Objectives Re-infection with chlamydia may increase subsequent reproductive morbidity in women. The authors sought to identify characteristics associated with re-infection.
Methods A cohort of all women aged 10–49 years with a notification of genital chlamydia in the Australian state of New South Wales during 1999–2008 was defined. Probabilistic linkage was used to identify women with repeat notifications in the same period. The risk of repeat notification was examined according to age and other characteristics using proportional hazards regression.
Results Among 40 936 women in the cohort, 3236 had at least one repeat chlamydia notification over an average of 3.5 years of follow-up. The incidence of repeat notification was greatest in the first year after index notification (4.5 per 100 person-years) and decreased thereafter. The RR of repeat notification increased by 8% (95% CI 7% to 9%) for each year decrease in age. Compared with women aged 20–21 years at index chlamydia notification, women aged <16 years were twice as likely to have a repeat notification (adjusted HR 2.12, 95% CI 1.75 to 2.56), while women aged 26–27 years were half as likely (adjusted HR 0.53, 95% CI 0.43 to 0.66). Year of index notification, parity and concurrent or past gonorrhoeal infection were also significantly associated with the risk of repeat notification, but socioeconomic status and area of residence were not.
Conclusions Younger age is a strong predictor of chlamydia re-infection in women. The results support targeting interventions to prevent re-infections to very young women.
- repeat infection
- record linkage
- chlamydia infection
- epidemiology (general)
- hormonal contraception
- epidemiology (clinical)
- bacterial infection
- primary care
- C trachomatis
- primary HIV infection
Statistics from Altmetric.com
- repeat infection
- record linkage
- chlamydia infection
- epidemiology (general)
- hormonal contraception
- epidemiology (clinical)
- bacterial infection
- primary care
- C trachomatis
- primary HIV infection
Re-infection with genital chlamydia in women has been shown to significantly increase the incidence of serious adverse reproductive health outcomes such as pelvic inflammatory disease and ectopic pregnancy compared with single infections.1 The prevention of re-infection through the follow-up and treatment of sexual contacts, coupled with re-screening of index cases, are therefore important elements of chlamydia control strategies in many countries.2 As these clinical interventions can be time consuming and expensive, their effectiveness may be improved if they could be targeted at those at highest risk of re-infection.
Younger age3–5 and concurrent infection with gonorrhoea have consistently been found to be associated with an increased risk of chlamydia re-infection,3 whereas behavioural factors, such as the number of recent sexual partners or condom use, and testing factors, such as the type of clinic and test type, have not.3 ,6–8 Studies also suggest that female sex hormones may play a role in susceptibility to chlamydia infections,9 but there is limited information on the effects of hormonal changes, such as parity on re-infection risk.
In Australia, opportunistic annual testing for chlamydia is recommended for all sexually active adults aged <25 years, although there are no specific recommendations regarding the type of specimen used for testing nor are there specific guidelines for pregnant women.10 Testing rates in Australia have increased three to fourfold over the last 10 years,11 but overall coverage is low. An estimated 12.5% of sexually active young women are tested annually with the highest rates in women aged 20–24 years.12 Guidelines for re-testing are inconsistent, but generally advise that this should be conducted 3–6 months following a positive test.10 ,13 Data on re-testing women for chlamydia are more limited, but one study suggests this ranges from 25% to 40%.14 Between 1999 and 2002, nucleic acid amplification tests took over from antigen detection or culture as the method of choice for diagnosing chlamydia.15
In Australia, there is a statutory requirement for all laboratories to notify genital chlamydia diagnoses to public health authorities.16 Collation of midwifery data on births is also required by statute. Chlamydia notifications and birth records can be linked providing an opportunity to examine repeat notification rates and characteristics associated with repeat notification, including parity, in a large registry-based cohort of Australian women.
This study was conducted in New South Wales (NSW), Australia's most populous state with over 7 million inhabitants. We used two statutory databases, the NSW Notifiable Diseases Database,16 which contains a record of all notified laboratory diagnoses of genital chlamydia in NSW since 1998 and gonorrhoea since 1993, and the NSW Midwives Data Collection, which contains a record of all births (defined as 20+ weeks gestation or birth weight 400+ g) in NSW since 1994. Each notification record in the NSW Notifiable Diseases Database includes identifying details of the individual including name, date of birth, residential address as well as information on the condition notified, the onset date (for chlamydia and gonorrhoea, this is equivalent to the specimen collection date) and the type of specimen. This database does not record information on negative tests for chlamydia. Each record in the NSW Midwives Data Collection contains identifying details of the woman who gave birth, including her name, date of birth and residential address and details of the birth including the date, and whether she had previous births.
