Computational protocol: Haemodynamic Optimization by Oesophageal Doppler and Pulse Power Wave Analysis in Liver Surgery: A Randomised Controlled Trial

Similar protocols

Protocol publication

[…] Because of limited sample sizes and non-normal distributions of the observations, data were expressed as median [25%, 75% quartiles], or frequencies [%], respectively. For the same reason, differences between the regarded groups in terms of interesting clinical parameters were univariately tested by using non-parametric exact Mann-Whitney tests for independent groups or exact Wilcoxon tests for pairwise comparisons within the intervention groups, respectively. Frequencies were (univariately) tested by the exact Mantel-Haenszel test (ordered categories) or the exact Chi-square test. Changes in interesting clinical outcomes with respect to time were analysed using multivariate nonparametric analysis of longitudinal data in a two-factorial design (1st (independent) factor: groups, 2nd (dependent) factor: repetitions in time). Therefore, all the time points were simultaneously compared on the corresponding response curves.The following hypotheses have been tested with those analyses: differences between groups (over time) [Group], systematic changes in time (over groups) [Time overall], Interactions between groups and time [Group x Time] as well as systematic changes in time separately for every group [Time course].After global testing, post-hoc analyses were carried out to detect specific differences between groups at fixed times (Mann-Whitney tests) or within groups with respect to interesting pairs of time points (Wilcoxon tests). A two-tailed p-value < 0.05 was considered statistically significant. All tests have to be understood in the area of exploratory data analysis. Therefore, no adjustments for multiple testing have been made.Bland-Altman analysis for repeated measurements per patient were used to assess the agreement between ODM and PPA in terms of absolute values of stroke volume during the course of surgery. The bias was defined as the mean of differences between the two methods. The differences were regressed on the average to check for a constant or non-constant bias and variance. A linear mixed model with random effects was used to adjust for patient times replicate and method times patient interaction, resulting in a common standard deviation (SD) to calculate the limits of agreement (LOA’s) with upper (mean bias+1.96SD, ULOA) and lower (mean bias-1.96SD, LLOA) limits []. The percentage error (PE) was calculated as [1.96×SD of the bias/(mean(SVIODM + SVIPPA)/2)] []. The limit for an acceptable PE was redefined according to the precision of the monitors []. The precision of the ODM was previously reported ranging from 4.7% to 8.5% [–]; because there is no data on the precision of the PPA, we set the precision of the PPA at 20% []. Implying the worst reported precision of the ODM (8.5%), the redefined PE was therefore determined at < 21.7%.Trending between ODM and PPA was analysed using the polar plot method [, ]. Trending between the two methods is shown by the angle from the polar axis and the magnitude of the change in SVI by the distance from the origin. In calculating the polar statistics, negative changes were converted to positive changes by rotating through 180°; central zone data (10% changes) were excluded due to intrinsic random error of the measurement of the ODM and PPA. Subsequently the mean polar angle (or angular bias) and the radial limits of agreement (RLOA’s, radial sector that contains 95% of the data points) with an upper (R-ULOA) and lower (R-LLOA) limit were determined. Acceptable trending is generally defined as an angular bias < ±5° and RLOA’s lying within radial limits of ±30° (boundary limits) []. Similar to Bland-Altman analysis the boundary limits for acceptable trending were adapted according to the precision of both monitors from ±30° to ±21.7° around the angular bias. The angular concordance rate was calculated as percentage of data points lying within ±21.7° limits.All numerical calculations were performed with IBM SPSS Statistics, Version 21, Copyright 1989, 2010 SPSS Inc., SAS, Version 9 1, Copyright by SAS Institute, Inc., Cary, NC, USA, and the R project for Statistical Computing, Version 3.0.1. […]

Pipeline specifications

Software tools SPSS, The R Project for Statistical Computing
Application Miscellaneous
Organisms Homo sapiens
Diseases Stroke