Abstract

Background

New indicators of potential human immunodeficiency virus (HIV) transmission are being actively explored. We aim to categorical testing of the viral load (VL) of persons living with HIV听(PLWH) in order to explore new indicators to measure the intensity of the epidemic and the effectiveness of the response in the community.

Methods

A dynamic cohort study was conducted in Yining to monitor the VL of all persons living with HIV from 2017 to 2019. Different population VL (PVL) surrogate indicators were measured and the strength of the associations of different PVL surrogates with HIV incidence, antiretroviral therapy (ART) coverage, virus unsuppression, and viremia prevalence was assessed. PVL surrogate indicators were used to describe the current status of HIV transmission in different populations and communities.

Results

All PVL indicators decreased from 2017 to 2019 (P鈥�<鈥�0.05). Arithmetic mean community viral load (CVL) (r鈥�=鈥�1.000, P鈥�=鈥�0.006) and geometric mean CVL (r鈥�=鈥�1.000, P鈥�=鈥�0.001) were positively associated with HIV incidence, ART coverage and viral unsuppression (P鈥�<鈥�0.05). CVL was higher in the male, 鈮も��25 years of age, primary school or below, other household registration, other medical insurance types, other source of sample, nonmarital and noncommercial heterosexual contact, and nonmarital and commercial heterosexual contact subgroups. Community-based cross-sectional analyses showed that CVL in community 10 was positively correlated with viral unsuppression rate and viremia prevalence but negatively correlated with ART coverage rate, suggesting that the community was a hotspot for HIV epidemics.

Conclusions

CVL can be used as an indicator for assessing HIV transmission and identifying high-risk populations and hotspot communities.

Introduction

The epidemiology and modes of transmission of human immunodeficiency virus (HIV) have changed dramatically over the past few decades, and interventions for preventing HIV transmission are constantly being explored and updated [1]. In recent years, acquired immunodeficiency syndrome (AIDS) prevention worldwide has been based on the Treatment as Prevention strategy, in which patients receive antiretroviral therapy (ART) at an early stage to control viral load (VL) and achieve viral suppression, thereby reducing the risk of HIV transmission [2,3,4]. Previous HIV prevention and treatment efforts have focused more on individual VL monitoring, and ART has been shown to reduce individual HIV VL and transmission [5]. Theoretically, higher ART coverage in the community should lead to lower total and average population VL (PVL), higher viral suppression rate, and lower HIV transmission in the community. HIV VL, which objectively reflects the level of HIV replication in the body, is a major factor affecting HIV transmission [6,7,8]. The incidence of HIV infections serves as a better indicator for the current epidemiological status of the disease and is crucial for monitoring HIV transmission and control. However, the incidence of infection is difficult and time-consuming to measure and is somewhat limited by the need for frequent testing of entire populations [9].

PVL or community viral load (CVL) has been proposed as an indicator for HIV incidence [10, 11]. Arithmetic mean or geometric mean PVL is an indicator for HIV transmission in a particular geographic region or population during a specific period of time [12, 13]. However, accurate measurement of PVL indicators is challenging due to undiagnosed cases, confirmed but untreated cases, cases with undetectable VL under treatment, and individuals who are unaware of their VL status. The US CDC guidelines for CVL [11], Solomon et al. [14] and Jiang et al. [15] have proposed 4 realistic and feasible surrogate indicators for PVL based on different levels of PVL monitoring, including monitored viral load (MVL), in-care viral load (ICVL), aware viral load (AVL), and CVL. CVL has been recognized in several studies as an important indicator for evaluating the risk of HIV transmission in a population or region [11, 14, 16]. Das et al. [17] showed that mean CVL was correlated with HIV incidence in San Francisco. A Colombian study [18] reported strong correlations among increased ART coverage, decreased PVL, and decreased number of new HIV cases. A study of injected drug users in San Francisco indicated that mean CVL was significantly associated with HIV incidence [19], validating the utility of CVL as a measure of HIV transmission. Furthermore, we have conducted a preliminary analysis focusing on the temporal aspect of the substitution effect of geometric CVL for PVL, and the initial findings were quite encouraging [20].

