Sexual risk behaviour relates to the virological and immunological improvements during highly active antiretroviral therapy in HIV-1 infection

Objectives: To evaluate the effect of highly active antiretroviral therapy (HAART) on the sexual behaviour of homosexual men, we conducted (i) an ecological study of time trends in sexual behaviour and sexually transmitted diseases; (ii) a HAART-effect study focused on the practice of unprotected anogenital sex.

Design: Subjects were participants in the ongoing Amsterdam Cohort Studies (ACS) among homosexual men, initiated in 1984. Data for (i) represented all ACS visits by HIV-1-positive and -negative participants who entered ACS at or below 30 years of age and were followed until 35 years (n = 1062). Data for (ii) represented all ACS visits of HIV-1-positive men from 1992 to 2000 (n = 365), of whom 84 were HAART recipients with at least 2 months of behavioural follow-up.

Results: (i) After HAART became generally available in July 1996, unprotected sex was practised more frequently and the incidence of gonorrhoea was higher compared to March 1992-June 1996 among HIV-1-negative and -positive men, respectively. (ii) Among HIV-1-positive men, a higher level of unprotected sex with casual partners was observed after HIV-1 RNA became undetectable and CD4 cell counts increased with the use of HAART. Notably, in individuals who did not receive HAART, high HIV-1-RNA levels (above 10⁵ copies/ml) were likewise related to unprotected sex with casual partners.

Conclusion: Data support the need for the reinforcement of safe sex prevention messages among HIV-1-negative men, and our data also provide a lead for redirecting and tailoring current prevention strategies to the needs of HIV-1-positive men.

Introduction

Several years after reporting a drastic reduction in high-risk sexual behaviour and a substantial decrease in sexually transmitted diseases (STD), centres in several industrialized countries now report the more frequent practice of unprotected anogenital sex and increasing rates of gonorrhoea and syphilis, especially among homosexual men ^[1-6] (I.G. Stolte, N.H.T.M. Dukers, J.B.F. de Wit, H.S.A. Fennema, R.A. Coutinho, in preparation). It is striking that these increases occur in an era in which highly active antiretroviral therapy (HAART) became available in industrialized countries. Perhaps these increases in unprotected sex and STD reflect reduced concern regarding HIV-1 because of the positive effects of HAART ^[7,8], which has substantially improved survival in HIV-1-infected individuals ^[9,10].

The aim of our study was to investigate the relationship between HAART and sexual risk behaviour among homosexual men participating in the Amsterdam Cohort Studies (ACS). Therefore, we first examined trends in STD incidence and sexual behaviour among both HIV-1-negative and HIV-1-positive young men, to gain insight into changes before and after the general introduction of HAART in July 1996. We will further refer to this research as the 'Ecological study'. Moreover, we examined whether actually receiving HAART was related to the practice of unprotected anogenital sex in HIV-1-positive homosexual men. During successful treatment with HAART, HIV-1-RNA levels often drop to undetectable levels and CD4 cell counts tend to increase ^[11]. We hypothesized that such virological and immunological improvements affect the practice of unprotected anogenital sex, because successfully treated individuals may now feel better (physiologically or psychologically) and may perceive their infectiousness as being diminished ^[7,8,12]. In the second part of our study, further referred to as the 'HAART-effect study', we examined the effect of HAART and its accompanying virological and immunological improvements on sexual risk behaviour.

Methods

The prospective ACS on HIV-1 seroconversion and AIDS among HIV-1-seronegative and HIV-1-seropositive homosexual men was initiated in 1984 ^[13]. Over time, entry criteria have changed with respect to HIV-1 status and age. Before 1995, both young and older men were allowed to enter the study, but from that year onwards only young participants (aged ≤ 30 years) were recruited. Also since 1997, older HIV-1-negative men were no longer followed. As a result of these procedures, HIV-1-negative men in active follow-up in recent years have been relatively young (maximum age by December 1999 was 34 years). In contrast, HIV-1-positive men were followed after 1995 and therefore include both young and older participants.

