Studies of the sexual behaviour of gay men have reported that rates of unprotected anal intercourse (UAI) with steady partners are higher than with casual partners [1–8]; H. J. Hospers, J. B. F. de Wit, M. W. Ross, G. Kok, Maastricht University, manuscript submitted]. It has been argued that the elevated rates of UAI in steady relationships are not necessarily indicative of high-risk behaviour as some steady partners establish HIV seroconcordance and reach agreements regarding sexual practice outside the relationship . These strategies are commonly defined as ‘negotiated safety’ and regarded as a proper context for the practice of low-risk UAI between steady partners [9,10]. However, recent studies have shown that the practice of UAI among gay men is higher with steady partners than with casual partners, independent of negotiated safety strategies [8,11]. One of these studies  reported that after the rates of UAI with steady partner were corrected for negotiated safety, these rates remained higher than the rates of UAI with casual partners (39 versus 20%, P = 0.001)
If risky UAI is so prevalent in steady relationships, the question arises whether these relationships may be a significant source of HIV transmission. Unfortunately, little is published about the rates of HIV transmission that can be attributed to steady partners and which reflect the consequences of high levels of risky behaviour in this group. The purpose in this study was therefore to examine the share of HIV transmission that can be attributed to steady partners in cases of documented seroconversion. To do so, we analysed data collected by the Amsterdam Cohort Study among Homosexual Men from 1984 to 2000. We established the rates of seroconversion that can be attributed to either steady or casual sex partners and examined the effect that year of seroconversion might have had on the source of seroconversion. We hypothesized that a later stage of the AIDS epidemic would be associated with a higher likelihood of infection by a steady partner, basing this assumption on the principle that, prior to the AIDS era, anal intercourse with either casual or steady partners was by definition unprotected. Subsequently, in order for the rates of risky behaviour with steady partners to surpass that of casual partners, as indicated in the literature, UAI rates with casual partners should have dropped more strongly over time than UAI rates with steady partners. Such time-based differences in sexual behaviour would have consequences for seroconversion and could have resulted in steady partners becoming a more dominant source of HIV infection in the later years of the AIDS epidemic.
Furthermore, based on reports that young gay men differ in their sexual risk behaviour from older gay men [12–16], we examined whether age at seroconversion plays a role in predicting the source of HIV transmission. This examination was exploratory, as we did not have a formal hypothesis of how age would influence the source of HIV transmission.
Behavioural data was available for 1846 of the participants in The Amsterdam Cohort Study among Homosexual Men. Overall, the cohort is comprised of gay men living in and around the city of Amsterdam, enrolled since 1984. Their median age at recruitment was 29 years (range, 17–65 years); 82% were born in the Netherlands, and 86% were of Dutch nationality; 42% had high educational levels (college degree or equivalent), 37% had medium educational levels (attended school up to the age of ± 18) and 21% had low educational levels (attended school up to the age of ± 16); 66% had paid jobs, 19% were students and 15% were unemployed. Most participants defined themselves as exclusively homosexual. On a Kinsey scale for sexual preference ranging from exclusively heterosexual (1) to exclusively homosexual (7), the mean score was 6.5 (SD, 0.83).
Our working sample consisted of 144 cohort participants who seroconverted within the study between 1984 and 2000. Of these, 10 seroconversions had to be excluded as, in these particular cases, the source of HIV transmission could not be established (see Measures section for more detail). The final working sample consisted therefore of 134 participants. There were no significant differences between the total cohort and the working sample of seroconverters in any of the characteristics noted above.
Since October 1984, participants have completed, at 6-month intervals, self-administered questionnaires as to sexual behaviour with steady and casual partners. At each 6-month measurement wave, HIV-negative participants were tested for HIV using enzyme-linked immunosorbent assays (ELISA; Abbott Laboratories, North Chicago, Illinois, USA; Vironostika Teknika, Organon, Boxtel, The Netherlands). Seroconversion was confirmed with Western blot analysis. Seroconverters co-operated in a structured interview with a nurse soon after notification of seroconversion. The interview investigated the assumed source of infection, as well as risk behaviour with steady and casual partners in the preceding 6 months. For a more detailed description of the procedures, the reader is referred to Davidovich et al.  and de Wit et al. .