So far, there have been few studies in China that directly compare the VL monitoring levels among different populations and jointly analyze the strength of the association between PVL surrogate indicators and HIV transmission from both temporal and spatial perspectives. Xinjiang is one of the regions with the highest HIV incidence in China [21]. Therefore, it is particularly important to reduce the persistently high rate of HIV transmission in the population as early as possible, to implement targeted interventions for key populations, and to rationally allocate financial and human resources. The current state of knowledge regarding the HIV transmission potential in Yining City, Xinjiang is limited, and the present study aims to explore the association between PVL proxy indicators and HIV transmission potential through a cohort study. The objective is to identify new and convenient indicators, with the aim of better identifying the key HIV populations and communities in the region. This will help to set up a targeted HIV prevention and control programme and reduce the HIV transmission potential.

Methods

Subjects and test parameters

We established a prospective cohort to expand VL testing in Yining City, Xinjiang. The cohort started on January 1, 2017. Informed consenting persons living with HIV (PLWH) aged 13 years or older from the Yining City HIV outbreak pool were recruited as study participants, including both treated and untreated individuals, as well as those with newly reported HIV infection in the current year. The inclusion criteria were as follows: (a) HIV confirmatory test positive; (b) current address in Yining City, Xinjiang, China; (c) age鈥夆墺鈥�13 years; (d) informed consent. Exclusion criteria: those who did not meet the inclusion criteria. Following the recruitment of subjects and the acquisition of informed consent (for participants under 18 years of age, informed consent was required from both the subjects and their guardians), laboratory tests and questionnaires were administered. The questionnaires collected baseline information, HIV testing status, antiretroviral therapy (ART) status, previous immunological and virological history, behavioral data, and clinical information. Laboratory tests and questionnaires were administered annually and prospectively followed for three years, collecting data on newly reported HIV infections each year. The on-site follow-up concluded on 31 December 2019.

The cohort follow-up was based on routine HIV testing, with grassroots staff responsible for the accurate monitoring of HIV-infected patients. The baseline status of PLWH was categorized as follows: patients on treatment, who were sampled annually at sentinel hospitals for VL and CD4 testing; patients not on treatment, who were sampled annually at local Centers for Disease Control and Prevention (CDC) for CD4 testing; and newly reported HIV infections within the same surveillance and testing population, who underwent testing for new infections, as well as VL and CD4 testing following diagnosis. Follow-up visits and laboratory testing were conducted in accordance with the classification status of PLWH. Among these, newly reported HIV-infected individuals were classified as treated or untreated based on their engagement with ART. Treated individuals were followed up by the outpatient physicians they regularly consulted, while untreated individuals received regular follow-up from local CDC staff. All study participants completed a questionnaire, which was filled out by their follow-up physician or CDC staff, designed with reference to the national sentinel surveillance questionnaire. The outpatient doctors and local CDC staff were researchers at the study sites, with whom the group had established long-term collaborations. Prior to the commencement of the study, standardized training was organized centrally for the staff at each site, and appropriate compensation was provided to ensure high-quality completion of the follow-up work.

Laboratory tests

VL in blood samples was measured using the COBAS TaqMan HIV-1 test v2.0 HIV-1 kit (Roche) on an automated instrument (Roche COBAS AmpliPrep COBAS TaqMan48). HIV screening was performed using an HIV ELISA diagnostic kit (Invitrogen). Subjects who were positive at initial screening were retested using the HIV antibody ELISA diagnostic kit from Beijing Jinhao. All samples (including positive and negative samples) were stored for at least 12 months after the end of study.

Indicator definition and calculation

VL monitoring indicators

According to the CVL guidelines published by the US CDC [11] and findings of Solomon et al. [14] and Jiang et al. [15], VL surveillance indicators of different populations were defined as follows: (1) MVL: VL of subjects currently on treatment and surveillance; (2) ICVL: VL of subjects receiving therapy regardless of monitoring; (3) AVL: VL if subjects who were aware of their own VL status, regardless of treatment; (4) CVL: VL of subjects on treatment and those with confirmed infection but untreated.

