Description of studies

Twenty one potentially eligible studies were screened. Seven studies were excluded from the review: full details are given in the Characteristics of Excluded Studies. Fourteen randomised controlled trials were identified on the surgical management of pelvic organ prolapse. These were conducted in seven countries (five from Italy, four from the United States of America, and one each from Australia, Netherlands, Great Britain, Taiwan and Finland) evaluating 1004 women, all of whom received a surgical intervention. All but one trial (Meschia 2003) reported median follow up of greater than one year but only two trials reported outcomes at greater than five years (Colombo 1997; Colombo 2000).

Given the diverse nature of pelvic organ prolapse, to allow a meaningful analysis of the data, the review was divided into subgroups related to the site of the prolapse: upper vaginal; anterior vaginal wall; posterior vaginal wall; and to continence issues following prolapse surgery in continent women.

Upper vaginal prolapse (Hypotheses 1, 2, 3 and 4)
Five trials compared the management of upper vaginal prolapse (Benson 1996; Lo 1998; Maher 2004; Meschia 2003; Roovers 2004).

• Abdominal sacral colpopexy versus vaginal sacrospinous colpopexy

Benson's trial reported data for 80 of 101 randomised women with uterovaginal or vault prolapse: the women with uterovaginal prolapse all underwent hysterectomy (Benson 1996). Lo's trial reported follow up of 118 of 138 continent women who had at least Stage 3 prolapse: some underwent anterior or posterior repairs or abdominal or vaginal hysterectomy in addition to repair of the prolapse actually being compared in the trial (Lo 1998). Maher's trial included 89 women with posthysterectomy vaginal vault prolapse (Maher 2004). In Benson and Maher's trials, the abdominal group underwent sacral colpopexy with colposuspension, paravaginal repair or a vaginally performed posterior vaginal wall repair as required. In the vaginal arm of Benson's trial, a bilateral vaginal sacrospinous colpopexy was performed as compared to a unilateral sacrospinous colpopexy in Maher's trial: in Lo's trial this was not specified but Nichols' method was referenced.

Women with stress urinary incontinence were treated with a needle suspension in the vaginal arm (n=20) of Benson's trial and a colposuspension in the abdominal arm (n=14), as compared to women with stress urinary incontinence or occult incontinence (n=14 and 15 in the abdominal and vaginal arms respectively) receiving an abdominal colposuspension in both arms of Maher's trial. In Maher's trial, 27 women had symptoms of overactive bladder at baseline (n=13 and 14 respectively). Simple costs were calculated by Benson and Maher incorporating length of stay and operating theatre cost. Formal cost effectiveness was not reported in either study. All three trials were considered to be similar enough to be combined in a meta‐analysis. However, there was significant variation in the outcome measures (Benson and Lo: incomplete site specific prolapse reporting; Maher and Lo: failure to report time to recurrent prolapse; Lo: optimal surgical cure of prolapse was considered to be Stage 2 prolapse or less). These factors contributed to heterogeneity and prevented complete meta‐analysis of all variables.

• Abdominal sacrohysteropexy versus vaginal hysterectomy and repair

Roovers' trial evaluated only women with uterine prolapse who underwent sacrohysteropexy (with uterine preservation) in the abdominal group (n=41) and vaginal hysterectomy and vaginal repair with the vault being fixed to the uterosacral cardinal ligament complex in the vaginal group (n=41) (Roovers 2004). Roovers' trial is analysed as a separate subcategory in the analyses as the vaginal arm did not include a sacrospinous colpopexy and the abdominal group included uterine preservation.

• Posterior intravaginal slingplasty versus vaginal sacrospinous colpopexy

Meschia compared two vaginal procedures: the infracoccygeal sacropexy (posterior intravaginal slingplasty, n=24) and the sacrospinous colpopexy (n=23) for uterine or vault prolapse (Meschia 2003). Again, it was analysed as a separate subcategory.

