Anterior sphincteroplasty for fecal incontinence: predicting incontinence relapse

Abstract

Purpose

This study was designed to evaluate the efficiency of anterior sphincteroplasty in preventing fecal incontinence relapsing in 85 female patients.

Methods

This observational study followed individuals for up to 10 years after intervention. Fecal incontinence relapse was analyzed using Generalized Linear Models and Kaplan-Meier tables. Bias due to informative censoring and missing data were assessed. Two postoperative cutoff Wexner scores (4 and 8) were used to classify individuals into continent or incontinent, and their model implications were examined.

Results

The hazard of relapsing appeared constant over time. This led to exponential time-to-relapse functions, and a linear increase of cumulative hazard over time. Predicted median relapsing time was 33 years, and overall risk 0.09 ± 0.03, when using a cutoff Wexner score of 8 (moderate), and 5 years, overall risk 0.45 ± 0.05, when using a cutoff of 4 (mild). There was a potential underestimation in parameters (bias) due to informative censoring, i.e., individual with better prognoses were more likely to drop out before relapsing compared to those with worse prognoses. Thus, true relapsing times may be longer than our current estimates.

Conclusions

The predictive model can be used in practice for individual prognosis after intervention, based on preoperative Wexner scores. The effect of anterior sphincteroplasty on fecal incontinence does not seem to deteriorate over time. A consensus Wexner cutoff is necessary to compare data and interventions.

Appendix

Assuming a constant hazard (h) over time, the time-to-relapse function can be written as S(t) = e ^H(t) = e ^−ht, where H(t) is the cumulative hazard up to year t, and S(t) is the probability of survival or feacal continence (FC). Thus, the probability of FC after AS decreases exponentially at a rate of e ^−h per year [18]. An estimate of h can be obtained by regressing H(t) on t, where H(t) is the KM estimate given in Table 1. The resulting equation is H(t) = 0.12 + 0.12t, so S(t) = e ^{−(0.12+0.12t)} = 0.89e^−0.12t, where t = 0,…,10. The factor 0.89 in S(t) accommodates the fact that S(0) = 0.93, and not 1, i.e., there were six mild FI relapses within the first year after AS, when t = 0. When the critical postoperative WS was 8, the regression of H(t) on t rendered equation H(t) = 0.04 + 0.02t, and thus S(t) = 0.96e^−0.02t.

The model in Table 3 for critical WS = 4, predicts h as a function of preoperative WS as $h=1-{e}^{-{e}^{\eta }}$, where η = −3.55 + 0.13WS, and WS is the preoperative WS. The time-to-relapse function can be shown to be $S(t)={e}^{-ht}=\frac{e^{t{e}^{\omega }}}{e^t}$, where ω = −0.03 × 1.14^WS.