Determining the source of HIV transmission
In order to determine the origin of the HIV transmission in each individual case of seroconversion, risk behaviour was examined in the period between the last HIV-negative and the first HIV-positive test result (± 6 months). The purpose was to examine whether risk behaviour during that period could be exclusively attributed to either casual or steady partners. Two separate measurement tools were used for this assignment: the self-administered questionnaires completed at 6-month intervals and the structured post-seroconversion interviews.
In both the questionnaires and the interviews, risk behaviour was defined as UAI, which was assumed by us to be the mode of HIV transmission. In the questionnaires, UAI was measured by asking participants whether they had engaged in anal intercourse in the preceding 6 months and how often they had used condoms during such intercourse. These variables were always measured for both insertive and receptive anal intercourse with the exception of the first four data waves (1984–1986), in which condom use was measured for receptive intercourse only.
During the post-seroconversion interviews, the interviewing nurse inquired about the practice of insertive and receptive UAI with steady and casual partners in the 6 months following the last HIV-negative test result and prior to the first HIV-positive test result. In addition to the practice of UAI, the nurse inquired into the efficacy of condom use, that is, whether condom breakage or condom slip-off had occurred. Cases in which such incidents were reported were also labelled as UAI.
Unprotected anal intercourse with steady versus casual partners
From 1991 to 2000, whether UAI was performed with steady or casual partners could be determined directly from the questionnaires, because after 1990, sexual behaviour was measured separately for steady and casual partners. During 1991 and 1992, a definition of ‘steady’ partner was provided to participants. The requirements for such a partner were a minimum relationship duration of 6 months, seeing each other on average once or more per week, and having regular sexual contact with each other. Participants were asked to regard all other types of partners as casual contacts. From 1992 onwards, no a priori definition of steady partners was provided. The motivation behind this change was to allow participants to subjectively determine their partner's status. It was assumed that the perceived relationship status would be an important determinant of behaviour without necessarily being mediated by external criteria such as a minimum relationship duration.
During the initial nine data waves of the cohort (1984–1990), risk behaviour was not measured separately for steady and casual partners. In that period, because of other research priorities, participants were not asked to report the practice of anal intercourse according to the type of partner but only to indicate the total incidence of UAI with all partners. However, other items were included in these questionnaires (e.g., number of sexual partners in the preceding 6 months, degree of anonymity of sexual contacts, and whether participants were in a steady relationship) which helped us to determine whether UAI was practised with steady or casual partners. Details on the classification methods for these years (1984–1990) are available from the first author.
Whether UAI was performed with steady or casual partners could also be directly inferred from the post-seroconversion interviews since, during the interviews, data on UAI was always obtained separately for steady and casual partners.
In several cases, both the questionnaires and the post-seroconversion interviews indicated that UAI occurred with both steady and casual partners, which made it impossible to narrow down the source of infection to one type of partner. Fortunately, through follow-up contacts with the nurse, we had knowledge in some cases of negative HIV-test results for the steady partners of participants that were obtained after the seroconversion of these participants. These test results indicated that casual partners had been the source of infection in these particular cases. Eventually, 10 cases of seroconversion had to be excluded from the analysis. Of these, seven cases were excluded because UAI was reported with both types of partners without the possibility of identifying which type of partner was the source of infection; three cases were excluded because no UAI was reported with any type of partner.
