Sexual Behaviours of Homosexual and Bisexual Men in France: A Generational Approach

Objective In high-income countries, the social and epidemiological contexts surrounding homosexuality and AIDS have changed profoundly in recent decades. This work sought to examine key indicators of the long-term sexual trajectories of successive generations of men who have sex with men (MSM) in France. Methods We performed a longitudinal analysis of the French Gay Press surveys, which were self-administered socio-behavioural questionnaires, repeated from 1985 to 2011 in the gay press, and on the internet in 2004 and 2011. An age-cohort analysis using graphical representations and multivariate logistic regressions was conducted among participants aged 18-59 (N=38 821). Results First sexual intercourse occurred more often with a male partner in younger generations than in older ones: 76.0% in MSM who turned 18 in 1956-1959, 75.6% in 1980-1983, 83.7% in 2008-2011, poverall=0.0002). Every generation showed the same pattern of sexual trajectory between 1985 and 2011: globally, the frequency of masturbation increased from the 1985 survey to the early 1990s and then decreased from the late 1990s to the end of the study period. Inversely, the frequency of oral and anal sex decreased in the mid-1980s and increased from 1990 to 2011. The frequency of both oral sex and anal intercourse is currently quite high, regardless of generation (>95% and around 80%, respectively). Compared to their predecessors, recent generations of young MSM reported more frequent oral and anal sex, but fewer male partners in the previous 12 months. Discussion While the increased frequency of first intercourse with a man over successive generations since the 1970s may be related to reduced social pressure for heterosexuality, there is evidence that sexual norms among MSM are widespread, with practices spreading across age groups and generations. Although AIDS profoundly affected sexual practices in the 1980s, further AIDS-related events (discovery of HIV antiretroviral drugs and their use in prevention) do not appear to have accentuated ongoing trends in sexual practices.

evidence that sexual norms among MSM are widespread, with practices spreading across age groups and generations. Although AIDS profoundly affected sexual practices in the 1980s, further AIDS-related events (discovery of HIV antiretroviral drugs and their use in prevention) do not appear to have accentuated ongoing trends in sexual practices.
This kind of approach has a unique set of supporting documentation in France: the Enquêtes Presse Gay (Gay Press Surveys) [33][34][35], which are repeated nationwide surveys that collected data on how MSM adapted to the HIV/AIDS epidemic from 1985 to 2011. The data permit a retrospective analysis of trends in sexual behaviour over the past six decades, starting at first intercourse, and provide more than 25 years of hindsight concerning sexual practices, from the first signs of the AIDS epidemic up to 2011. Furthermore, the questionnaires are a source of detailed individual information on MSM and their lifestyles, health, sexuality and preventive behaviours.
Here we aim to use this series of French surveys to study the sexual trajectories of successive generations of MSM, from those who started their sexual life before the AIDS epidemic to those who began theirs in the current context of a new HIV prevention paradigm that includes the use of antiretroviral drugs. These trajectories are described in terms of conditions of sexual debut, number of male partners, and sexual practices such as masturbation, oral sex and anal intercourse.

The Gay Press surveys and their design
The Gay Press surveys are anonymous self-administered socio-behavioural questionnaires (convenience samples) that were repeated from 1985 to 2011 (annually from 1985 to 1993, then in 1995, 1997, 2000, 2004 and 2011). The media recruitment plan changed over time. From 1985 to 1992, a gay magazine with a large nationwide readership (Gai Pied Hebdo) included a card containing the questionnaire. When this journal ceased printing, several magazines were included in the survey from 1993 to build up the same variety of profiles [36]: 6 gay magazines in 1993, 10 in 1995, 9 in 1997, 20  All questionnaires were completed anonymously. The French data protection committee (Commission nationale de l'informatique et des libertés) approved the surveys and data collection and storage.

