Diverse blood exposures associated with incident HIV infection in Calabar,Nigeria

Participants

We examined correlates of incident HIV infection in Calabar, Nigeria, between August 2007 and February 2008. On days during this period when interviewers were available, we invited all clients at the HIV voluntary counselling and testing (VCT) centres at the University of Calabar Teaching Hospital (UCTH)/University Medical Centre to participate in a study about factors related to HIV infection. We estimate that approximately 75% of clients invited to participate decided to participate in the study. Participating clients gave informed consent to participate, and participation or non-participation did not affect any services or care clients received. Most of our analyses focus on the participants who had previously tested HIV negative at the VCT centres.

Procedures

Participants were interviewed face-to-face in private rooms prior to HIV pre-test counselling. Interviewers were medical students and clinical support staff trained by UCTH faculty and were assisted by interpreters when necessary. Interview questions focused on demographics, reasons for testing, sexual behaviours and potential exposures to blood. Participants reporting a particular blood or sexual exposure were asked when the most recent exposure of that type occurred; for some blood exposures, they were also asked whether the sharp involved had been used previously by another person. Blood exposures that typically occur only during childhood in Nigeria (e.g. circumcision and scarification) were not assessed. No identifying information was collected from participants, and interview responses were kept strictly confidential and not shared with clinical providers or staff.

As part of their routine VCT visit, participants provided blood specimens drawn by venipuncture. HIV status was determined with two rapid tests of each specimen performed by the UCTH haematology department. A third test was performed to resolve serostatus if the first two tests showed discordant results. We used anonymous codes to link participants' interview responses to their HIV test results from their prior and current visits. Participants were offered compensation for transportation (up to approximately US$5) and refreshments, but most participants declined these incentives. All study procedures were approved by the UCTH ethical committee before the study began.

Statistical analysis

We assessed the representativeness of participants who had tested multiple times at the VCT centres (serial testers) relative to those who were testing for the first time at the VCT centres (single testers) in terms of their characteristics and lifetime exposures. For these comparisons, we used chi-square tests of independence and multiple logistic regression (when adjusting for age). We estimated HIV incidence in serial testers by dividing the number of seroconverters by the sum of serial testers' person-years of observation (sum of intervals between most recent HIV-negative test and the current test visit). To assess correlates of incident HIV infection, we computed odds ratios, Goodman and Kruskal's tau ⁸ (proportional reduction in error measure of association for nominal scale variables), gamma correlations (for ordinal scale variables), point biserial correlation coefficients (for ratio scale variables), and multiple logistic regressions and the associated 95% confidence intervals (or inferential tests) between serial testers' characteristics and incident exposures and their serostatuses (seroconversion/HIV positive versus HIV negative). We maintained at least a 15∶1 case:variable ratio for all logistic regression analyses. We also summarized the perceived reuse of sharps for some blood exposures with simple percentages. All analyses were performed with SPSS 7.5 (SPSS Inc., Chicago, IL, USA).

The incident exposure variables refer to reported exposures that occurred between a serial tester's two HIV tests. When dates of exposures were reported crudely (e.g. year only), these were converted to the midpoint of that period. We coded participants who reported that their last exposure occurred within three weeks before testing as unexposed (because of the HIV tests' window periods). Two exposures (enema and shaving with a razor used previously by someone else) appeared to occur regularly, given the recency of many reported events. Consequently, we created alternate codings of these exposures such that participants with very recent (<3 weeks) enema or shaving exposures were recoded as exposed. We excluded phlebotomy as an incident exposure, because we were often unable to distinguish serial testers' most recent reported phlebotomy events from their prior HIV tests.

Participant characteristics

Three hundred and twenty-one clients participated, and 45 of these were serial testers. Single and serial testers were similar in terms of many demographic characteristics, sexual and blood exposures and HIV prevalence (Table 1). Overall, the sample was highly educated. Serial testers tended to have more education and were somewhat more likely to be female than single testers. Serial testers were also more likely than single testers to report a desire to know their serostatuses for personal reasons and less likely to report illnesses, symptoms or doctor's request as reasons for testing. However, single and serial testers were similar in age (single testers: mean = 32 years, median = 29, standard deviation [SD] = 12, range = 11–83; serial testers: mean = 30, median = 29, SD = 8.0, range = 20–65). Many types of blood exposures were common.

In both single and serial testers, more female subjects had had induced abortions than had given birth. Among female subjects who had given birth, serial testers were more likely to report having had a caesarean section than single testers. Serial testers reported significantly (P < 0.05) more lifetime vaginal sex partners (female mean/median = 2.9/2, male mean/median = 8.3/4) than did single testers (female mean/median = 2.3/2, male mean/median = 4.2/3), but these differences were statistically non-significant (P > 0.05) after adjusting for age.

