The HIV epidemic in Yunnan first appeared in 1989 with the discovery of 146 infected injecting drug users (IDUs) in Ruili, a town bordering Myanmar and situated along the drug trafficking routes channeling heroin into China from Southeast Asia's opium-producing “Golden Triangle” region.1 Although this region has a long history of opium use, the advent of using injection needles as a common form of drug administration in the late 1980s created an efficient transmission mode that spread HIV among IDUs. Surveillance data shows HIV had appeared in geographically disparate groups of IDUs beginning around 1995 and has since spread with increasing momentum. Within several years, 8 counties were reporting HIV prevalence rates of more than 40%. By 1999, all 16 prefectures of Yunnan had reported HIV infections, and despite making up only 3% of the national population, Yunnan's reported cases of HIV accounted for 34.8% of total HIV cases in China.2 Moreover, prevalence rates among some high-risk groups have already reached crisis proportions in certain areas; rates as high as 74.5% were reported among IDUs2 and was approximately 10% among female sex workers (FSWs).3-4
In 1992, the provincial government began using a sentinel surveillance system to track the epidemic. As a result, Yunnan's HIV epidemic is one of the most consistently and thoroughly tracked in China, providing valuable insight into long-term trends and characteristics.4-6 Additionally, many of China's earliest pilot studies testing intervention methods occurred in Yunnan.7-9
Yunnan has been described as a “key HIV epicenter in China,”10 and an in-depth analysis of its long-term epidemiological trends provides a lens of examining how HIV is spread in Chinese settings. Yunnan can also act as a regional case study to which other regions can compare their own epidemiological trends and use Yunnan's experience to inform their choice of intervention methods.
Two types of data were used in this article: sentinel surveillance data and reported cases of HIV infection via multiple strategies of HIV testing.
In 1992, a standardized HIV sentinel surveillance system was established to consolidate information at the provincial level.10 By 2007, 89 sentinel sites were set up conducting regular surveillance among IDUs (18 sites; sample size: 100-200 per site), FSWs (33 community-based sites and 1 women re-education center site; sample size: 400 per site), male clients of FSWs (1 male re-education center site; sample size: 300-400 per site), sexually transmitted disease (STD) clinic attendees (12 sites; sample size: 300-400 per site), men who have sex with men (MSM) (one site; sample size: 300), male migrant workers (4 sites; sample size: 400 per site), pregnant women (18 sites; sample size: 800 per site), and tuberculosis (TB) patients (1 site; sample size: 800). Subjects of every category provided a 2-3 mL of blood sample for HIV antibody testing and participated in an anonymous interview to collect information on demographic details and HIV transmission-related risk behaviors.
Other Forms of HIV Case Detection
Testing in Medical Settings
All positive HIV cases found in any medical setting were reported directly to the provincial Centers for Disease Control and Prevention (CDC). Tests were conducted under any of the following conditions: presurgical screening, antenatal screening for pregnant women, physician-recommended testing for individuals exhibiting AIDS-related symptoms, the screening of blood donors, or as part of standard medical exams.
Testing in Other Settings
Voluntary testing and counseling (VCT) in China is offered by the national, provincial, prefectural, and county-level CDCs. CDC-certified hospitals and specialized medical centers also provided VCT services. VCT is also recommended for couples applying for marriage certificates.
In 2004, the national Ministries of Health, Justice, and Public Security passed a joint resolution requiring HIV testing of all individuals admitted to a detention center for the first time. By definition, “detention centers” include prisons, drug detoxification centers, re-education through labor camps, and detention centers for commercial sex offenders (commercial sex workers and clients) and for other types of offenses.
Key Changes in Testing Strategies
The introduction of new testing strategies in the mid-2000s resulted in dramatic changes to reported HIV cases as the number of people tested increased significantly, explaining the sharp increase in the number of reported HIV/AIDS cases beginning during that period (Fig. 1). The initiation of testing for detainees increased the number of tested individuals but inevitably also sampled groups particularly vulnerable to HIV infection, resulting in an upward bias in the reported number of HIV/AIDS cases. In addition, timeliness of case reporting improved in 2005, dictating all detected cases from various health facilities be directly reported to the National CDC within 24 hours, in contrast to previously aggregating HIV cases on a quarterly basis.11 Finally, since 2005, all Yunnan couples who registered for marriage were recommended premarital screening.