Variables for analysis
We compared the data from surveys from which common indicators could be created, restricting analysis to participants aged 18-59 (N = 38821) due to the relatively sparse participation of respondents aged >59. Year of the 18 th birthday (= year of birth + 18) was used as the variable defining the cohort/generation, as reference to the starting point of sexual life: 18 was the median age of first sexual intercourse with a man (see Results section) in our samples. This variable allowed us to study MSM who had begun sexual activity (i.e., turned 18) from 1944 to 2011 (1944 corresponding to the oldest MSM included, i.e., those aged 59 in 1985) ( Table 1). Other socio-demographic variables were education level, marital status, self-identification and size of the city of residence. We also took into account the source of the questionnaire (press or internet).
The sexual debut was described by the sex of the first sexual partner and the age at the first intercourse with a man, from data in surveys from 1995 to 2011. First intercourse is used here to translation for premier rapport sexuel in French. No definition was provided in the surveys; it refers to first sexual experience. Individuals interviewed with this question and then asked about their interpretation of the questions generally understood the term as indicating penetrative sexual relations [37]. Questions used for the analysis of sexual debut were: "How old were you when you had your first sexual intercourse with a man?" and "How old were you when you had your first sexual intercourse with a woman?". When reported age was the same for both, the sex of the very first partner could not be determined and was considered as missing. This concerned only 4.6% (723/15759) of the respondents of the 1995-2004 surveys. In 2011, a supplementary question provided this information.
We examined the percentages of men who had more than 10 male partners in the previous six months from 1985 to 1990 and in the previous 12 months from 1991 to 2011. We also analysed the rates at which reciprocal masturbation, oral sex and anal intercourse were reported as frequent practices with any partner (steady or casual) from 1985 to 2011. The practice was coded frequent if the respondent answered "always" or "often" (vs. "sometimes" and "never"). Questions referred to different recall periods according to survey year: i) no recall period before 1989; ii) recall period of six months for 1989-1990 and iii) 12 months after 1990.

A preliminary analysis: Comparability of the samples
The Gay Press surveys are convenience samples obtained from varying sources. Trends observed can thus result from real changes in sexual behaviours or from changes in the socio-demographic structure of the respondents. One way to verify the comparability of the samples is to test if indicators that are not supposed to change over time within a given generation are actually stable. For instance, age at first intercourse reported by MSM of the 1980-1983 generation should be the same in the 1995 survey when they are 30-33 years old as in the 2004 survey when their age is 39-42. This means that, after controlling for generation, no significant effect of the survey year should remain. We used linear and logistic models to test this assumption for education level (higher education in respondents aged 20), sex of the first partner and age at first intercourse with a man; survey years were included as dummy variables. Models were also adjusted for age at survey to take into account the participation bias linked to age in this sort of convenience survey [38,39]. Age and generation were included in the models as continuous variables after fractional polynomial transformation when it improved model fit [40].
Results of this preliminary analysis showed that while the survey year had a highly significant effect (p year <0.0001) on whether subjects had at least a university degree in a bivariate logistic regression including age, it was no longer significant after further adjustment for generation (p year = 0. 20). No year effects remained significant concerning first sexual intercourse with a man (p year = 0.80) or on age at that event (p = 0.93). These results support the stability of the reporting at the generational level and the comparability of the samples.

Statistical analysis
The categorical variables were reported as percentages, and distributions were compared with Pearson's Chi-squared test. We tested linear trends in bivariate logistic models with the survey year included as a continuous variable. Concerning quantitative variables, mean age differences between surveys were assessed with a bivariate linear model in which the survey year variable was included as a categorical variable. The median age at first intercourse with a man was estimated by linear interpolation.
Graphical representations were used to compare sexual itineraries of the successive generations. To take age-related participation bias (see above) into account, percentages of first intercourse with a man by generations were calculated from an age-adjusted logistic model (Fig 1), and median age at first intercourse with a man was represented by age group (Fig 2). Standard age-cohort representations on raw data were used for the number of male partners (Fig 3) and sexual practices (Fig 4).  Finally, a multivariate logistic regression was conducted to examine factors associated with reporting more than 10 partners in the previous 12 months from the 1991 to the 2011 surveys. Age and age squared were included in the model to adjust for the quadratic effect of age on the dependent variable. We used the Wald test to check for interactions between each of the confounding variables (questionnaire source, self-defined sexual orientation, size of city of residence and level of education) and age (linear term) and generation.
Stata 13 was used to manage the data of the 13 available surveys and statistical analyses.

Results
Profiles of MSM in the surveys (Table 1) The mean age of the respondents varied between 31.   internet samples than in the press samples: in 2004, 6.1% overall, 4.8% from the press, 10.5% from the internet (p<0.0001); in 2011, 10.0% overall, 4.1% from the press, 10.6% from the internet (p<0.0001). The proportion of participants who refused to define themselves decreased throughout the 1985-2011 period (linear trend: p<0.0001).
Sexual debut Fig 1 represents the age-adjusted percentages of respondents who reported that their first sexual intercourse was with a man, by generation. It shows that younger generations had their first intercourse with a man more frequently than the older ones did: 76.0% in the 1956-1959 generation, 75.6% in the 1980-1983 generation, and 83.7% in the 2008-2011 generation (p overall = 0.0002). Fig 2 shows median age at first sexual intercourse with a man according to year of 18 th birthday. The median age at this event remained stable at around 18, beginning with the 1970 generation. As explained above, younger respondents (aged 18-24) had a lower median age at first intercourse (around 17) due to participation bias; after adjustment for age, age at first sex did not differ between the 1968-1971 generations and those that followed them (p = 0.21).