Different VL metrics

The limit of VL detection in this study was 20 copies/mL, and all VL values below the limit of detection were documented as 10 copies/mL. The VL measures assessed were total VL, mean VL, and geometric mean VL. Median VL has been used as a VL index in previous studies [6, 11, 12] but was not selected for this study due to its value being lower than the limit of detection (10 copies/mL) in the population. (1) Total VL: Sum of VLs of all HIV/AIDS cases (copies/mL); (2) Mean VL: the sum of VL of all HIV/AIDS cases divided by the total number of HIV cases (copies/mL); (3) Geometric mean VL: VL of all HIV/AIDS cases were log transformed (base 10 logarithm), then summed and divided by the total number of HIV cases (log₁₀ copies/mL).

Indicator definitions

(1) Viral unsuppression was defined as VL鈥夆墺鈥�1000 copies/mL; Viral suppression was defined as VL鈥�<鈥�1000 copies/mL. (2) ART coverage was defined as the number of ART cases/total number of HIV cases 脳 100%. (3) Virus unsuppression rate was defined as the number of virus unsuppression/total number of HIV cases 脳 100% ; Viral suppression rate was defined as the number of VL suppression/total number of HIV cases 脳 100%. (4) Viremia prevalence was defined as the number of cases with VL鈥�>鈥�20 copies/mL/total number of HIV cases 脳 100%. (5) We estimated the rate of new HIV infections in the entire population of Yining City by utilizing annual data on testing for recent infections. To estimate HIV incidence, we employed the HIV incidence estimation formula proposed by McWalter and Welte (2010) [22,23,24,25,26]. The HIV incidence rate was calculated as: $Ir = \frac{{R - \varepsilon P}}{{(1 - \varepsilon )\omega N}}$, N鈥�=鈥塶umber of HIV-negative subjects; P鈥�=鈥塶umber of PLWH; R鈥�=鈥塶umber of new infections; 蠅鈥�=鈥塎ean time to new infection; 蔚鈥�=鈥塮alse reject rate for new infections. HIV incidence in this study was estimated based on the values (蠅鈥�=鈥�130d, 蔚鈥�=鈥�2.3%) provided by the ELISA kit (Jinhao, Beijing).

Statistical analysis

Time-based analysis

Sensitivity analysis: There were 55 subjects with missing VL values, which was lower than the maximum allowable limit (25%). Missing values were imputed using the Markov Chain Monte Carlo (MCMC) [27] algorithm in SPSS 22.0, and indicators with missing VL value were analyzed before and after imputation. In addition, sensitivity analysis was performed on the distribution of subjects with and without missing VL (Table S1). Measurement data were compared between groups using the Wilcoxon rank sum test in SPSS 22.0. A P鈥�<鈥�0.05 was considered statistically significant. Graphs were plotted using GraphPad Prism 8.3.0.

Geo-community-based analysis

Yining has 9 townships, 8 sub-district offices, and 4 districts (1 economic zone and 3 field districts), totaling 21 community units (collectively referred to as communities). Considering that the 2018 data had the lowest number of missing VL and were obtained from a large cohort, the cross-sectional analysis was performed on the 2018 data. One of the communities did not have an existing HIV infection, and hence only 20 communities (communities 1 to 20) were included in this analysis. PVL-related indicators were selected as the best indicators in the longitudinal-based analysis.

Data from the 20 communities were collated and analyzed, and the lollipop chart were generated using R4.3.0. Correlation analysis was performed using the Pearson鈥檚 correlation coefficient for data with normal distribution and the Spearman鈥檚 rank correlation coefficient for data without normal distribution. Association among viral unsuppression rate, viremia prevalence, untreated rate and CVL was determined using linear regression. The adjusted R² represented the proportion of variance explained by the covariate and was used to evaluate the advantages and disadvantages of the model. Model fitting was done using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) [28]. A P鈥�<鈥�0.05 was considered statistically significant.

Results

Demographic characteristics of PLWH in Yining between 2017 and 2019

A total of 3920, 4167, 4080 subjects were monitored in 2017, 2018, and 2019, respectively. There was no change in the distribution trends of sex and routes of transmission between 2017 and 2019 (P鈥�>鈥�0.05). HIV infection was more prevalent in males, and the routes of transmission were mainly injected drugs, non-marital and non-commercial heterosexual contact, and Positive for spouse/fixed partner. Age, education level, household registration and medical insurance type of the surveyed subjects were significantly different across the 3 years (P鈥�<鈥�0.05). Our data showed that HIV infection was most common at the age of 36鈥�45 years, followed by 46鈥�55 years, which indicates that HIV infection is becoming increasingly prevalent at an advanced age. The education level of the subjects was mainly primary school or below, and most subjects had a household registration in urban areas. The main types of medical insurance were social insurance and new rural cooperative medical insurance (Table 1).