Anterior vaginal wall prolapse (Hypotheses 5 and 6)
Seven trials included various surgical procedures for treating anterior vaginal wall prolapse, with or without stress urinary incontinence (Bump 1996a; Colombo 1996a; Colombo 1997; Colombo 2000; Meschia 2004; Sand 2001; Weber 2001). Due to clinical heterogeneity in stage of prolapse, types of operations and whether women with urinary incontinence or occult incontinence were included, only four of these could be combined for meta‐analysis (Sand 2001) with (Weber 2001); and (Bump 1996a) with (Colombo 1997).

• Cystopexy without pubourethral ligament plication versus cystopexy with pubourethral ligament plication

Colombo enrolled only continent women with cystocele Stage 2 or more (Colombo 1996a). None of the women had preoperative detrusor overactivity. The trialists studied the prevention of de novo stress urinary incontinence after cystopexy with (n=50) or without (n=52) pubourethral ligament plication.

• Cystopexy with posterior pubourethral ligament plication versus cystopexy with needle suspension

Columbo enrolled women with cystocele Stage 2 or more and either occult (n=73) or symptomatic (n=36) urinary incontinence (Colombo 1997). None of the women had preoperative detrusor overactivity. The trialists compared cystopexy with posterior pubourethral ligament plication (n=55) versus cystopexy plus Pereyra bladder neck suspension (n=54).

• Anterior repair with urethrovesical plication versus anterior repair with needle colposuspension

In Bump's trial, women were all continent but had bladder neck hypermobility in addition to Stage 3 or 4 prolapse. (pelvic organ prolapse quantification recommended by the International Continence Society (ICS)) (Bump 1996a). Six women had detrusor overactivity successfully treated before operation. All women had an anterior vaginal wall repair for anterior vaginal wall prolapse ICS Stage 3 or 4. The trialists compared the effects of needle colposuspension (n=14) with plication of the urethrovesical junction endopelvic fascia (n=15) on postoperative development of stress incontinence. They analysed 29 women; 10/15 in the fascia plication group and 10/14 in the needle colposuspension group had potential stress incontinence (defined as a mean pressure transmission ratio of less than 90% for the proximal three quarters of the urethra or a positive stress test during barrier testing). This trial was considered to be sufficiently similar to the previous one (Colombo 1997) to allow the data to be combined in meta‐analysis.

• Anterior vaginal wall repair versus Burch colposuspension

In a third trial from Italy, women were studied who had primary Stage 2 or 3 cystocele and concomitant urodynamic urinary stress incontinence (Colombo 2000). None of the women had preoperative detrusor overactivity. The 68 women were randomised to receive either Burch colposuspension (n=35) or anterior vaginal wall repair (n=33).

Prolapse repair and urethrovesical endopelvic fascia plication versus prolapse repair and TVT
In a recently published fourth Italian trial, women with severe genital prolapse and occult stress urinary incontinence were enrolled (Meschia 2004). None of the women had preoperative detrusor overactivity. The women were randomised to receive either urethrovesical endopelvic fascia repair (n=25) or tension‐free vaginal tape (TVT) (n=25) in addition to vaginal hysterectomy and prolapse repair. Most also had a posterior repair (23/25 and 20/25 respectively).

Two other trials enrolled women with cystoceles with or without incontinence (Sand 2001; Weber 2001): one of the trials excluded women who needed a concomitant continence procedure such as colposuspension, sling or needle suspension (Weber 2001). Both trials compared traditional anterior vaginal wall repair with anterior vaginal wall repair supplemented by the use of absorbable mesh overlay (polyglactin) for cystocele. The trials were considered similar enough to combine in meta‐analysis. To enable meaningful comparison between the trials the standard and ultralateral anterior vaginal wall repair groups in Weber's trial (Weber 2001) were combined mimicking Sand's groups (Sand 2001) when comparing anterior vaginal wall repair with and without polyglactin mesh overlay.

• Anterior vaginal wall repair versus anterior vaginal wall repair with mesh overlay

Sand randomly allocated women with cystocele to or beyond the introitus to anterior vaginal wall repair alone (n=70) or anterior vaginal wall repair and polyglactin mesh overlay (n=73) (Sand 2001). The surgery was for primary cystocele in 85% of cases. Concomitant surgery was performed as required including vaginal hysterectomy, vaginal sacrospinous colpopexy, posterior vaginal wall repair (n=67/70 and 65/73) and continence surgery. The women who underwent posterior vaginal wall repair and were assigned to the polyglactin mesh overlay for the cystocele also had their posterior vaginal wall repair augmented with polyglactin mesh.