Return visits are scheduled every 3 (HIV-1-positive men) or 6 (HIV-1-negative men) months. At each visit a medical history, including self-reported information on gonorrhoea and syphilis, is taken by a trained nurse and blood is drawn and stored for virological and immunological testing. At entry and every 6 months thereafter, participants complete a standardized behavioural questionnaire. After participants develop AIDS, they remain in follow-up, but no longer provide information on sexual behaviour and STD, because participants with AIDS are seen at a different location (university hospital).

Analysis of HIV-1 antibodies was performed with two commercially available enzyme-linked immunosorbent assays (Abbot Laboratories, North Chicago, Illinois, USA; Vironostika, Organon Teknika, Boxtel, the Netherlands) and confirmed by Western blot analyses. Analysis of CD4 cell counts was determined by cytofluorometry and prospectively performed in all HIV-1-positive men. Results have been available to the participants since the start of the ACS. For analyses of HIV-1 RNA, serum samples from the earlier years were tested with nucleic acid sequence-based amplification assay (Organon Teknika), with a quantification threshold of 1000 HIV-1-RNA copies/ml. From 1997 onwards, use has shifted to the more sensitive NucliSens test (Organon Teknika) with a quantification threshold of 400 HIV-1-RNA copies/ml. For some men, participating in clinical trials has led to additional HIV-1-RNA tests, with quantification thresholds ranging from 5 HIV-1-RNA copies/ml (Ultra NucliSens, Organon Teknika) to 500 HIV-1-RNA copies/ml (Quantiplex bDNA, Organon Teknika). HIV-1 RNA was retrospectively determined in stored serum samples from HIV-1 seroconverters taken at the first seropositive visit and at one year intervals until the end of follow-up. Routine prospective HIV-1-RNA testing of all HIV-1-positive men was applied from July 1996 onwards and since this date test results (whether HIV-1 RNA is below or above the detection threshold) have been available for the participants within one month after testing.

Ecological study

Subjects

To evaluate ecological trends from 1984 to 2000 in STD and sexual behaviour, we selected entry and follow-up visits for men aged 30 years or younger at study entry who were followed until age 35. This selection was made to ensure data comparability over time and between HIV-1-negative and HIV-1-positive men. The resulting study population consisted of 1062 young men with 10 988 visits, recruited between the start of ACS until 1 January 2000.

Variables

The entry questionnaires provided us with sociodemographic variables, the lifetime number of sexual partners and a 5 year history of anogenital gonorrhoea and syphilis (see Table 1). Furthermore, we used information from follow-up visits regarding anogenital gonorrhoea and syphilis (since last visit) and sexual behaviour (over past 6 months). Participants were asked whether they engaged in anogenital intercourse (both insertive and receptive) as well as the frequency of condom use. When, at least once, no condom had been used when engaging in anogenital sex in the past 6 months, we defined this practice as 'unprotected sex'. 'Protected sex' was defined as being when condoms were always used when practising anogenital sex.

Table 1

Before HAART was generally available, other therapy methods (mono and double therapy) were applied in the cohort. According to the timing of the introduction of these different therapy methods, we defined four 'therapy-periods': (i) no antiretroviral therapy (October 1984 to April 1987); (ii) Zidovudine monotherapy (May 1987 to February 1992); (iii) double therapy including zidovudine (March 1992 to June 1996); and (iv) HAART, a combination of three or more antiretroviral agents, mostly including a protease inhibitor (July 1996 onwards).

Statistical analyses

Characteristics at ACS entry of the newly recruited participants were compared among the four therapy periods, using χ² and Kruskall-Wallis tests. Regarding STD incidence and sexual behaviour, we considered the period March 1992-June 1996 as the reference period, because we were especially interested in comparison with the period after the general introduction of HAART. The proportion of engagements in sexual behaviour over a specified time period was calculated by dividing the number of visits at which sexual behaviour was reported by the total number of visits at which behavioural information was provided. The sexual behaviour reported by a participant within 3 months after the cut-off dates of each period was allocated to the previous period to reflect the proper calendar period (as behaviour was asked over the previous 6 months). In calculating the incidence of gonorrhoea and syphilis (expressed as number of self-reported episodes per 100 person-years), participants were considered to be 'at risk' during the entire study period. To calculate HIV-1 incidence, the HIV-1 seroconversion date was estimated as the midpoint between the last seronegative and the first seropositive visit. For some STD and sexual practices we were interested in a more detailed time-trend evaluation, and therefore we evaluated changes over single calendar years in addition to changes over therapy periods.

