Initially designed to treat urge urinary incontinence, sacral nerve stimulation (SNS) has been shown to be effective for the treatment of severe fecal incontinence (FI) when conservative measures have failed.1,2 It is universally accepted to screen patients for response with a trial temporary SNS before deciding on a permanent stimulation through an implanted pulse generator (IPG).3
A temporary external SNS screening test is considered successful when the number of FI episodes measured on a standardized bowel diary is reduced by ≥50% compared with baseline.3 No other effective selection criteria predictive of successful SNS has been identified thus far.4–6 Furthermore, changes in bowel habits, time to defer defecation, and urgencies, also recorded in standard diaries, are poorly evaluated in the decision to implant an IPG.
Commonly the time period for an SNS test phase is 3 weeks, but it varies according to national guidelines.7 A 3-week test phase is an empirical compromise between variability in FI frequency and the risk of lead-related infection, which is expected to increase with a longer testing duration.
Experience has shown that some patients respond rapidly to SNS so that a longer test phase might not be necessary for them. In our routine practice, patients undergoing SNS screening tests are asked to keep a bowel diary before and during the test phase over a 3-week period, in compliance with guidelines issued by the French health authorities.7,8 The purpose of this study was to assess patient response to an SNS screening test on a week-by-week basis to determine whether an early response during the test phase could predict permanent IPG implantation.
MATERIALS AND METHODS
From January 2006 to December 2012, an SNS screening test was performed in 164 patients with severe FI who were recruited in 3 French specialized centers (Nantes, Grenoble, and Paris) and enrolled in a prospectively maintained database (Fig. 1). Four items were recorded daily before and during the test phase: number of bowel movements, number of fecal leaks, time to defer defecation in minutes, and number of urgencies. To be considered for inclusion in this retrospective analysis, the number of fecal leaks had to be available every day for each patient before and during the test phase. The other items had to be available ≥4 days per week during the 3-week pretest period and the test phase. Items were analyzed as mean number/time per week. Changes during the SNS screening test were assessed week by week for each patient in comparison with baseline, defined as the mean value of the pretest diary. SNS was performed using a standard 2-stage procedure.3,9 For the screening test, a permanent quadripolar tined lead (Medtronic, Minneapolis, MN) was inserted into the S3 or S4 sacral foramen with x-ray guidance and under local or general anesthesia according to center preference. Lead placement was guided by the best sensory or motor response to lowest stimulation. A test was considered as positive leading to IPG implantation when the number of FI episodes measured over the 3-week period was reduced by ≥50% compared with baseline. The improvement of other parameters was also considered by the investigators who were in charge of the final decision to implant an IPG.
Continuous and qualitative variables were described as a mean (± SD) or median (quartile 1–3) and percentages. Intergroup comparisons were performed using Student t tests or Wilcoxon tests for quantitative variables and exact Fisher tests for qualitative variables. Subsequently, a multivariate logistic regression model was constructed with the 4 variables of the first screening test week to create a predictive score. The 4 variables were selected a priori and were tested with an univariable analysis as difference from baseline and as absolute value. The most predictive between relative and absolute for every variable (lowest p value) were selected for the final model. The probability of being a responder was computed to be 1/(1 + e-logit(score)), with the logit(score) constructed using the β coefficients from the regression model. The performance of the model was assessed by studying discrimination (receiver operating characteristic (ROC) analysis) and calibration (the goodness of fit of the final model was evaluated using the Hosmer–Lemeshow χ2 statistic). To validate the model and because of the small sample size, a bootstrap method was subsequently applied.10 The small data set was repeatedly resampled to create 200 replicated data sets, each the same size as the original. The predictive model was then fitted to each of 200 replicated data sets. Each fitted model was then applied both to the resampled data set from which it was generated and to the original data set; the area under the ROC curve (c-index) was calculated for both, and the difference between these 2 statistics was calculated. The 200 differences were then averaged to give an estimate of the optimism. The optimism corrected estimate of the c-index was then calculated as the naïve c-index minus the estimated optimism.
Ethical approval for data collection on the treatment of FI by SNS was obtained in accordance with national legislation on data protection and civil liberties (Commission Nationale de l’Informatique et des Libertés, Paris, France). Informed consent was obtained from all of the patients entering the study.
A total of 144 patients (130 women) with a mean age of 61.3 years (range, 29.3–85.0 y; SD ± 10.9) who were experiencing severe FI (baseline Cleveland Clinic incontinence score, 14.2 ± 2.9) provided workable bowel diaries. Causes of FI included neurogenic (n = 36), postobstetric (n = 29), surgical trauma (n = 39), idiopathic (n = 32), and other including IBD, radiation, sacral agenesis, anorectal malformation, and systemic sclerosis (n = 8). Patient characteristics and bowel diary data at baseline, first, second, and third week of the SNS screening test are presented in Table 1. Leak rate fell ≥50% within 1, 2, or 3 weeks for 81 (56%), 96 (67%) and 93 patients (70%). When a 50% leak rate reduction was obtained within the first week (n = 81 patients), the response was maintained during the 2 subsequent weeks in 65 (80%), rose to a higher rate (86.0% and 90.1% reduction of FI episodes at the second and third weeks), fluctuated (remained higher than 50% in only 1 of the 2 following weeks) in 12 (15%), and vanished (change from baseline during the 2 following weeks <50%) in 4 patients (5%). The 65 patients with a high response rate during the first week and sustained response during the 2 subsequent weeks were defined as fast responders.
