Things go better with Burch

Trial design standardized key elements

Many types of bias confound the results of nonrandomized studies, particularly selection bias (eg, when surgeons select which procedure to perform on the basis of patient characteristics), and valid conclusions of cause and effect cannot be drawn. However, with a randomized trial, subjects are separated into groups by chance and no other factor. Thus, the groups are equivalent at baseline—provided the sample size is large enough (and allowing for random differences)—and therefore any changes measured after the experimental intervention can be confidently attributed to the intervention itself.

Another strength of the trial is standardization. The subjects were “standardized” by rather broad inclusion and exclusion criteria to constitute an important clinical population and to ensure they were sufficiently similar so that the treatment (abdominal sacrocolpopexy) was appropriate for all. In addition, surgeons at the multiple participating sites agreed to standardization of the technical details of the Burch colposuspension so that the subjects received the same intervention regardless of site. And data collection in follow-up was performed in a standard way by research staff who were blinded to the subjects’ group assignment (intervention versus control), so the data were as free of bias as possible.

Homogeneous study population may be a weakness

KOHLI: I agree that the methodology of this well-designed study is its major strength. What are its weaknesses?

WEBER: No doubt there are several, only some of which may be apparent at this time. For example, most women in the study were Caucasian, and very few were Hispanic, Asian, or black. Although we have no scientific reason to believe that Burch colposuspension has different responses in women of different racial and ethnic backgrounds, the trial’s subjects are not diverse enough to analyze the data by subgroups to confirm or refute the hypothesis that response to the Burch procedure is independent of race or ethnicity.

BRUBAKER: Another weakness: Because this study was closed after the first interim analysis, some of our secondary analyses will be underpowered, although we clearly demonstrated a difference in our primary endpoint.

It is important to remember that this study is not “finished.” Our participants are still in active follow-up for 2 years following surgery. It will be interesting to see what happens during the longer follow-up, especially with regard to prolapse and incontinence. We are also doing additional in-depth analyses of urodynamic and other parameters.

KOHLI: Again, I think the study design and analysis were well thought out. It would have been interesting to see how the results broke down according to site, to see if there was variation—which could indicate variation in surgical technique.

BRUBAKER: We have not done this analysis and do not plan to at this time.

Why paravaginal repairs were allowed

KOHLI: What about the decision to include surgeries that involved paravaginal repair?

WEBER: That generated a fair amount of discussion during trial design, as there was no clear “right” answer. Perhaps it would have been “cleaner” to eliminate the option of performing paravaginal repair, but when the trial was designed, we lacked unequivocal evidence that paravaginal repair at the time of abdominal sacrocolpopexy provides additional support for the anterior vagina. Therefore, we decided to allow the decision to be based on surgeon judgment.

Some surgeons perform paravaginal repair with abdominal sacrocolpopexy in almost all women because they believe quite strongly that this reduces the risk of recurrent anterior vaginal prolapse. Others never perform paravaginal repair with abdominal sacrocolpopexy and feel just as strongly that their patients are adequately treated and protected from subsequent anterior vaginal prolapse.

Investigators feared paravaginal repairs could dilute Burch effects

Study surgeons did agree that paravaginal repair reduces the likelihood of postoperative stress incontinence, although not as effectively as Burch colposuspension. Thus, our dilemma: If paravaginal repairs were performed in a large number of subjects, thereby improving their postoperative continence status regardless of whether Burch was performed, the effect of Burch could be so diluted as to be lost. On the other hand, if paravaginal repairs were completely excluded, that would restrict some surgeons’ practices and potentially reduce the number of women who would be offered participation in the study if their surgeons felt their anterior vaginal prolapse would be potentially undertreated.

We resolved the dilemma as follows:

1. A relatively low proportion—about one quarter—of surgeons performed paravaginal repairs regularly with abdominal sacrocolpopexy, so the potential impact in the trial would not be great.
   
2. Paravaginal repairs were allowed, but only when declared necessary by the surgeon before randomization; this step prevented surgeons from changing their minds about the necessity of paravaginal repair if the subject was assigned to the Burch group (ie, the woman would be receiving additional anterior vaginal support by way of the Burch).
   
3. We stratified for paravaginal repair in the randomization, so women with paravaginal repair were equally distributed between the intervention and control groups.