Factors influencing HIV seroconversion rate in Chinese heterosexual discordant couples: A systematic review and meta-analysis CURRENT STATUS: POSTED

At present, there are few studies focusing on the factors affecting the seroconversion rate of serodiscordant couples, but the number of serodiscordant couples is increasing, which has become a serious global public health problem.We sought to identify factors contributing to HIV seroconversion of the uninfected spouse in couples serodiscordant for the virus. We used a systematic literature search to identify studies describing serodiscordance and seroconversion among couples in which one partner was infected with HIV. We conducted a meta-analysis of 18 published articles in which we combined the random-effects models to identify factors contributing to seroconversion.Forest plots, sensitivity analysis and Egger’s regression asymmetry test were performed by STATA SE-64 (version 12) to measure the overall results, stability of results and bias. Heterogeneity was tested by I2 statistic.

2 Abstract Background At present, there are few studies focusing on the factors affecting the seroconversion rate of serodiscordant couples, but the number of serodiscordant couples is increasing, which has become a serious global public health problem.We sought to identify factors contributing to HIV seroconversion of the uninfected spouse in couples serodiscordant for the virus.

Methods
We used a systematic literature search to identify studies describing serodiscordance and seroconversion among couples in which one partner was infected with HIV. We conducted a metaanalysis of 18 published articles in which we combined the random-effects models to identify factors contributing to seroconversion.Forest plots, sensitivity analysis and Egger's regression asymmetry test were performed by STATA SE-64 (version 12) to measure the overall results, stability of results and bias. Heterogeneity was tested by I2 statistic.

Results
The overall seroconversion rate was lower in the ART-receiving group odds ratio (OR) = 0.34, 95% confidence interval (CI) (0.16; 0.71), demonstrating that ART is effective in preventing HIV transmission. Farmers had greater odds of HIV transmission than those who were not farmers OR = 1.28, 95% CI(1.07; 1.52). Further, the serodiscordance rate was higher in intravenous drug abusers OR = 0.76, 95% CI(0.66; 0.87). However, HIV patients and AIDS patients had no statistical differences in serodiscordance rates OR = 1.38, 95% CI(0.81; 2.35).

Conclusions
These findings suggest that promoting early and long-term access to ART treatment, developing a rural HIV prevention model, and focusing on families of drug addicts may be beneficial interventions for reducing HIV transmission. Further research should explore the effects of ART intervention in different periods (HIV/AIDS).

Background
Human Immunodeficiency Virus(HIV) seriously impacts global public health through the pandemic's 3 influence on morbidity and mortality. By the end of 2017, there were 78 million reported HIV/AIDS(Acquired Immune Deficiency Syndrome) infections worldwide, and the mortality caused by AIDS or its complications reached 35 million [1]. In China, the situation is not optimistic-China had about 780,000 living HIV/AIDS patients by the end of 2011 [2]. Research has shown that heterosexual transmission, superseding injection drug use, has played a dominant role in all modes of transmission, and around 1/3 of those infected by this method were infected by a spouse [3,4]. HIV serodiscordant couples, defined as one partner who is HIV-negative and the other who is HIV-positive, are at a high risk of becoming infected without any intervention of protective measures [5].
Unprotected sex between discordant couples provides a chance for transmission and has become a predominant source of new infections [6], necessitation the identification of factors associated with seroconversion rates. Use of antiretroviral therapy (ART), condom use, gender of index case, HIV viral load, and age of index case are associated with transmission between discordant couples [7]. One study showed that the seroconversion rate is higher in couples who had not received ART and lower in couples who had received ART, and that the effect of ART is remarkable in the first year but then stalls with the extension of the follow-up period [8]. However, less attention has been given to additional factors, such as disease stage and occupation (e.g., farmer).
Because there is high risk of acquiring HIV from an HIV-positive partner, clear effective prevention strategies and interventions are particularly significant to not only protect the HIV-negative partner, but also to reduce the society-wide AIDS prevalence. Therefore, we evaluated the effectiveness of potential factors associated with HIV seroconversion in serodiscordant couples using a systematic review and meta-analysis of published articles, applying subgroup analysis to eliminate the effects of confounding factors.

Search strategy and selection criteria
We defined HIV serodiscordant couples as heterosexual couples in which one partner was HIV-infected while the other was not infected. The definition for couples was a partner who had a legal effect or those in a traditional marriage or stable, long-term sexual relationship. Legal marriage referred to 4 marriage registered at the Civil Affairs Bureau and protected by the constitution. Traditional marriage referred to extremely remote areas where there was no capacity to register. HIV-positive was defined as an individual who tested positive for HIV antibody by western blot.
We systematically searched the literature published between June 1981 and May 2018 using the following Chinese and English databases: China National Knowledge Infrastructure (CNKI), Wanfang, Weipu and Pubmed, Embase and SinoMed. The following keywords were used to search for relevant articles: HIV/AIDS, serodiscordant, seroconversion rate, ART, and transmission. Articles whose study area were not in China were not included. We excluded articles that presented only qualitative data, or presented quantitative data in a manner that did not allow calculation of seroconversion rate.
Articles that did not identify areas of study, methods of study, or demographic baseline characteristics or those with anti-HIV antibody tests not verified by western blot were excluded. For different articles from the same study, we only used the most complete data. We also did a manual search to address unpublished and gray literature.In addition, we included the most recent study if the integrity of data from articles was similar. A flow diagram of preferred reporting items for systematic reviews and meta-analyses describing the article selection process is shown in Figure 1.

