Complications following vaginal mesh procedures for stress urinary incontinence: an 8 year study of 92,246 women

Complications of surgical mesh procedures have led to legal cases against manufacturers worldwide and to national inquiries about their safety. The aim of this study was to investigate the rate of adverse events of these procedures for stress urinary incontinence in England over 8 years. This was a retrospective cohort study of first-time tension-free vaginal tape (TVT), trans-obturator tape (TOT) or suprapubic sling (SS) surgical mesh procedures between April 2007 and March 2015. Cases were identified from the Hospital Episode Statistics database. Outcomes included number and type of procedures, including those potentially confounded by concomitant procedures, and frequency, nature and timing of complications. 92,246 first-time surgical mesh procedures (56,648 TVT, 34,704 TOT, 834 SS and 60 combinations) were identified, including 68,002 unconfounded procedures. Peri-procedural and 30-day complication rates in the unconfounded cohort were 2.4 [2.3–2.5]% and 1.7 [1.6–1.8]% respectively; 5.9 [5.7–6.1]% were readmitted at least once within 5 years for further mesh intervention or symptoms of complications, the highest risk being within the first 2 years. Complication rates were higher in the potentially confounded cohort. The complication rate within 5 years of the mesh procedure was 9.8 [9.6:10.0]% This evidence can inform future decision-making on this procedure.


Methods
Data source. Records from English National Health Service (NHS) database of Admitted Patient Care (including day-cases) 19 were extracted (on 26 th November 2015) from HES. All data extraction and analyses were carried out in accordance with relevant guidelines and regulations. Pseudonymised episode-level HES data were provided by NHS Digital to the Newcastle upon Tyne Hospitals NHS Foundation Trust under Data Sharing Framework Contract CON-313204-B3P1Y and Data Sharing Agreement HDIS-DSC-0109. In accordance with the terms of these agreements, only aggregated totals of patients and procedures are reported and no identifiable information was available for analysis. This research involved only previously collected, non-identifiable information and did not require review by a UK Research Ethics Committee. The study is registered on clinicaltrials. gov; NCT02850120.
Study design and population. This was a retrospective cohort study of all women discharged from hospital in England between 1 st April 2007 and 31 st March 2015 after surgical mesh procedures for treatment of SUI. Unique pseudo-anonymised patient identifiers were extracted from finished episodes of care which contained at least one of the (Office of Population Censuses and Surveys Classification of Interventions and Procedures, OPCS4.7) procedure codes M53.3 (Introduction of tension-free vaginal tape, TVT), M53.6 (Introduction of trans-obturator tape, TOT) or M52.1 (Suprapubic sling operation, a mid-urethral sling introduced via a suprapubic approach, SS), Supplementary Table S1. All finished inpatient episodes during the study period for this population were extracted from HES.

Data preparation & cleaning.
Episodes were removed which were exact duplicates (of patient identifier, admission date and method, discharge date, destination and method, hospital, gender, age, all procedure codes and all diagnostic codes (International Statistical Classification of Diseases 10 th revision, ICD-10), Supplementary  Table S1) and for patients with: an admission not coded as female; age missing or under 18 years; an invalid or missing admission method; a missing admission date; episodes after a reported date of death.
In this study we define insertion to mean the introduction of a mesh for the treatment of SUI, repair to mean a further procedure on a previously inserted mesh; renewal to mean a procedure to remove a previously inserted mesh and to replace it with a new mesh; and removal to mean the complete or partial removal of a previously inserted mesh. The OPCS-4 codes which correspond to each of these definitions are given in Supplementary  Table S2.
Index spells were considered "unconfounded" if they included no concomitant procedures, or if concomitant procedures were considered unlikely to affect outcomes (Supplementary Table S3), or if they were rescue procedures associated with the mesh insertion procedure (Supplementary Table S4). The remaining index spells were considered "confounded" by concomitant procedures which potentially influenced outcomes. Two cohorts of patients were defined: those with an unconfounded index spell and those with a confounded index spell. Subgroups of patients within each cohort were defined by the type of mesh procedure (one of TVT, TOT or SS).
Analysis was conducted using the programming language R 21 . Analysis software and its saved output 22 are available in Supplementary Files S5 & S6 respectively. HES data are available from NHS Digital via formal application 18 . Outcomes. Age, admission method, length of stay, frequency of endoscopic examination of the bladder and/or urethra (procedure codes: M45, M77), frequency of urinary or suprapubic catheter intervention (M30.2, M38.2, M47.1/4/8/9, M48.1) and rate of peri-procedural complications were reported for each mesh procedure 23 . Peri-procedural complications were classified, based on coding, as attributable to the procedure; attributable to the device; involving urinary symptoms (e.g. immediate acute urinary retention); or "other" (Supplementary Table S7).
All-cause readmissions, readmissions due to a complication 23 , or further mesh surgery (Supplementary Table S2) within 30 days of the index mesh insertion were recorded. 30-day readmissions for which the primary diagnostic code was not a complication code, or qualified by one, were considered "routine" readmissions.
Each patient was followed from their index procedure until 31 st March 2015 or their in-hospital death (if earlier). During follow-up particular events were recorded: in-hospital deaths; admissions due to complications of mesh implanted previously (using diagnostic codes intended for that purpose, from section T83 "complications of genitourinary prosthetic devices, implants or grafts", or main codes qualified by codes from section Y73 "gastroenterology or urology devices associated with adverse incidents" (Supplementary Table S8)); admissions for further mesh surgery (excluding those due to a complication recorded within 30 days of the index procedure -to avoid double counting).
Statistical analysis. Crude incidence rates for long-term complications were calculated as the number of mesh-related readmissions per 1000 person-years of follow-up. Instantaneous hazard rates were calculated using kernel-based smoothing 24,25 . Kaplan-Meier analysis was applied to the time from index procedure to the time of the first mesh-related readmission. Patients who suffered no mesh-related readmission and were alive at the end of the study were considered censored. Data availability. The Hospital Episode Statistics (HES) data used in this study were provided under license by NHS Digital to Newcastle Hospitals NHS Foundation Trust. The terms of the license do not permit the authors to make the data publicly available, but the data can be requested via formal application to NHS Digital using its online Data Access Request Service (http://content.digital.nhs.uk/dars).
The analytical software developed for this study is licensed by the authors under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (Supplementary File S5).