Serial testers were less likely to report receiving a therapeutic injection or an incision for administering medicines/herbal remedies than single testers. Serial testers were more likely to report prior phlebotomy than single testers (96% versus 67%), as might be expected, because they at least had prior blood draws for their previous HIV tests. No participants reported illicit injection drug use.

HIV seroincidence

The 45 serial testers had a median of 629 days between tests (mean = 793, inter quartile range = 258–1038, range = 89–2900), for a total of 97.8 person-years of observation. The 10 incident infections represent a 10% annual incidence rate among serial testers. The annual incidence rate declined with increasing test intervals: 17% for intervals <1 year, 14% for intervals 1–2 years long and 8% for intervals >2 years.

Correlates of incident HIV infection

Among the serial testers, seroconverters and seronegatives did not differ significantly or meaningfully in terms of most demographic factors or reason for testing (Table 2). However, seroconverters had non-significantly longer test intervals (mean = 972 days, median = 921) than seronegatives (mean = 742, median = 485; P > 0.05 on Mann-Whitney and t-tests). Similarly, seroconverters (mean = 37 years, median = 34) tended to be older than seronegatives (mean/median = 28; point biserial r = 0.47, P < 0.01).

Vaginal sex during the test intervals was almost universally reported by both seroconverters and seronegatives (Table 3); no serial testers reported anal sex during the test intervals. After adjusting for age, there were no differences between seroconverters and seronegatives in their reported lifetime number of sex partners (Table 3). Seroconverters and seronegatives were also equally likely to have had symptoms of genital ulcer disease (GUD) during the test intervals (Table 4). However, the seroconverter who had GUD also had injection treatment for the symptoms, while only one of the three seronegatives with GUD had such treatment.

Because of the differences in test intervals between seroconverters and seronegatives, we adjusted for test interval in all analyses of the association between incident blood exposures and HIV infection (clients with longer test intervals would be expected to be more likely to have any kind of exposure). In each of these analyses, seroconversion was the outcome variable, the particular exposure measure was the predictor variable, and test interval was the covariate. During the test intervals, seroconverters were more likely than seronegative participants to have shaved with a razor used previously by another person and to have had medical surgery, blood transfusions, enemas, vaccinations (apart from tetanus toxoid vaccination during pregnancy) and infusions (Table 4). Seronegatives were slightly more likely to have had finger prick phlebotomy and therapeutic injections than seroconverters during the test intervals. None of the serial testers reported dental surgery, incisions for administering medicines/herbal remedies or incarceration during their test intervals. Seroconverters were also more likely than seronegatives to have blood exposures involving sharps not specified in the other exposures assessed.

Among female serial testers, seroconverters were much more likely than seronegatives to have had tetanus vaccinations (recoded), given birth, and had a caesarean section during their test intervals. Seroconverters and seronegatives had similar likelihoods of induced abortion during their test intervals, and very few serial testers had ear piercings or episiotomies during their test intervals. No female serial testers reported the use of injectable contraception during the test intervals.

For all summary measures of blood exposure – whether a participant had any blood exposures, the diversity (number of types) of blood exposures and alternate codings of such – seroconverters were substantially more likely to have been exposed than seronegatives. The most conservative summary measure of blood exposure refers to any of three blood exposures (vaccination [excluding tetanus vaccination during pregnancy], shaving with a razor used previously by another and medical surgery) that cannot be explained as a possible consequence of unprotected vaginal sex or diagnosis of/treatment for symptoms of HIV infection. (The only medical surgery reported by a serial tester was unrelated to pregnancy, childbirth, or potential HIV symptoms or complications.) On this measure of blood exposure, seroconverters were significantly more likely to be exposed than seronegatives.

Perceived sharps reuse

We restricted our analysis of perceived sharps reuse to recent exposures (those reported as occurring in 2007 or 2008). The percentages of all participants who reported that a syringe or needle had been reused or did not know whether reuse occurred for their most recent exposure in this period were 14% (10/73) for phlebotomy, 50% (4/8) for tetanus vaccination and 11% (5/46) for therapeutic injection. The question for phlebotomy asked whether the participant saw the health-care provider take the syringe and needle from a sealed sterile package; responses other than ‘yes’ were coded as ‘possible reuse’. (Fewer than 3 participants reported injectable contraception or incisions for administering medicines/herbal remedies in the 2007–2008 period, thus their reports are not summarized here.)