Table 1 Distribution of general demographic characteristics of PLWH in Yining from 2017 to 2019

Comparison of different PVL monitoring indicators for 2017鈥�2019

Total VL, mean VL, and geometric mean VL were calculated separately for MVL, ICVL, AVL, and CVL. There was a significant annual decrease in different VL values (total VL, mean VL, and geometric mean VL) for each PVL indicator (P鈥�<鈥�0.05). In particular, CVL was higher than MVL, ICVL, and AVL in all 3 years (Table 2).

Since there were missing values for AVL and CVL, sensitivity analyses were performed to determine the robustness of the results before and after imputation. Our results showed slightly higher values after imputation, but the difference was not statistically significant (Table S2).

Table 2 Comparison of VL monitoring indicators for different populations from 2017 to 2019

Correlation between different PVL monitoring indicators and HIV incidence

The incidence of HIV was 0.002685% in 2017, 0.002252% in 2018, and 0.001246% in 2019, showing a trend of annual decrease. MVL, ICVL, AVL, and CVL were decreased from 2017 to 2019, but only the decrease in CVL was correlated with the decrease in the incidence of HIV. Only mean CVL (r鈥�=鈥�1.000, P鈥�=鈥�0.006) and geometric mean CVL (r鈥�=鈥�1.000, P鈥�=鈥�0.001) were positively correlated with HIV incidence (Fig. 1). Sensitivity analyses of imputed missing VL values yielded the same results as above for mean CVL* (r鈥�=鈥�1.000, P鈥�=鈥�0.003) and geometric mean CVL* (r鈥�=鈥�0.999, P鈥�=鈥�0.028) (Figure S1).

Fig. 1

Trends in different PVL indicators and HIV incidence between 2017 and 2019

Correlation of different PVL indicators and ART coverage, viral unsuppression rate, and viremia prevalence

ART coverage increased from 74.01 to 87.04% (P鈥�<鈥�0.001), viral unsuppression rate decreased from 35.56 to 23.58% (P鈥�<鈥�0.001), and viremia prevalence decreased from 47.35 to 23.58% (P鈥�<鈥�0.001) in the overall population from 2017 to 2019. These results indicated an overall better HIV status in the city. Sensitivity analyses showed that the PVL indicators changed after imputation of missing values, with ART coverage increasing from 74.06 to 87.11% (P鈥�<鈥�0.001), viral unsuppression rate decreasing from 35.61 to 23.63% (P鈥�<鈥�0.001), and viremia prevalence decreasing from 47.60 to 36.89% (P鈥�&濒迟;鈥�0.001).

Correlation analysis revealed a significant association of HIV incidence with ART coverage and viral unsuppression rate (P鈥�<鈥�0.05). In addition, arithmetic mean and geometric mean CVL were significantly associated with ART coverage and viral unsuppression rate (P鈥�<鈥�0.05) (Table 3). Sensitivity analysis indicated that the results were robust (Table S3). Taken together, arithmetic mean CVL and geometric mean CVL are better surrogates for PVL for describing the overall HIV transmission in the region.

Table 3 Correlation of PVL indicators with ART coverage, viral unsuppression rate and viremia prevalence

Comparison of CVL between different treatment, viral suppression, and viremia statuses

CVL was significantly different between treatment statuses, viral suppression statuses, and viremia statuses (P鈥�<鈥�0.05). In addition, CVL was significantly different across the 3 years in all groups except in the viral suppression group (P鈥�<鈥�0.05) (Table 4). Sensitivity analysis indicated that the results were stable (P鈥�>鈥�0.05) (Figure S2 and Table S4).

