Burchell, Ann N. MSc*†; Calzavara, Liviana M. PhD*; Orekhovsky, Victoria MEd*; Ladnaya, Natalia N. PhD‡; for the Canada AIDS Russia Project and the Russian HIV Response Network
THE RUSSIAN FEDERATION IS CURRENTLY experiencing one of the fastest growing HIV epidemics worldwide.1 The annual number of reported HIV cases increased nearly 100-fold from 3623 cases on March 31, 1997 to 318,394 in May 2005.2 The true number infected is likely to be far greater, and was estimated to be between 560,000 and 1.6 million by the end of 2005.3 The HIV prevalence rate among adults aged 15 to 49 is now believed to be 1.1%.3
To date, the majority of cases have been diagnosed among injection drug users (IDU). This was most pronounced in 2001, when 93% of reported HIV cases were in that exposure category (Natalia Ladnaya, Russian Federal AIDS Center, personal communication, 2004). HIV surveillance data suggests that HIV transmission may be shifting to the noninjecting population. The proportion of cases reporting heterosexual contact as their only exposure increased from 6% in 2001 to 25% in 2004.1
Several conditions may have facilitated a heterosexual HIV epidemic in the Russian Federation. The political, social, and economic changes since the break-up of the Soviet Union have resulted in reduced life expectancy,4,5 increased unemployment,6 and a growing criminal economy including drugs and the sex trade.4,7 Injection drug use has increased and the epidemic is now firmly established in networks of injectors. In 2004, there were over 340,000 Russians who were registered as IDU; the true number is likely to be far greater.1 Epidemiologic surveillance suggests that by 2004 HIV prevalence was approximately 50% among IDU in the 5 cities with the highest rate of transmission (Natalia Ladnaya, Russian Federal AIDS Center, personal communication, 2007). IDU tend to be young and sexually active, which may lead to transmission to their sexual partners.1,8
The increase in drug use has also been accompanied by a liberalization of sexual values, without improved sexual and reproductive health education in schools.8,9 Preceding the emerging HIV epidemic were increasing rates of other sexually transmitted diseases.8 For example, the rate of syphilis infection increased 66-fold from 4.2 cases per 100,000 population in 1988 to 277 cases per 100,000 in 1997.9
Relatively little HIV-related research has been conducted in the Russian Federation, despite the obvious need. Studies to date have been primarily cross-sectional surveys10,11 and rapid assessments.7,12 Although risk factors for HIV are well documented, the relative importance of various risk behaviors may vary across populations, and the unique factors influencing the unfolding of the Russian epidemic require understanding. Therefore, the objectives of this article were to determine risk factors for incident HIV infection via the sexual route among Russians who do not report recent injection drug use and to determine the relative importance of identified risk factors. These were addressed using data collected for a case-control study of recent HIV infection among injecting and noninjecting men and women in 4 regions of the Russian Federation, with restriction of the analysis to noninjecting participants.
Materials and Methods
In November 2001, the Canada AIDS Russia Project invited all 89 administrative regions of the Russian Federation to submit applications to participate in a training project to enhance the capacity of key Russian governmental and nongovernmental organizations working in the HIV/AIDS field, and to support the establishment of an integrated, multidisciplinary network to respond more effectively to the spread of HIV/AIDS at the federal and regional level. Sixteen regions applied and 4 were selected. These were the regions of Altaiskiy Krai, Krasnoyarskiy Krai, Saratovskaya Oblast, and Tverskaya Oblast. Criteria for selection were adequate numbers of new HIV diagnoses; sufficient and highly qualified personnel to direct the project; commitment of institutional resources and scope for scale-up; and capacity to serve as a model pilot region for other regions. The project provided training, information and support in the areas of epidemiologic surveillance, clinical care, laboratory diagnostics, community development, and public policy. Training included an epidemiologic demonstration project which was the source of data for the present article.