• Anterior vaginal wall repair versus anterior vaginal wall repair with mesh overlay versus ultralateral anterior vaginal wall repair

Weber evaluated the efficacy of standard anterior vaginal wall repair (n=33), ultralateral anterior vaginal wall repair (n=24) and standard anterior vaginal wall repair plus polyglactin mesh overlay (n=26) in women who underwent surgery for anterior vaginal wall prolapse (Weber 2001). Other concomitant prolapse surgery was performed as required.

Posterior vaginal wall prolapse (Hypotheses 7, 8 and 9)
Two trials (Kahn 1999; Nieminen 2004) compared the vaginal and transanal approach for the management of rectoceles. In addition, another trial provided data for women with rectoceles undergoing posterior repair with and without mesh (Sand 2001).

The trials involving transanal repair have been published as abstracts at ICS Annual Meetings, and the authors have provided additional data. Each trial had slightly different inclusion criteria. Kahn included women who had symptoms of prolapse or symptoms of impaired rectal evacuation with incomplete emptying on isotope defecography and normal compliance on anorectal manometry (Kahn 1999). Nieminen included women with symptomatic rectoceles not responding to conservative treatment (Nieminen 2004). Importantly, women with compromised anal sphincter function and other symptomatic genital prolapse were excluded. In both trials the vaginal repair was performed by gynaecologists and the transanal repair by colorectal surgeons. In Kahn's trial the posterior vaginal wall repair was performed using levator plication and in Nieminen's trial the rectovaginal fascia was plicated. The trials were considered to be similar enough to be combined in a meta‐analysis.

In the third trial, the women were included if they had a central cystocele with or without urinary incontinence, for which they required an anterior repair. The majority of the women were also having a posterior repair for rectocele (132/143, 92%) (Sand 2001). The women allocated to the mesh augmentation arm for their anterior repair also had their posterior repair augmented with mesh, and recurrence rates of rectocele were reported separately. However, no clinical outcomes relating to urinary, bowel or sexual function were reported.

Any type of prolapse (Hypotheses 10, 11, 12, 13, 14 and 15)
There were no trials which compared surgery with either mechanical devices or conservative treatment, nor open surgery with laparoscopy, nor did any trials address the choice of suture types.

Occult urinary incontinence (Hypothesis 13)
Occult urinary incontinence is diagnosed when women have no symptoms of stress urinary incontinence but do have demonstrable stress urinary incontinence when the prolapse is reduced. One trial included women with occult stress urinary incontinence and provided data separately for their urinary outcomes (Meschia 2004); four trials included only continent women or reported outcomes separately for a continent subsample (Colombo 1996a; Colombo 1997; Lo 1998; Maher 2004); and one other trial included both continent women and those with 'potential' incontinence (the term 'potential' was interpreted as 'occult') (Bump 1996a).

• In one Italian trial, all the women were continent but a continence procedure was only performed in one arm (pubourethral ligament plication in addition to a standard colpopexy) (Colombo 1996a).

• In a recent Italian trial, all the women were continent but demonstrated to have occult stress urinary incontinence on preoperative prolapse reduction (Meschia 2004).

• Another included a mixed sample of women, with and without incontinence (Colombo 1997). However, data were presented separately, allowing assessment of prolapse surgery on urinary outcomes in the 73 continent women.

• In Bump's trial, 29 women (10/15 in the fascia plication group and 10/14 in the needle colposuspension group) had urodynamically defined potential stress incontinence (defined as a mean pressure transmission ratio of less than 90% for the proximal three quarters of the urethra or a positive stress test during barrier testing) (Bump 1996a). However, all the women were symptomatically continent and both arms included a continence procedure. Data from this trial were aggregated with those from Colombo 1997.

