Page 22 - EVALUATION OF TREATMENT FOR HEAVY MENSTRUAL BLEEDING by Herman, Malou

Chapter 7
of IPD of randomised controlled trials (RCT) regarding second generation endometrial ablation for HMB. We refer to that paper for a full report of selection procedure and data processing of participating RCT’s.8 In short, they included 19 trials published between inception and 2011. We updated this database by looking for the newest endometrial ablation studies published between May 2010 and November 2013. Three new trials were added to this database.
We first selected the studies that evaluated second-generation ablation in at least one randomization arm. We only included data from patients that had undergone a second generation endometrial ablation who had PBAC scores collected at baseline and follow-up. Data until 12 months of follow-up (the most popular follow-up time point) were used for analysis, or results at 6 month if data were not available at 12 months.
Outcome measures were defined as satisfaction (yes/no) and surgical re-intervention (yes/no). Satisfaction and reintervention rates were collected at 12 months off follow-up. Re-intervention could be a hysterectomy or a re-ablation. Women who had a surgical re-intervention were defined as dissatisfied after their initial ablation treatment. The authors used different satisfaction scales in their papers, so the trial unit recoded these results to a simplified scale; satisfied or dissatisfied.
Statistical analysis
We compared the PBAC scores and measures of effectiveness. Effectiveness was measured as satisfaction with the treatment effect (yes/no) or surgical re- intervention (yes/no). We provided box-and-whisker plots to express PBAC scores against both satisfaction (yes/no) or re-intervention (yes/no).
A logistic regression analysis, with correction for study, was performed to predict the probability for satisfaction and re-intervention after 12 months of follow-up, using absolute PBAC scores at 12 months of follow-up (PBAC12m) and percentage change, i.e. 100% * (PBACb - PBAC12m)/PBACb.
We then calculated the 75th percentile of the PBAC score of each study and repeated the analysis for the absolute PBAC scores using multiples of the 75th percentile. We performed this analysis to normalise data from the different studies so that individual test results could be compared. We used the P75, as patients can also score a zero on the PBAC scale, which means amenorrhea, and some studies could have a median of zero.
84