The estimated total population of the Russian Federation was 142.4 million in 2004.13 The population is largely urban (73%) and 99.6% of adults are literate.13 Altaiskiy Krai has a population of 2.6 million and is located in southern Siberia, and borders Kazakhstan. Krasnoyarsky Krai is the second largest region of the Federation, comprising 2,340,000 km2 in central Siberia, with a population of 3.5 million. Saratov Oblast is situated in the southeastern portion of the East European plain, and borders Kazakhstan to the south; its population is 2.7 million. Tverskaya Oblast has a population of 1.4 million; its capital city, Tver, is situated 167 km northeast of Moscow.
As of 2004, approximately 19,574 HIV infections were diagnosed in these 4 regions, representing 6% of the total diagnosed in the Russian Federation (Natalia Ladnaya, Russian Federal AIDS Center, personal communication, 2007). Similar to the country as a whole, the majority of infections were among IDU, with a growing number of heterosexually-acquired infections.
A case-control study of incident HIV infection was undertaken. The goal of the study was to identify factors associated with recent infection among injecting and noninjecting men and women. Persons with documented recent HIV infection (cases) and 1 HIV-antibody-negative control per case were enrolled and interviewed between March 2003 and September 2004. The study protocol, consent form, and survey instruments received ethical approval from the University of Toronto Human Subjects Review Committee.
Cases were defined as men and women with documented recent infection. Their first positive HIV antibody test must have occurred within 6 months of enrollment. Laboratory evidence of recent infection was considered present if (1) cases had an HIV-antibody negative test result within 12 months of their first positive result; or (2) cases had an indeterminate HIV-antibody test (i.e., positive or indeterminate HIV antibody enzyme-linked immunosorbent assay, and/or a negative or indeterminate HIV antibody Western blot) followed by a positive Western blot result, indicating detection during the seroconversion window period. It was anticipated that few eligible cases would arise among men who have sex with men in these regions precluding a meaningful analysis of infection within that exposure category. Men reporting this activity were considered ineligible a priori.
Eligible cases for the study were identified by AIDS Center staff using nominal HIV-antibody test results at the regional AIDS Centers. All persons testing HIV-positive in these regions are referred to the AIDS Centers for treatment. Seventy-one percent of eligible cases who were contacted enrolled in the study. The response rate ranged from 42% to 85% across regions.
HIV-negative controls were sampled among persons testing HIV-negative at the regional AIDS Centers and at other HIV test sites. All had evidence of an HIV-antibody negative test within 6 months of enrollment. Controls were matched to cases by region, gender, and IDU status. The response rate for controls was 76% (range 61%–96% across regions).
A total of 129 cases (85 women, 44 men) and 126 controls (74 women, 52 men) were recruited. Recruitment was highest in Tverskaya Oblast (101), followed by Altaiskiy Krai (81), Krasnoyarskiy Krai (42), and Saratovskaya Oblast (31). The present analysis focused on risk factors for HIV infection via the sexual route. (Risk factors for infection among recent injectors will be analyzed in another article.) Therefore, the sample was restricted to participants who reported sexual activity but not injection drug use in the 12 months before the HIV-antibody test that lead to study entry, the period in which HIV infection occurred for cases. Eighty-six participants (42 cases, 43 controls) were eliminated because of recent injection drug use; an additional 3 participants (1 case, 2 controls) did not report recent sexual activity and were excluded. The final sample size for analysis was 166 (19 male cases, 67 female cases, 22 male controls, 58 female controls).
HIV antibody screen testing was conducted in each region using HIV test kits provided by the Federal Ministry of Health. EIA screening test kits included CombiBest Anti-HIV1 + 2 (ZAO Vector-Best, Novosibirsk, Russia), DS IFA-Anti-HIV-UNIF (NPO Diagnostic systems, Nizhny-Novgorod, Russia), and Avicenna-HIV1,2 EIA (Medical Center Avicenna, Moscow, Russia). Initially reactive specimens were repeated in duplicate at the local laboratories and if repeatedly reactive, they were forwarded to the Regional AIDS Center Laboratory for supplemental testing with an alternate test kit including Vironostika HIV Uniform II (Organon Teknika, Uppsala, Sweden), or Genscreen HIV 1/2 Version 2 (Bio-Rad Laboratories, France) and confirmatory testing using Western blot (Bio-Rad Laboratories). Repeat specimens were requested and confirmed for all first time positive cases before a diagnosis of HIV infection was made.