HIV Testing Laboratory Procedures
In 1992, the China State Food and Drug Administration approved the use of enzyme-linked immunosorbent assay-testing method, confirmed by Western Blot analysis.12 Rapid testing has also since become available for testing populations who benefit the most from fast notifications of test results, such as pregnant women for the prevention of mother-to-child-transmission, and for mobile populations who are less likely to return for their test results (eg, FSWs, STD clinic patients, migrant workers, and MSM).
In 1995, Yunnan began using alternative HIV testing methods (without using Western Blot for confirmation) recommended by the Joint United Nations Program on HIV/AIDS and the World Health Organization for developing countries.13-15 Alternative testing strategies are most commonly used for testing IDUs in Yunnan.
All statistical analysis for this report was conducted using the SPSS 12.0 statistical analysis software package (SPSS Inc. Chicago, IL, USA). Comparisons of characteristics between groups were conducted using χ2 tests and results with P values of 0.05 or less were considered statistically significant. Cumulative results from the first year of available sentinel surveillance data were used for systematic comparisons. Surveillance sites with the longest operating and continuous data available were included.
By the end of 2007, Yunnan had recorded 57,325 cases of HIV infection (including 2077 foreign nationals) among which 7630 had progressed to AIDS. A cumulative total of 4525 deaths among HIV/AIDS-infected individuals was recorded, of whom, 1949 died of AIDS-related causes. Most reported HIV infections (29,458 cases, 51%) were identified in medical setting, followed by VCT (10,087 cases, 18%) (Table 1).
Trends Reflected in Reported Cases of HIV
Data collected from nonsurveillance settings between 1989 and 2007 revealed the following trends:
* The severity of the epidemic varies widely across regions, with reported cases in the prefectures of Dehong, Honghe, Dali, Lincang, Wenshan, and the municipality of Kunming, making up 78.2% of all reported cases in the province by the year 2007 (Fig. 2).
* The proportion of women among people living with HIV/AIDS (PLWHA) is increasing. The ratio of men to women has fallen from 40:1 in 1989 to 1.7:1 in 2007 (Fig. 3).
* The age distribution of PLWHA individuals has gradually increased over time. The proportion of 20-year to 29-year olds has fallen from 52.4% in 1989 to 36.6% in 2007; whereas the proportion of 30-year to 39-year olds has increased from 28.8% in 1989 to 43.2% in 2007.
* HIV transmitted through needle-sharing behaviors has drastically decreased from 100% in 1989 to 42.5% in 2007. Sexual contact and mother-to-child-transmission transmission now account for 47.4% and 1.3% of total infections, respectively.
* In 1989, the share of PLWHA among rural residents was 93.8%, among unemployed persons was 4.1%, and among persons with other forms of employment was 2.1%. By 2007, this had changed to 46.4%, 24.3%, and 29.3%, respectively.
* The ethnic background among PLWHA has also changed since 1989, from 75.3% Dai, 16.4% Jingpo, 5.7% Han, and 2.3% of other ethnicities, to 10.5%, 5.7%, 63.2%, and 20.6%, respectively in 2005.
Trends in Sentinel Surveillance Data
Between 1992 and 2007, a total of 268,978 subjects were tested for HIV in sentinel surveillance program. The number of subjects tested in each subgroups breakdowns as follows: 31,138 IDUs; 34,049 FSWs; 4944 male migrants; 4501 male clients of FSWs; 68,038 STD clinic patients; 123,036 pregnant women; 2963 TB patients; and 309 MSM. The prevalence of HIV infection change over time among IDUs, FSWs, male clients of STD clinics, and pregnant women, as given in Figure 4.