Details of ethics approval.
Only aggregated totals of patients and procedures are reported. No identifiable information was available for analysis. No ethical approval was required.

Results
Participants. During the study period, 101,081 patients had at least one surgical mesh insertion for SUI; 564,463 inpatient episodes of care were recorded for these patients. 2832 episodes were excluded from analysis because they were duplicates, or from patients with missing or ineligible demographic information (Fig. 1).
Index procedures. An index spell was identified for each of the 100,795 remaining patients (101,364 episodes). 6884 admissions were excluded due to no documented diagnosis of SUI and 1665 were excluded due to evidence of earlier or concomitant mesh surgery (e.g. surgical mesh repairs, removals or pelvic organ prolapse surgery). Of the remaining 92,246 index spells, 68,002 (73.7%) were unconfounded and 24,244 (26.3%) were confounded.
Yearly totals of index surgical mesh insertions for SUI are shown in Fig. 2. These were carried out in the majority of NHS trusts providing acute care in England, and almost all admissions were elective (Table 1) Table S11). In the confounded cohort, 2345 (9.7%) women were re-admitted within 30 days; including 738 (31.5%) whose readmission was due to a complication (Supplementary Table S12). The 30-day complication rate for the unconfounded cohort was 1.7 [95% CI 1.6-1.8] % and 3.0 [95% CI 2.8-3.3] % for the confounded cohort.
Longitudinal outcomes. The aggregate follow-up in the unconfounded cohort was 286,273 years (mean 4.2 years; 90.2% were followed >1 year; 66.5% >3 years; 40.9% >5 years, 13.6% >7 years). The frequencies of mesh-related readmissions, either for further mesh surgery or for symptoms indicating a complication of a previous mesh, are shown in Table 2. Table 3 shows the unadjusted number of readmissions and crude incidence rates for the unconfounded cohort (overall 16.0 per 1000 patient-years of follow-up); Fig. 3 shows the estimated hazard rate of readmission during 8 years after a first-time mesh implant for SUI.
Adjusting for different lengths of follow-up, 5.9 [95% CI 5.7 to 6.1] % of women having unconfounded mesh procedures were readmitted for a further mesh intervention or for symptoms of mesh complications within 5 years of their first-time mesh procedure, Fig. 4 (Supplementary Figs S13-S15).
In the unconfounded cohort, 1212 patients (1.8%) were readmitted for complications of earlier mesh surgery (based on their diagnostic codes) and for further mesh surgery (based on their procedure codes), in some cases during the same admission. Overall, 3541 (5.2%) were readmitted for either reason (Supplementary Table S16). The majority of patients readmitted for complications of an earlier mesh surgery (1212/1541, 78.7%) required further mesh surgery.
The aggregate follow-up in the confounded cohort was 102,870 years (mean 4.3 years). Frequencies of mesh-related readmissions, total number of readmissions and crude incidence rates are reported in Supplementary Tables S17 and S18. The estimated hazard rate of readmission during 8 years is shown in Supplementary Fig. S19. Adjusting for different lengths of follow-up, 6.4 [95% CI 6.1 to 6.8] % of women undergoing potentially confounded mesh procedures were readmitted for a further mesh intervention or for symptoms of mesh complication within 5 years of their first-time mesh procedure (Supplementary Fig. S20).
The proportion of patients experiencing a complication during the mesh insertion procedure, within 30 days or within 5 years, was 9.

