Women aged 16 to 25 years were recruited from general practice, family planning and sexual health clinics in Victoria, New South Wales and the Australian Capital Territory, three jurisdictions in South-Eastern Australia. The methods have been previously described [17]. In brief, written informed consent was obtained by all participants in the study. All eligible women were assessed for competency by research staff prior to being invited into the study. The research staff worked closely with the clinical staff to ensure only competent women were approached. All ethics committees approved the inclusion of participants over the age of 16 without parental or guardian consent. Women were recruited when they attended the clinic for any reason during the recruitment period and were eligible if they had ever had vaginal sex with a man, were not pregnant, could comprehend written English and were contactable by post during the 12-month period of the study. Women were tested at the time of recruitment (baseline) and at 6 and 12 months; women who tested positive at any stage were also re-tested three months after a positive test. All follow-up testing was conducted using self-collected vaginal swabs sent through the mail.

Women who tested positive at baseline were managed by their clinician. The research team provided free treatment of 1 gram stat of oral azithromycin. Clinicians were given an education pack with the latest treatment guidelines [20], partner notification support and a telephone number they could call to discuss clinical management of STIs with a sexual health physician [17]. The research team contacted each clinic after recruitment to confirm that the azithromycin was prescribed to women testing positive for chlamydia at baseline. If a participant tested positive during the course of the study, she was contacted by the research team who coordinated a telephone consultation with a sexual health physician. Azithromycin was sent to the participant free of charge along with partner notification letters and information sheets about chlamydia. When appropriate, treatment was provided to the woman's current sex partner. If the sexual health physician decided more clinical intervention was required, arrangements were made so that the woman had a face-to-face consultation with a clinician.

A prevalent infection was defined as a positive chlamydia result at the time of recruitment (baseline). An incident infection was defined as a positive chlamydia result diagnosed on a follow up test during the study period. Women were classified as having a re-infection if they tested positive for chlamydia again during follow up after a previous positive result earlier in the study, with or without a negative test in between positive test results. The subsequent positive results were defined as re-infections. Any re-infections diagnosed were also considered incident infections. To differentiate between chlamydia re-infection, treatment failure or persistent infection, we used a modified version of the chlamydia re-infection algorithm developed by Batteiger and colleagues (Figure 1) [6]. If a woman had two infections with different genotypes, then the second infection was considered a re-infection. If there had been a confirmed negative test between two positive results, the second infection was also considered a re-infection. If women had taken the correct antibiotics but had unprotected sex with her current partners or new partners, the second infection was also considered a re-infection. Treatment failure was defined as a positive chlamydia result following treatment with azithromycin if a woman reported she had either abstained from sex between the two tests or always used condoms with sex. An infection was defined as persistent if the woman had two consecutive positive chlamydia test results and was not treated following the first positive result.

At recruitment, women were asked to complete a self-administered paper-based questionnaire which collected demographic data. Paper-based questionnaires were also sent every three months and asked about sexual behaviour (number of sex partners, new sex partners and condom use), recent antibiotic and contraceptive use, and any pregnancies including termination and/or miscarriages. It also included questions about the presence of any recent genital symptoms, including abnormal vaginal discharge, abnormal vaginal odour, dysuria, abdominal pain and abnormal vaginal bleeding.

Calculations assuming a design effect of 2 indicated that a sample size of 1,000 would be sufficient to generate a 95% confidence interval incidence of 2.8 to 6.2 events per 100 person-years if the estimated rate is 4.5 events per 1000 person-years.

Chlamydia incidence and re-infection rates and 95% confidence intervals (95% CIs) based on the estimation of robust standard errors were calculated using Poisson regression. The rate of re-infection was calculated using only data from those who tested positive at least once during the study and had two or more tests. The Kaplan-Meier method was used to estimate the cumulative risk over time (equivalently the probability of remaining infection-free to any given time) of incident chlamydia infection or re-infection after initial infection. Follow-up was censored at 18 months after enrolment. The association between demographic, behavioural and clinical factors and the chlamydia incident or re-infection was investigated using a discrete-time version of the proportional-hazards regression model described by Carlin et al [21] from which rate ratios and robust standard errors were generated. For the purpose of the Kaplan Meier and proportional hazards regression modeling, re-infections were also considered as incident infections and included in the analysis of incident infections.

Factors associated with infection risk in a univariate analysis or that were clinically important in the view of the investigators were included in the multivariate model; if two factors were highly correlated, only one was included in the model.

The chlamydia organism load values were logarithm transformed for the purpose of analysis. Box plots were used to display the interquartile range, median value and overall range of organism load values. For the purpose of the box plots, data were stratified by infection type – prevalent, incident (where re-infections are excluded) and re-infection. T-tests or Mann-Whitney U-tests were used to compare organism load between groups where appropriate. Data were analysed using STATA version 11.1 [22].

Ethics approval to conduct this study was obtained from ten Human Research Ethics Committees throughout Australia.

