We found consistent evidence of a strong association between HIV and LGV across the literature. In thirteen descriptive studies at least two-thirds of MSM with LGV were co-infected with HIV. In a meta-analysis, MSM with LGV were over eight times more likely to have HIV (OR 8.19, 95% CI 4.68-14.33) than those who had non-LGV chlamydia infection. This association is stronger than for other sexually transmitted infections which have also re-emerged in MSM over the past decade . As the descriptive studies presented results mainly in aggregate form, it was not possible to further investigate the effect of LGV genotype or rectal and non-rectal infection on the association between HIV and LGV. However, in the majority of studies rectal symptoms and proctitis were the dominant manifestation of LGV, as shown by results from the three largest studies: in the UK 90% of confirmed cases had proctitis , in France only rectal swabs were included in the surveillance system , and 91% of cases in Netherlands had proctitis . Similarly only aggregate information was available for LGV genotype, but Canada , France  and the Netherlands  found only L2b genotype in the samples that were sequenced. Other descriptive studies selected for this study found mainly L2 serotype with predominantly L2b variant in the sequenced samples (such as 426/470 of samples in UK surveillance that were confirmed L2b ). Primary data sources would be needed to analyse differences between HIV positive and HIV negative LGV cases.
A number of detailed reviews have been published on the emergence of LGV [14, 51]. However to our knowledge this review is the first to systematically identify and quantitatively analyse the association between HIV and LGV in the re-emergence of LGV, and includes data from 17 reports. However, it is limited by the inclusion of a number of surveillance reports. Surveillance of LGV is said to suffer from the lack of consistency between countries giving rise to different case-definitions and surveillance methods, some of which are based on voluntary and potentially partial reporting . Surveillance reports formed a large proportion of publications used for the HIV-prevalence estimates, but the studies in this review all defined detection of LGV serovars as a requirement for confirmed cases. Nevertheless it is possible that cases have been identified with different accuracy between countries and the international data may not be comparable. Previous literature has noted that the variation in the structure of surveillance systems limits cross-country comparison . The geographical differences may be a result of variation in sexual partnerships, differential sexual behaviours (such as seroadaptive behaviour), testing and case finding policies, time of introduction of LGV into the network and the impact of local public health interventions .
In this review we did not apply strict quality criteria. This is due to the relatively small number of published studies and our desire to be inclusive in order to increase sample size. This means that we cannot exclude the possibility that sampling and measurement differences may account for some of the heterogeneity. Outcome-level assessment for LGV was performed by including only confirmed cases (identification of LGV serovars required for a confirmed case) which assures the validity of the data for the main outcome. Due to the small number of studies conducted to date and the recent emergence of LGV, publication bias was not thought to be relevant in this analysis. The search and article retrieval was conducted by the first author alone, but the analysis and interpretation was done by both authors.
The strong association between LGV and HIV in an unadjusted model is interesting, and some association would be expected to remain after adjusting for confounders, especially since the association is strong even though the control group represents another high risk group of MSM with a non-LGV CT infection. It is possible that there may be a biological interaction between HIV-infection and LGV. HIV-infection has been reported to cause abnormalities in the structure and function of the gastrointestinal tract as well as abnormal lymphocyte trafficking  that may facilitate LGV acquisition. In a South African cohort of patients with GUD, LGV infection was diagnosed more often among the HIV-infected participants compared to HIV-negative participants with relative risk 1.3 (95% CI 1.2-1.4) . Also in the Bahamas, an LGV epidemic was associated with crack cocaine use and HIV infection . Two of the case-control studies included in this review compared median CD4 count between cases and controls, but neither found a significant difference [57, 58]. van der Bij et al.  also hypothesised that immunorestoration inflammatory syndrome could explain the emergence of this highly symptomatic LGV. They analysed the association between having symptomatic LGV and start of highly active antiretroviral therapy (HAART) but no association was found. Time-series analysis may be a more effective way to measure the potential interaction between HIV and LGV.
The strongest confounder for this hypothesis is high-risk sexual behaviour. There have been behavioural changes in MSM communities as a response to HIV epidemic, and serosorting may make HIV-positivity an intermediating factor: individuals participate in certain type of practices partly because of their HIV-infection . Seroadaptive behaviour can reduce the number of new HIV infections but it also facilitates the spread of other STIs and can create dense sexual networks where disease propagation is fast . Exploring this further was not possible in this review because of the lack of comparable data on sexual behaviour.