Number of male partners
The percentage of respondents reporting more than 10 male partners in the previous six months decreased between the 1985 and 1987 surveys: 27.2% in 1985, 15.7% in 1987, then 17.1% in 1988 and 17.0% in 1989. Similar trends were observed in every generation (data not shown). Fig 3 shows that age had a global effect on the number of sexual partners: those at the beginning of sexual life had more than 10 male partners less often than older men (30 or older). A difference between generations also existed: comparison of the curves shows that, at a given age, from the oldest to the 1980-1983 generation, the number of sexual partners increased in each successive generation (i.e., the curve is higher for each newer generation). After the 1980-83 generation, however, this trend tended to reverse: recent generations of young MSM reported fewer male partners than older generations at the same age.
The multivariate analysis of data from the 1991-2011 surveys (Table 2)    The most marked change in these trajectories occurred around the year 1990, for masturbation as well as for oral sex and anal intercourse. The corresponding trends are, however, reversed between masturbation, on the one hand, and oral sex and anal intercourse on the other: as frequent masturbation increased in the 1980s, frequent oral sex and anal intercourse decreased; and as frequent masturbation decreased in the 1990s and 2000s, frequent oral sex and anal intercourse increased.
The change in the recall period might have played a role, although it is hard to argue that the frequency of a practice is likely to differ much "in the past six months" versus "in the past 12 months". In any case, even if we excluded the 1989 and 1990 surveys, which included the "past six months" questions, these trends would still be present.
Increases in oral sex and anal intercourse in each successive generation. At any given age, the most marked differences between generations concerned anal intercourse, and, to a lesser extent, oral sex. Their frequencies increased in each successive generation. For instance, at around 32 years old, frequent anal intercourse (Fig 4c) was reported by 60.3% of the MSM of the 1972-1975 generation, 68.6% of the 1980-1983 generation, 81.7% of the 1988-1991 generation, and 84.1% of the 1996-1999 generation. Contrary to our observations about the number of sexual partners, for which the younger generations reported fewer partners than older ones, MSM of the generations 1996-1999 reported frequent oral sex and anal intercourse more often than older generations at the same age. Frequent oral sex has reached a very high level (> 95%) and frequent anal intercourse a high level (about 80%) today. They are both reported by MSM even at the beginning of their sexual life.