Discussion
We estimated that 9.8% of patients undergoing surgical mesh insertion for SUI experience a complication peri-procedurally, within 30-days or within 5 years. The risk of readmission is highest within the first 2 years and some women required up to 6 mesh-related readmissions (mean follow-up of 4.2 years). We estimate that 5.9 [5.7-6.1] % of women (6.4 [6.1-6.8] % in the confounded cohort) required readmission for further mesh intervention within 5 years of their index procedure, more than double the cumulative incidence rate of mesh revision and removal (2.3% at 5 years) reported previously 26 . These findings provide evidence in this area of concern particularly considering that the prevalence of stress urinary incontinence is between 10-40% in community-dwelling women and higher in the elderly 27 .
The NHS England Mesh Working Group found insufficient evidence to determine the extent of longer term complications 14 . An observational cohort study used insurance claims from 188,454 women over a 9-year period, but included approximately 98,000 women with concomitant POP surgery 28 . The study with the longest follow-up was a prospective multi-centre cohort study of 90 women following TVT mesh insertion for a mean of 7.5 years 29 . Although the mean follow-up of our study was 4.2 years, approximately 9240 women (13.6%) were followed for over 7 years.
An independent review commissioned by the Scottish government also analysed administrative data, reported as an interim results 10 and as a full paper 30 . The proportions of unconfounded TVT, TOT and SS patients experiencing peri-procedural complications (2.9%, 1.4% and 4.9% respectively) reported here are lower than those estimated by the Scottish Inquiry (4%, 2% and 6%) and reported in their full paper 30 . This may be a consequence of the different methods used to identify complications. We used the method recommended by the NHS Classification Service 23 . The Scottish Inquiry considered the presence of prospectively defined diagnostic codes (grouped as haemorrhage, infection, pain and procedure-related complications) to indicate a complication, but these included other diagnoses (e.g. ICD-10: N94.1 "Dyspareunia" and R52.2 "Other chronic pain") which, without a qualifying supplementary code, may indicate comorbidities rather than procedural complications.
The Scottish Inquiry reported crude readmission rates 30 almost double that of our study. This may be explained by the different methodologies. For example, the Scottish Inquiry included pelvic organ prolapse (POP) procedures, additional non-mesh intervention and any diagnosis of general infection, chronic pain, and haemorrhage as relevant readmissions. However, we found it difficult to attribute these to an earlier mesh insertion: subsequent POP or non-mesh procedures may indicate efficacy rather than safety. Furthermore the presence of diagnostic   Table 1. Outcomes from index admissions to introduce tension-free vaginal (TVT), trans-obturator (TOT) and suprapubic sling (SS) surgical mesh products in the treatment of stress urinary incontinence. Key: *Percentage of procedures with a complication, † Index admissions without concomitant procedures which may influence outcomes, ‡ Index admissions with concomitant procedures which may influence outcomes.
codes for infection or pain were not necessarily due to a previous mesh insertion. In summary our methods were more specific, but perhaps less sensitive than that applied by the Scottish Inquiry, with our study providing a more conservative but robust estimate of long-term complications 31 . This is the largest study to date of SUI mesh procedures (92,244), with 100% coverage of NHS patients (including private patients treated in an NHS setting) in England over an 8-year period. The British Society of Urogynaecology (BSUG) established a database to record data relating to any anti-incontinence and/or prolapse procedure, including those conducted in a private setting. However results from this audit are, as yet, unpublished 32 .
Case ascertainment was optimised by using recommended coding practice to identify peri-procedural and 30-day complications, and choosing codes directly attributable to mesh complications to identify long-term events. Around two thirds of first-time mesh insertions for SUI had no concomitant procedures likely to affect outcomes; these cases were analysed separately from potentially confounded ones, increasing confidence in attributing complications to mesh intervention. A limitation is the dependence on clinical coding. For example, outcomes were not compared between TVT, TOT and SS, or between the confounded and unconfounded cohorts because severity of incontinence is not available from ICD-10 diagnostic codes for baseline matching. Analysis of procedural complications was limited by lack of detail (e.g. "T81: Complications of procedures, not elsewhere classified" was the most common complication reported peri-procedurally); the introduction of additional complication codes in ICD-11, expected in 2018 33 , will increase the utility of future administrative data to assess the safety of interventional procedures. The outcome measures chosen reflected the concerns expressed by patient representatives, and were available from the HES database of inpatient and day case episodes; however complications managed entirely in an outpatient setting were not captured in this study. Because emigration and out-of-hospital death are not recorded in HES more women were assumed to be at risk, in the long-term analysis, than was actually the case. Both these limitations suggest that the true hazard rates may be higher than reported. The accuracy of clinical coding is an important factor to consider in the potential limitations of this study. The accuracy of coding in HES data, as in all routinely collected datasets derived from administrative sources, is often questioned, however the Audit Commission in England have shown that HES coding accuracy has improved over time, when auditing Payment by Results (PbR) assurance programme 34 .   Table 3. Reasons and frequency of readmissions during follow-up. The total number (%, percentage of cohort) of patients readmitted is also given (some patients being readmitted on multiple occasions).

Conclusions
This is the largest study to date of surgical mesh insertions for SUI. It includes all NHS patients in England over an 8-year period. We estimate that 9.8% of patients undergoing surgical mesh insertion for SUI experienced a complication peri-procedurally, within 30-days or within 5 years of the initial mesh insertion procedure. This is likely a lower estimate of the true incidence. Given concerns about the safety of these procedures, this study provides robust data to inform both individual decision-making and national guidance.