Table 4 Comparison of CVL between ART, viral suppression and viremia statuses

Comparison of CVL in the general population between 2017 and 2019

We selected the pre-imputation data for this analysis as there was no significant difference in population distribution before and after imputation for missing VL values. A comparison of CVL values showed significant decrease in sex, age, education level, and medical insurance type (P鈥�<鈥�0.05) in 2019 compared with 2017. Furthermore, CVL was significantly different among subgroups of each demographic characteristic (P鈥�<鈥�0.001) except for medical insurance types (P鈥�>鈥�0.05). In particular, CVL was higher in men, 鈮� 25 years old, primary school or below, other household registration place, other medical insurance types, other sample sources, nonmarital and noncommercial heterosexual contacts, and nonmarital and commercial heterosexual contacts (Table 5).

Table 5 Comparison of CVL in the general population between 2017 and 2019

Distribution of CVL, ART coverage, viral unsuppression rate, and viremia prevalence by community in Yining

Cross-sectional analysis of the 20 communities in 2018 showed that the CVL, ART coverage, viral unsuppression rate and viremia prevalence ranged from 1.64 to 2.49 log₁₀ copies/mL (median 2.18 log₁₀ copies/mL), 68.21鈥�89.47% (median 78.51%), 15.79鈥�39.74% (median 30.35%), and 21.05鈥�56.90% (median 44.53%), respectively. Community 10 had the highest CVL and viral unsuppression rate, lowest ART coverage, and second lowest viremia prevalence after community 6 (Fig. 2). On the other hand, community 18 had the lowest CVL, viral unsuppression rate and viremia prevalence as well as the highest ART coverage. There was a positive correlation among CVL, viral unsuppression rate and viremia prevalence, and a negative correlation between ART coverage and the other three indicators (Figure S3).

Fig. 2

Distribution of CVL, ART coverage, viral unsuppression rate, and viremia prevalence in various communities of Yining

Linear regression of CVL with viral unsuppression rate, viremia prevalence, and ART coverage

Linear correlation analysis and regression estimation showed that viral unsuppression rate had the strongest correlation with CVL (R² adj: 0.982, AIC/BIC: -80.91/-77.92), with each one-unit increase associated with a 0.036 log₁₀ copies/mL increase in CVL (95% CI: 0.033鈥�鈭�鈥�0.038), followed by viremia prevalence and ART coverage in the 20 communities (Table 6).

Table 6 Linear regression analysis of CVL with ART coverage, viral unsuppression rate and viremia prevalence

Discussion

Our study found that PVL indicators decreased annually in Xinjiang, which were consistent with findings reported in other countries [29, 30]. Although the PVL surrogate indicators MVL, ICVL, AVL, CVL, including their total VL, mean VL, and geometric mean VL, showed the same trend of change in HIV incidence, only CVL was positively associated with HIV incidence. Furthermore, we found that HIV incidence was significantly correlated with ART coverage, viral unsuppression rate, while arithmetic mean CVL, geometric mean CVL, and mean ICVL were strongly correlated with ART coverage and viral unsuppression rate.

Arithmetic mean CVL and geometric mean CVL are better PVL surrogate indicators for measuring HIV transmission and burden in the Xinjiang population, which is in line with findings from India [14], South Carolina [30], and Rhode Island [31], and confirms the hypothesis that the increase in ART coverage is linked to decreased viral unsuppressio rate, CVL and HIV incidence. However, Solomon et al. [14] reported that other PVL indicators were also correlated with HIV incidence, albeit the values were lower than CVL. This discrepancy may be related to the included population size for different PVL indicators in this study. In our study, the MVL population largely coincided with the ICVL population, which consisted of subjects with suppressed or undetectable VL and slow VL decline over the 3-year period. In previous studies [14, 18, 32], MVL and ICVL have been used to monitor VL in the on-treatment population, the easy observation of personnel and availability of data allowed the use of MVL and ICVL as surrogate PVL indicators to monitor HIV transmission potential in most studies [11]. However, both indicators are not applicable to the confirmed but untreated cohort, which consists of individuals with high VL [33], thus the PVL is prone to underestimation. Although AVL covers a portion of the untreated population on the basis of MVL and ICVL surveillance populations, population coverage is still inadequate as not all diagnosed persons are aware of their VL. We found that CVL was higher than MVL, ICVL, and AVL in 2017鈥�2019. CVL is a more comprehensive and representative estimate of PVL, encompassing all populations in whom PVL can be monitored (those who are undiagnosed cannot be monitored), including untreated individuals. This is in agreement with the US CDC guidelines for CVL [11] and the studies by Rozhnova et al. [34] and Farahani et al. [35]. Therefore, we believe that CVL is a better indicator of monitoring HIV transmission and burden in the population compared with other PVL surrogate indicators.