The personal identifying details in each database were used to link all records within both data sets to individual women. The linkage was conducted by the NSW Centre for Health Record Linkage (CHeReL),17 which uses both computerised probabilistic linkage methods and manual audit. Records belonging to individual women were assigned a unique number by the CHeReL. The de-identified records from each database (but including date of birth) were then provided to the researchers, and these were linked to the one individual using the unique number. Statutory records collated up until 31 December 2008 were available at the time of linkage. The study was approved by the NSW Population and Health Services Research Ethics Committee and the University of New South Wales Human Research Ethics Committee.
As 1999 was the first full year in which a diagnosis of genital chlamydia was legally required to be notified in NSW, we defined a cohort of women who had at least one genital chlamydia notification between 1 January 1999 and 31 December 2008. We assigned the index chlamydia notification to be the first recorded chlamydia notification in the database and included only women aged between 10 and 49 years at the time of first notification. A repeat infection was defined as a subsequent chlamydia notification >30 days after the index notification. To examine the rate of repeat notification, follow-up time was calculated from the date of index chlamydia notification to the date of first subsequent chlamydia notification or for women with no subsequent notification, the last day which we had complete records (ie, 31 December 2008). Only the first repeat notification was considered in analyses. Age-specific rates of repeat notification over the entire length of follow-up and in intervals of 6, 12, 12–24 and 24–36 months following index notification were estimated.
The proportional hazards model was used to assess the association between the risk of repeat notification and various characteristics of the women at the time of index chlamydia notification including age (in nine categories and as a continuous variable), year of index notification, type of specimen (vaginal swab or urine), past or concurrent gonococcal notification (based on a linked gonococcal notification on or earlier than the date of index chlamydia notification) and parity (based on information obtained from a linked record of the most recent birth preceding the index chlamydia notification). For both gonococcal notifications and parity, we assumed that if a woman had no linked record in the database that she had, respectively, no gonococcal infection and no previous birth. As socioeconomic status may be associated with testing patterns and risk of chlamydia, we used a woman's residential postcode at the time of index chlamydia notification to categorise her into a quintile of socioeconomic disadvantage (using a standard Australian index that ranks residential postcodes based on 2006 census data for 17 indicators including income, education, household overcrowding, low skill occupations and Indigenous status18). We also used postcode to derive categories of geographic accessibility to health services (based on another standard Australian index of distance from services19). Each characteristic was examined individually and then in a mutually adjusted (multivariate) model.
As women's characteristics collected at index notification may have changed over time, for the main analyses, we censored follow-up at 2 years after index notification. However, as the interval of follow-up in record linkage studies of routinely collected notification data has been shown to affect incidence estimates,20 we conducted sensitivity analyses by examining the risk of repeat notification over different time intervals (12 and 36 months) following the index notification. As our data did not include information on testing, we also examined the effect of age on repeat notification risk in subgroups of socioeconomic disadvantage, year of index chlamydia notification, geographic accessibility index and parity, as all these factors may be associated with differences in testing rates.11 ,21 ,22
From 1999 to 2008, 40 936 women aged 10–49 years had at least one chlamydia notification in NSW. Table 1 shows the characteristics of these women according to their age at index chlamydia notification. Most notifications (71%, 29 109/40 936) occurred in women aged 25 years or younger and 30% (12 166/40 936) occurred in women aged <20 years. Women who were younger at the time of their index chlamydia notification were more likely than older women to be socioeconomically disadvantaged, resident in geographic areas with less access to services and to have had a concurrent or past gonorrhoea notification (71%, 241/340 of gonorrhoea notifications were diagnosed at the same time as the index chlamydia notification). Younger women were also more likely than older women to have had their index chlamydia diagnosis made in more recent years and through providing a urine sample rather than a vaginal swab. They were less likely to have given birth than the older women with a chlamydia notification.