Interviews and Risk Factor Measures
Enrollees participated in a face-to-face, structured, quantitative interview with staff of the regional AIDS Centers. Interviewers were taught research interviewing techniques, research ethics, and sensitivity toward HIV/AIDS issues and affected communities. The median interview length was 50 minutes (range 10 minutes to 2.5 hours). Participants were given an honorarium of 150 rubles (approximately 5 US dollars) on completion.
A theoretical model of HIV infection via the sexual route directed the construction of measures in the survey instrument. It was based on known risk factors in heterosexual populations elsewhere in the world. Three main categories of risk factors were measured: partner characteristics; specific sexual behaviors (including unprotected sex); and concomitant sexually transmitted infections (STIs). Questions focused on the 12-month period before the HIV-antibody test that lead to enrollment, the period in which cases were infected.
A regular sexual partner was defined as a partner, spouse, boyfriend, girlfriend, friend, or someone else with whom one had an on-going sexual relationship. A casual sexual partner was defined as a one-night stand, someone one just met, or someone else with whom one did not have an ongoing sexual relationship. Paid sex was defined as having sex with someone in exchange for money, drugs, or other goods or necessities; receiving payment and providing payment were measured separately.
Markers of a possible exposure to an HIV-infected sexual partner were having a regular partner who the participant reported as HIV-positive or HIV status unknown; having a regular or casual sexual partner who was known to be an IDU; unprotected sex with a partner with a diagnosed STI or signs/symptoms of an STI; and the total number of sexual partners.
Unprotected sex was defined as not always using condoms for intercourse. Respondents were also asked whether partial condom use had occurred (i.e., any breakage/slippage, application of the condom after initially unprotected intromission, or removal of the condom and continuing with unprotected sex). Respondents who reported always using condoms but who also reported one or more types of partial condom use were coded as having had unprotected sex because at least some unprotected exposure occurred.
Concomitant STI were measured by self-report of an STI diagnosis in the 12-month period of interest (i.e., a diagnosis of syphilis, gonorrhea, genital/anal warts, chlamydia, trichomonas, pubic lice, herpes, HBV, HCV, urea plasmosis, and/or candidiasis). Respondents were also asked whether they had signs/symptoms of an STI in the same period but did not see a doctor (i.e., were undiagnosed). Signs/symptoms were painful urination, difficulty urinating, frequent urination, itching/burning sensation while urinating, blood in urine, sores in the genital area, abnormal genital discharge, and/or vaginal itching or burning.
Risk factors for HIV infection were assessed and results are reported as odds ratios (OR). Statistical analysis was conducted using SAS, version 8 (SAS Institute, Cary, NC). Preliminary descriptive analysis involved a review of distributions of risk factors and their crude associations with HIV status using Pearson chi-squared tests, Wilcoxon rank sum test, and unadjusted logistic regression. A Mantel-Haenzel analysis, stratified by gender, was conducted with all risk factors because gender was expected to be an important effect modifier a priori. Continuous variables were categorized based on the pattern of risk when the linearity assumption was violated. Only risk factors for which data were sparse were eliminated from further analysis. Remaining risk factors were combined in one multiple unconditional logistic regression model; all pairwise interactions with gender were examined. Independent variables that were nonstatistically significant (P ≥0.05) and had 0.5< OR <2 were then removed from the model in a backward elimination fashion. The Hosmer-Lemeshow goodness-of-fit test was used to determine model fit.
The population attributable risk percentage (PAR%) was calculated for each independent risk factor. We used the equation [PAR = (RR − 1)/RR × proportion exposed among cases], where RR is the relative risk of infection for that risk factor.14 The OR was used as an estimate of the RR. When PAR varied by gender, a weighted average was calculated using the equation [PARtotal = ∑pi × PARi] where pi is the proportion of cases in gender stratum i.14
Characteristics of the Sample
Characteristics of cases and controls are shown in Table 1. Men were somewhat older than women (P = 0.07, t test). Age differences by HIV-status were not statistically significant, but the data suggest that the youngest were HIV-positive females, and the oldest were HIV-positive males, with HIV-negative males and females of intermediate age. Although the majority had completed postsecondary education, a sizable proportion reported having difficulty meeting their basic living expenses.
