This cross-over study is to test the hypothesis that the cold temperature of the dialysate may cause the drop in patient's blood pressure. The patients were randomized into two groups. One group was treated with room temperature dialysate in period 1 and followed by warmed dialysate in period 2, while the other group was first treated with warmed and then with room temperature dialysate. the systolic and diastolic blood pressure per patient was recorded before the dialysis and was recorded four times after it.
The following boxplots show the patients' systolic and diastolic blood pressure before and after the dialysis with the warmed(hot) dialysate and the room temperature(cold) dialysate. The boxplots in the second row show the change from the pre-dialysis to the following four dialysis blood pressure readings.
From the boxplots, we can see that the changes in systolic and diastolic blood pressure seem a little bit greater with warmed dialysate. In other words, when treated with warmed dialysate, patients seem to have greater drop in blood pressure. We can also see this from the table below. The mean is calculated from the difference between the pre-dialysis and the four after-dialysis blood pressure. For both systolic and diastolic blood pressure, the mean is greater under warmed temperature. This suggests that the hypothesis may be wrong.
| BP | temp | mean | variance |
|---|---|---|---|
| Systolic | hot | 16.7007042253521 | 486.975193302167 |
| Systolic | cold | 12.827380952381 | 389.307485019324 |
| Diastolic | hot | 7.3784355179704 | 223.184612219217 |
| Diastolic | cold | 3.54389880952381 | 146.412069970393 |
We know that for a cross-over study, we should not only consider the treatment effect, the effect of the temperature of the dialysate to the drop in blood pressure, but we should also consider the period effect and the treatment-period interaction. The reason is that there may be a carry-over of treatment effect from the period 1 to 2. Using two sample t test to compare the differences between the two groups of patients, we can get the following table showing the p-values, based on the pre-dialysis and the first after-dialysis blood pressure records. With a significant level of 0.05, only the period effect of systolic blood pressure is statistically significant. This suggests that the temperature of the dialysate is not statistically significantly related with the drop in blood pressure.
| BP | Temp | Period | TempandPeriod |
|---|---|---|---|
| Systolic | 0.1502 | 0.01303 | 0.8223 |
| Diastolic | 0.1365 | 0.06316 | 0.4329 |
Using the analysis of covariance approach, we include the pre-dialysis blood pressure(baseline) into our model: Blood pressure~temp*period+predialysis BP. The result shows that period is no longer significant, but temperature and termperature-period interaction are significant, based on a significant level of 0.05. This suggests that the temperature of the dialysate is statistically significantly related with the drop in blood pressure. This result is different from the previous analysis, since in the previous analysis, we only consider the change in blood pressure and this may lead to a bias result. This is becouse baseline values are correlated with change. Patients with high blood pressure at baseline generally drop more, having greater changes than those with low blood pressure. In this way, we prefer the analysis of covariance, and we can conclude that the temperature of the dialysate is statistically significantly related with the drop in blood pressure.