Study design and data collection

Participant recruitment took place from March 2008 through May 2010. 72 study sites offering HIV testing in 35 cities from different regions in Germany (Figure 1), accounting for approximately 70% of all reported HIV diagnoses from 2000–2006 were included. Study sites included private practices (74%), outpatient clinics (13%), local health authorities (11%), and other agencies (0.2%).

Participants were recruited through convenience sampling, and participation was offered right before or sometimes after receiving an HIV test result. Eligible persons were informed about the study by test counsellors or physicians, and written informed consent was obtained from each participant.

Although we tried to recruit persons from all sub-groups with higher risk of HIV infection (MSM, migrants, intravenous drug users, persons with heterosexual transmission), sub-samples other than MSM were too small for analysis.

Cases were defined as men who accessed HIV testing services, reported having sex with other men, and being diagnosed with HIV. Controls were male individuals who accessed HIV testing services, reported having sex with other men, and having received a negative test result.

Inclusion criteria for cases and controls were i. age 18 or older. ii. signed informed consent. iii. having received an HIV test result. For cases inclusion criteria was extended to: iv. confirmed HIV-diagnosis within the last 12 weeks (ELISA reactive and Immunoblot positive). Cases and controls were matched for age group, and WHO-region of birth/origin.

Test counsellors or physicians completed a questionnaire for cases and controls (“physicians questionnaire”), covering: i. demographic information, ii. transmission risk, iii. last negative test result, iv. probable source of transmission, v. symptoms, vi. in cases information on anti-retroviral therapy.

When accessing pre-test counselling and before receiving the HIV-test result, cases and controls were asked to complete a standardized questionnaire (“patients’ questionnaire”), containing questions on KABP and other information. The following sections were covered in the patients’ questionnaire:

Physicians’ and patients’ questionnaire, as well as the informed consent form, were pre-tested in a pilot study in Berlin in 2007 [26, 27] and adapted afterwards. The patient questionnaire was available in German, English and French. After the first year of data collection, some items were modified: The response category “with person met online” was introduced into the questions on condom use with different partners, and the response option “I was certain that sex partner was HIV negative, because we talked about our serostatus” was added as possible category to the questions on reasons for not using a condom at last sex. The latter was added, because a high number of study participants listed this as a reason for not using condoms in the “other reasons” option.

A unique identifier was assigned to all patients’ questionnaires and it was impossible to trace patients’ identities. Patients participated voluntarily and did not receive compensation. Study sites received an allowance of €30 for every case and €20 for a matched control. Questionnaires were sent to Robert Koch Institute (RKI) and entered into an Access-database. Ethics approval was received from the ethics board at the Charité Berlin (medical university and clinic).

Laboratory tests

Capillary blood was obtained from cases by finger prick and then applied onto filters (Whatman #903). Filters were dried for at least three hours and sent to the HIV-laboratory at RKI via regular mail. The BED IgG-capture ELISA (BED-CEIA, Calypte Biomedical Corp., Lake Oswego, USA) was performed from dried blood spots (DBS) to distinguish recent (acquired within the last 5 months) from longstanding (> 5 months) infections. The evaluation of the test and validation for its use from DBS is described elsewhere [28, 29]. Sensitivity and specificity of BED-CEIA were acceptable (82% and 85% respectively) and comparable to values published earlier [30]. We increased specificity and positive predictive value by using a lower cut-off for the normalized optical density (from 0.80 to 0.65) as described previously [27] and increased specificity and positive predictive values to 97.6% and 93.6% respectively. Consequently, samples classified as “recent” were very likely to truly be recently acquired infections. Results of BED-CEIA were linked with responses from the physicians’ and the patients’ questionnaires.

Data analysis

Data was analyzed using SPSS 20.0. We compared reported knowledge about HIV, partnership status, duration of partnership, condom use and sexual risk behaviour between recently infected MSM and HIV-negative controls. We created the variable “always safe” which was defined as always using a condom when having sex outside of a relationship, with an acquaintance, with a person met online, a stranger, or with an HIV-positive person or not engaging in these situations. To detect differences between cases and controls chi-square test and unadjusted Odds Ratios (OR) were calculated. A multivariate analysis was subsequently conducted including factors found to differ between cases and controls using a significance level of p ≤ 0.05 of all results in bivariate analyses.

Study population

The study population consisted of 105 cases with recent HIV-infection and 105 matched controls.

Knowledge on HIV and HIV testing

Only 8% of cases and controls agreed with the statement that an HIV positive person with undetectable viral load cannot transmit HIV sexually, whereas the majority (66% cases and 69% controls) did not concur or did not know (26% vs. 24%). The proportion of MSM testing for HIV at least twice within the last twelve months was similar in cases and controls (85% vs. 81%).

Relationship status and duration of partnership

Number of sex partners

94% of cases and 93% of controls indicated that they had sex within the last 6 months. Information on number of male sex partners in that time period was available for 76 cases and 78 controls. There was a significant difference in number of sex partners between cases (mean = 11.8; median = 5) and controls (mean = 6.6; median = 4) (p = 0.03 (t-test)). 18% of cases and controls reported being monogamous over the last six months, whereas 12% of cases and 4% of controls reported more than 20 partners) (p = 0.06, 95%). Most cases (70%) and controls (78%) had between 2 and 20 sex partners (Table 3).