Trends Among IDUs
In 1989, IDU infections accounted for 70% of total HIV cases. The provincial average prevalence rate among IDUs increased from 1992 to 2004 (2.7% in 1992, 15.0% in 1995, 30.0 % in 1999, 32.4% in 2004) and decreased to 28.4% in 2007. In 7 counties (Yingjiang, Longchuan, Luxi, Kaiyuan, Dali, Lincang, and Yanshan), HIV prevalence rates among IDUs has surpassed 40%.
Characteristics of HIV infection among IDUs could be summarized as follows. First, HIV prevalence rates among male IDUs' have remained significantly higher than those of female IDUs (22% vs. 15%; χ2 = 52.83, P < 0.01). Second, there is a recent downward trend in HIV prevalence among IDUs who were 24 years or younger. Third, HIV prevalence among IDUs with only primary schooling or less were significantly higher than those who had completed secondary schooling or more (22.7% vs. 15.9%; χ2 = 59.7, P < 0.01). Finally, HIV prevalence among IDUs was higher for ethnic minorities than for the Han ethnic group, even after controlling for different levels of education across ethnicities.
Trends Among FSWs
FSW surveillance data from the Kunming detention center and the community-based testing sites were analyzed separately. Until 1994, FSWs in the Kunming detention center had a 0% HIV prevalence rate, but this rate steadily increased over the years: 0.5% in 1995, 1.5% in 1997, 2.9% in 2006, and 4.0% in 2007. For FSWs tested in community sites, average provincial prevalence rates in 2005, 2006, and 2007 were 1.5% (range: 0%-4.5%), 1.5% (range: 0%-3.7%) and 1.9% (range: 0%-5.3%), respectively. In 2007, 14 additional surveillance sites were established across 10 prefectures to improve monitoring HIV epidemic among this group. Regional diversity characterized infections in this group; sites in Honghe, Dehong, and Lingchang prefectures have all recorded HIV prevalence rates >5%, whereas Zhaotong (Zhaoyang District) has not found any cases of HIV infection for the 3 consecutive years, 2005-2007.
Drug use (DU) had a profound impact on HIV among FSWs; DU FSWs had HIV prevalence rates as high as 35.5% compared with non-DU FSWs (1.9%, χ2 = 897.4, P < 0.01). Further analysis among non-DU FSWs found that those with primary schooling or less had significantly higher HIV prevalence than those with secondary schooling or more (2.7% vs. 1.6%; χ2 = 14.0, P < 0.01) or as compared with DU FSWs with high school education or more (1.5%; χ2 = 7.1, P < 0.01). Similarly, average HIV prevalence rates among non-Han FSWs are higher than Han FSWs (2.3% vs. 1.7%; χ2 = 5.3, P < 0.05); ethnic minority groups rank in the following descending order of prevalence rates: De'ang 20.6% (χ2 = 135.7, P < 0.01), Jingpo 5.9% (χ2 = 8.86, P < 0.01), Wa 4.2% (χ2 = 4.4, P < 0.05), Miao, 3.8% (χ2 = 2.9, P < 0.05), Zhuang: 3.1% (χ2 = 1.2, P > 0.10), and Hani 2.8% (χ2 = 2.29, P > 0.10). HIV prevalence rates among local FSWs (registered residents of Yunnan) were significantly higher than those from other provinces (2.2% vs. 0.9%; χ2 = 23.1, P < 0.01). Street-based FSWs were also found to have higher HIV prevalence rates than establishment-based FSWs (3.7% vs. 1.7%; χ2 = 23.6, P < 0.01).
Trends Among MSM
HIV was first detected in MSM in 1999; recent surveillance in Kunming recorded prevalence rates of 4.0% in 2005 and 13.2% in 2007.
Trends Among Male Clients of FSWs
Surveillance sites targeting male clients of FSWs in Kunming have recorded the following prevalence rates: 0%, between 1995 and 1997; 0.3% in 1998; 0.3%-1.8% between 1999 and 2004; and 0.8 %-1.8% between 2005 and 2007.
