A prospective observational study examining the association between fluid therapy and outcome reports the following results:

"Crude 90-day mortality of patients who received colloids was higher than in  patients  treated  exclusively with crystalloids; (25.5% vs. 15.4%, odds ratio (OR) 1.84, 95% confidence interval (Cl) 1.56 to 2.18). After multiple logistic regression analysis, the adjusted OR was 0.923, 95% Cl (0.87 to 1.19), p = 0.09."

a)  Interpret these results.    (30% marks)

[Click here to toggle visibility of the answers]

College answer

There was a significantly higher mortality in patients who received colloids compared to those who received crystalloids. However, when other factors likely to influence mortality were taken into account by multiple logistic regression analysis, the difference was no longer statistically significant. The interpretation is that fluid choice is not significantly associated with 90-day mortality. (3 marks)


The data here comes from Ertmer et al, 2011

The crude odds ratio here appears statistically significant as the CI is well away from 1.0. The effect size there is also significant. There is no p value reported, which is unhelpful. The adjusted OR is very different and is in fact the opposite of the crude OR, which raises major concerns. The "multiple logistic regression analysis" would have to be more carefully scrutinised to determine which variables they threw into the soup. Usually, the investigators just choose whichever variables had a p-value below 0.05 in the first univariate analysis. The more intelligent method would be to test the independent variables in pairs and in groups to understand the meaning behind their interaction, and then pick only the meaningful variables for your multivariate analysis. In short, there was no difference in mortality, according to the presented fragments of data.