How were the studies combined?
Scatterplots of the baseline and cycle 6 means for pairs of the parameters (total cholesterol, triglycerides, HDL-C, and LDL-C), and for each parameter independently, were examined.
Combined estimates of the differences between the means at the 3 time points (baseline, cycle 3 and cycle 6) were obtained for the following pairs of study subgroups:
1. studies with maximum age greater and equal to or less than 35 years;
2. studies with a washout period of less than and equal to or greater than three cycles;
3. the use of randomisation;
4. the use of additive or nonadditive statistical models; and
5. the presence or absence of a report of the standard deviation (or standard estimate of the mean) at cycles 0 and 6.
Both weighted and unweighted statistical methods were used. Descriptions of the statistical procedures used are provided.
How were differences between studies investigated?
The following study characteristics were examined for homogeneity:
1. the maximum age of participants;
2. the duration of the washout period before initiation of treatment with the study;
3. the timing of plasma sampling during the menstrual cycle; and
4. whether the studies were randomised.