Consistency of measuring tools
We examined the concordance between the questionnaires and the interviews in assigning the source of HIV infection to either casual or steady partners. In 69% of the cases (n = 92), there was full agreement between the two measuring tools. In 27% of the cases (n = 36), one measuring tool had missing data or was inconclusive whereas the other provided conclusive data. In only 4% of the cases (n = 6), were there discrepancies between questionnaire and interview. In four of these cases, we considered the interview to be the more valid source of information since a ‘confession’ element was documented: participants acknowledged during the interview to having had UAI with a particular partner and having withheld that information in the questionnaire. In the remaining two cases, the interview was considered more valid because it allowed the nurse to clarify vague replies. It was also assumed that the fragile timing of being interviewed soon after notification of seroconversion would increase the reliability of self-report. Nevertheless, when these six conflicting cases were omitted from the sample for control purposes, the patterns of the results of the analyses reported below remained unchanged.
Basic seroconversion data
The mean age at seroconversion in the sample was 36 years (SD, 7.6; range, 23–58 years) and the median age was 35 years. Between 1984 and 2000, 76% (102 of 134) of the new HIV infections could be attributed to risky behaviour with casual partners, and 24% (32 of 134) to risky behaviour with steady partners. The median year of seroconversion was 1987.
Year and age effect
This study had two main goals. The first was to examine whether seroconversion later in the AIDS epidemic would be more often associated with infection by a steady partner than early in the epidemic. The second was to examine whether age at seroconversion would have an effect on the source of HIV transmission. To answer these questions, the source of infection, which was the outcome variable, was dichotomized as infection via steady partners versus infection via casual partners. We then entered year of seroconversion, age at seroconversion, and their interaction as continuous variables into a logistic regression model. Wald tests indicated that the choice of a linear model was justified since no non-linear terms were significant.
Results showed a significant effect of the overall model (P < 0.01). We further found a significant interaction effect between the year of seroconversion and age at seroconversion on the source of HIV infection (P < 0.05). To interpret the interaction effect, both of the predictor variables were categorized. The time of seroconversion was divided into three categories: 1984–1987 (time 1, n = 67), 1988–1993 (time 2, n = 40), and 1994–2000 (time 3, n = 27). The cut-off point for time 1 was set at 1987, because the highest rates of seroconversion in the cohort were recorded up to that year. In the remaining years (1988 to 2000), seroconversion dropped to lower levels and stabilized . This long period was arbitrarily divided into two equal time intervals. The age at seroconversion was split into two age groups: younger seroconverters (≤ 30 years) and older seroconverters (> 30 years). The younger group had a mean age at seroconversion of 27 years (n = 36; SD, 2.5; range, 23–30 years). The older group had a mean age at seroconversion of 39 years (n = 98; SD, 6.2; range, 31–58 years). The choice to divide younger and older seroconverters using a cut-off point of 30 years is in accordance with previous studies investigating the behavioural effect of age differences in gay men .
The stratification of the source of HIV transmission according to year of seroconversion and age at seroconversion is presented in Figure 1. It can be seen that the interaction effect reflects the growing probability of only young gay men to be infected by a steady partner as time progressed from 1984 to 2000. To further interpret this effect, the year of seroconversion was entered as a categorical variable into two separate logistic regression models, per age group. As expected, it was found that year of seroconversion was a significant predictor for the change in the source of HIV transmission among young seroconverters only (P < 0.05). Younger participants who seroconverted between 1994 and 2000 had greater odds to be infected by their steady partner than did those who seroconverted between 1984 and 1987 [odds ratio, 11.33; 95% confidence interval, 1.77–72.13]. For older seroconverters, no statistically significant change was found in their HIV infection pattern over time.
We further examined, within each time category, whether the probability for younger seroconverters to be infected by steady partners differed from the probability to be infected by casual partners (Table 1). Proportion tests showed that young seroconverters had higher probability to be infected by casual partners than by steady partners between 1984 and 1987 (P < 0.01). In contrast, between 1988 and 1993 and between 1994 and 2000, their probability to be infected by casual partners did not significantly differ from the probability to be infected by steady partners even though between 1994 and 2000 the rate of infection by a steady partner reached 67%. This is probably due to the relatively small sample size in these cells. For older seroconverters, proportion tests showed a higher probability to be infected by casual partners than by steady partners within all time categories.