• Although Lo did not report the total number of women who developed new urinary incontinence after surgery, he did report how many women required surgery for incontinence (Lo 1998).

• In one other trial, Maher performed additional Burch colposuspensions for all women with urodynamically proven or occult stress urinary incontinence in women randomly allocated to abdominal sacral colpopexy (14) or vaginal sacrospinous colpopexy (15) for vaginal vault prolapse (Maher 2004). Women undergoing concomitant colposuspension were stratified to ensure equal representation in the groups. Occult stress urinary incontinence was detected in 5/14 (11% of 46 in whole arm) of the abdominal group and 6/15 (13% of 43) of the vaginal group, but urinary outcomes were not available separately according to this baseline diagnosis. However, data were provided about the occurrence of new urinary incontinence in women previously continent (n=22 and 24 respectively) and new overactive bladder symptoms in women previously unaffected by urgency, detrusor overactivity or overactive bladder syndrome (n=33 and 29).

Use of mesh for prolapse surgery (Hypothesis 14)
Two trials included mesh for prolapse repair, as described above (Sand 2001; Weber 2001). The data from two similar arms in one trial (standard anterior vaginal wall repair and ultralateral anterior vaginal wall repair) were combined for the purpose of comparing with the arm including mesh (Weber 2001). In the other trial, data were available separately for women who also underwent a posterior vaginal wall repair (Sand 2001).

Full details of the included trials are given in the Characteristics of Included Studies Table.

Risk of bias in included studies

Sufficient detail was provided in seven trials to confirm that secure concealment of the randomisation process was used, e.g. allocation by remote person or computer (Benson 1996; Bump 1996a; Maher 2004; Meschia 2003; Meschia 2004; Roovers 2004; Weber 2001). Of the remainder, six stated that they used computer generated number lists but it was unclear whether the allocation was concealed before assignment. The last trial stated that a computer‐generated but open number list was used, and it was therefore classified as a quasi‐randomised trial (Colombo 2000).

Women and surgeons could not be blinded as to the procedure when different surgical routes were being compared (Benson 1996; Colombo 2000; Lo 1998; Maher 2004; Meschia 2003; Roovers 2004). Blinding of patients was not performed in any other trials. Outcome assessments were conducted by non‐surgeons in three trials (Benson 1996; Roovers 2004; Weber 2001). Only the Roovers trial data were reported according to the CONSORT guidelines. In four trials, data were analysed on an intention to treat basis (Lo 1998; Maher 2004; Roovers 2004; Weber 2001).

Loss to follow up was a variable problem ranging from zero (Colombo 1997; Kahn 1999; Meschia 2003; Meschia 2004) to 24% (26/109) (Weber 2001). Weber also reported a statistically significant higher loss to follow up in one arm of the trial (ultralateral anterior vaginal wall repair).

Baseline demographics were reported in all trials and were equally distributed except in one trial (Sand 2001) which reported that previous hysterectomy was more common in the mesh overlay group and a second trial (Kahn 1999) which reported a difference in menopausal status and previous hysterectomies between the groups. Preoperative prolapse status was reported in all trials but equal distribution and severity of prolapse between groups was not specifically reported in four trials (Benson 1996; Bump 1996a; Meschia 2004; Sand 2001). One trial included 7% of women with stage 1 anterior vaginal wall prolapse preoperatively (at time of inclusion) which would also have been classified as a postoperative success (Weber 2001). Length of follow up was less than one year in one trial (Meschia 2003) and greater than five years in another trial (Colombo 1996a), with all other trialists reporting results between one to five years.

Effects of interventions

A. Upper vaginal prolapse (uterine and vaginal vault) (Comparison 01)
Five trials provided data regarding the outcome of prolapse surgery for upper vaginal prolapse (Benson 1996; Lo 1998; Maher 2004; Meschia 2003; Roovers 2004).