Data extraction
We used a standard form to extract the following information from each article: 1) name of first author; 2) publication date; 3) study area; 4) study region; 5) study date; 6) number of couples; 7) ART; 8) routes of infection; 9) seroconversion rate; 10) occupation; and 11) disease stage. Basic characteristics of the articles are shown in Table 1.

Statistical analysis
We chose the following items as the main output indicators: 1) seroconversion rate in couples receiving or not receiving ART; 2) seroconversion rate in couples with different occupations (farmer or non-farmer), disease stage (HIV or AIDS) and infection route (intravenous drug use or heterosexual intercourse). Study methods contained both retrospective and prospective analyses, so odds ratio (OR) values were used to select effect quantities. Subgroup analysis was performed if heterogeneity was high. We used Cochrane Collaboration's tool for randomized controlled trials (RCTs) and Newcastle-Ottawa scale for non-RCTs to assess study quality. We estimated the possibility of heterogeneity between studies with I2 statistic and used random-effects models for meta-analysis whether heterogeneity was high or not. Given that the number of included articles was not large, Egger's regression asymmetry test was chosen to assess potential for publication bias. We used sensitivity analysis to determine whether results were substantially altered or not. Statistical analyses were performed using STATA SE-64 (version 12) .
Research subjects were HIV serodiscordant couples, and the HIV prevalence rate was available in each article. In total, 15 articles included data on ART, 11 articles provided data on occupation, 6 articles included infection route of the index partner, and 3 articles indicated disease stage of index partner.

Effect of ART
Whether infected individuals received ART and their respective seroconversion rates were provided in 15 articles detailing 110,098 HIV serodiscordant couples. Seroconversion rate was higher in couples who did not receive ART in 14 articles, while 1 article reported the opposite result. Heterogeneity test showed high heterogeneity (I2 =99.0%; P < 0.001) between articles, so we used a random effects model. Meta-analysis showed that ART was effective in preventing HIV transmission, as overall Sensitivity analysis was conducted to measure stability of the results. After each article was neglected, analysis showed no significant difference in the results ( Figure 4A). Further, Egger's regression asymmetry test revealed no significant publication bias (t = 0.59; P = 0.564) ( Figure 5A).

Effect of occupation
Occupation of couples, classified as farmer or non-farmer, was evaluated in 11 articles with data on 60,284 couples. In those 11 articles, 9 reported that the serodiscordant rate among farmers was higher than that for non-farmers, while 2 articles found a lower serodiscordant rate in farmers. The heterogeneity test indicated no significant heterogeneity (I2 =21.9%; P = 0.234). Mean meta-analysis showed farmers had more opportunity for HIV transmission than non-farmers [OR = 1.28; 95% CI (1.07, 1.52)] ( Figure 2B). When any individual article was ignored in sensitivity analysis, the results did not change ( Figure 4B). However, Egger's regression asymmetry test indicated strong publication bias (t =1.91; P < 0.1) ( Figure 5B).

Effect of infection route
Infection route of index partner (heterosexual sex or intravenous drug use) was reported in 6 articles with data on 52,206 couples. Of these 6 articles, 5 reported increasing serodiscordant rate in patients infected by heterosexual sex. Articles showed no heterogeneity (I2 = 0.0%; P = 0.478). The serodiscordant rate was higher for those infected by intravenous drug use [OR = 0.76; 95% CI (0.66, 0.87)] ( Figure 2C). Sensitivity analysis revealed that 1 article had so much weight that results [OR = 0.88; 95% CI (0.70, 1.11)] showed no difference between transmission routes ( Figure 4C). Further, Egger's regression asymmetry test found publication bias (t =0.04; P < 0.1) ( Figure 5C).

Effect of disease stage
Disease stage data were offered in 3 articles containing data on 2400 couples of which 2 articles  Figure 2D). Sensitivity analysis showed that results were stable ( Figure   7 4D). However, Egger's regression asymmetry test indicated the presence of publication bias (t = 6.45; P < 0.1) ( Figure 5D).