Cases and controls reported similar HIV test histories, with a mean total number of times tested of 3.2 for cases and 3.4 for controls. The HIV antibody test that lead to enrollment in the study took place at the regional AIDS Centers for the majority of participants. Test sites differed between cases and controls (P = 0.001, χ2 test) and by gender (P = 0.002, χ2 test) (Table 1).
All men reported a heterosexual orientation and none reported ever having a male partner. Among women, 98% identified as heterosexual and 2% (2 of 125) as bisexual with 2% (3 of 125) ever having a female sex partner.
Over one quarter (27%, 44 of 160) had ever used illicit drugs, with little difference by HIV status (29% of cases, 26% of controls). Twelve percent (20 of 165) had ever injected drugs. Ever injecting was more common among cases (14%, 12 of 85) than controls (10%, 8 of 80), but this was not statistically significant. Ever injecting was reported more often from participants who were living in Altai (22%) versus 14% in Krasnoyarsk, 8% in Saratov, and 6% in Tver; were aged 25 to 34 (17% vs. 10% <25, 8% 35–44, and 0% 45+); male (27% vs. 7% female); were neither married nor living common-law (18%) versus 9% among the married/common-law; had not completed postsecondary education (17%) versus 8% among those who had; or had difficulty meeting basic living expenses in the 12-month period (19%) versus 9% among those who did not report such difficulty.
Risk Factors and Risk Behavior
The proportion reporting potential risk factors in the 12 months before their HIV-antibody test are reported in Table 2, stratified by gender. Over half reported a diagnosis of an STI. Male cases were far more likely than male controls to report STI diagnosis (OR 8.0, 95% CI 1.9–33). Specific STI diagnoses that were most strongly associated with HIV infection among men were herpes (OR 3.9, 95% CI 0.4–41), hepatitis B (OR 3.6, 95% CI 0.6–21), and gonorrhea (OR 2.5, 95% CI 0.21–30), but none reached statistical significance. Conversely, female cases and controls were equally likely to report an STI diagnosis, whether this was analyzed as a grouped variable or separately (data not shown). More female cases (12%) than controls (2%) reported a diagnosis of hepatitis C, although this did not reach statistical significance (OR 7.7, 95% CI 0.9–64).
Women were more likely than men to report signs or symptoms of an STI, and both male and female cases were more likely than controls to report signs or symptoms of STI (Table 2). Among women, specific signs/symptoms that were significantly more commonly reported among cases than controls were itching or burning when urinating (22% vs. 3%, OR 8.1, 95% CI 1.8–37), abnormal discharge (43% vs. 17%, OR 3.7, 95% CI 1.6–8.4), and vaginal itching or burning (28% vs. 10%, OR 3.4, 95% CI 1.3–9.3).
More women then men reported regular partners (94% vs. 76%, P = 0.001) (Table 2). Nearly half (46%) of female cases reported an HIV-positive regular partner as compared with only 3% of female controls. Men were more likely than women to report casual partners, sex worker partners, or 5 or more sexual partners in total (P <0.0001). Few women (1%) reported receiving payment for sex. The majority of HIV-positive female cases (85%) reported unprotected sex with a regular partner who was HIV-positive or of unknown status, compared with 41% of female controls. Nearly half of men reported unprotected sex with a casual partner. Condom use at least sometimes was reported by 68% of men and 65% of women, and frequency of use was not statistically significantly different by HIV status. Few reported always using condoms.
Independent Risk Factors for Recent HIV Infection
Independent risk factors for HIV infection in multiple logistic regression modeling are shown in Table 3. These were unprotected sex with an HIV-positive/status unknown regular partner (among women only: OR 5.4); a regular sexual partner who was an IDU (OR 3.6); having a total of 5 or more sexual partners (among men only: OR 2.7); unprotected sex with a partner with diagnosed STI or signs or symptoms of an STI (OR 6.4); and undiagnosed signs or symptoms of an STI (OR 3.4). In multiple logistic regression modeling, there was a gender interaction for diagnosis of any STI that approached significance (P = 0.06, not shown in Table 3). The effect of any STI diagnosis was nearly 5-fold for men (adjusted OR 5.0, 95% CI 1.0–24) but was absent for women (adjusted OR 0.89, 95% CI 0.36–2.2). No specific STI diagnoses were independently associated with HIV infection among men or women (data not shown).