STIs and unprotected anal intercourse (UAI)

23% of cases and 22% of controls reported an STI within the last 6 months. STIs that occurred among the study population were syphilis (11 cases/5 controls), gonorrhea (4 cases/10 controls), venereal warts (5 cases/6 controls), herpes (4 cases/2 controls), chlamydia infection (3 cases/1 control) and three other STIs (1 case/2 controls).

Reasons for not using condoms during last unprotected (vaginal/anal) sex

Condom use

We compared frequency of condom use in various situations. Fifty-six percent of cases and 66% of controls indicated always using condoms in the beginning of a new relationship, whereas 44% of cases and 34% of controls stated using condoms inconsistently. This difference was not statistically significant. There was also no difference in condom use during relationships, with strangers or an HIV positive partner (Table 3). Cases reported lower rates of condom use outside of primary relationships (42%/30%; OR = 2.3; p = 0.012), with acquaintances (62% vs. 43%; OR = 2.2; p = 0.016), and with persons met online (55%/22%, OR = 4.4; p = 0.005). Controls reported more often than cases to have been “always safe” (29% vs. 9%, OR = 4.6; p = ^< 0.000) (Table 4).

Results from multivariate analysis

Conversation about serostatus: “Serotalking”

The practice of serosorting, which is defined as having sex without condoms exclusively or preferentially with partners of concordant HIV status and of using condoms with HIV-discordant or HIV status unknown partners, has become increasingly common among MSM [12, 14, 36, 37]. Research has shown that this practice can increase the risk of acquiring HIV and other STIs [13, 14, 36, 38, 39], whereas others have found that this practice might decrease risk of infection [11]. Whether serosorting might or might not work as a risk management strategy is influenced by various factors, such as the prevalence of HIV in the population; the explicitness of communication; the proportion of people living with HIV who are aware of being infected; the incidence of new HIV infection among people previously testing negative for HIV and practicing HIV serosorting, and their position in sexual networks; the willingness to disclose HIV infection to a potential sex partner; the context of serostatus communication; and the proportion of people living with HIV under effective ART [14, 37].

Our results suggest that the way serosorting is performed might have an impact on HIV transmission risk. One of the key findings was that having an explicit conversation about HIV serostatus before sexual activity reduces the risk of acquiring HIV. This might be attributable to the mode of serostatus communication: namely a direct and explicit conversation might be protective, whereas other ways of serosorting, such as relying on online profiles, or guessing/assuming HIV status based on appearance, might be much less effective. This interpretation is supported by the finding that “assuming my partner was HIV-negative” was associated with an increased risk of being diagnosed with HIV in bivariate analysis. Direct communication compared to non-verbal perceptions reduces the potential for misinterpretations such as “he must be HIV-positive/HIV-negative because he doesn’t insist on condom use”. MSM diagnosed with HIV rarely report UAI with non-steady partners known to be serodiscordant, but relatively frequently report UAI with HIV serostatus unknown partners.

Further research should investigate the utilization of serotalking by persons with different relationship status and thus examine whether serotalking is more likely used among MSM in steady relationships than by MSM engaging in casual sex with non-primary partners. According to former research the duration of partnership might have an impact on truthfully disclosing risk behaviour and serostatus [15]. Also, further research is needed to determine what information and topics serotalking should entail to become an effective risk management strategy.

Of course, a direct and explicit conversation about serostatus might not always be possible or feasible before having sex, e.g., in anonymous sex venues, and consequently “serotalking” cannot be promoted as the “one and only” risk management strategy. In our study, sex with acquaintances or with partners met online was rather common, and in these situations communication about serostatus might be an appropriate risk reduction strategy. Ideally, such communication should contain not only, but also information on recency of test results and number of potential HIV exposures since the last test [40, 41].

Limitations

There are some limitations to consider when interpreting our results. Our study was conducted in 2008–2010, an era when late antiretroviral treatment initiation was still common in Germany (CD4 threshold for treatment initiation was increased from 200 CD4 cells/μl to 350 CD4 cells/μl in a treatment guideline update in September 2008) [42]. Meanwhile, among MSM close to 90% of men diagnosed with HIV are receiving ART. Most MSM who could effectively transmit HIV today are likely to be unaware of being infected [43]. Serostatus communication as a HIV risk management strategy might therefore be less efficient today compared with the situation a few years ago.

With our sampling strategy we only reached MSM who accessed HIV testing services in the respective study sites. Therefore, our study population might include a high number of persons who have a concrete reason to get tested, namely a previous high risk situation. Also, we did not reach MSM who do not access testing. The KABP-questionnaire was not developed specifically for MSM, but we initially tried to recruit participants from all populations relevant for HIV-transmission in Germany. Consequently, some of the questions might not reflect the language or the specific situation of MSM (e.g., no questions on specific sex venues; substance use; in questions on sex with persons with unknown and positive serostatus, first response choice was vaginal sex). Because we reached predominantly MSM with the study, we changed some items of the questionnaire during the study period, which might have created limitations to comparability. Further, some of the questions had many response options, making the questionnaire more complicated and potentially causing missing answers. We did not include questions on UAI in new partnerships in order to understand what happens in the early stages of a relationship. We also did not add questions on substance use and alcohol consumption (except for intravenous drug use), even though this might affect risk behaviour and condom use. Lastly, the sample is rather small (105 cases and controls) and with a bigger sample size we might have been able to detect and identify other relevant risk or protective factors.