Hypothesis 1: abdominal sacral colpopexy is better than vaginal sacrospinous colpopexy:
Three trials were considered to be similar enough to allow combination of data for comparison of abdominal sacral colpopexy and vaginal sacrospinous colpopexy (Benson 1996; Lo 1998; Maher 2004). Abdominal sacral colpopexy was better than vaginal colpopexy in terms of:

• a lower rate of recurrent vault prolapse (3/84 versus 13/85; RR 0.23, 95% CI 0.07 to 0.77, Comparison 01.03.01) (Benson 1996; Maher 2004);

• the number of women failing to improve to Stage 2 or better (3/52 versus 13/66; RR 0.29, 95% CI 0.09 to 0.97, Comparison 01.02.02) (Lo 1998);

• less postoperative dyspareunia (7/45 versus 22/61; RR 0.39, 95% CI 0.18 to 0.86, Comparison 01.13.01) (Benson 1996; Lo 1998; Maher 2004);

• less postoperative stress urinary incontinence (14/47 versus 28/81, RR 0.55, 95%CI 0.32 to 0.95, Comparison 01.05.01) (Benson 1996; Maher 2004). However, caution should be exercised when evaluating these data due to significant variation in the methodology of the two papers as described above. There was no statistically significant difference in reoperation rates for stress urinary incontinence (RR 0.6, 95% CI 0.21 to 1.73, Comparison 01.22.01) (Benson 1996; Lo 1998; Maher 2004);

• The lower reoperation rate for prolapse after abdominal surgery did not reach statistical significance (6/84 versus 14/85, RR 1.46, 95% CI 0.19 to 1.11, Comparison 01.21.01) (Benson 1996; Maher 2004).

The results for intraoperative blood loss were inconsistent in two studies with a mean difference of 298 ml less blood loss in the abdominal group in Lo's study (Lo 1998) and 33 ml more blood loss in Maher's trial (Maher 2004) (Comparison 01.15.01). Benson did not report blood loss but the postoperative change in haemoglobin was not statistically different (Benson 1996).

Women treated abdominally took significantly longer to present with recurrent prolapse (WMD for months to recurrence ‐10.90, 95% CI ‐17.12 to ‐4.68, Comparison 01.24.01) in one trial (Benson 1996). On the other hand, the sacral (abdominal) colpopexy was associated with a longer operating time (WMD 21 minutes, 95% CI 12 to 30, Comparison 01.17.01) (Benson 1996; Lo 1998; Maher 2004), longer time to recover (WMD 8.3 days, 95% CI 3.9 to 12.7, Comparison 01.19.01) (Maher 2004) and was more expensive (WMD US$1334, 95% CI 1027 to 1641, Comparison 01.20.01) (Benson 1996; Maher 2004) than the vaginal approach.

There was no statistically significant difference in subjective outcome between the two approaches (subjective failure after abdominal surgery: 9/84 versus 18/85, RR 0.53, 95% CI 0.25 to 1.09, Comparison 01.01.01) (Benson 1996, Maher 2004). On the limited evidence available, patient's satisfaction (RR 0.82, 95% CI 0.32 to 2.06, Comparison 01.04.01) (Maher 2004) and objective failure at any site (any pelvic organ prolapse: RR 0.77, 95% CI 0.39 to 1.53, Comparison 01.02.01) (Maher 2004) were not different in both groups. Although data were available for bowel outcomes (Comparisons 01.10 and 01.11) and adverse events (Comparison 01.25), they were too few to provide sufficiently precise estimates to identify or rule out clinically important differences.

