In the univariate analysis contrary to the result from the multiple analysis, there was no significant difference in constipation syndrome between the intervention group and control group. This can be explained by the difference in constipation syndrome at baseline between the intervention and control group (Table 2). To adjust for this difference, baseline value of constipation syndrome or value at week 4 was added in the multiple analysis; this showed a significant relation between constipation syndrome and abdominal massage. The analysis of diarrhoea syndrome shows an opposite pattern. There was a significant difference between intervention and control group concerning diarrhoea syndrome in the univariate analysis at week 8, but not in the multiple analysis. This can be explained by different degrees of severity ofdiarrhoea syndrome in the two groups, although this is stable over time, which makes no significant difference in the multiple analysis.