Generalized estimating equations were applied, to correct for dependency between measurements within an individual ^[14]. We assumed a Poisson distribution for evaluating STD incidence (using a first order autoregressive covariance matrix) and a binomial distribution for evaluating sexual behaviour (using an unstructured covariance matrix). We controlled for potential confounders including age, education, nationality, number of sexual partners (past 6 months) and individual ACS follow-up time. Analyses were performed separately for HIV-1-negative and HIV-1-positive men.

Highly active antiretroviral therapy-effect study

To examine the effects of HAART, CD4 cell counts and HIV-1-RNA levels on the practice of unprotected sex, we selected all study visits with behavioural data in the period January 1992 to January 2000 from all HIV-1-positive men, independent of age. This selection resulted in a population consisting of 365 HIV-1-positive men with 1762 visits yielding behavioural data. Data obtained before 1992 were excluded, as before that year data regarding sexual practices were not collected according to partner type (steady or casual), which was of particular interest.

In 1995, the first man in the ACS received HAART and of all 365 HIV-1-positive patients (of whom 251 were still in follow-up after 1995), 174 had received HAART by 1 January 2000. Of these 174 men, 91 men had behavioural follow-up after they started HAART and 83 men had no behavioural follow-up after that point. Of the 91 men with behavioural follow-up data, seven men had less than 2 months of follow-up after initiating HAART, which we considered to be insufficient follow-up time. The 84 men with behavioural data were comparable with the 90 men with insufficient behavioural follow-up with respect to all studied entry characteristics (data not shown).

Variables

In addition to the variables considered in the ecological study, we used information on the individual HIV-1-RNA levels and CD4 cell counts. To investigate the effect of these virological and immunological parameters on sexual practices we used the HIV-1-RNA levels and CD4 cell counts obtained in the period preceding the visit at which sexual behaviour was evaluated. If a person had more than one measurement of CD4 cells or HIV-1 RNA between two behavioural visits, we calculated the mean of the CD4 cell counts or the geometric mean of HIV-1 RNA over that period from the preceding visit until the present visit. When an individual had in this period at least one undetectable HIV-1-RNA measurement, we defined the HIV-1-RNA level as 'undetectable'. To assess the effects of HIV-1-RNA levels and CD4 cell counts on sexual behaviour, we categorized the former as: detectable HIV-1-RNA levels without HAART (reference category); undetectable levels without HAART; switch to undetectable level during HAART; and continued undetectable levels with HAART. We categorized the CD4 cell counts as: less than 350 × 10⁶ CD4 cells without HAART (reference category); 350 × 10⁶ cells or greater without HAART; recent increase to 350 × 10⁶ cells or greater during HAART; and continued 350 × 10⁶ cells or greater during HAART.

Statistical analyses

In the presented analyses, we included the total group of HIV-1-positive men (n = 365) to provide a substantial control group to use as a reference category. Furthermore, the large number of visits at which men did not receive HAART enabled us to investigate the effects on sexual risk taking of different HIV-1-RNA levels, including very high levels, which are uncommon in those taking HAART. However, one could argue that the results are biased, because these 365 men included men who never received HAART and men who left the study before HAART became available. Therefore, analyses were repeated using only the 84 men receiving HAART or using only those 251 men who were still in follow-up after 1995 (and thus were 'at risk' of receiving HAART). The resulting risk estimates were comparable to those presented for the total group of 365 HIV-1-positive men in Table 3, indicating that a bias by using the total group of HIV-1-positive men was limited. Therefore, results based on the total group of 365 men are reported.

Table 3

In statistical analyses, generalized estimating equations were used with an unstructured covariance matrix ^[14]. As we were interested in the effect of HAART, HIV-1-RNA levels and CD4 cell counts on the practice of unprotected sex, regardless of any time trends in this sexual practice, we presented risk estimates for the factors controlling for calendar time (by including calendar years as a continuous variable). In multivariate models, we controlled for further potential confounders including age, nationality, education, number of partners, and individual follow-up time. As our newly constructed variables regarding the use of HAART, HIV-1-RNA levels and CD4 cell counts are strongly related, they were not included together in the final models. Categorical variables with a considerable amount of missing data (such as HIV-1 RNA) were multivariately modelled with a separate 'missing values' category.

