THE IDEA THAT A SMALL proportion of the population, the core group, is responsible for the maintenance and spread of sexually transmitted diseases (STDs) has become an established paradigm in STD epidemiology. 1 The potential usefulness of this concept stems from the implication that interventions targeted at the core group members could reduce their STD rates to the point that disease would cease to be sustainable in the larger population. This potential has yet to be fully realized, however, because precise identification and targeting of core group members has proven extremely difficult. One reason for this difficulty is that the mathematical definitions used to identify core group members are based on the amount of disease transmission each individual is responsible for, which is difficult to ascertain. Such definitions also do not provide information about easily identifiable characteristics (such as occupation) that might be used by an intervention program to identify core group members. 2 For this reason, a number of surrogate definitions of the core group have been suggested, which are less specific but facilitate identification of possible core group members. Clinical–epidemiologic definitions include STD clinic attendees and their contacts or geographic clusters of cases. Sociocultural definitions encompass occupational high-risk groups such as prostitutes, truck drivers, or military personnel. The core group can also be approximated behaviorally based on the number of sexual partners within a given time period. 2
A second factor complicating the effective targeting of core group members is that core sexual networks can be fluid and evolve before an intervention can be implemented. 3 In particular, it is currently not well understood whether the core group consists mainly of people who engage in high-risk sexual behavior over prolonged periods or if there is a larger group of people whose behavior places them in the core group only occasionally. 4 The latter scenario implies continual migration in and out of the core group. The only previous calculation of the migration rate out of the core group was by Stigum and colleagues, 5 who estimated an annual migration rate of 12% in a Norwegian population. They also showed that migration out of the core significantly affects the prevalence levels of various STDs.
The aim of our study was to shed more light on both these issues, by assessing sexual behavior longitudinally to measure the amount of migration, and by examining the socioeconomic and behavioral factors associated with core group membership over time. The surrogate for the core was defined by the rate of partner change. Currently, there is no consensus on what rate of partner change, over what period, best approximates the core. Laumann and Youm 6 defined the core group as those with 4 or more partners in 1 year. Stigum et al. 5 defined the core group as those with an average of 5 or more partners per year over a number of years, and also showed that this definition corresponded to a core with a reproductive number (or rate) Ro>1 for gonorrhea, chlamydia, and HIV, thus fulfilling the mathematical definition of a core group. Because these 2 types of definitions have the potential to identify core groups with different memberships, we chose 2 analogous definitions (as described in Methods) and compared their migration rates and the socioeconomic and behavioral characteristics of their members.
We report results from a cohort study of both men and women, which collected data on the sexual behavior of nearly 1000 participants at ages 18, 21, and 26.
Participants were enrolled in the Dunedin Multidisciplinary Health and Development Study, a large longitudinal study of a cohort born in Dunedin, New Zealand, between April 1, 1972, and March 31, 1973. Sample members were first organized into a cohort at 3 years of age, when 1037 of 1139 eligible children were seen. 7 That group was followed up every 2 years until age 15, and again at ages 18, 21, and 26 years. By age 26, there were 1019 known survivors of the original cohort.
The primary variable of interest was the number of heterosexual partners reported by respondents. We decided to exclude homosexual partnerships from the analysis, because there were not enough of them to analyze separately, and we judged that analyzing homosexual and heterosexual partnerships together could complicate interpretation of the data.
Participants were first asked about their number of sexual partners at the age 18 assessment, during 1990 to 1991. The specific question asked was, “How many people have you had sex with in the last 12 months?” Although this question does not explicitly exclude same-sex partners, 2 subsequent questions were asked of the men: “Were any of your partners male?” and “How many male partners have you had in the last 12 months?” Only 2 men reported having any male partners, and the number of male partners was subtracted from the total number of partners initially reported. For women, no questions about same-sex partners were asked at age 18. The questions asked at age 18 were based on the World Health Organization's partner relationships questionnaire 8 and were presented by computer. An interviewer was present, placed so she could not see the sample members’ responses to the questions. She provided instructions on the use of the computer and aided those with very low reading ability, if requested.