The analysis presented thus far excluded participants who reported injection drug use in the 12 months before their HIV-antibody test, but included participants who may have ever injected drugs. Inaccurate recall of injection drug use (i.e., not remembering having done so within the time frame of interest) may have introduced error in the estimates of sexual risk factors if some HIV infection was via injection practices. A sensitivity analysis was conducted by excluding participants who had any history of injection drug use in their lifetime. Results are shown in Table 3, model 2. Results were similar, with one exception. The OR for sex with 5 or more partners among males approximately doubled from 2.7 in model 1 to 5.2 in model 2, although the wide confidence intervals overlap.
The inclusion of age (either continuous or categorical treatment) or the site for the last HIV test (AIDS Center vs. other) in the model did not meaningfully alter the magnitude of the ORs (data not shown).
PAR% were calculated using the adjusted ORs from model 1 (Table 3). In descending order, these were unprotected sex with an HIV-positive/status unknown regular partner (54%); undiagnosed signs/symptoms of an STI (37%); a regular partner who was a known IDU (31%); sex with a partner known to have an STI or who had signs/symptoms of an STI (12%); and having 5 or more sexual partners (7%). The latter ranked last primarily because this only posed a risk for males; as few cases were male, it received low weight.
In this case-control study of heterosexually-acquired HIV infection in 4 regions of the Russian Federation, the predominant risk factors were unprotected sex with an HIV-positive or status unknown regular partner, undiagnosed signs and symptoms of STI, a regular partner who was an IDU, unprotected sex with someone who had an STI or signs or symptoms of an STI, and sex with 5 or more partners.
A challenge for all case-control studies is the selection of an appropriate control group for comparison. The ideal control group would consist of a sample of the underlying source population that gave rise to the incident cases.15 Although eligibility criteria for controls were as consistent as possible with that for cases, there were differences in the location of the HIV antibody test site between cases and controls such that controls were more likely to have been tested at a regional AIDS Center. Adjustment for the location of the test site did not materially affect the point estimates of the ORs, suggesting an absence of strong selection confounding due to some latent variable related to test site.
To be eligible, participants had to have undergone HIV testing before their enrolment. HIV testing is routinely done in the Russian Federation such that the majority would have been tested at least once. In 2007, is it estimated that 34% of the general population underwent testing (Natalia Ladnaya, Russian Federal AIDS Center, personal communication, 2008). In this sample, there was no difference in the number of times cases and controls had been tested. Should bias have been present, it may have resulted in over- or underestimation of risk factor effects. Those who were at greatest risk of HIV infection may have been more likely to undergo testing either because they requested it, or because their health care provider considered testing to be appropriate (e.g., patients with symptoms or an STI diagnosis). Conversely, those most at risk may have avoided testing out of fear of stigmatization.
All risk factor information was self-reported and collected in face-to-face interviews conducted by trained staff at the Regional AIDS Centers. Although respondents were assured of confidentiality, it is possible that participants underreported potentially stigmatizing behavior resulting in social desirability bias. If underreporting was equally common among cases and controls, this would have lead to nondifferential misclassification and bias toward the null, suggesting that ORs were underestimated.
Among women, the strongest risk factor was unprotected sex with a regular partner who was HIV-positive or of unknown status. Because this is reported retrospectively, recall bias must be considered. A woman who was just diagnosed may be more inclined to say that her partner was HIV-positive or of unknown status, particularly if she had only 1 partner. Therefore, it is likely that the OR is somewhat inflated. Nevertheless, the data are persuasive for the existence of an association, even if it is of lesser magnitude, because no evidence of recall bias was observed among men, women with multiple partners were least at risk, and research in other countries has found that women are commonly infected from their regular, stable partner.1
Contrary to the findings among women, men who had many partners had the highest risk of HIV infection. This did not reach statistical significance despite a large OR; however, power was low as only 41 men were analyzed. These results provide tentative evidence that HIV is transmitted heterosexually to men primarily through sex with nonregular (predominantly casual) partners, or from regular partners who injected drugs; whereas among women, sex with multiple partners was not necessary for infection.