Trends Among Male STD Clinic Attendees
The average HIV prevalence rate among male STD clinic attendees were: 0% in 1992; 0%-0.6% between 1993 and1995; 1.2%-2.3% between 1996 and 2000; 1.7%-3.1% between 2001 and 2004; and 2.0%-2.1% between 2005 and 2007. Non-Han ethnicities have higher prevalence rates than Han (3.1% vs. 1.9%; P < 0.01), among whom Jingpo and Dai had the highest rates at 6.3% (χ2 = 10.2, P < 0.01) and 5.1% (χ2 = 50.6, P < 0.01), respectively.
Trends Among the Male Migrant Population
Beginning in 2005, 4 HIV sentinel surveillance sites among male migrant workers were established, and they monitored workers on a road project, at a hydropower plant, at a coal mine, and at a building construction site. In 2005, 2006, and 2007, the average HIV prevalence rates among these 4 groups were 0.55% (range: 0%-1.8%), 0%, and 0.13% (range: 0%-0.25%), respectively. Of these migrants, non-Han had a higher HIV prevalence than Han (0.45% vs. 0.12%, χ2 = 4.8, P < 0.05).
Trends Among Pregnant Women
The provincial average HIV prevalence rate among pregnant women was: 0.16% in 1992; 0.14%-0.25% between 1993 and 2002; 0.37% in 2003; 0.38% in 2004; 0.39% in 2005; 0.43% in 2006; and 0.50% in 2007. In the counties of Ruili, Yingjiang, Longchuan, Kaiyuan, Gejiu, Linxiang, Dali, and Wenshan, HIV prevalence in antenatal clinics has surpassed 1%. Non-Han pregnant women have a higher HIV prevalence than Han (0.33% vs. 0.24%, χ2 = 157.8, P < 0.01).
Trends in Other Populations
HIV surveillance testing for TB patients began in 2002, with the following prevalence rates: 1.6% in 2002; 1.5% in 2003; 0.8% in 2004; 1.2% in 2005; 1.6% in 2006; and 0.7% in 2007.
Screening of potential blood donors began in Kunming in 1992 and recorded HIV prevalence rates since are: 0%-0.0075% between 1992 and 2000; 0.015% in 2001; 0.029% in 2002; 0.056% in 2003; 0.072% in 2004; 0.127% in 2005; 0.098% in 2006; and 0.084% in 2007.
More cases of HIV infection have been reported from Yunnan than any other province in China, showing both the severity of the epidemic and the strength of the long established HIV surveillance system. The defining characteristics of the Yunnan HIV epidemic are the key role played by injecting drug use (IDU) and the overall makeup of the infected population: young, male, rural, and unemployed. Wide regional variations in the severity of the epidemic correlate strongly with local IDU habits; 78.2% of cumulative reported cases have come from the prefectures of Dehong, Honghe, Lincang, Dali, Wenshan, and the city of Kunming, where heroin use is high. Several conclusions can be drawn from the trends detected in surveillance data and in the changing profile of the HIV-infected populations, as listed below.
Interpreting the Trends: Older IDUs
HIV prevalence is increasing among older IDUs and falling among younger IDUs. This may be due to lower IDU initiation by younger people or at least safer injection behaviors practiced by younger IDUs. Alternatively, testing rates among older IDUs at greater risk of infection may have increased with the introduction of a nationally subsidized antiretroviral therapy treatment program in 2004, thus inflating the number of infected IDUs from previously undetected infections. Furthermore, the mass roll out of treatment services has kept older, infected IDUs alive longer, resulting in an increase in the average age of infected IDUs.
