In accordance with the protocol
, we identified cities in Peru with more than 50 000 inhabitants, and excluded those that were involved in another STI prevention trial. Geographically separate cities (classed as separate when more than 30 minutes travel by land between the cities) eligible for inclusion were matched into pairs that were randomly assigned to intervention or standard care.
Of 30 Peruvian cities with more than 50 000 inhabitants, we excluded Lima because its large population of 9 million precluded matching with another city, two cities because effects could be altered by proximity to Lima, and three cities involved in another STI prevention trial,7
leaving 24 cities. Population-based baseline surveys of young adults occurred from August to November, 2002 (). We then selected 20 cities, matched them into pairs, and randomly assigned them to intervention or standard care. Outcome surveys in 2005 included random samples of young adults in these 20 cities. In 2006, a data and safety monitoring board recommended continuation of the interventions, with final outcome surveys in late 2006.
Baseline surveys in 2002, and subsequent outcome surveys in 2005 and 2006, were done in random household samples of young adults and in FSWs. Young adults aged 18–29 years who had lived in the city for 6 months were eligible for inclusion in our analysis. For the baseline survey, we also did a comprehensive census of commercial sex venues. FSWs working at these venues, or attending screening clinics, were non-randomly, consecutively selected until all venues had been visited or the quota of 200 FSWs per city was reached. Subsequent FSW surveys used sampling11
of individuals at randomly selected venues and times.
Institutional review boards at the University of Washington, Universidad Peruana Cayetano Heredia, and US Naval Medical Research Center Detachment approved the protocol, consent forms, and instruments. Eligible FSWs older than 14 years and survey participants provided verbal consent. All review boards exempted pharmacy workers and clinicians from informed consent. Young adult survey participants received study caps and T-shirts; FSW survey participants received free condoms.
All surveys of young adults used identical sampling methods, with paper questionnaires for the baseline survey and computer-assisted self-interviews with personal digital assistants in the later surveys.12
Men were asked to provide blood and urine samples, and women to provide blood and self-obtained vaginal swabs, or urine if unwilling to provide swabs. Participants received information about STIs and HIV, and referrals for test results, counselling, and treatment when indicated. FSWs answered face-to-face questionnaires in the baseline survey, followed by collection of blood and one endocervical and two vaginal swab specimens during speculum examination. All FSW participants completed questionnaires and provided three self-obtained vaginal swabs in 2005, and four in 2006.
After completion of a census of all pharmacies, and of pharmacy workers, clinicians, and midwives in each city, the PREVEN network was created in intervention cities; STI training was undertaken for each group, involving STI syndromic management, and cross-referral of clients and patients across the network. Individuals who completed the programme of training were included in the network. Six key activities for improvement of syndromic management were adapted from our early programmes.8,9
First, workshops and materials were provided for pharmacy workers, emphasising recognition of four STIs (male urethritis, vaginal discharge, genital ulcer disease, and pelvic inflammatory disease). Second, physicians and midwives were trained in STI syndromic management. Third, so-called prevention salespersons made monthly visits with educational materials to network members and pharmacy clients. Fourth, clinicians (including those certified by the network) were given a year 2 booster online STI course.13
Fifth, health-communication campaigns for the general population were run in 2003–04, and in 2005, to improve recognition of STI symptoms and seeking of early appropriate STI health care. Finally, social marketing of low-cost condoms and inexpensive STI treatment packets for urethral discharge (containing ciprofloxacin, 500 mg, and azithromycin, 1·0 g) and for vaginal discharge (containing metronidazole, 2·0 g) were placed in pharmacies in 2003–04. Network training began in July, 2003, with all trainees certified by January, 2004. Prevention salesperson activities continued throughout 2006.
We created mobile teams and laboratory support systems in intervention cities to deliver clinical and preventive services to FSWs from July, 2003, to December, 2006. Each mobile team was made up of a nurse or midwife and an FSW peer educator. Mobile teams' activities included two visits to each sex venue in each of 20 cycles of 8 weeks to provide periodic presumptive treatment14
with metronidazole for trichomoniasis and bacterial vaginosis to FSWs who were not pregnant or breastfeeding, and willing to forego alcohol consumption for 72 h. Self-obtained vaginal swabs were collected for local T vaginalis
culture and for nucleic acid amplification in Lima for N gonorrhoeae
and C trachomatis
. The teams returned 1 week later, providing test results and treatment for specific infections identified (ciprofloxacin for gonorrhoea, azithromycin for chlamydia, and metronidazole for positive T vaginalis
cultures not treated a week earlier). FSWs were encouraged to visit local government clinics for periodic syphilis and HIV testing, and for interim STI symptoms. Laboratory technicians joined mobile teams from February, 2005, to December, 2006, and did rapid syphilis testing.15
Separate mobile team outreach to male sex workers by a nurse or midwife and a peer occurred from April to December, 2006. Initial presumptive treatment was 1 g azithromycin for infection with C trachomatis and 500 mg oral ciprofloxacin for N gonorrhoeae, with screening for syphilis and HIV. Seropositive male sex workers were referred to government clinics for care.