The results suggest that young gay seroconverters today have a higher likelihood than earlier in the AIDS epidemic to have contracted HIV from their steady partner. Steady relationships seem to have become a more substantial source of HIV infection for young gay seroconverters. In recent years, the proportion of HIV infections that can be attributed to steady partners even surpassed that of casual partners, reaching 67%. This is in strong contrast with the first years of the AIDS epidemic when the steady partner constituted only 15% of the source of seroconversion among young gay men. Our results suggest also that at any time during the AIDS epidemic, older gay seroconverters were more likely to be infected by their casual sexual contacts than by their steady partners.
The limitations of the present investigation should be discussed before we further debate the results. In general, a working sample size of 134 seroconversions should not be regarded as small, considering the difficulty in obtaining reliable seroconversion data. However, since the number of seroconversions in the cohort declined over time, the findings for later calendar years are based on fewer cases than in early years and are therefore more sensitive to error. Replication of the present findings using larger samples is required to further validate the pattern we have encountered. An additional limitation to the present study is the possible bias in the establishment of the source of HIV infection on the basis of self-report measurements. However, the consistent use of the same measuring methods over the total study period increases the reliability of the trend that was encountered over time, as does the use of questionnaires alongside interviews to determine the source of HIV infection. Combining interview and questionnaire methods with DNA sequencing methods  would have been the best option for determining the source of seroconversion. However, DNA sequencing methods are difficult to implement and were technically infeasible in the present study as we typically had access to the participants only, and not to their steady or casual partners. Finally, some limitations with respect to the generalization of the findings should be noted. Our sample can be considered strongly gay-identified and, as such, may not represent the general population of men who have sex with men. Furthermore, our participants were based in the Amsterdam metropolitan area and had mostly a medium to high level of education. Therefore, caution is required in generalizing results to different settings or groups with other characteristics.
The most important and intriguing issue raised by our findings is the difference between the infection patterns found among older and younger gay seroconverters during the later years of the AIDS epidemic. For younger seroconverters today, infection by the steady partner seems to play a more dominant role than it does for older seroconverters. For the latter group, casual partners seem to have been, and remain, the major source of HIV infection. Our data can not readily account for the differences in seroconversion patterns found between older and younger seroconverters. The most obvious explanation would be that young gay men practice more risky behaviour with their steady partners than older gay men do. As young gay men engage in more high-risk behaviour with steady partners than with casual partners , it appears that young gay men have adopted, over time, safer sex practices with casual partners but to a lesser extent with steady partners. In contrast, older gay men seem more successful in adopting safer-sex practices with steady partners.
We can at this stage only speculate how these conditions have emerged. It could be that young gay men practice less negotiated safety than older gay men do, or that they lack efficacy in the practice of negotiated safety, which exposes them to more risk. It could also be that young gay men have more romantic or infatuated attitudes towards relationships, which facilitate risk-taking behaviour.
Another possibility is that young gay men are more likely to change steady partners more frequently than older gay men do. Such change combined with the frequency with which young gay men engage in high-risk behaviour with their steady partners might put them at higher risk. Finally, it should be examined whether young gay men have, in general, fewer casual partners than older gay men have and, therefore, have fewer chances to be infected by casual partners. Future research is needed to reveal which of these attitudinal and behavioural factors do in fact underlie the observed differences in HIV infection patterns between younger and older gay men. These factors could have important implications for HIV-prevention efforts. There is no doubt that prevention should continue to target high-risk behaviour with casual partners, as our data confirm that this route of transmission is still a prominent source of infection for both older and younger seroconverters of the cohort. However, these findings also emphasize the need for health practitioners and prevention workers to pay specific attention to the sexual behaviour of young gay men with their steady partners. The dominant role that steady partners have acquired as a source HIV infection among young gay men could mean that this route of infection has not been sufficiently addressed in past prevention efforts. It should therefore receive adequate attention if we wish to further reduce the overall spread of HIV in the population of gay men.
