On the basis of self-report measurements and post-seroconversion interviews, the primary partner is the source of infection in 50% of cases of HIV seroconversion that occurred within the Amsterdam Young Gay Men Cohort Study between 1995 and 1999 (manuscript in preparation). This percentage should serve as a reminder that the primary relationship should be recognized as a notable source of HIV infection among young gay men. Initially, most of the attention given to the measurement of sexual risk-taking behaviour among young or older gay men did not differentiate between steady and casual partners. Studies that specifically investigated risk-taking behaviour within steady relationships usually reported higher rates of unprotected anal intercourse (UAI) with steady partners in comparison to casual partners [1–8] [H.J. Hospers, J.B.F. De Wit, M.W. Ross, G. Kok, in preparation].
In sexual contacts with casual partners, almost any incident of UAI can be regarded as risky. The assumption is that one does not know or can not know the valid serostatus of one's casual partner and therefore can not engage in a correct risk appraisal. Such is not the case with steady partners. Knowledge of the steady partner's sexual history [3,9,10], familiarity with the steady partner in general [3,11], monogamy , trust [3,10,12], love [9,10,12], and intimacy [10,11,13] might all influence risk appraisal and the decision-making process regarding engagement in UAI. However, aspects such as monogamy, trust, love, and intimacy are subjective heuristics that do not necessarily guarantee safety. Only few objective criteria are commonly accepted, although not without debate [14,15], to be valid premises for allowing UAI with steady partners: (i) both steady partners tested HIV-negative; and (ii) both partners are either monogamous or have agreed to have only safe sex (e.g. protected anal intercourse) outside of the primary relationship. The fact that some couples intentionally use the above conditions to create a context that allows them to have low-risk UAI with each other, has received growing attention since Kippax et al.  coined the term `negotiated safety' (NS) to describe this strategy.
The use of NS strategies in the decision making of couples who consider having UAI with each other brought about the distinction between low-risk UAI and high-risk UAI. It was our goal to measure the rate of risky UAI with steady partners among the participants of the Amsterdam Young Gay Men Cohort Study. To do so, the rates of UAI, HIV-seroconcordance, and sexual agreements among steady partners were measured so we could determine how many of them practised NS. We then corrected UAI rates for NS. In addition, we took into consideration compliance with NS agreements to reach a final and accurate estimation of the prevalence of risky UAI within steady relationships. NS agreements that have not been complied with were regarded as a risky setting for the practice of UAI. Correction for NS compliance has been, up until now, seldom applied in the measurement of risk behaviour in steady relationships.
In addition, we wanted to compare the rates of risky UAI during sexual contacts with steady versus casual partners. It was assumed that, after correcting UAI rates for NS and NS compliance, the rates of risky UAI with steady partners would be lower than the initial UAI rates. We were interested to know whether the corrected rates of UAI with steady partners would be higher than UAI rates with casual partners. The assumption is that after correction for NS and NS compliance, the remaining UAI rates can not be explained by any correct form of risk-reduction strategy. If the corrected rates of UAI with steady partners are indeed higher than those with casual partners, this can add support to the assumed existence of a risk-facilitating context, which is specific for steady relationships.
For the investigation, we used the first data wave of the Amsterdam Young Gay Men Cohort Study dating from May 1995 to December 1996. The Amsterdam Municipal Health Service initiated the study to provide epidemiological, behavioural and psychosocial data regarding the population of young gay men in the Amsterdam region in relation to the AIDS epidemic. Only the first wave data were used in order to measure the `naïve' rates of UAI and NS behaviour. It was assumed that sexual risk behaviour and the motivation to engage in NS might have changed in the subsequent data waves because of the effect of repeated HIV testing. Repeated HIV testing is required for participation in the cohort study but has no impact on the behaviour captured in the first data wave.