Enhanced VL surveillance, along with comprehensive utilization of VL data, facilitates the transition from monitoring individual VL to PVL, which transforms PVL from a mere observational parameter of the epidemic into an actionable indicator for improving HIV disease burden [33]. Based on the above results, we selected arithmetic mean CVL and geometric mean CVL to quantify HIV burden in Yining in order to identify the HIV hotspot populations. We found that CVL was relatively high in the male, 鈮� 25 years old, primary school or below, other household registration place, other medical insurance types, other sample sources, non-married and commercial heterosexual contacts, and non-married commercial heterosexual contacts subgroups. Targeted interventions should be implemented with respect to individuals with these characteristics, with the aim of effectively lowering HIV burden in the population.

Viremia prevalence adjusted for the number of HIV-negative individuals was proposed by Solomon et al. [14] as a better indicator for HIV transmission. However, we only found an association between total CVL and viremia prevalence but not HIV prevalence, ART coverage or virus unsuppression rate, suggesting that caution should be taken when using this indicator for monitoring HIV burden in communities in Xinjiang. In addition, it is important to note that the VL threshold used to define viremia differed among studies, such as 20 copies/mL in our study compared to 150 copies/mL in the study by Patel et al. [36]. Sensitivity analysis should be performed in future studies to determine the impact of different VL thresholds on the results within the same sample.

To analyse the burden of HIV disease and the effectiveness of HIV prevention and treatment in each community based on geographic communities. By observing the distribution of HIV-related indicators in each community, we can intuitively identify 鈥渉otspot communities鈥� that require more attention, accurately locate the high CVL community, and measure the HIV epidemic status in the city. HIV hotspot communities are important for large-scale intervention in the future [37] as they can be used to predict the risk of infection in HIV-negative individuals and the potential of sustained HIV transmission [38, 39]. This study supports the use of CVL for the geographical identification of high VL populations and hotspot communities as well as underscores the necessity of expanding HIV testing within these communities and populations in order to discover new infections and facilitate effective epidemic monitoring. Integrating geospatial analysis into routine public health planning will enable interventions to be targeted at specific geographical areas, maximizing epidemiological impact and ensuring efficient resource allocation. This approach also allows for a more nuanced assessment of the effectiveness of HIV treatments.

Several limitations should be noted in this study. First, HIV transmission was analyzed without taking into account the impact of HIV prevalence and behavior of the population. For example, in different populations with the same CVL, those with higher HIV prevalence and incidence of high-risk behaviors are likely to have higher HIV transmission [13]. Based on this, we analyzed the CVL of different characteristic populations. Second, the duration of our longitudinal analysis was relatively short and the change in viremia prevalence was small. Therefore, future CVL studies with extended follow-up time and attention to HIV prevalence and high-risk behaviors will be better suited for analyzing the HIV epidemic and assessing the effectiveness of prevention and treatment efforts at the population level. Additionally, we were unable to produce a visualization map for each indicator due to the irregular shape of community distributions and the presence of individual communities composed of multiple, non-contiguous regions. This complexity resulted in suboptimal visualizations. To address this limitation, we opted to represent the higher levels of each indicator using a lollipop chart.

Taken together, our findings suggest that CVL is an effective indicator for assessing the risk of HIV transmission and reflects the potential for HIV transmission in the population. It can be used to dynamically monitor the HIV epidemic and assess the effectiveness of HIV prevention and treatment measures. Currently, the CVL in Yining has decreased annually, affirming the effectiveness of the current prevention and control efforts.

Conclusions

CVL can be used as an indicator for evaluating HIV transmission and the effectiveness of HIV interventions in a population. Efforts should be dedicated to the development of targeted interventions for key populations and hotspot communities to reduce HIV transmission and burden.