Figure 1 shows the probability of repeat notification by time since index notification for the entire cohort. The incidence of repeat notification was greatest in the first 12 months following the index notification (4.5 per 100 person-years, 95% CI 4.3 to 4.7) and gradually decreased with increasing time since index notification. In the 12–24- and 24–36-month intervals following index notification, the incidence of repeat notification was 2.4 (95% CI 2.3 to 2.6) and 1.6 (95% CI 1.5 to 1.8) per 100 person-years, respectively. The incidence of repeat notification was highest among the youngest women and fell with increasing age. This age-related decrease in incidence was consistent across all intervals of follow-up (table 2).
In the primary analysis, with follow-up censored to the 2 years following the index chlamydia infection, 2388 women had a repeat notification. For these women, the median time between index notification and repeat notification was 0.67 years (IQR 0.36–1.13). The time to repeat notification did not appear to be associated with age at index notification (table 3). In the unadjusted regression analyses, younger age, lower socioeconomic group, later year of index chlamydia notification, residence in areas with less access to services, urine specimen (compared to vaginal swab) and a concurrent/past gonococcal infection were all found to be associated with a greater risk of a repeat chlamydia notification. After adjustment for all these factors as well as past births, a woman's age, year of index chlamydia notification and a gonococcal notification remained significant (table 3). Having a previous birth also became significant (HR 1.39, 95% CI 1.23 to 1.58). In sensitivity analyses when we limited follow-up to intervals of 1 and 3 years after index notification, we found the results were similar (see online appendix).
Figure 2 shows how the relationship between age and risk of repeat notifications follows a pattern of increasing risk with decreasing age. When age was analysed as a continuous variable, the estimated increase in the risk of repeat notification was 8% (95% CI 7% to 9%) for each 1-year decrease in age (HR 1.08, 95% CI 1.07 to 1.09, p<0001). This increase in risk with decreasing age did not change substantially when we estimated separately in subgroups defined by socioeconomic status, region of residence, year of index chlamydia notification or by a record of previous birth (table 4).
We found a striking increase in the relative risk of a repeat chlamydia infection with younger age. Women aged <16 years were twice as likely as women aged 20–21 years to have a repeat infection within 2 years, while the risk halved for women aged 26–27 years. We also found that the risk of repeat infection increased with a history of previous or concurrent gonorrhoea infection, with a history of previous births and with later year of chlamydia notification. Although previous studies have found that chlamydia re-infection occurs more often at a younger age, none have examined the relationship in such narrow age groups as has been analysed here.3–5 The very large difference in repeat notification rates with such small increments in age that persisted following adjustment for multiple other risk factors such as previous gonorrhoea notification, socioeconomic status, accessibility to services and year of index notification suggests that the relation between young age and the risk of re-infection is not due to residual confounding from factors measured in our study.
The strong association observed between age and chlamydia re-infection is difficult to interpret, given the lack of information in the cohort on chlamydia testing, sexual behaviours and the fact that overall testing rates for chlamydia in Australia are low.12 However, based on other studies and data sources, we can speculate that possible explanations may include differences in re-testing rates, contact tracing, sexual risk behaviours or biological susceptibility.
With respect to testing rates, information at the population level are available from the Australian Medicare health insurance programme, which includes all chlamydia tests billed to this system by health practitioners. While the data do not distinguish between first and subsequent tests and do not include the relatively limited number of tests conducted by hospital laboratories, they show that testing rates per capita in 15–34-year-old women increased three to fourfold between 1999 and 2008.11 This is somewhat consistent with the difference in risk of repeat notification that we observed in relation to later year of index chlamydia notification, although our data showed little effect of year after 2005. The Medicare data and other studies also show that while there is about 20% more testing per capita conducted in women aged 15–24 years than in those aged 25–34 years,11 there is little difference between those within the 15–24-year age group.12 Specific data on re-testing rates according to age are more limited. However, if younger women were more likely to be re-tested to the extent that they would appear to have higher rates of repeat infections, we might expect to find that the time to re-notification should also differ by age, with the more actively re-tested populations having a shorter time to re-notification. However, in our data, this was not the case. The median time to repeat notification was 0.75 years in women aged <16 years and 0.60 years in women aged 20–21 years (table 3).