A total of 435 men participated in the first measurement. To qualify for inclusion participants had to be under the age of 30 years and living in the Amsterdam area. If living outside of Amsterdam, participants had to make regular use of the Amsterdam bar and club scene. Furthermore, participants had to have had at least one sexual encounter with a male partner during the 6 months before enrolment. Seventy-two per cent of the participants were obtained via active recruitment strategies. Active recruitment took place among support groups and organizations of young gay men, the sexually transmitted diseases clinic of Amsterdam, and during special events and manifestations where volunteers targeted and approached young gay men. Nineteen per cent of the participants were recruited via the chain referral method, in which participants were asked to recruit their personal contacts. Five per cent of the participants were recruited via advertisements in the (gay) press and on the Internet, 3% via handout distribution in gay bars, clubs and meeting places in Amsterdam, and 1% via non-participant referral.
The participants' mean age at enrolment was 25 years (SD = 3, range 17–30 years). Five per cent of the participants (n = 20) were found to be HIV positive at entry. Eighty-three per cent of the participants were native Dutch and 17% were born outside of the Netherlands. Sixty per cent had a high educational level (college degree or equivalent), 30% had a medium educational level (attended school up to the age of ±18 years), 10% had a low educational level (attended school up to the age of ±16 years). Fifty-two per cent had paying jobs, 36% were students, and 12% 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 = 1, range 1–7).
All contacts with the participants were conducted via a team of nurses. During the intake interviews, the nurses informed candidates of the future procedures and the different aspects involving participation in a long-term cohort study. The advantages and disadvantages of performing an HIV test, which is an integral part of the procedures, were thoroughly discussed. For those apprehensive of the needle, the nurses offered an HIV test based on a saliva sample. Participants were, however, informed that they would not receive the results based on the saliva test because the method was used for epidemiological purposes only.
After intake, candidates were given time to reflect and decide whether they wanted to enrol in the study, and whether they wanted to know the results of their HIV test. Those who decided to join were presented with an informed consent form, which was explained and given for reading and signing. After these procedures, participants were ready for the first measurement. Participants went through a short medical anamnesis for epidemiological purposes, and completed a self-administered questionnaire regarding sexual behaviour with casual and steady partners, HIV status, steady relationships, related cognitive-behavioural and psychosocial aspects, and demographic data.
Relationship status and sexual agreements
Participants were asked to state whether they had had a steady relationship in the 6 months preceding measurement, and the length of that relationship. Participants were asked to indicate whether they had reached an agreement with their steady partner regarding sex outside the relationship. If they had, participants were then asked to indicate what the agreement was. Participant could choose between different types of agreement that can be clustered into two main groups: (i) agreements to be monogamous; (ii) agreements that allowed sex with others as long as anal sex was protected.
Unprotected anal intercourse
Participants were asked whether they engaged in anal intercourse, as well as the frequency of condom use during anal intercourse. These variables were measured for both insertive and receptive anal intercourse, and for contacts with casual and steady partners.
Participants were asked to indicate the results of their last HIV test and the last HIV test of their steady partner. We did not use the results of the cohort HIV testing because they reflected only the HIV status of the participants at enrolment and not of their perceived HIV status in the preceding 6 months.
NS in this study was defined as follows: (i) partners engaged in UAI with each other; (ii) both partners know they are HIV negative based on an HIV test; and (iii) both partners have agreed to be monogamous or have no UAI outside the primary relationship. It is important to note that no public campaign has been initiated in the Netherlands to raise awareness of NS at any time before the measurements. Considering that the term NS was, at that time, fairly unknown outside social scientific circles, it is unreasonable to assume that our participants treated NS as a generally recommended health policy or course of action. Therefore, we never referred to NS by name or as a strategy in the questionnaire.
Negotiated safety compliance
NS compliance was defined as the rate of adherence to NS agreements. Non-compliance with NS agreements was established if participants reported incidents of UAI with casual partners while practising NS. This was done by crossing data on NS agreements with sexual behaviour data. By establishing non-compliance directly from the sexual behaviour data we aimed at obtaining valid measurements, while reducing the social desirability response bias that might have had a stronger effect had we asked participants directly whether they had complied with agreements.
