Six RCTs (381 women) were included.
Significant placebo effects were seen in all studies at 3 months.
Surgical treatment (1 RCT).
The trial compared surgical laparoscopic treatment with expectant treatment (74 women entered; results from 63 analysed; stage IV endometriosis excluded). At 3 months, there was no significant difference between the treatment groups; the relative attributable experimental effect was 14%. At 6 months, significantly less pain was reported by the surgically-treated women than the control; the relative control effect was -2.0% and the relative experimental effect was 45%; the relative attributable experimental effect was 47%. Repeat laparoscopy for continued pain was less common in those treated surgically (16%) than in those with expectant treatment (52%). Where the follow-up was judged incomplete, reanalysis using the worst-case scenario significantly changed the results.
Medical treatment (2 RCTs).
Leuprolide versus placebo for 3 months (1 RCT; 48 women entered): there was no significant difference between the treatment groups at 3 months. The follow-up was judged incomplete (70% of the placebo group were 'lost' after 3 months). Blinding was broken at 3 months and women with significant pain were switched to leucoprolide. The analysis was not conducted on an intention to treat basis.
Medroxyprogesterone versus danazol versus placebo for 6 months (1 RCT; 59 women with stage I or II endometriosis; 14 women also received electrocoagulation of endometriosis implants). At 12 months, women who received medroxyprogesterone or danazol had significantly less pain than those receiving placebo. The relative attributable experimental effect was 50% (95% CI: 30, 71) for medroxyprogesterone and 74% (95% CI: 54, 95) for danazol. There was no significant difference between medroxyprogesterone and danazol. The analysis was not conducted on an intention to treat basis.
Combination of surgery and medical treatment (3 RCTs, all women had surgical debulking).
Medroxyprogesterone versus danazol versus placebo for 6 months (60 women with stage I to IV endometriosis; 51 women completed 12 months' follow-up). At 6 months, pain was reduced on average by 54 and 32% in the medroxyprogesterone and danazol groups, respectively, compared with placebo. At 12 months, the pain levels had returned to baseline in the control group (who had undergone surgical debulking at baseline).
Nafarelin versus placebo nasal spray (2 RCTs; 184 women). One RCT was of treatment with nafarelin versus placebo nasal spray for 3 months (75 women with stage III or IV endometriosis; 61 women completed 12 months' follow-up;). At 12 months, there was no significant difference in pain between the intervention groups.
The other RCT was of treatment with nafarelin versus placebo nasal spray for 180 days (109 women with stage I to IV endometriosis; results from 93 reported). Only 50% of the patients had significant pelvic pain, and a combined dysmenorrhoea-dyspareunia-pelvic pain score was presented. At 6 months, nafarelin was associated with significantly less pain than placebo. However, at 12 months, 6 months after stopping nafarelin, there was no statistically-significant difference in the pain scores between the treatment groups; the absolute attributable experimental effect was 0.4 (95% CI: -0.4, 1.2). The follow-up was judged incomplete and reanalysis using the worst-case scenario significantly changed the results. The analysis was not conducted on an intention to treat basis.
The following side-effects were observed in patients receiving medical treatment with medroxyprogesterone or danazol:
acne, in 1 of 4.5 and 1 of 2.4 women, respectively;
oedema, in 1 of 1.8 women, in both groups;
muscle cramps, in 1 of 9.1 and 1 of 3.8 women, respectively;
abnormal bleeding, in 1 of 2 and 1 of 1.8 women, respectively. The side-effects of medical treatment with leuprolide included hot flushes, headaches, and 'total adverse effects'.
The side-effects were similar in the combined treatment and medical treatment only groups. Surgical treatment alone: either the data were not reported or the sample sizes were too small to calculate the number-needed-to-harm.
Medical treatment: the number-needed-to-harm ranged from 2 to 5.
The quality scores of the included studies ranged from 4 to 6. Problems with the studies included: unsatisfactory description of the method of randomisation (2 RCTs); incomplete follow-up (3 RCTs); analysis not conducted on an intention to treat basis (3 RCTs); and a lack of CIs around the outcome of interest. All studies were considered to have similar groups at baseline, equal ancillary treatment, and adequate blinding.