The ethical review board of the Municipal Health Service of Amsterdam has approved the study and all participants in the study have signed an informed consent form upon enrolment. We wish to express our gratitude to the participants in the Amsterdam Cohort Studies for their continuous contribution to the battle against HIV and AIDS, Dieuwke Ram for the cohort management, Anneke Krol for the data-management and Lucy D. Phillips for the review of the English.
1. Hospers HJ, Kok G. Determinants of safe and risk-taking sexual behaviour among gay men: a review. AIDS Educ Prev 1995, 7: 74–96.
2. Misovich SJ, Fisher JD, Coates TJ. Close relationship and elevated HIV risk behaviour: evidence and possible underlying psychological processes. Review Gen Psychol 1997, 1: 72–107.
3. Hays RB, Kegels SM, Coates TJ. Unprotected sex and HIV risk-taking among young men within boyfriend relationships. AIDS Educ Prev 1997, 1: 314–329.
4. Buchanan DR, Poppen PJ, Reisen CA. The nature of partner relationship and AIDS sexual risk-taking in gay men. Psychol Health 1996, 11: 541–555.
5. Hof CC, Coates TJ, Barrett DC, Collette L, Ekstrand M. Differences between gay men in primary relationships and single men: implications for prevention: a review. AIDS Educ Prev 1996, 8: 546–559.
6. de Wit J, Hospers H, Janssen M, Stroebe W, Kok G. Risk for HIV infection among young gay men: Sexual relations, high-risk behaviour, and protection motivation.XI International Conference on AIDS.
Vancouver, July 1996 [abstract WE C 3486].
7. Bosga MB, de Wit JBF, Vroome EMM, Houweling H, Schop W, Sandfort TGM. Differences in perception of risk for HIV infection with steady and non-steady partners among homosexual men. AIDS Educ Prev 1995, 7: 103–115.
8. Elford J, Bolding G, Maguire M, Sherr L. Sexual risk behaviour among gay men in a relationship. AIDS 1999, 13: 1407–1411
9. Kippax S, Crawford J, Davis M, Rodden P, Dowsett G. Sustaining safe sex: a longitudinal study of a sample of homosexual men. AIDS 1993, 7: 257–263.
10. Kippax S, Noble J, Prestage G. et al
. Sexual negotiation in the AIDS era: negotiated safety revisited. AIDS 1997, 11: 191–197.
11. Davidovich U, de Wit JBF, Stroebe W. Assessing sexual risk behaviour of young gay men in primary relationships: the incorporation of negotiated safety and negotiated safety compliance. AIDS 2000, 14: 701–706
12. Greenland S, Lieb L, Simon P, Ford W, Kerndt P. Evidence for recent growth of the HIV epidemic among African-American men and younger male cohorts in Los Angeles County. J Acquir Immune Defic Syndr Hum Retrovirol 1996, 11: 401–409.
13. McKusick L, Coates TJ, Morin SF, Pollack L, Hoff C. Longitudinal predictors of reductions in unprotected anal intercourse among gay men in San Francisco: the AIDS behavioural research project. Am J Public Health 1990, 80: 978–983.
14. Vincke J, Bolton R, Miller M. Younger versus older gay men: risks, pleasures, and dangers of anal sex. AIDS Care 1997, 9: 217–225.
15. Imrie J, Fenton KA, Billington A. HIV prevention and homosexual men: should we be optimistic about the new millennium? Sex Transm Inf 1999, 75: 369–376.
16. Mansergh G, Marks G. Age and risk of HIV infection in men who have sex with men. AIDS 1998, 12: 1119–1128.
17. de Wit JBF, van Griensven GJP, Kok G, Sandfort TGM. The effectiveness of the use of condoms among homosexual men. AIDS 1993, 7: 1113–1118.
18. Geskus RB. On the inclusion of prevalent cases in HIV/AIDS natural history studies through a marker-based estimate of time since seroconversion. Statist Med 2000, 19: 1753–1769.
19. Leitner T, Escanilla D, Franzen C, Uhlen M, Albert J. Accurate reconstruction of a known HIV-1 transmission history by phylogenetic tree analysis. Proc Natl Acad Sci USA 1996, 93: 10864–10869.