Hypothesis 2: vaginal hysterectomy is better than uterine preservation:
In the fourth trial, Roovers compared abdominal sacral hysteropexy against vaginal hysterectomy and repair with vault fixation to the uterosacral‐cardinal ligament complex (Roovers 2004). Although more women had subjective prolapse symptoms at one year after abdominal surgery (RR 3.2, 95% CI 1.29 to 7.92, Comparison 01.01.02), this did not reach statistical significance on the prolapse domain of the urinary distress inventory (UDI) (mean difference 4.1, 95% CI ‐5.4 to 13.6); nor was the score for urinary incontinence significantly different (mean difference 6, 95% CI ‐2 to 14). However, at one year after surgery the vaginal group scored significantly better (lower) scores on the discomfort /pain domain (7.1, 95% CI 1.1 to 13.2), overactive bladder domain (8.7, 95% CI 0.5 to 16.9) and the obstructive micturition domain (10.3, 95% CI 0.6 to 20.1) as compared to the abdominal group. More women in the abdominal group required repeat prolapse repair (RR 9.00, 95% CI 1.19 to 67.85, Comparison 01.21.02): in the abdominal group, five women (13%) had a reoperation for recurrent cystocele and four women (10.5%) for recurrent uterine prolapse, whereas in the vaginal group only one patient required surgery in the first year for vaginal vault prolapse. The operating time was less for the abdominal group (WMD ‐10 minutes, 95% CI ‐12 to ‐8, Comparison 01.17.02), possibly reflecting the less invasive nature of the abdominal procedure in this trial (the uterus was preserved in the abdominal group as opposed to removed in the vaginal group).

Hypothesis 3: vaginal hysterectomy with McCall culdoplasty is better than vaginal hysterectomy and sacrospinous colpopexy:
No trials identified.

Hypothesis 4: vaginal McCall culdoplasty and uterosacral ligament plication is better than vaginal sacrospinous colpopexy and repair:
No trials identified.

Other comparisons for upper vaginal prolapse:
In the fifth trial (Meschia 2003) there were too few data to demonstrate a difference between posterior intravaginal slingplasty and vaginal sacrospinous colpopexy in the objective prolapse rate at median follow up of 11 months. Cuff (vault) prolapse recurred once in each group. The RR for recurrence or persistence of any type of prolapse was 0.80, 95% CI 0.26 to 2.45, Comparison 01.02.03 (Meschia 2003). However, the operating time was shorter (WMD ‐14 minutes, 95% CI ‐25 to ‐3, Comparison 01.17.03) and blood loss less (‐52 ml, 95% CI ‐77 to ‐27, Comparison 01.15.03) with the former approach. One woman in the vaginal sacrospinous colpopexy group had dyspareunia at six months, and one in the intravaginal slingplasty group had faecal incontinence.

B. Anterior vaginal wall prolapse (cystocele, urethrocele, paravaginal defect) (Comparison 02)
Seven trials included a variety of surgical procedures to treat anterior vaginal wall prolapse, with or without stress or occult stress urinary incontinence. (Bump 1996a; Colombo 1996a; Colombo 1997; Colombo 2000; Meschia 2004; Sand 2001; Weber 2001). Combination of data was possible for two sets of trials: two were comparable in terms of type of population (women with prolapse only) and types of operation (anterior repair with and without mesh) (Sand 2001; Weber 2001); and the other two in terms of types of operation (endopelvic fascia plication versus needle suspension) (Bump 1996a; Colombo 1997). The trials were all small and the data reported for outcomes were therefore mostly inconclusive.

Hypothesis 5: anterior vaginal wall repair is better than the abdominal paravaginal repair in the management of cystocele:
No trials identified.

Hypothesis 6: for midline cystocele defects, a traditional anterior vaginal wall repair is better than anterior vaginal wall repair with mesh reinforcement:
Data from two small trials suggested that traditional anterior repair may be followed by higher objective failure rates than after polyglactin mesh reinforcement of anterior repair (RR 1.48, 95% CI 1.07 to 2.04, Comparison 02.03.03) (Sand 2001; Weber 2001), but data on reoperation rates were not given and complication rates were similar. Weber did not find significant differences in cure rates for cystocele between the standard cystocele repair (30%), ultralateral repair (46%) and standard plus polyglactin mesh overlay (42%) at mean follow up of 24 months, but the trial was only powered to detect a 30% difference between the groups (Weber 2001).

Other comparisons for anterior vaginal wall prolapse:
Five other trials were identified which compared different operations for anterior vaginal wall prolapse or different continence procedures for women with urinary incontinence or occult urinary incontinence as well as anterior vaginal wall prolapse (Bump 1996a; Colombo 1996a; Colombo 1997; Colombo 2000; Meschia 2004).

