Comparison of Repeated Measurement Design and Mixed Models in Evaluation of the Entonox Effect on Labor Pain

In many medical studies, the response variable is measured repeatedly over time to evaluate the treatment effect that is known as longitudinal study. The analysis method for this type of data is repeated measures ANOVA that uses only one correlation structure and the results are not valid with inappropriate correlation structure. To avoid this problem, a convenient alternative is mixed models. So, the aim of this study was to compare of mixed and repeated measurement models for examination of the Entonox effect on the labor pain.
This experimental study was designed to compare the effect of Entonox and oxygen inhalation on pain relief between two groups. Data were analyzed using repeated measurement and mixed models with different correlation structures. Selection and comparison of proper correlation structures performed using Akaike information criterion, Bayesian information criterion and restricted log-likelihood. Data were analyzed using SPSS-22.
Results of our study showed that all variables containing analgesia methods, labor duration of the first and second stages, and time were significant in these tests. In mixed model, heterogeneous first-order autoregressive, first-order autoregressive, heterogeneous Toeplitz and unstructured correlation structures were recognized as the best structures. Also, all variables were significant in these structures. Unstructured variance covariance matrix was recognized as the worst structure and labor duration of the first and second stages was not significant in this structure.
This study showed that the Entonox inhalation has a significant effect on pain relief in primiparous and it is confirmed by all of the models.