Results

Ecological study

Over the study period, 1062 homosexual men who were 30 years or younger were enrolled in the ACS. These men had a median age of 26.4 years [interquartile range (IQR) 23.8-28.5]; 92.9% were of northern or central European nationality and 54.5% had a college degree. Compared with early enrollees, those of more recent time periods more often had a college degree (HIV-1-negative men) and less often reported a history of STD (Table 1). In addition, the number of reported sexual partners decreased after the start of the HIV-1 epidemic in the 1980s. In comparison with HIV-1 negative men, positive men were older, had a lower educational level, were less often of northern or central European nationality, and reported more STD and sexual partners (all P < 0.001).

Incidence of HIV-1 and sexually transmitted diseases

Of the 1062 young homosexual men, 185 were seropositive for HIV-1 and 877 were seronegative, of whom 64 seroconverted during follow-up. The incidence of HIV-1 seroconversion strongly decreased in the early years of the ACS, but has fluctuated in recent years (Fig. 1). In 1999, five young men seroconverted (incidence 2.0/100 person-years).

Fig. 1

Among HIV-1-negative men, the incidence of gonorrhoea increased slightly after July 1996, but this increase was not statistically significant in univariate analyses (data not shown), and the relative risk decreased with correction for potential confounders in multivariate analyses (Table 2). Among HIV-1-positive men, the incidence of self reported syphilis declined after the early 1980s, with only two young men reporting a syphilis episode in 1996-1997. However, the incidence of gonorrhoea in the period after July 1996 was significantly higher than in the period before, nearly reaching the level of October 1984 to April 1987.

Table 2

Sexual behaviour

Among the 1062 young men, the practice of anogenital sex (Fig. 2), started to increase again in the early 1990s among both HIV-1-negative and HIV-1-positive men. This increase continued after July 1996 and was observed for both steady and casual partners (data not shown).

Fig. 2

For HIV-1-negative men, after July 1996 the odds of having unprotected sex were higher than in the period March 1992 to June 1996 (P = 0.02, see footnotes Fig. 2). Examination of this sexual practice by calendar year (data not shown) already revealed in 1992-1993 an increase in unprotected sex followed by an equally strong decrease in this practice. Again, an increase was observed from 1996 onwards. The rate of unprotected sex, as reported at HIV-1-seronegative ACS visits, increased for steady partners from 70.1% in 1996 to 77.8% in 1999, and for casual partners from 27.5 to 33.3% (linear trends after July 1996, all P < 0.05).

For HIV-1-positive men, apart from the decrease in the early years, no statistically significant trend was observed in the rate of unprotected sex with either partner type. Overall, HIV-1-positive men less often reported unprotected sex than did HIV-1-negative men. However, after July 1996, unprotected sex with casual partners was reported more often by HIV-1-positive men (at 44.6% of all visits) than by HIV-1-negative men (31.5%) (P = 0.05) Unprotected sex with steady partners was more often reported by HIV-1-negative men (73.7%) than by HIV-1-positive men (50.8%) (P < 0.001).

Highly active antiretroviral therapy-effect study

The 365 HIV-1-positive men in our study group had a median age of 35.7 years (IQR 31.4-42.8) at their first ACS visit in the period 1992-2000. Participants were predominantly (88.4%) of northern or central European nationality, and 39.8% had a college degree. The 84 participants who received HAART, and for whom we had concurrent behavioural data, had a median follow-up time after the initiation of HAART of 0.9 years (IQR 0.6-1.1) and a median age of 39.9 years (IQR 34.7-47.6) at initiation of HAART. Of the 84, 78 achieved HIV-1-RNA levels below the detection threshold after the initiation of HAART. Of these, 57 had switched from a detectable level before HAART to an undetectable level after starting HAART, whereas 21 started HAART with undetectable levels and maintained those levels during HAART. During HAART, CD4 cell counts increased to 350 × 10⁶ cells or greater in 38 men, of whom approximately half (n = 20) had switched to undetectable HIV-1-RNA levels in the same period. In this study, we investigated the effect of receiving HAART and its accompanying virological and immunological improvements on having unprotected sex. Of the 365 men in our study 271 men (74.2%) engaged in anogenital sex, 207 men (56.7%) had anogenital sex with steady partners and 201 men (55.1%) with casual partners.