At ages 21 and 26, the presentation format was the same, but the questions on sexual activity were based on those used in the 1990 British National Survey of Sexual Attitudes and Lifestyles, 9 which had been developed after extensive piloting. The survey asked about the number of heterosexual partners, condom use, and concurrent (ie, simultaneous) partnerships. The specific questions asked were: “Altogether in your life so far, with how many men/women have you had sexual intercourse?” Similar questions were asked about “in the last 3 years” (age 21), “in the last 5 years” (age 26), and “in the last 12 months” (ages 18, 21, and 26). At age 21 and 26, we also asked, “The last time you had sexual intercourse with a man/woman was a condom used?” and The first time you had sex with your current partner, were you having a sexual relationship with anyone else?” (concurrent partnership). At age 21, participants were asked about their age at first sexual intercourse.
The participants were asked at age 21 if they had ever, and at age 26, if in the last 5 years, they had had an STD. They were asked to identify the condition (or conditions) from a list of common STDs. If they indicated they did not know the name, they were asked what problems it caused. People were considered to have had an STD if they reported bacterial or viral sexually transmitted infections, trichomoniasis, or an unnamed condition that caused pain on passing urine or abdominal pain. Cases of genital warts and/or herpes in the 5 years up to age 26 in people who had also reported the same infection at age 21 were assumed to be recurrent cases.
At age 26, data were also collected on the highest level of school reached (ie, excluding all post-high school education). In Table 4 for “Highest School Education,” participants having achieved a diploma or below were categorized as “low,” sixth form certificate/higher school certificate only as “medium,” and University entrance/bursary/scholarship as “high.”
The family's socioeconomic status (SES) at age 15 was derived from the reported occupations of the sample members’ parents using the Elley Irving socioeconomic index for New Zealand occupations, condensing the 6 levels in the index into high, middle, and low categories. 10 When more than one parent was working, the highest classification was chosen.
The results were analyzed using SPSS (Chicago, IL). EPI-INFO version 6 statistical software (Atlanta, GA) was used for the chi-squared tests to check for differences in behavioral and socioeconomic distributions between core and noncore group members for men and women.
Two core group definitions were chosen. The first was those people with 5 or more sexual partners in the year before the assessment. This core was thus defined at ages 18, 21, and 26. The second definition was those people with an average of 5 or more sexual partners per year between sexual debut and age 21 and/or between ages 21 and 26. The migration rates for both core definitions were calculated as by Stigum et al., 5 except with core group size not assumed to be constant. The migration rate was defined as the proportion of people who left the core group each year as a result of a decrease in partner frequency to less than 5 partners per year. If the size of the core group at the first assessment equals nc, the number of people who have migrated out by the next assessment is equal to mt, and t is the number of years over which migration is possible; then κ, the proportion of people who migrate out per year, can be calculated according to the formula:MATH
The value of t varied according to the core definition for which migration was being calculated. For the first core definition, t was the number of years between assessments. For the second, it was the average number of years over which the earlier annual number of partners was calculated.
Of the original 1037 cohort members who were seen at age 3, 1027 were believed to be still alive at age 18, 1020 at age 21, and 1019 at age 26. Of these, 862 completed the questions on sexual relationships at age 18 (83.9% response rate), 935 did so at age 21 (91.7% response rate), and 966 did so at age 26 (94.8% response rate). A total of 822 responded at all 3 ages (80.7% response rate) and 991 (97.3%) at at least one age. Similar response rates were seen for both men and women. At age 26, 21.3% of cohort members were living outside New Zealand but completed the assessment. There were no statistically significant differences between those living in and outside New Zealand in reporting of 5 or more partners in the preceding year.
There was substantial overlap between people grouped into the core group under each definition. Of the 13 women with an annual average of 5 or more partners between sexual debut and age 21, 8 (62%) had 5 or more partners in the year before age 21. Of the 6 women with an annual average of 5 or more partners between ages 21 and 26, 3 (50%) had 5 or more partners in the year before age 26. Of the 23 men with an annual average of 5 or more partners between sexual debut and age 21, 16 (70%) had 5 or more partners in the year before age 21. Of the 26 men with an annual average of 5 or more partners between ages 21 and 26, 21 (81%) had 5 or more partners in the year before age 26. However, the substantial overlap is partially because the core group defined by 5 or more partners in the last year was approximately twice as large as the equivalent core defined by an annual average of 5 or more partners. Thus, the core group defined by 5 or more partners in the last year included most of the participants with an annual average of 5 or more partners as well as a large number of others.