One single trial comparing anterior repair with Burch colposuspension showed statistically significant lower rates of cystocele recurrence (RR 0.09, 95% CI 0.01 to 0.64, Comparison 02.03.05) (Colombo 2000), but higher rates of persisting urinary incontinence (RR 3.39, 95% CI 1.40 to 8.22, Comparison 02.06.03) (Colombo 2000). However, this was not reflected in differences in reoperation rates for either prolapse or incontinence (Comparisons 02.18.03 and 02.19.03) (Colombo 2000). Another small trial reported that more women were incontinent after endopelvic fascia plication than after TVT supplementing prolapse surgery (RR 9, 95% CI 1.23 to 65.85, Comparison 02.07.08) (Meschia 2004) but the data were too few to comment on the effect on prolapse or other clinical outcomes. However, there was a shorter operating time for the former operation (WMD ‐19 minutes, 95% CI ‐29 to ‐9, Comparison 02.10.08) (Meschia 2004).

C. Posterior vaginal wall prolapse (rectocele) (Comparison 03)
Two small trials compared vaginal and transanal approaches to the management of rectoceles (Kahn 1999; Nieminen 2004), and a third examined posterior repair with and without mesh reinforcement (Sand 2001).

Hypothesis 7: posterior vaginal vaginal wall repair is better than a transanal repair:
Many of the important outcome parameters were recorded in only one trial thus limiting the data available and the ability to perform meta‐analyses. The results for posterior vaginal wall repair were better than for transanal repair in terms of subjective (RR 0.36, 95% CI 0.13 to 1, Comparison 03.01.01) and objective (RR 0.24, 95% CI 0.09 to 0.64, Comparison 03.02.03) failure rates (persistence of rectocele and/or enterocele) (Kahn 1999; Nieminen 2004). Analysing rectocele alone showed that recurrent rectocele occurred in 2/39 in the vaginal group and 7/48 following the transanal repair, a difference that did not reach statistical significance (RR 0.32, 95% CI 0.07 to 1.34, Comparison 03.02.01). Postoperative enterocele was, however, significantly less common following the vaginal surgery as compared to the transanal group (RR 0.23, 95% CI 0.07 to 0.83, Comparison 03.02.02).

Postoperative hospital stay was longer after vaginal surgery than after transanal surgery in one trial (mean difference (MD) 1 day, 95% CI 0.5 to 1.5, Comparison 03.15.01) despite a shorter operating time (MD ‐7 minutes, 95% CI ‐12 to ‐2) (Kahn 1999). The operating times in the other trial (Nieminen 2004) were the same for both groups (35 minutes). When data for operating time were combined (WMD ‐3.6 minutes, 95% CI ‐7.4 to 0.2, Comparison 03.11.01), there was significant heterogeneity (P = 0.07, I‐squared = 69%). The vaginal approach was associated with a significantly higher blood loss (79 ml, 95% CI 40 to 119, Comparison 03.09.01) (Kahn 1999; Nieminen 2004) and postoperative narcotic use (Comparison 03.12.01, (Kahn 1999) only) as compared to the transanal approach.

Nieminen reported that the mean depth of rectocele on postoperative defecography was 4.13 cm in the transanal group and this was significantly larger than the 2.73 cm in the vaginal group (WMD ‐1.43, 95% CI ‐2.86 to 0, P=0.05, data not shown). Postoperative difficulties in bowel evacuation were seen in 9/31 in the vaginal group as compared to 14/34 in the transanal group, a difference that was not significantly different (RR 0.73, 95% CI 0.37 to 1.42, Comparison 03.06.01). No significant differences were seen in the rate of incontinence to flatus or faeces postoperatively between the groups, nor in rates of postoperative dyspareunia but the trials were too small for these data to be reliable. There were differences between the trials for the outcome postoperative complications: in one trial, four women had a haematoma and one needed a blood transfusion in the vaginal arm (Kahn 1999) whereas in the other, one woman had a wound infection after transanal operation (Nieminen 2004) (Comparison 03.13.01).

Hypothesis 8: posterior vaginal wall repair is better than an abdominal posterior repair:
No trials identified.