Effect of receiving highly active antiretroviral therapy

Although, particularly with steady partners, there was a tendency towards a lower frequency of unprotected sex during HAART, these associations were not statistically significant in either bivariate or multivariate analyses (Table 3 a). Also, when the duration of receiving HAART treatment was evaluated, no statistically significant effect on having unprotected sex was observed (data not shown).

Effect of HIV-1-RNA levels

Both in bivariate and multivariate analyses, the odds of having unprotected sex with casual partners were significantly higher after a recent switch to undetectable HIV-1 RNA levels following initiation of HAART, compared with having detectable HIV-1-RNA levels without HAART (Table 3 b). Notably, all participants who received HAART had access to their HIV-1-RNA test result. Restricting our analyses to the 84 men who had received HAART yielded comparable results on the effect of switching HIV-1-RNA levels on having unprotected sex (data not shown).

Another effect of HIV-1 RNA was that unprotected sex with steady partners was associated with having undetectable HIV-1-RNA levels while not receiving HAART. This association was stronger for men who did not receive HAART and had access to their HIV-1-RNA test result [odds ratio (OR) 2.2, 95% confidence interval (CI) 1.1-4.5] than for men who did not receive HAART and had no access to their test result (OR 1.4, 95% CI 0.9-2.3).

In addition, we investigated the effect of a high HIV-1-RNA level on unprotected sex, because the risk of HIV-1 transmission increases greatly with higher HIV-1-RNA levels ^[15]. We found that unprotected sex with casual partners occurred more frequently at higher HIV-1-RNA levels (Fig. 3), although with steady partners no such association was observed (data not shown). To test the effect of high HIV-1-RNA levels statistically, we subdivided the category 'detectable HIV-1-RNA level (without HAART)' into three categories: 'detectable HIV-1-RNA level below 10⁵ copies/ml (new reference category), 'recently increased HIV-1-RNA level above 10⁵ copies/ml' and 'continued (i.e. not recently increased) HIV-1-RNA level above 10⁵ copies/ml'. Interestingly, multivariate analyses indicated that a recent increase in HIV-1-RNA levels above 10⁵ copies/ml (in individuals not on HAART) was borderline significantly (P = 0.07) related to having unprotected sex with casual partners (adjusted OR 1.9, 95% CI 0.9-3.8). For having a continued HIV-1-RNA level above 10⁵ copies/ml, the adjusted OR for having unprotected sex with casual partners was 1.5 (95% CI 0.7-3.1), which was not statistically significant (P = 0.37).

Fig. 3

Effect of CD4 cell counts

In bivariate analyses, we observed a statistically significant association between either high CD4 cell counts (without HAART) or increasing CD4 cell counts (with HAART) and unprotected sex with casual partners (Table 3 c). After controlling for several factors in multivariate analyses, these risk estimates decreased slightly and were no longer statistically significant. However, the effect of increasing CD4 cell counts was still of borderline significance (P = 0.06), and is in concordance with the effect observed for reaching undetectable HIV-1-RNA levels during HAART.

In additonal analyses we investigated the effect of HAART, HIV-1-RNA and CD4 cell counts on the practice of anogenital sex. A (borderline) statistically significant association was observed between having anogenital sex with steady partners and a switch to undetectable HIV-1-RNA (OR: 1.9 (0.9-3.8), p = 0.06). In line with what was found for unprotected sex, having undetectable HIV-1 RNA levels (without HAART) was also related to having anogenital sex with steady partners (OR: 1.7 (1.2-2.5, p < 0.01). All other investigated effects were not statistically significant (although they were in the same direction as was observed for having unprotected sex).

