We are grateful for the opportunity to respond to the letter submitted by Drs Raassen and Hancock.
First, Drs Raassen and Hancock's claim that the ‘whole article is based on incorrect indications for abdominal repair’ is greatly overstated. Drs Raassen and Hancock seem to have misunderstood our aims, the first of which was to empirically assess the indications for the abdominal route of repair using our dataset.
Second, we would like to reiterate that extensive scarring and tissue loss were included in our composite measure representing indications for abdominal repair because vaginal stenosis is a published indication for abdominal route of repair. To allay their concerns regarding urethral involvement, we highlight that both the multivariate analyses examining the predictors of abdominal route of repair and those examining the influence of route of repair on repair failure controlled for the degree of urethral damage. Furthermore, their implication that vesicouterine location was omitted is simply untrue.
Nonetheless, to respond to their concerns, we conducted analyses excluding extensive scarring and tissue loss from the composite variable, and our findings were similar. In terms of our first aim, published indications for abdominal route remained highly predictive [adjusted relative risk (ARR), 8.68; 95% confidence interval (95% CI), 3.73–20.17], with no meaningful differences in the final multivariate model. With regard to our second aim, the ARR of the vaginal approach on failure to close the fistula was 1.26 (95% CI, 0.90–1.75). Stratified analyses, conducted as part of the third aim, show the same trend as that using the more liberal definition of ‘abdominal repair indicated’: women meeting the indications had a lower risk of failed fistula closure (RR, 1.53; 95% CI, 0.95–2.44) than women who did not (RR, 2.52; 95% CI, 1.67–3.80).
Drs Raassen and Hancock claim that missing data on the mode of delivery and fistula cause might have revealed noncomparability between the groups. However, in any observational study, comparability is achieved by controlling for potential confounding variables. When we examined predictors of route of repair, we controlled for fistula characteristics, such as location and type, rather than more distal factors, such as fistula cause or mode of delivery. The latter factors may influence fistula location and type, but would not be expected to influence directly the route of repair undertaken. Similarly, analyses examining the effect of route of repair on fistula closure controlled for indications for abdominal route of repair (including intracervical fistulas) among other measures of fistula location and type.
It is true that many women who met the indications for abdominal repair in our dataset did not have an abdominal repair, and we would like to refer Drs Raassen and Hancock to the first paragraph of the Discussion section where we discuss this finding. Indeed, we did not claim that these were absolute indications.
We are very clear in the article about its limitations; nonetheless, previous research on this subject has suffered from methodological weaknesses on which we improved through the use of robust statistical methods.