The proportion of all women reporting 5 or more partners in the past year was similar at ages 18 (6.9%) and 21 (7.0%) and then decreased to 4.9% at age 26 (Table 1). For men, the proportion increased from ages 18 (9.6%) to 21 (14.0%) and then also decreased at age 26 (12.6%). For both men and women, the proportion of participants with no reported sexual partners in the previous year dropped sharply between 18 and 21, and dropped further to age 26. In contrast, the proportion of participants with one sexual partner in the previous year increased steadily from ages 18 to 26 for both genders.
In an analysis confined to those who reported one or more partners in the previous year, selected behavioral characteristics are shown in Table 2. For the core defined by 5 or more partners in the previous year, the proportion of participants who had concurrent partnerships at the start of their last partnership was significantly higher for core group members at ages 21 and 26 for both men and women. However, the proportions that reported using condoms at their last intercourse were similar for core and noncore group members, with the exception of women at age 26, among whom core group women were more likely to have used a condom, although this did not reach statistical significance.
For the core group defined by an annual average of 5 or more partners, a greater proportion of core members reported concurrent partnerships at the start of their last partnership, but this was significant only for men between sexual debut and age 21. Condom use was similar between the core and noncore groups, except that core group men between sexual debut and age 21 reported less use, although this was not statistically significant. Reported STD rates were higher for the core group for each time period and gender, and these differences were statistically significant except for men up to age 21.
Socioeconomic and behavioral characteristics of the core and noncore groups, using 2 different core group definitions, are shown in Tables 3 and 4. A significantly higher proportion of men who had first sexual intercourse before age 16 were in the core group at both 21 and 26 years of age (Table 3). This pattern was also seen for women but did not reach statistical significance. In relation to the family's socioeconomic status at age 15 and education acquired by age 18, some nonsignificant patterns emerged for women. Those in the highest SES and education group were less likely to be in the core at age 18 but more likely to belong to the core at age 21. For men, again those with the highest education were less likely to be in the core group at age 18 (P <0.001) but equally likely to be in it at ages 21 or 26.
For the core defined by an annual average of 5 or more partners (Table 4), the numbers were too small for women in the core from ages 21 to 26 (6 women) to permit separate analysis. Significantly higher proportions of men and women who reported sex before age 16 were core group members, in both time periods for men. Less educated men were more likely to be in the core group between sexual debut and age 21 (P <0.05). A similar effect was seen for women but did not reach statistical significance. No significant differences were seen for family SES at age 15.
For the core defined by 5 or more partners in the previous year, 71 women (14.7%) and 133 men (26.0%) were core group members at one or more assessments (Table 5). Of the 29 women in the core at age 18, 7 were still in the core at age 21 (24.7% annualized migration rate). For men, of the 42 in the core at age 18 years, 13 were still in the core group at age 21 (21.3% annualized migration rate). As a consequence of the high migration rate, the long time period between age 18 and 26, and the small initial number of people in the core at 18, only 2 women and 3 men who were in the core group at ages 18 and 21 remained in the core group at 26. Of the 32 women in the core group at age 21, 5 were still in the core group at age 26 (16.8% annualized migration rate). For men, of the 66 in the core group at age 21, 16 were still in the core group at age 26 (15.2% annualized migration rate). Of the 20 women who were in the core group at age 18 but left at age 21, none returned to the core group at age 26. In contrast, of the 23 men who left the core group between ages 18 and 21, 6 returned to the core group at age 26, showing that those who migrate out of the core group could still return later.
For the core group defined by an annual average of 5 or more partners, 18 women (3.8%) were in the core group for at least 1 of the 2 periods: between sexual debut and age 21, or between ages 21 and 26. This includes the 13 women who were core group members between sexual debut and age 21, and the 6 who were in the core between ages 21 and 26. Only one woman was in the core group at both periods. In contrast, the number of men in the core group increased slightly from 23 between sexual debut and age 21 to 26 between ages 21 and 26. Only 6 men were in the core group at both periods, thus 43 men (8.9%) were in the core group for at least 1 of the 2 time periods.
This longitudinal study of sexual behavior at ages 18, 21, and 26 revealed substantial turnover in the core group, indicating that it is not a static group of individuals, but rather a group with constantly changing membership. For this reason, a substantial fraction of the population became core group members at some point (15% of females and 26% of males, using the wider definition), but very small numbers maintained high levels of sexual activity for long enough to persist as core group members for several assessments.