Discussion

In the present study, we examined the relationship between the introduction of HAART and engagement in sexual practices among HIV-1-negative and HIV-1-positive homosexual men, using self-reported STD and (unprotected) anogenital sex as outcome measures. Our ecological study demonstrated that after July 1996, unprotected sex was practised more often, and that the incidence of anogenital gonorrhoea was much higher among HIV-1-negative and -positive young men, respectively, compared with the years before 1996. These findings are in concordance with reports from cohort studies and STD clinics in several other industrialized countries ^[1-6] (I.G. Stolte, N.H.T.M. Dukers, J.B.F. de Wit, H.S.A. Fennema, R.A. Coutinho, in preparation). However, although the increases coincide with the timing of the introduction of HAART, direct evidence for such a relationship cannot be derived in this way, and further research is needed to examine the factors underlying these increases. In a more direct approach taken by our HAART-effect study among HIV-1-positive men, we demonstrated that unprotected sex was related to HAART, as was found in a study by Miller et al. ^[16]. However, in our study, HAART-induced immunological and virological improvements appeared to be the crucial factor in the relationship between treatment and unprotected sex: after HIV-1-RNA levels switched from detectable to undetectable levels and after CD4 cell counts increased as a result of HAART, a higher level of unprotected sex with casual partners was observed.

Study limitations

Our study has several limitations that must be taken into account when interpreting the findings. First, information on STD and sexual behaviour was self-reported by participants, which could lead to information bias. It is unknown to what extent such bias was present in our study and if it changed over time. Second, the men receiving HAART who were included in the HAART-effect study (n = 84) comprised a relatively small group of the total number of ACS participants receiving HAART (n = 174). The remaining 90 men ceased to fill in behavioural questionnaires before or soon after starting HAART, and their reasons for doing so are unknown. Although, it is possible that the 84 men in our HAART-effect study are not completely representative of the total group of HAART recipients in the ACS, this is not likely because the 90 men with no behavioural data after HAART were otherwise comparable to the 84 men in our study. Moreover, the 84 HAART recipients had only a limited behavioural follow-up time during treatment, so no inferences can be drawn about the sexual practices of men who are treated with HAART for an extended period. The limited follow-up time also did not allow us to examine sexual behaviour among individuals who failed treatment and in whom HIV-1-RNA levels were again increasing. Third, the subjects of the ecological study were younger than those of the HAART-effect study. As a result of possible age-related differences in sexual behaviour ^[17,18], caution must be exercised in extrapolating results from the HAART-effect study to young men. Finally, HIV-1 RNA was not measured in every HIV-1-seropositive individual at each visit before July 1996, and when it was measured, tests varied and detection thresholds ranged from 5 to 1000 copies/ml. However, similar results (Table 3 b) were obtained by taking into account the different assays, standardizing all detection thresholds at 1000 copies/ml (threshold of the most commonly used test).

Impact on the HIV-1 epidemic

The first increases in HIV incidence were recently reported by the city of San Fransisco ^[19]. Among young homosexual men in our study, such a clear upward trend was not observed. However, along with the generally noted increasing pool of HIV-1 infections due to the prolonged life expectancy of people infected with HIV-1, the combined results from our two studies clearly demonstrate a change in factors known to enhance HIV-1 transmission. As shown in Fig. 2, the practice of anogenital sex has increased, and the practice of unprotected sex (not always using condoms) has either increased (HIV-1-negative men) or remained stable (HIV-1-positive men). There is thus an overall and absolute increase in unprotected sex (versus protected or no anogenital sex) among the total population. That the increase was not only present for steady partners but also for casual partners is of particular importance, assuming that one does not or cannot know the valid serostatus of one's casual partner. The increase in anogenital gonorrhoea among young HIV-1-positive men does not necessarily imply an increase in sexual practices favouring HIV-1 transmission ^[20], but infection with gonorrhoea may enhance both HIV-1 susceptibility and infectiousness ^[21-24]. A particular threat for the HIV-1 epidemic is the high frequency of having unprotected sex with casual partners (but not with steady partners) by persons whose HIV-1 RNA had increased to very high levels, at which they were undoubtedly highly infectious ^[15].