Sexual risk behaviour
In the category `risk behaviour', all incidences of UAI with casual partners were included and all incidents of UAI with steady partners with whom NS was not practised or not complied with.
Out of the 435 participants, 66% (n = 285) reported having had a steady partner in the 6 months before measurement. The average length of relationship was 17 months (SD = 21, range 1–120 months). Forty per cent (113/285) of the participants in a steady relationship had had UAI with their steady partner, and 23% (66/285) reported being in an HIV-negative seroconcordant relationship on the basis of a previously conducted HIV test. A total of 33% (37/113) of the participants who had had UAI with their steady partner reported being in an HIV-negative seroconcordant relationship.
Negotiated safety agreements
Twelve per cent (n = 35) of the 285 participants in a steady relationship practised NS. More specifically, 31% (35/113) of the participants who had UAI with their steady partners practised NS. These 35 participants represent 95% of the 37 participants who were in an HIV-negative seroconcordant relationship and who engaged in UAI with each other. Sixty per cent of those who practised NS (21/35) agreed to be monogamous, and 40% (14/35) allowed sex outside the relationship, as long as anal intercourse was protected.
Negotiated safety compliance
To measure compliance with NS agreements, our examination was limited to the participants who practised NS and had had a steady relationship for longer than 6 months—that is, for the complete wave period. This was done in order to avoid overestimation of non-compliance because it was not possible, for technical reasons, to differentiate between sexual contacts with casual partners that occurred before the beginning, during, or after the end of a relationship of less than 6 months' duration. As a result, the sub-sample consisted of 21 participants. The total rate of risky non- compliance, that is, engagement in UAI with a casual partner, was 10% (2/21).
Risk behaviour in steady relationships
We measured the rates of UAI with steady partners before and after correction for NS and NS compliance. In order to correct for compliance, the group examined had to be limited to the participants who had had a relationship for longer than 6 months. Fifty-five per cent (n = 65) of the 118 men who had had a steady relationship for longer than 6 months had UAI with their steady partner. After correction for NS, the risky UAI rate dropped to 37% (44/118). When corrected for NS compliance, the risky UAI rose to 39% (46/118), which is the final estimate of the rate of risky UAI with steady partners.
Risk behaviour with steady versus casual partners
We wanted to examine whether the rates of UAI with casual partners would differ from the rates of UAI with steady partners before and after correction for NS and NS compliance. For this purpose, we compared the group of 143 participants with only casual partners to the group of 118 participants with a relationship of longer than 6 months' duration. The rate of UAI with casual partners for the group of participants with only casual partners was 20% (28/143).
Chi-square tests with continuity correction showed that the uncorrected rate of UAI with steady partners (55%) was significantly higher than that with casual partners (20%), χ2 (1) = 34.0, P < 0.001. After correction for NS and NS compliance, the difference between the rate of UAI with steady versus casual partners remained statistically significant (39 versus 20%), χ2 (1) = 11.0, P = 0.001.
Our findings demonstrate that in order to obtain an accurate estimation of risk-taking behaviour within steady relationships, one has to incorporate NS and NS compliance in the assessment process. Only after correction for NS and NS compliance were we able to obtain the actual rate of risky UAI that poses a real threat of HIV transmission in steady relationships. Furthermore, the corrected rates of risky UAI were almost a third lower than the initial rates of UAI reported by steady partners (39 versus 55%). This suggests that the actual rates of risk-taking behaviour within steady relationships might be lower in reality than is sometimes assumed.
Rates of risky UAI with steady partners remained significantly higher than UAI rates with casual partners, even after correction for NS and NS compliance. After correction for NS, the remaining high rates of UAI with steady partners could not be justified by any other objective risk-reduction strategy. These findings support the assumption that steady relationships create a specific context that facilitates risk-taking behaviour and therefore potential HIV transmission. This assumption is supported by an experimental study of a heterosexual sample by De Wit and Schutten . The authors used two vignettes describing a situation in which participants met a new person and wanted to have intercourse with that person. No factual information regarding the other person was provided. The only difference between the two vignettes was the depiction of the person as a potential casual partner versus a potential steady partner. De Wit and Schutten  found that the mere perception of the new person as a steady partner brought about lower risk perceptions and lower intentions to use condoms. The authors suggested that emotion-based decision processes are related to risk behaviour in steady relationships, whereas information-based processes are related to the same behaviour in contacts with casual partners. It seems therefore that research and prevention efforts need to target steady relationships specifically as a notable source of HIV transmission.