Discussion
In our study, since the 1970s, the more recent the generation of MSM, the more often their first sexual experience was with a man. Age at first sexual intercourse with a man has nonetheless remained stable, with a median age of about 18. Every generation showed the same pattern of sexual trajectory between 1985 and 2011: while masturbation increased from 1985 to the early 1990s and decreased thereafter, the number of male partners and the frequency of oral and anal sex decreased in the mid-1980s and have increased since 1990. Recent generations of respondents were more likely to report frequent oral sex and frequent anal intercourse than older ones, but also fewer male partners in the past 12 months. As community-based convenience samples, these surveys are not representative of the French MSM population [41,42]. A non-probability sampling scheme is however used most often to survey MSM because they are a so-called "hard-to-reach" population and they represent too small a percentage of the general population to be recruited in nationally representative surveys [43]. Nonetheless, comparison of a series of convenience surveys remains possible as long as there are no differential biases between samples. In our study, comparability is strengthened by the intra-generational stability over time of some indicators concerning past events (university degree, sex of first partner and age at first intercourse with a man). The larger proportion of bi and heterosexual men in recent surveys (mainly in 2011) is likely due to the greater propensity of bisexuals today to answer this kind of survey because they are closer to the media that relay it (internet in particular) rather than to a strong change in sexual behaviours.
Furthermore, convenience sampling methods are known to induce participation bias linked to age [38,39]. This is a major issue in this work for two reasons: first, self-selection of participants according to their age is particularly obvious in gay surveys: participating in such surveys involves buying a gay journal or spending time at gay websites, and answering questions about sexual behaviours. Younger participants are likely to be the most self-assured of their generation about their own homosexuality. This point is illustrated by the finding that the youngest participants (18)(19)(20)(21)(22)(23)(24) enrolled in the surveys had a median age at first intercourse with a man lower than that of older participants. We may also suppose that older MSM who continued to participate in the surveys were the most involved in sexuality and/or invested in gay life. The second reason is that we conducted an age-generation analysis, for which knowledge of the age effect is essential for interpreting the sexual trajectories of the successive generations of MSM. Accordingly, we used bi-and multivariate models that allowed us to adjust for age in addition to other potential confounding factors, such as educational level, size of city of residence, selfdefined sexual orientation and source of the questionnaire.
Even though these methodological issues require that our results be interpreted cautiously, this work presents an original longitudinal analysis that offers a new perspective on sexual behaviour trends.
Our results describe changes in sexual itineraries of MSM associated with trends in the social context surrounding homosexuality. An earlier study of the Gay Press surveys, from 1985 to 1995, showed that MSM who had their first sexual experience before 1970 (i.e., before the emergence of social movements for the recognition of marginal sexualities) were more likely to have lived a heterosexual life (for example, had married) than more recent generations [44]. Our finding that since the 1970s the first sexual experience has progressively occurred more and more frequently with a man shows that improved social acceptance of homosexuality and lesser social pressure for heterosexuality have also translated into sexual behaviours from the beginning of the sexual life. In France, the decriminalisation of homosexuality in 1982, the legal recognition of same sex couples in 1999 [45] and civil marriage in 2013 are all major legislative steps in the progress of sexual minorities. They followed the social movements of the 1970s and the gay mobilization and advocacy that has accompanied the increase in social acceptance since the mid-1980s. Nevertheless, this more favourable social context has not eliminated the difficulties that some MSM still experience in living with their homosexuality [30,[46][47][48].
Our results also show evidence of widespread community norms in sexual practices. The trends observed in the pre-ART era for number of sexual partners and sexual practices were the same for all generations, regardless of age. AIDS had a deep impact on the sexual practices of all MSM, with a drop in the number of male partners in the 1980s, less frequent oral sex and anal intercourse, and more frequent masturbation. This can be seen as a safety reaction during a period of uncertainty and lack of scientific knowledge about AIDS. These changes in sexual behaviour were in accordance with the first community-based preventive interventions conducted in France by gay associations in the mid-1980s to inform MSM about AIDS [49]. In 1987, the first national campaign directed at the general population promoted the use of condoms [50]. Meanwhile, studies agreed that the risk associated with oral sex was low. Thus, MSM adopted condoms for anal intercourse and reengaged in penetrative practices, with an increasing number of partners in the 1990s. This occurred during a period of AIDS mortality rates that remained high (the number of AIDS deaths in France peaked in 1994 [51]).
Similar trends in the frequency of sexual practices were also observed among all generations in the post-ART era. In particular, we observed no differences during that period between MSM who experienced the dark years of AIDS in the 1980s and 1990s and those with sexual debuts after the efficacy of antiretroviral drugs was demonstrated in 1996. Practices seem to have spread in the community across age groups and through successive generations, although the community landscape had been changing profoundly for several decades and the "gay community" is a complex and protean concept [52]. Social community norms in terms of sexual practices remain strong, regardless of the development of new areas of sexual socialization, such as the internet [53].
Although the youngest generations ( 1996-1999) reported more frequent oral sex and anal intercourse than older ones at the same age, they reported more than 10 male partners in the past 12 months less often. This generational trend remained after adjustment for self-identification and source of questionnaire. This result seems quite concordant with the decrease in the number of partners observed in the US since the beginning of the 2000s [16,20] and may reflect differential evolutions in sexual networks, which may mostly concern the young generations of MSM.
These trends also revealed that epidemiological events and preventive concerns have not necessarily had an immediate and sustainable impact on sexual behaviours. Contrary to the emergence of AIDS and the first HIV preventive policies in the 1980s, it does not appear that either the onset of the ART era in 1996 or the new preventive strategies based on the use of antiretroviral drugs (such as Treatment as Prevention and Pre-Exposure Prophylaxis) in the late 2000s have modified old trends or generated new ones in the sexual practices studied here. If anything, they might rather have sustained emerging trends that had started earlier.
This longitudinal analysis also revealed an effect of age different from that usually seen in transversal analyses. In particular, we found that the frequency of sexual practices may increase lifelong, even at older ages.
Finally, similar trends observed in every generation from 1985 to 2011 may be interpreted in terms of strong period effects and sexual behaviours mostly influenced by contextual factors.
On the one hand, past experience of some generations, such as those profoundly touched by AIDS in the 1980s and early 1990s, appeared not to affect their subsequent sexual practices differentially. On the other hand, the sexual practices of the younger generation showed no especially strong features, except fewer male partners. For instance, they reported more frequent anal intercourse than the older generations at the same age, but this was true for each generation compared with the older ones.
This analysis of large nationwide surveys showed that global trends in sexual behaviours may not necessarily respond to and indeed may precede changes in the epidemiological and preventive context. It underlined how difficult it can be for preventive policies to deal with the social dimensions of sexuality. A more detailed analysis of preventive practices of MSM will provide further information on these itineraries.