Figure 1 shows the numbers of women attending the clinic who were potentially eligible for the study, those recruited into the study, and those excluded with reasons. Study recruitment was done by 42 different clinicians, including both doctors and nurses, between March 2009 and January 2010. Of 8401 women who attended the clinic during this time and who were potentially eligible for recruitment, 4931 (59%) returned study forms—that is, recorded evidence of the study having been considered or discussed. Study forms were not completed by 3470 (41%) women—that is, there was no documented evidence of study inclusion being considered by the clinician.
Study flowchart of participants
Full demographic and clinical data and patient infected statuses were available for 3973 women. Several test results were missing owing to problems with sample collection; 33 (0.8%) endocervical samples for the AC2 assay and four (0.1%) samples for culture could not be processed because of staff errors in labelling. After sampling a site for the AC2 assay, the swab must be placed in a screw top tube containing liquid medium, and the stem of the swab must be broken off before the lid is replaced. During this process, handlers may accidentally spill the liquid or perforate the foil top. Of the self taken vulvovaginal swabs, 40 (1%) could not be processed because of participants’ handling errors with the sample tube, and 37 (0.9%) could not be processed owing to staff errors in labelling. As a result, we had to exclude 114 (3%) women before statistical analysis using the paired McNemar’s test (fig 1).
Of the 3973 study participants with complete data, the mean age was 25 years (range 16-59); self reported ethnicity was white in 3171 (80%), black in 362 (9%), mixed in 297 (7%), and other in 143 (4%). A previous diagnosis of STI was reported in 1478 (37%), and 292 (7%) reported contact with a partner recently diagnosed with an STI. At least one symptom suggestive of a bacterial STI was reported by 1671 (42%) participants. A clinical diagnosis of cervicitis was made in 218 (5%) women, and 169 (4%) had a clinical diagnosis of pelvic inflammatory disease. One hundred of the 3973 women were infected with gonorrhoea (prevalence 2.5%), and 55 (55%) of these were coinfected with C trachomatis. Of the 3873 women without gonorrhoea, 355 (9%) had chlamydia infection (overall prevalence 10.3%). Table 1 shows the factors associated with gonorrhoea infection.
Table 1 Factors associated with gonorrhoea infection in study cohort (n=3973)
Gonorrhoea infection was significantly associated with black and mixed ethnicity, younger age, symptoms suggestive of a bacterial STI, contact with an STI, and clinical diagnosis of cervicitis and pelvic inflammatory disease. However, 23 (23%) women with gonorrhoea had none of these risk factors.
Test sensitivities and specificities
Table 2 shows test results of the 3859 women who had complete results of the three diagnostic tests. The sensitivities of culture, clinician taken endocervical NAATs, and self taken vulvovaginal NAATs for gonorrhoea detection were 81%, 96%, and 99%, respectively. The AC2 tests were significantly more sensitive than culture (P<0.001). The sensitivities between NAATs for endocervical and vulvovaginal swabs did not differ significantly (P=0.375).
Table 2 Results of women with complete diagnostic tests for gonorrhoea
The specificity of gonococcal culture with biochemical confirmation was acknowledged to be 100%. All positive results from the AC2 assay were further analysed using the Aptima GC before being reported clinically positive. One self taken vulvovaginal swab positive by the AC2 assay was unconfirmed on the Aptima GC assay because there was insufficient fluid. This result was reported as indeterminate. Therefore, we could not designate this AC2 result as either true positive or false positive, and we treated it as a missing result. No endocervical samples gave indeterminate results for gonorrhoea. Therefore, the specificities and positive predictive values of all tests in all sites were 100%, and the negative predictive values of all tests were 99% or greater.
Negative results from the AC2 assay were not further analysed using a second NAAT, but each woman had three different diagnostic tests for gonorrhoea. Since we defined the patient infected status as any one of the three tests being positive, each women regarded as non-infected was negative for all three tests. No participants with positive results from gonococcal culture had negative results from the AC2 assays.
Diagnostic accuracy of tests in women with and without symptoms
Of 1625 (42%) participants with complete results who had at least one symptom suggestive of a bacterial STI, 56 (3.4%) were infected with gonorrhoea. The sensitivities of culture, AC2 for clinician taken endocervical swabs, and AC2 for self taken vulvovaginal swabs for gonorrhoea detection were 84%, 100%, and 100%, respectively. The AC2 assays were significantly more sensitive than culture (P=0.004, for both endocervical and vulvovaginal swabs). The sensitivities between AC2 assays for endocervical and vulvovaginal swabs did not differ significantly (P=1.0).
Of 2234 women who did not have symptoms suggestive of a bacterial STI, 40 (1.8%) were infected with gonorrhoea. The sensitivities of culture, AC2 for clinician taken endocervical swabs, and AC2 for self taken vulvovaginal swabs for gonorrhoea detection were 78%, 90%, and 98%, respectively. The vulvovaginal swab NAAT was significantly more sensitive than culture (P=0.008). There was no significant difference between the endocervical and vulvovaginal AC2 assays (P=0.375) or between the endocervical AC2 assay and culture (P=0.125).
Comparing the sensitivities of tests and samples in women with symptoms versus those in women without symptoms suggestive of a bacterial STI, culture performed equally well in both groups (sensitivities 84% v 78%; P=0.59). The AC2 assay for vulvovaginal swabs also performed equally well between the groups with and without symptoms (100% v 98%; P=0.42). However, the AC2 assay for endocervical swabs was significantly more sensitive at detecting gonorrhoea in women with symptoms than in those without (100% v 90%; P=0.028, Fisher’s exact test).