Some limitations to this study must be noted. The relatively small sample size used, particularly in the sub-samples examining NS and NS compliance, calls for caution. The resulting conclusions should, therefore, be considered preliminary, with the need for replication with larger sized samples. Furthermore, the validity of the HIV status concordance established between steady partners and, consequently, the validity of the NS practice rate which was detected, could be limited. We established the HIV status of the participants and their steady partners according to the participants' report of the results of previously conducted HIV tests. The fact that cohort HIV testing could not be used to confirm these reports limits their validity. This may result in an underestimation of the true level of risk-taking behaviour in the cohort. Nevertheless, we should assume that a hypothetical false HIV status reported by any of the participants was not premeditated. It is better to assume that the reported HIV status, even if invalid, did reflect the perceived HIV status at the time, and therefore did not equal sexual risk-taking behaviour that involved clearly unknown or discordant HIV status. Another limitation that might result in the underestimation of sexual risk behaviour is that we did not take into consideration the possible engagement in risk activities other than UAI (e.g. oral sex with ejaculation).
Some limitations with respect to the generalization of the findings in this study should also be noted. Our sample can be considered strongly gay-identified, and as such, might not be representative of the more broad population of men who have sex with men. Furthermore, our participants were based in the Amsterdam metropolitan area, had a medium to high level of education, and had a young mean age of 25 years. Therefore, caution is required in generalizing the results to different settings or group characteristics.
As for compliance with NS agreements, we found a 10% risky non-compliance rate with NS agreements. Our findings are similar to findings inferred from data published by Kippax et al.  regarding 144 gay men who practised NS. Kippax et al.  found an 11% risky non-compliance rate with the NS agreement to always have protected anal intercourse outside the relationship. Our findings are also in agreement with findings by E.M.M. De Vroome, W. Stroebe, T.G.M. Stafford, J.B.F. de Wit, G.J.P. van Griensven (in preparation), who detected a slightly higher percentage of 17% of risky non-compliance with similar NS agreements in a sample of 242 gay men, with an average age of 41 years, from the Amsterdam area.
It was further assumed that the reported rates of non-compliance with NS agreements are an underestimation of the actual non-compliance rates. We believe that there could be a social desirability bias in reporting engagement in UAI with casual partners, especially if it suggests non-compliance with agreements reached with the steady partner. We also typically see only one member of the steady couple and cannot examine directly the non-compliant behaviour of the steady partner who is not an active participant in our study. The non-participants' behaviour could cause the rates of non-compliance for the same relationship sample to rise. In addition, our data cover a period of 6 months only. This is just a fraction of the time in which couples might have been practising NS. The rates of UAI, measured over longer periods, might be significantly higher, as suggested by De Wit and Griensven . More research is needed to determine the prevalence and dynamics of non-compliance with NS agreements. Samples should be larger than our present sample, involve both partners in the relationship, and examine longer periods in the relationships.
We do, however, join Kippax et al.  in noting that those who practise NS, or advocate its use, should take rates of non-compliance into consideration. NS can provide the necessary protection from HIV infection, but incidences of non-compliance do occur and involve risk. We believe that those who practice NS would do well to add an additional condition to their agreement: the obligation to report incidents of non-compliance to the steady partner. It must be made clear that the disclosure of incidents of risky non-compliance is an inseparable part of the NS agreement. Furthermore, it should be emphasized that after an incident of risky non-compliance, the NS process must start anew if the steady partners wish to continue to engage in UAI with each other. The re-establishment of HIV-negative seroconcordance and the affirmation of an existing or a new NS agreement must take place.