It is possible that higher levels of sexual risk behaviour or more limited contact tracing and treatment may account for the observed increase in the risk of repeat infection with younger age.7 However, this cohort only included women who already had a chlamydia notification, so all women in the cohort would be more likely than the general population to be at increased sexual risk. School surveys provide no indication that sexually active adolescents aged about 15–16 years are more likely than their slightly older peers, aged about 17–18 years, to engage in riskier sexual behaviours such as having greater numbers of sexual partners in the last year and not using condoms.23 On this basis, it is difficult to attribute the strong age differential solely to differences in sexual risk behaviours. With respect to partner notification, there is limited information on differences in compliance between age groups, but it is possible that younger women may be less able to negotiate effective partner notification, contributing to the age differential in risk of repeat notification.
Biological differences (anatomical and immunological) are also possible explanations for the age difference in repeat notification rates. A study of over 8000 women aged 20–90 years demonstrated a linear relationship between decreasing age and increasing presence of cervical ectopy such that in women aged <25 years, 94% displayed cervical ectopy compared with 2% of women aged over 64 years24; however, this relationship may not extend to adolescents.25 As Chlamydia trachomatis infects columnar and not squamous epithelium, it is possible that some of the predisposition to repeat infection, particularly in women aged over 20 years, may result from this anatomical difference.
Greater immunity to chlamydia infection from prior exposure may also explain lower infection rates in older women.26 In this study, all women had had at least one prior infection and so all could have developed some immune response. While we could not know if older women had more chlamydia infections prior to the index notification than younger women, when we restricted the analyses to women with an index notification in the year 2005 onwards (ie, women who had no record of a notification in any of the 7 years preceding this), we found the same strong age gradient (table 4). This observation, as well as the lack of an age differential in the median time to repeat notification (table 3), suggests that at best the immunity hypothesis would only partially explain the strong age association. Finally, we did not have data on men. However, other studies that have reported on age and chlamydia re-infection in both men and women have found either no relationship in men5 or that there is an association, but it is substantially reduced in magnitude compared with women.27 ,28
Parity, or a history of having given birth, has not previously been described as a factor that may increase the risk of repeat chlamydia infection. Possible reasons for this finding are that past birth, particularly among adolescents, may be a surrogate marker of riskier sexual behaviour. Women who have had a birth may also be more likely to be tested for chlamydia. Furthermore, parity and exposure to oestogens may be associated with increased cervical ectopy and risk of genital infection,9 ,24 although this relationship is complex. We also found that prior to 2005, an earlier year of index notification was associated with an decreased risk of repeat notification. This finding may be partly explained by the large increases in testing which occurred in Australia over the time of this study.11
The strengths of our study are the large sample size and whole of state cohort, ensuring minimal selection bias and better generalisability of our findings. As chlamydia notification by laboratories is required by law,16 we would expect essentially all new diagnoses in the study period to be included in our data set. Chlamydia, gonorrhoea and parity information were linked independent of study investigators by the CHeReL system, and audits have demonstrated very low false-positive linkage rates (0.4%).17 In addition to the lack of testing and behavioural data, a limitation to our study is the use of passive surveillance data, which are likely to underestimate true re-infection rates,3 especially given that testing12 and re-testing rates in Australia29 are low overall. We also had limited data on Indigenous status so despite the fact that Indigenous Australians are known to have much higher rates of chlamydia and other sexually transmitted infections than non-Indigenous Australians,30 we were unable to comment specifically on repeat notifications in this population.
In conclusion, we found younger age to be a very strong predictor of genital chlamydia re-infection in women. This difference in risk may result from higher levels of re-testing, differences in behavioural risk factors or potentially biological differences, such as hormonal changes in the female reproductive tract or increased immunity in older women. Overall, our results support targeting interventions to prevent re-infections to very young women and further research on the possible mechanisms of susceptibility to chlamydia re-infection.
The relative risk of having a repeat notification of chlamydia was greatest in women aged <16 years and decreased with increasing age.
Interventions to prevent re-infections should focus on very young women.
The NSW Centre for Health Record Linkage conducted the linkage of the data sets. NSW Health provided de-identified data to the researchers.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Download Supplementary Data (PDF) - Manuscript file of format pdf
Funding This work was supported by the Australian National Health and Medical Research Council (NHMRC) grant numbers 568971 and 573122. The funder had no input into the study design, analysis, interpretation, writing nor decision to submit for publication. BL, RG, BD and JK are supported by NHMRC fellowships.
Competing interests None.
Ethics approval Ethics approval was provided by NSW Population and Health Services Research Ethics Committee and University of New South Wales Human Research Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.