Discussion
HIV is one of the most influential infectious diseases in the world, and the spread of HIV virus through HIV serodiscordant couples is a transmission route that cannot be ignored. Risk of HIV transmission is determined by factors such as the possibility of exposure to the infected individuals and explicit sexual activity [24]. Some studies have even carefully described the specific mechanisms of HIV sexual transmission [25]. In this study, we analyzed a total of four factors related to HIV transmission among HIV serodiscordant couples: whether or not they received ART, occupation (farmers or nonfarmers), infection route of index case (intravenous drug abuse or heterosexual intercourse), and disease stage (HIV or AIDS). Among these factors, ART, occupation and infection route of index case had a significant relationship with the HIV seroconversion rate in HIV serodiscordant couples.
Our meta-analysis shows that the HIV seroconversion rate in couples who received ART was lower than in couples who did not receive ART. However, not all studies confirmed the effect of ART. One study showed that after 2 years of ART, 33% of the patients experienced virological treatment failure [26]. Factors affecting ART outcomes are increasingly being reported. Several studies included in our analysis explored the effects of ART interventions in the HIV infection and AIDS stages,but due to different definitions of disease staging, we could not conduct a meta-analysis of this outcome. A 8 large-scale clinical trial of ANRS 12249 in South Africa found that early initiation of ART had no effect on HIV incidence [27], which is contrary to existing theoretical knowledge. During the follow-up of patients receiving ART therapy, we found that they have further improved their awareness of AIDS and become more aware of the safety of sexual behavior, such as the reduction of sexual intercourse and the increase of condom use. Therefore, early and long-term acceptance of ART can not only reduce the viral load in patients, but also carry out health education in the process of intervention.
Although the effectiveness of ART needs to be subdivided and explored from different angles to achieve long-term suppression of HIV in patients who need lifetime ART, it is still considered necessary to promote ART therapy in general.
The disease stage of AIDS doesn't always mean the higher seroconversion rate than HIV infection.
According to general theoretical knowledge, patients with AIDS have fewer CD4 cells, more copies of the virus, and a greater probability of HIV transmission. However, the clinical symptoms of patients with AIDS could be more serious. Such couples likely use more measures to protect the uninfected partner from HIV infection, partially offsetting other adverse effects of AIDS. In addition, related studies have found that not all HIV-infected individuals are identified, especially patients with acute and early infections, which are the most infectious [26]. If one individual of a couple was infected with HIV and the symptoms had not yet manifested, the infectiousness of HIV may be underestimated.
Indeed, because HIV has an incubation period, patients vary in the length of time from diagnosis of HIV infection to development of AIDS. Evidence to describe the difference in risk of seroconversion between these periods could guide medical resources more effectively.
The HIV-infected population in rural central China is likely the largest known HIV-infected cohort in the world [28]. Due to the limitations of the medical environment and the challenges in identifying these cases, this subset of HIV serodiscordant couples may have an increased HIV seroconversion rate.
Meta-analysis of occupation also showed that HIV seroconversion rate was significantly higher in farmers. A less educated index case is more likely to pass the HIV virus to others due to a lack of awareness [29]. At present, rural regions that represent a major portion of HIV cases are important considerations for disease control. However, due to the low level of education and lack of health knowledge, farmers are resistant to HIV interventions and the vast majority of rural residents do not understand government policies and social health service resources. Our results highlight the need for an AIDS prevention model specifically targeted at farmers, with a focus on increasing awareness of and education about AIDS.
Analysis by infection route of the index case revealed that those who injected drugs were more likely to transmit the HIV virus than those having heterosexual intercourse. Further, the effect of ART in patients infected through heterosexual intercourse is better than in those infected by injection drug use [26]. Patients infected by drug injection have poor cognitive ability and treatment compliance, which may partially account for these results [30]. This suggests that drug injection has become a significant factor in seroconversion. Some people addicted to drugs choose to share syringes to meet their needs, despite knowing the risk of HIV infection. Spouses of injecting drug users may also be willing to take the risk despite not injecting drugs themselves. Relevant institutions should pursue interventions that reduce injection drug use and educate drug users on the risk of spreading HIV to their spouses.
Our results are novel; we have not found any other studies that explore factors that may influence HIV seroconversion rate. Further, our research content is relatively comprehensive and studies not only the effects of ART (the most popular topic of research), but also factors that are often overlooked, such as infection route of the index case, which is sometimes related to the HIV seroconversion rate. Other factors that should theoretically be related to HIV seroconversion rate, such as disease stage, may have no effect. In addition, our inclusion of external intervention factors and patients' own conditions make the results more accurate.
This study also had some limitations. We found large heterogeneity in ART studies. Subgroup analysis indicated heterogeneity may arise from the study area, which was over-dispersed. Sensitivity analysis of infection route of index case indicated no difference between the two routes when we excluded a heavily weighted article, indicating that this article likely influenced overall results. Therefore, more research is needed on the infection route of index case. Egger's regression asymmetry tests showed different degrees of publication bias in all results except ART. In addition, it was impossible to require all the included articles to be prospective studies because few articles examined factors affecting the HIV seroconversion rate. The varied research methods of included articles may also have complicated our results. Yet, after combining analysis of the existing literature with the test of heterogeneity and sensitivity, we consider these results persuasive.

Conclusions
This study explored four influencing factors related to HIV seroconversion rate in heterosexual discordant couples. We found positive results of ART, occupation and infection route and negative results of disease stage. We conclude that promoting early and long-term access to ART treatment,

Consent for publication:Not applicable.
Availability of data and material:The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Competing interests:The authors declare that they have no competing interests.