A recent study by Quinn et al. ^[15] demonstrated that no HIV-1 transmission occurred within couples when HIV-1-RNA levels were low (below 1300 copies/ml). Therefore, the implications for the HIV-1 epidemic of the higher level of unprotected sex at undetectable HIV-1-RNA levels found in our HAART-effect study are less obvious. However, although greatly suppressed by HAART, the seminal shedding of cells harbouring the HIV-1 provirus still continued after treatment ^[25]. Furthermore, in a study by Lampinen et al. ^[26], HIV-1 DNA was present in the anorectal canal of approximately half of the patients who received HAART. Therefore, the possibility of transmitting HIV-1 RNA at low virus levels cannot be ruled out. Moreover, the practice of unprotected sex is still of concern for the spread of STD other than HIV.

The role of virological and immunological factors

HAART-induced improvements in levels of HIV-1 RNA and CD4 cells appeared to play an important role in predicting the practice of unprotected sex. Our study does not provide us with the explanation for such behaviour, but hypothesizing we would like to interpret these findings as a consequence of treatment optimism ^[7,8]. After learning their newly improved HIV-1-RNA level, participants may well feel optimistic about their life expectancy ^[9,10], perhaps also believing that their infectiousness has diminished. This possible psychological effect could lead to a reduction in condom use ^[27,28]. Furthermore, a physiological component (physically feeling better) might play a role.

Just as switching HIV-1-RNA levels during HAART related to having unprotected sex with casual partners, so did increasing CD4 cell counts during HAART (although borderline significantly). It was striking that after HIV-1-RNA levels remained undetectable and CD4 cell counts remained high with HAART, the odds of having unprotected sex with casual partners were lower, although not statistically significantly lower (Table 3). It may be that the observed higher level of unprotected sex in those men having HAART-induced virological and immunological improvements reflects a temporary and short-lived effect of the psychological and physiological consequences of HAART.

Among men not receiving HAART, those having undetectable HIV-1-RNA levels had greater odds of practising unprotected sex with steady partners. This association was strongest among participants aware of their HIV-1-RNA test result, indicating at least a partial psychological effect.

Strikingly, men whose HIV-1-RNA levels were very high and had recently risen reported more unprotected sex with casual partners but not with steady partners. One could speculate that these men experience fatalism ^[29], and they suspend caution with casual partners while retaining consideration for steady partners. However, it is clear that from this study we cannot determine the underlying reasons for having unprotected sex, and specific individual motives must be carefully examined.

Therefore, as it appears, the level of HIV-1 RNA plays an important role in the practice of unprotected sex by either psychological or physiological mechanisms, or both, and moreover has distinct effects with different types of partners.

Recommendations

Although our findings need to be replicated by other (longitudinal) studies, our results alone indicate a need for the rethinking and renewal of prevention activities. Conventional methods propagated a general safe sex message directed towards the total group of homosexual men. Our results suggest that more targeted approaches would be more successful, because sexual risk behaviour was observed in different subgroups and distinct settings. Among HIV-1-negative men prevention messages of safe sex need to be reinforced. Further findings also indicate suggestions for developing prevention messages among HIV-1-positive men. Special attention should be given to those men who are about to start HAART, especially with regard to the moment when the first virological and immunological improvements will take place. These individuals should be aware that although their HIV-1-RNA load will probably become undetectable in their blood, they might still infect other people when having unprotected sex, and on top of that there is a risk of contracting other STD. Another group that may need special care and attention consists of those HIV-1-infected men who have high and increasing HIV-1-RNA levels and often engage in unprotected sex. Such individuals should be counselled on how to cope with their situation.

Conclusion

Among young homosexual men in the ACS, a recent increase in unprotected sex (for HIV-1-negative men) and anogenital gonorrhoea (for HIV-1-positive men) was observed, coinciding with the introduction of HAART. Among HIV-1-positive HAART recipients, treatment-induced immunological and virological improvements were associated with having unprotected sex with casual partners. Higher levels of unprotected sex with casual partners were also observed among men who did not receive HAART and had very high HIV-1-RNA levels. Therefore, prevention messages need to be targeted and tailored to the needs of these specific groups.

Acknowledgements

The authors wish to thank Nel Albrecht and Dieuwke Ram for interviewing participants and taking blood samples; Ronald Geskus, Udi Davidovich and Ineke Stolte for critically reading the manuscript, and Lucy Phillips for editing.