Mobile teams also provided motivational interviewing to promote condom use by sex workers, and gave up to 15 condoms to each FSW in each 8 week cycle in the first 1·5 years, then increased to 50 condoms per cycle. For the general population, the local non-governmental organisation APROPO implemented social marketing of a low-cost condom, the OK condom, through pharmacies in intervention cities only, from October, 2003, to October, 2004, then more widely.
For all surveys, urine specimens, two polyester swab specimens stored in cryovials, and blood samples were promptly placed into coolers and transported to local laboratories. Trichomonas vaginalis culture media (InPouch, Biomed Diagnostics, White City, OR, USA) inoculated with cotton swabs were transported at ambient temperature to local laboratories. For the first survey, an additional swab was collected using the APTIMA Specimen Collection kit (Gen-Probe, San Diego, CA, USA). Serum, urine aliquots, and dry swabs in cryovials were frozen to −20°C at local laboratories and shipped weekly to Lima, where 1 mL of 2-sucrose-phosphate was added to each cryovial, and serological testing and nucleic acid amplification were done. Initial analysis of results identified a need for additional confirmatory testing for Chlamydia trachomatis and Neisseria gonorrhoeae infections because prevalences were high and inconsistent in some cities, and to improve N gonorrhoeae test specificity. Duplicate specimen collection allowed such testing. On the basis of the interim report from the data and safety monitoring board, nucleic acid amplification testing for T vaginalis was also added. Final analyses followed completion of these tests.
Baseline and 2005 detection of Chlamydia trachomatis was with Amplicor assays (Roche Diagnostics, NJ, USA). Indeterminate specimens were later tested by Aptima Combo 2 assay (Gen-Probe, San Diego, CA, USA). Results were counted as positive if Amplicor was positive, or if indeterminate and Aptima positive. The 2006 survey used Aptima assays. To exclude possible specimen contamination during collection or testing, positive tests were confirmed with concurrently collected urine or swab specimens suspended in 2-sucrose-phosphate.
For N gonorrhoeae, Amplicor assays were used for samples collected in the baseline survey. Positive specimens were later confirmed by Aptima Combo 2 assay, with urine or swab samples suspended in 2-sucrose-phosphate. In 2006, we defined infection with N gonorrhoeae by positive Aptima results in initial specimens, with repeat confirmatory tests on concurrently collected specimens.
Culture specimens incubated at local laboratories were examined every day for 5 days for motile trichomonads. Urine and vaginal swabs collected during baseline and 2006 surveys were later tested in Lima with T vaginalis
analyte-specific reagents and Aptima General Purpose Reagents (Gen-Probe, San Diego, CA, USA).16
Specimens consistently yielding relative light unit values of more than 100 000 were classed as positive. We counted specimens as positive for syphilis if they were seroreactive at titres of 1:8 or more by rapid-plasma-reagin test and confirmed by Treponema pallidum
particle agglutination assay. Specimens that twice tested positive on the enzyme-linked immunosorbent assay were confirmed by western blot assay in Lima; indeterminate specimens were retested at the University of Washington (Seattle, WA, USA).
The coefficient of variation across cities in the 2002 young-adult survey was 0·26 for the primary endpoint. Further calculations suggested matching could reduce this number to about 0·10.17
With ten city-pairs, 500 individuals sampled per city, and a composite STI prevalence of 0·10 in the control group, we anticipated 90% power to detect a relative risk (RR) of 0·75 (α=0·05, two-tailed; formula 7·14).18
Similarly, 200 FSWs surveyed per city provided 87% power to detect a 30% reduction in the combined STI outcome, with the assumption of a control community prevalence of 14%. A data and safety monitoring board reviewed the trial in 2006, with an analysis that would stop the trial if the intervention was highly effective (α <0·01) on the combined STI endpoint in both FSWs and young adults.
Continuous measures were summarised by medians and IQR, and categorical measures by proportions. The composite STI outcome for each participant was defined as positive if any of the C trachomatis, N gonorrhoeae, vaginal T vaginalis (by GenProbe nucleic acid amplification tests), or syphilis seroreactivity results were positive; as negative if all were negative; and otherwise as missing. However, 563 missing syphilis serology values in young adults were classed as negative for composite endpoint calculation in the 2006 young adult survey (results were essentially unchanged by this classification).
The primary analysis of young adults regressed differences in STI prevalence between intervention and control communities of the ith pair in 2006 outcome surveys against corresponding differences in baseline surveys. In this regression, the intercept gives the adjusted difference in prevalence between intervention and control communities in 2006. We tested the null hypothesis that the intercept was zero. Standard linear regression methods were used to fit the model and generate 95% CIs for adjusted risk differences. We used one-way ANOVA to estimate intraclass correlation. We used generalised estimating equations with a log link to compute RR and 95% CIs, adjusted for city-specific baseline prevalences with a modified approach19
to account for variance bias that can occur with few clusters. RR analyses do not incorporate information on pairing and are, therefore, conservative. Methods giving RR for paired data18,20
often failed to converge for rare infections; for consistency, we present unpaired RR analyses for all outcomes. All analyses were intention to treat. We did all statistical analyses with Stata (version 10·1). This study is registered, number ISRCTN43722548.