<0.0001，95%置信区间（CI）=0.788-0.799），以及（B）0.814（p<0.0001，95% CI=0.809-0.819），在 Bland-Altman 分析中具有可接受的一致性。通过使用线性混合效应建模并考虑 PRx 的 ARIMA 结构，可以使用基于无创 TCD 的指数估算 PRx。我们描述了我们在这种建模和 PRx 估计方面的首次尝试，建立了脑血流和脉动脑血流容积两个方面之间的牢固联系。需要进一步的工作来验证。