At the end of the test phase, 115 patients were considered to be responders, leading to 114 IPG implantations (1 patient declined a permanent implantation). The decision to implant an IPG was essentially based on fecal leak rate reduction. Among the 81 responders at the end of the first week, 74 were implanted (64 of the 65 fast responders who had persistent response and 10 with fluctuated response). Forty other patients had an insufficient response strictly considering the criterion of 50% improvement during the first 2 weeks but achieved a good and sufficient response during the third week of screening test and were implanted with an average improvement of ≈45% considering the 3 weeks of test phase. In nonresponders, there was no change in leakage episodes compared with baseline (Fig. 1).
After univariate analysis (Table 2), a score including the number of leak episodes, bowel movements, and urgencies and the time to defer defecation expressed in minutes during the first screening test week was computed (Table 3). Setting the score cutoff at 84.8%, the predicted final outcome (implantation or not) had 72.6% sensitivity (95% CI, 59.8–83.1) and 100% specificity (95% CI, 78.0–100) and area under the ROC curve of 0.86 (95% CI, 0.78–0.94; De Long method; Table 4). This area represents the probability that a randomly selected responder will have a higher score than a randomly selected nonresponder. A 1% increase in the calculated score results in an increase of 5% of being implanted (OR = 1.05 (95% CI, 1.03–1.09)). The Hosmer–Lemeshow test eliminated a problem of score calibration with a p value at 0.244, and the optimism corrected estimate of the c-index was 0.81.
Considering the number of fecal leaks alone, score sensitivity and specificity were 54.8% (95% CI, 45.2–64.1) and 78.6% (95%% CI, 59.8–91.7), and area under the ROC curve was 0.71 (95% CI, 0.60–0.82; De Long method). The score was highly predictive of the final outcome for the fast-responder group in comparison with the nonresponder group (mean score ± SD = 0.91 (0.08) vs 0.59 (0.29); p < 0.001).
The present study with no equivalent in the present literature focused on temporary SNS test-phase results and assessed week-by-week data of bowel diaries collected during this time period and compared with baseline. The objective was to determine whether an early response could predict the response at the end of the 3-week test phase that is recommended in our guidelines, as in others.7 We demonstrated that a fast response to a temporary external SNS screening test, with a significant reduction in the number of FI episodes during the first week of stimulation, predicts the final result of the test and therefore the IPG implantation. The level of reduction in leakage episodes during the first week of the test phase appears highly predictive of the decision regarding implantation (Fig. 1). The majority of patients who responded during the first week maintained their response at a high level throughout the subsequent 2 weeks of the test phase. Thus, in fast responders a 1-week screening test could be long enough to proceed to IPG implantation without the need to wait 2 more weeks.
The concept of fast response could be an interesting approach to select those patients who would benefit from a shorter test phase to limit hospital stays and general costs of SNS. However, for surgical teams to adapt the duration of the test phase to the level of response would require greater operating room and hospital flexibility in scheduling, which might not be easily achieved in real life. In fact, as a positive response vanished for 4 patients (4.9%) and was not maintained for 2 others in the subsequent weeks, the concept of fast responders alone appears not reliable enough on which to base a decision regarding IPG implantation by the end of a first week of the test phase. However, the score computed on data recorded during the first week of the test phase has 100% specificity and would confidently allow for shortening of the test phase and hastening of IPG implantation. We are the first to propose a predictive score for IPG implantation based on bowel diary data recorded during the first screening test week. Bowel diaries provide crucial information for improving screening sensitivity and specificity when correctly fulfilled. Considering the cost of the therapy, a test allowing more confidence in decision-making for SNS implantation could be helpful for clinicians. We also outlined that other diary items and not only the number of FI leaks are useful in deciding IPG implantation as commonly observed in practice. A potential limitation of our study could be the potential applicability of this computed score to clinical daily practice.
For urinary incontinence, a 1-week testing phase has proved to be adequate.11 For FI, a shorter test phase would also essentially improve patient comfort and potentially limit the risk of lead-related infection. Reported infection rates ranged from 1.5% to 10.8% in the recent literature.4,5,9,12,13 Although there is no documented proof that the risk of infection increases with test duration, it is known that a majority of infections occur after implantation. Wexner et al12 reported a 10.8% infection rate, with all cases occurring after implantation, and Mellgren et al13 had 10% including at least 1 case during the test period. Dudding et al14 studied bacterial colonization of test electrodes at withdrawal in 13 patients and found that 54% of the electrodes were colonized despite the total absence of clinical signs of infection. These authors were unable to find any correlation between the duration of the test phase and the rate of electrode bacterial colonization.