Finally, only 12% of our couples practised NS. When we examined only the couples who engaged in UAI, we saw that a mere 31% practised NS. The question is why the rates of NS in our sample of young gay men are so low. The answer might rest in the fact that the low NS rates are related to the low percentage of couples in our sample that established HIV-negative seroconcordance (23%). It could be that there was a deterring aspect or simply a lack of motivation for many of the couples to establish seroconcordance with an HIV test. When we looked specifically at the steady partners who did establish negative seroconcordance and had UAI with each other, we saw that, indeed, 95% have had NS agreements. Further research should investigate the factors influencing the establishment of HIV seroconcordance in steady relationships. We must, however, be aware of the present dynamics in the field of HIV testing. These days, the motivation to be tested might become higher as antiretroviral therapy makes HIV testing more important as a health protective behaviour. As it becomes more evident that commencing antiretroviral therapy at an early stage after HIV contraction bears better treatment results , we may observe an increase in HIV testing. It will be intriguing to examine whether the availability of antiretroviral therapy does lead to more readiness for HIV testing, and if it does, whether this gives rise to more practice of NS.
The authors wish to express gratitude to the `AIDS-Fonds' for financing this research project, Professor Dr Roel Coutinho for his support and critical reading of the manuscript, Nel Albrecht and Dieuwke Ram for the cohort-management, Thaddeus C. Breen for the English review, and Anneke Krol and Marja Dekker for the data management.
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. Vroome EMM, Sandfort TGM, van den Bergen HSP, Keet IPM, van den Hoek JAR. Jonge homoseksuele mannen: psychosociale determinanten van onveilig seksueel gedrag (young homosexual men: psychosocial determinants of unsafe sexual behaviour). Tijdsch Seksuol 1995, 19: 4–18.
3. Misovich SJ, Fisher JD, Coates TJ. Close relationship and elevated HIV risk behaviour: evidence and possible underlying psychological processes. Rev Gen Psychol 1997, 1: 72–107.
4. 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.
5. 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.
6. 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.
7. 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. XIth International Conference on AIDS
. Vancouver, July 1996 [Abstract WE C 3486].
8. 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.
9. McLean J, Boulton M, Brookes D. et al
. Regular partners and risky behaviour: why do gay men have unprotected intercourse? AIDS Care 1994, 6: 331–341.
10. Boulton M, McLean J, Fitzpatrick R, Hart G. Gay men's accounts of unsafe sex. AIDS Care 1995, 7: 619–630.
11. Remien RH, Carballo-Diéguez A, Wagner G. Intimacy and sexual risk behaviour in serodiscordant male couples. AIDS Care 1995, 7: 429–438.
12. Carballo-Diéguez A, Dolezal C. HIV risk behaviors and obstacles to condom use among Puerto Rican men in New York city who have sex with men. Am J Public Health 1996, 86: 1619–1622.
13. McNeal LM. The association of idealization and intimacy factors with condom use in gay male couples. J Clin Psychol Med Settings 1997, 4: 437–451.
14. Kippax S, Noble J, Prestage G. et al
. Sexual negotiation in the AIDS era: negotiated safety revisited. AIDS 1997, 11: 191–197.
15. Ekstrand M, Stall R, Kegels S, Hays R, De Mayo M, Coates T. Safe sex among gay men: what is the ultimate goal. AIDS 1993, 7: 281–282.
16. 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.
17. De Wit JBF, Schutten M. Sexual risk with romantic partners, is it based on informational or emotional decision-making? XIIth International Conference on AIDS.
Geneva, July 1998 [Abstract 14367]
18. De Wit JB, Griensven GJ. Time from safer to unsafe sexual behaviour among homosexual men. AIDS 1994, 8: 123–126.
19. Rosenberg ES, Billingsley JM, Caliendo AM. et al
. Vigorous HIV-1-specific CD4+ T-cell responses associated with control of viremia. Science 1997, 278: 1447–1450.