Commercial sex workers were likely underrepresented in our sample. A rise in sex work since the early 1990s has been noted in the Russian Federation.7,9 The estimated HIV prevalence among Russian sex workers was 15% in 2005, as compared with 1.1% in the general population of adults aged 15 to 49.3 Nevertheless, in our study a low proportion of female participants reported sex work. Their underrepresentation may have been because of avoidance of nominal HIV testing, which would have limited the study’s ability to identify them as eligible.
Since the beginning of the HIV epidemic in Russia in 1997, the predominant risk factor has been IDU,16 although HIV surveillance data suggest that heterosexually-acquired infections are increasing.1 Our data provide evidence that this is due in part to bridging between the IDU and non-IDU population. Sexual activity with a regular partner who was a known IDU posed a 3.6-fold increased risk of HIV infection. Further, we estimated that 31% of excess heterosexually-acquired infections can be attributed to such exposure. These results are unlikely to be greatly affected by information bias because people’s perceptions of the IDU status of their regular partner have been shown to be generally accurate, at least in Western societies.17,18 Our results are also consistent with molecular epidemiologic results, which have shown that HIV subtypes acquired heterosexually are those found among those infected through IDU (predominantly subtype A), whereas men infected via homosexual activity tend to be infected with subtype B.16 Altogether, these findings are clear that HIV is spreading from the injecting to the noninjecting population via heterosexual activity. This implies that prevention efforts directed at IDU should continue and be intensified in the Russian Federation, as beneficial effects among IDU will indirectly reduce sexual transmission to partners.
Undiagnosed symptoms of an STI presented a major risk factor for HIV infection, although self-report of receiving an STI diagnosis did not, particularly among women. The use of self-report instead of laboratory-confirmed diagnoses likely introduced some error in these measures. If error was nondifferential it would have diluted the effects of STI, suggesting their true effects are larger than observed. What explains the phenomenon of a stronger effect of STI signs and symptoms in the absence of health-care seeking? We propose 2 possible explanations. The first is that undiagnosed signs and symptoms may be a marker for less severe infections; in the absence of treatment these would have a longer duration and a greater potential for an effect on HIV acquisition through increased susceptibility. An alternative explanation is that there was an overrepresentation of HIV-negative women who had recently received an STI diagnosis among female controls. This could have arisen if women who receive an STI diagnosis are more likely to be offered HIV testing. Such bias would have diluted the true effect of STI on HIV infection in women. An additional limitation is that the time of HIV acquisition is unknown within the 12 months preceding the last HIV test for many of the participants. Therefore, one cannot determine whether STI were present before exposure to HIV (thereby increasing the risk of infection) or if symptoms reflected cotransmission of STI (behavioral confounding).
Nevertheless, the potential role of STIs in the facilitation of HIV transmission is well founded. STIs other than HIV tend to increase infectiousness in the infected by their effects on HIV shedding, while increasing susceptibility in the uninfected through their effects on inflammation and disruption of mucosal barriers to infection.19,20,21 In this study the PAR% estimate for signs and symptoms consistent with STI suggests that 37% of ongoing heterosexually-acquired infections in the regions studied could be prevented by elimination of STI. This is likely an underestimate for the reasons mentioned above, and because in studies of seroconversion the PAR% reflects only the effect on susceptibility, and excludes the effect of STIs on HIV infectiousness.20 The contribution of STIs may well be generalizable to other regions of the Russian Federation given high rates of STIs since the early 1990s.9 Furthermore, the PAR% estimate is closer in line with those observed in the developing world than in the United States.20
At a minimum, effective STI policy must include access to high quality screening and treatment, the promotion of safer sex and other STI-related health behaviors, and surveillance of HIV and STI rates.20 STI control policy in the Russian Federation underwent significant change after the fall of the Soviet Union, and is now in a period of transition,22 suggesting that there is potential to implement practices and policies demonstrated to be effective elsewhere.23