It is likely that parameters related to electrode insertion have an influence on response.15 It has been reported that response is sustained better over time in patients who respond to low-level stimulation.9,15 The same stimulation protocol was applied for all of the patients along the study; the definitive tined-lead position was guided by the best motor and/or sensory response obtained at the lowest stimulation intensity. In all of the patients, a Medtronic 3093 quadripolar lead was implanted, because a report has showed that it was the only factor significantly associated with success at test phase.5
This study has clearly several limitations. It is retrospective; however, data collection was done prospectively, and it includes a large number of patients. In some cases, the external stimulation parameters had to be modified during the 3-week test period to readapt a nonadequate stimulation or in case of external pulse generator dysfunction. We could not eliminate a placebo effect/new-technique feeling effect or any change in eating and bowel habits during the first few days of external stimulation, which are known to disappear after some time during the test phase, but the positive response obtained at the end of the first week of the SNS test phase vanished in only 4 patients. We have made no cost analysis or projection based on our proposal to reduce the test phase. It might also be useful to determine on the basis of this retrospective study whether rapid response and magnitude of the response during the test period are predictive of the long-term response, as suggested in a recent study.16
This study demonstrates that a positive response during the first week of an SNS test phase with ≥50% reduction in fecal leaks is followed by sustained response during the subsequent 2 weeks of the test phase in a majority of patients The ≥50% reduction in fecal leaks during the first week of test phase is partly predictive of test success. A computed score based on all first-week, well-completed daily bowel diary data could help to select fast responders to SNS testing who could confidently be candidates for IPG implantation. A longer 3 week test phase also allows for implantation in slower responders. Additional studies are ongoing to ascertain whether a rapid response is also predictive of the long-term response.
The authors thank Laurence Salomon, M.D., and the URC–Fondation Ophtalmologique Adolphe de Rothschild (Paris, France) for their support in statistical analysis.
1. Thin NN, Horrocks EJ, Hotouras A, et alSystematic review of the clinical effectiveness of neuromodulation in the treatment of faecal incontinence. Br J Surg. 2013;100:1430–1447.
2. Maeda Y, Lundby L, Buntzen S, Laurberg SSuboptimal outcome following sacral nerve stimulation
for faecal incontinence. Br J Surg. 2011;98:140–147.
3. Maeda Y, O’Connell PR, Lehur PA, Matzel KE, Laurberg SEuropean SNS Bowel Study Group. Sacral nerve stimulation
for faecal incontinence and constipation: a European consensus statement. Colorectal Dis. 2015;17:O74–O87.
4. Gallas S, Michot F, Faucheron JL, et alClub NEMO. Predictive factors
for successful sacral nerve stimulation
in the treatment of faecal incontinence: results of trial stimulation in 200 patients. Colorectal Dis. 2011;13:689–696.
5. Altomare DF, Rinaldi M, Lobascio P, Marino F, Giuliani RT, Cuccia FFactors affecting the outcome of temporary sacral nerve stimulation
for faecal incontinence: the value of the new tined lead electrode. Colorectal Dis. 2011;13:198–202.
6. Dudding TC, Parés D, Vaizey CJ, Kamm MAPredictive factors for successful sacral nerve stimulation
in the treatment of faecal incontinence: a 10-year cohort analysis. Colorectal Dis. 2008;10:249–256.
8. Leroi AM, Damon H, Faucheron JL, et alClub NEMO. Sacral nerve stimulation
in faecal incontinence: position statement based on a collective experience. Colorectal Dis. 2009;11:572–583.
9. Dudding TC, Hollingshead JR, Nicholls RJ, Vaizey CJSacral nerve stimulation for faecal incontinence: patient selection, service provision and operative technique. Colorectal Dis. 2011;13:e187–e195.
10. Harrell FE Jr, Lee KL, Mark DBMultivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361–387.
11. Faucheron JL, Chodez M, Boillot BNeuromodulation for fecal and urinary incontinence: functional results in 57 consecutive patients from a single institution. Dis Colon Rectum. 2012;55:1278–1283.
12. Wexner SD, Hull T, Edden Y, et alInfection rates in a large investigational trial of sacral nerve stimulation
for fecal incontinence
. J Gastrointest Surg. 2010;14:1081–1089.
13. Mellgren A, Wexner SD, Coller JA, et alSNS Study Group. Long-term efficacy and safety of sacral nerve stimulation
for fecal incontinence
. Dis Colon Rectum. 2011;54:1065–1075.
14. Dudding T, Vaizey CBacterial colonization of stimulation electrode wires in patients undergoing temporary sacral nerve stimulation
. Colorectal Dis. 2010;12:141–143.
15. Dudding TC, Hollingshead JR, Nicholls RJ, Vaizey CJSacral nerve stimulation for faecal incontinence: optimizing outcome and managing complications. Colorectal Dis. 2011;13:e196–e202.
16. Altomare DF, Giuratrabocchetta S, Knowles CH, Muñoz Duyos A, Robert-Yap J, Matzel KEEuropean SNS Outcome Study Group. Long-term outcomes of sacral nerve stimulation
for faecal incontinence. Br J Surg. 2015;102:407–415.