Hypothesis 9: posterior vaginal wall repair is better than posterior vaginal wall repair with mesh reinforcement:
One trial compared posterior repair with and without mesh reinforcement (Sand 2001). Rectocele recurrence appeared equally common with and without polyglactin mesh augmentation (7/67 versus 6/65), but the confidence intervals were wide (RR 1.13, 95% CI 0.40 to 3.19, Comparison 03.02.04) (Sand 2001). No trial reported mesh erosion.

D. Any type of prolapse (Comparisons 04, 05, 06, 07, 08)

Hypothesis 10: surgical treatment is better than conservative treatment in the management of pelvic organ prolapse (Comparison 04):
No trials addressed this comparison.

Hypothesis 11: surgical treatment is better than mechanical devices in the management of pelvic organ prolapse (Comparison 05):
No trials addressed this comparison.

Hypothesis 12: open abdominal surgery is better than the laparoscopic approach for the management of prolapse (Comparison 06):
No trials addressed this comparison.

Hypothesis 13: potential stress urinary incontinence detected on reduction of prolapse prior to surgery is best treated with formal continence surgery at the time of prolapse surgery, rather than being left untreated (Comparison 07):
The effects of surgical management of pelvic organ prolapse on urinary symptoms were addressed in six trials which included data for women without urinary symptoms at baseline (Bump 1996a; Colombo 1996a; Colombo 1997; Lo 1998; Maher 2004; Meschia 2004).

The trials involved several different operations and different populations. Each single trial was too small to demonstrate differences in new urinary symptom outcomes between the two arms, in terms of new stress urinary incontinence, persistent or new urgency, detrusor activity or overactive bladder, or in postoperative voiding dysfunction. However, one trial showed a higher rate of new stress urinary incontinence after pubourethral ligament plication than after Pereyra needle suspension, although in only one outcome, objectively demonstrated stress urinary incontinence (RR 2.06, 95% CI 1.05 to 4.06, Comparison 07.02.02) (Colombo 1997).

Another small trial included continent women with occult stress urinary incontinence. More women were incontinent after endopelvic fascia plication rather than when TVT was used as a continence procedure to supplement prolapse surgery, in respect of both subjective stress urinary incontinence (36% versus 4%, RR 9, 95% CI 1.23 to 65.85, Comparison 07.01.08) and objective stress urinary incontinence (44% versus 8%, RR 5.5, 95% CI 1.36 to 22.32, Comparison 07.02.08) (Meschia 2004). However, the data were too few to demonstrate an effect on the need for subsequent continence surgery (Comparison 07.06).

Since all the women in Comparison 07 were continent before prolapse surgery, it was possible to provide estimates of the effects on subsequent urinary function. Overall, 37/300 (12%) of women reported new subjective stress urinary incontinence (Comparison 07.01), and 26/323 (8%) new symptoms of overactive bladder (Comparison 07.04). Long‐term voiding dysfunction (difficulty emptying the bladder) was reported by 32/365 (9%) of women (Comparison 07.05).

Hypothesis 14: utilisation of mesh or synthetic grafts is better than native (autologous) tissue in the management of pelvic organ prolapse (Comparison 08):
Two trials evaluated the effects of using mesh to augment prolapse repairs (Sand 2001; Weber 2001). The data were aggregated in meta‐analysis, and two non‐mesh arms from one trial (traditional anterior vaginal wall repair and ultralateral anterior vaginal wall repair) were also aggregated for comparison with the mesh arm in one of the trials (Weber 2001).

Standard anterior repair was associated with a significantly higher recurrence rate of cystocele compared with augmentation with polyglactin mesh overlay (RR 1.39, 95% CI 1.02 to 1.90, Comparison 08.01.01) (Sand 2001; Weber 2001). One vaginal polyglactin mesh erosion was reported in total from both trials (Weber 2001).

Rectocele recurrence appeared equally common with and without polyglactin mesh augmentation in the other trial, but the confidence intervals were wide (RR 1.13, 95% CI 0.40 to 3.19, Comparison 08.02.01) (Sand 2001).

Hypothesis 15: one type of suture is better than another type of suture:
No trials addressed this comparison.