Interpreting the Trends: Growth of a Sexual Epidemic
HIV surveillance data suggest sexual transmission through high-risk behaviors, particularly commercial sex, began in the mid-1990s. This trend had the most pronounced impact on FSWs and their male clients, MSM, and sexual partners of IDUs. All sexual epidemics originated in areas with prevalent IDU, suggesting an important link between the IDUs and sexual epidemics. A growing body of research supports a conceptual framework of a threshold prevalence rate among IDUs having a “seeding effect” in the development of larger sexual epidemics.16,17 In Yunnan, there is strong evidence that male IDUs who visit FSWs and FSWs who inject drugs have played a crucial role in the transition of the epidemic from being primarily IDU-driven to sexually driven.18
Two vulnerable groups of particular concern in the sexual epidemic are street-level FSWs and MSM. These groups are not only at risk for both primary and secondary infections to their partners but are also difficult to access because of their underground behaviors and mobile lifestyles. Future prevention efforts must identify new ways of interacting with these groups to better analyze their behaviors and develop measures for effective behavioral interventions.
Interpreting the Trends: Spread to General Population
Increasing prevalence among blood donors and pregnant women indicate that HIV has begun to spread among the general low-risk population. HIV prevalence rates among these groups in Yunnan are above the national average and are increasing annually. In 8 counties, prevalence rates reported by antenatal clinics have already reached or passed 1%.
Interpreting the Trends: Impact on Ethnic Minorities
The HIV epidemic in Yunnan has disproportionately impacted ethnic minorities. Surveillance data show that among IDUs, the Jingpo, Dai, and Yi minorities have the highest prevalence rates; that among FSWs, the De'ang, Jingpo, Zhuang, Hani, Miao, Yi, and Zhuang ethnicities have high prevalence, and that among male STD patients, the Jingpo and Dai groups have the highest prevalence rates. HIV prevalence rates of ethnic minorities among pregnant women and male migrants are also high. The significant impact on these populations is thought to be associated with these communities' long exposure to heroin through their proximity to the Golden Triangle region. This exposure, coupled with limited access to health care and lower public health awareness, has fueled the spread of HIV within these subpopulations.
LIMITATIONS AND CHALLENGES
This study faces several data limitations. First, analysis of trends in reported cases of HIV is useful for understanding the changing profile of HIV-infected individuals but is limited in its ability to capture changes in the actual infected population. The representativeness of the tested population is affected by policy changes that either encouraged voluntary testing, such as in 2004, when free VCT services and nationally subsidized ARV treatment became available, or required compulsory testing of previously undocumented populations (who are often more vulnerable), as in 2005 when China began testing persons admitted to incarceration centers. Sampling bias can be reduced by introducing policy changes across the administrative area, but it is impossible to remove the time impact which must at least be acknowledged in any analysis of case reporting data.
Sentinel surveillance data must also acknowledge biases introduced though sampling methods and assumptions made about the generalizability of certain groups. For instance, the earliest sentinel surveillance stations often oversampled groups from high prevalence areas, which in turn would overestimate actual prevalence rates. Furthermore, the behaviors of populations thought to represent the general low-risk populations, such as blood donors, STD clinic patients, and pregnant women, may have systematic differences from the general population. For example, STD patients and pregnant women may have come from a sector of the population with above average rates of unprotected sex, or people with a history of blood donation may have had lower socioeconomic backgrounds, less public health awareness and/or health-seeking practices than the general population.
Future efforts in the prevention of HIV in Yunnan must focus on several key areas. First, the epidemic is still spreading among IDUs and prevention efforts must continue to target this group to prevent primary and secondary infections among them and their sexual and needle-sharing partners. Second, the association between HIV infection, poverty, and low levels of education indicates the need for strengthening public health awareness among underserved communities and also highlights the relationship between economic/social development and long-term public health goals. Third, there is a critical need for a better understanding of the characteristics and behaviors of street-level FSWs and MSM who are particularly vulnerable to getting infected and are also difficult to be reached. Last but not least, proven intervention programs that reduce risk behaviors and the transmission of HIV must be scaled up to improve coverage rates. Such interventions may include harm reduction or condom use programs and those that address multiple risk behaviors or that utilize a comprehensive approach to controlling HIV through programs that link prevention and treatment efforts.
The authors thank Naomi Juniper and Adrian Liau for their editing assistance.