A total of 1116 women participated in the study with a recruitment response rate of 66%. The retention in the study over the 12 month period was 79% (877 women). There were 2,937 chlamydia tests conducted throughout the study. Overall, 738 women (66%) were recruited from general practice clinics and 378 (34%) were recruited from sexual health and family planning clinics. There were 55 cases of chlamydia diagnosed at the time of recruitment with a chlamydia prevalence of 4.9% (95% CI: 3.7%, 6.4%).

During the follow-up period, there were 47 cases classified as incident infections and 1056.34 person years of follow up with an incidence rate of 4.4 per 100 person-years (95% CI: 3.3, 5.9). Six women were diagnosed with two separate incident infections during the study period and 35 with a single incident infection. Chlamydia incidence was greatest 6 to 9 months after recruitment with 25 incident cases diagnosed during this time [rate = 9.2 per 100 person-years (95%CI: 6.2, 13.7)] (Figure 2). Chlamydia incidence was similar between women recruited from general practice clinics and those recruited from sexual health or family planning clinics [4.6 (95%CI: 3.3, 6.5) versus 4.1 (95%CI: 2.4, 7.0) per 100 person-years (p = 0.68)].

Univariate analysis found that being aged 16 to 20 years [RR = 3.5 (95%CI: 1.8, 6.8)], being less well educated [RR = 3.0 (95%CI: 1.4, 6.5)] and having two or more new sex partners [RR:4.8 (95%CI:2.4, 9.8)] were associated with incident chlamydia. Recent antibiotic use [RR:0.1 (95%CI: 0.0, 0.6)] was associated with reduced incidence. Multivariate analysis found that being aged 16 to 20 years [RR = 3.7 (95%CI: 1.9, 7.1)], being employed [RR = 2.4 (95%CI: 1.1, 4.9)] and having two or more new sex partners [RR = 5.5 (95%CI: 2.6, 11.7)] were associated with an increased incidence of chlamydia. Recent antibiotic use [RR:0.1 (95%CI: 0.0, 0.5)] was associated with a reduced incidence. Education level was highly correlated with age (p<0.01) and excluded from the model. Use of condoms, previously having chlamydia, and other demographic characteristics were not associated with incident infection (Table 1).

There were 81 women who were at risk of re-infection during the study and contributed 62.7 years of follow-up. There were 14 re-infections of which seven were considered ‘definite’ re-infections in women who had had either a negative test in between two positive tests (four cases) or different serovars for each test (three cases). The other seven were ‘probable’ re-infections in women who had had unprotected sex in between positive results. Three women had two episodes of re-infection and eight had one episode of repeat infection. The overall re-infection rate was 22.3 (95%CI: 13.2, 37.6) per 100 person-years and the cumulative risk of re-infection during the study period was 20.3% (95%CI: 11.6, 31.7). The median time to re-infection was 4.6 months with 50% of re-infections acquired between 3.5 and 6.6 months after initial infection (Figure 3).

There were no associations found between re-infection and participant characteristics (Table 1).

One persistent infection was diagnosed in a woman who failed to take the prescribed antibiotics between two positive results of the same serovar. One possible case of treatment failure was identified in a woman who stated she had always had protective sex with condoms and her partner had been treated concurrently. One false positive was found in a participant who tested positive at follow-up despite taking antibiotics and having no risk of re-exposure. This woman was sent another swab that tested negative, and we concluded from her low risk of re-infection that her first test had been a false positive result. She was prescribed azithromycin as a precaution.

Organism load was able to be measured for 50 (91%) prevalent infections, 24 (73%) incident (excluding re-infections) infections and 11 (79%) re-infections. Organism load was significantly higher for the 50 prevalent infections than for the 24 incident infections (excluding re-infections) (p<0.01) or the 11 re-infections (p<0.01). Organism load was not different between the 11 re-infections and the 24 incident infections (excluding re-infections) (p = 0.33) (Figure 4).

For those with a re-infection, organism load was significantly lower for the first repeat infection compared with the initial infection (p<0.01), but there was no difference in load between the first and second repeat infections (p = 0.49) (Table 2).

There were 63 women who tested positive during the study, had a follow up test and had organism load measured. This includes 37 women whose first positive result was detected at baseline and 26 women whose first positive result was detected on a follow up test. Among these 63 women, four (6%) were diagnosed with a repeat infection of the same serovar at their next test; the other 59 women tested negative (57) or tested positive with a different serovar (2).The organism load at the first test was higher for the four women who tested positive again with the same serovar than for those women who tested negative at their next test, although this was not a statistically significant difference (p = 0.06) (Figure 5).

Organism load was not associated with age (data not shown).

The chlamydia serovar was detected in the 39 of the 47 positive incident chlamydia samples; 60% were serovar E (n = 28), 11% were serovar F (n = 5), 6% were serovar D (n = 3), 4% were serovar G (n = 2) and 2% were serovar H (n = 1). The chlamydia serovars detected in the re-infection samples were primarily serovar E (n = 12, 86%), one was serovar D (7%) and one was serovar F (7%). Among prevalent infections, the majority were serovar E (Table 3). Overall, there were no associations observed between serovar and organism load or serovars and demographic characteristics.