The factor most closely associated with core group membership was early age of first sex for both men and women. A low education level was associated with core group membership at age 18, particularly for men, but at ages 21 and 26 those with higher education were equally or more likely to be core group members. The relationship between family SES and core group membership was unclear.
Comparison of the 2 core definitions revealed broad similarities in their memberships. The core group defined by 5 or more partners in the past year largely included that defined by an annual average of 5 or more partners. Both core groups showed similar trends in condom use and partner concurrency, with the core group defined by 5 or more partners in the last year showing slightly more partner concurrency, whereas the other core group showed less condom use among men before age 21. Although the condom use rates were similar between the core and noncore groups, the increased rates of concurrency in the core group are an additional mechanism through which the core group could disproportionately transmit STDs to the population. 11 Our results, showing that concurrency is associated with a greater number of partners, are unsurprising and similar to those of the recent study by Manhart et al. 12 Reported STDs were also considerably more common in the core group. Similar sociodemographic factors predicted membership of both core groups, with stronger associations for early age at first sex when the core was defined by a high annual average number of partners.
The high rates of participation and retention over time in this longitudinal study afforded a high degree of precision and lack of bias in measuring changing sexual behavior over time. The number of years over which we calculated the annual average number of partners to estimate migration was shorter (4.82 years for women, 4.63 years for men) than that of the other study, which made a comparable estimate (15.9 years). 5 Asking participants to recall sexual activity over a shorter period of time could have made our study less vulnerable to recall bias. A limitation of our data is that we did not ask if participants’ sexual partners were new partners in each time period, and we did not ask for an estimation of their partners’ number of partners.
Comparison With Other Studies
Compared with recent large-scale, population-based surveys on sexual behavior in the United States and Britain, our results showed somewhat higher proportions of the population belonging to the core group. The National Health and Social Life Survey (NHSLS), conducted in the United States in 1992 (79% response rate), found that 8.6% of the population 18 to 24 years of age (males and females) had 5 or more sexual partners in the previous year. Combining our data for males and females at ages 18 and 21, we found a rate of 9.9%. The NHSLS also reported that 5.7% of those aged 25 to 29 had 5 or more partners in the past year. In contrast, 9.1% of the 26 year olds in our study reported 5 or more partners in the previous year.
The differences are more striking when compared with the 1990 to 1991 National Survey of Sexual Attitudes and Lifestyles in Britain (63% response rate). The proportion of respondents reporting 5 or more partners in the previous year was 3.5% for men aged 16 to 24, 1.9% for men aged 25 to 34, 1.0% for women aged 16 to 24, and 0.3% for women aged 25 to 34. Although once again exact comparisons between age groups are not possible, clearly a much higher proportion of our sample was in the core group.
The higher number of partners reported in our study could partly be the result of greater disclosure, because of computer presentation, of stigmatized behavior, 13 although recent results from the year 2000 version of the British National Survey on Sexual Attitudes and Lifestyles showed no difference in reported sexual behavior between respondents answering computer-generated or paper-based self-administered surveys. 14 Our higher response rate also does not appear to account for the more extreme behavior we found, because respondents who were difficult to recruit showed no difference in their reported numbers of partners compared with those respondents who were easy to recruit (Poulton R, personal communication, December 2002). Nevertheless, greater disclosure of sensitive behaviors is likely in the Dunedin cohort because of trust built up over many years. Cohort effects are unlikely to be responsible for differences with other surveys, because the age 18 and age 21 phases of our survey were conducted in the early 1990s, at a similar time as the other surveys. On balance, it appears that the New Zealand population has higher numbers of sexual partners than populations in Britain and possibly the United States. Indeed, the cohort underrepresents Maori and those with low education levels, 7 so numbers of partners are likely to be biased downward compared with New Zealand as a whole, especially at age 18.
The migration rates out of the core group are slightly higher than those found in the Norwegian population sample. 5 This higher migration rate could be the result of the younger age of our population, because annualized migration rates were higher between 18 and 21 years than between 21 and 26 years.
The high degree of variability over time of sexual behavior, insofar as this approximates core group membership based on STD transmission, adds another degree of complexity to the identification of core group members. Indeed, it provides another reason for questioning the use of the term 15 because of the misleading connotations of a group of people whose behavior remains constant over time. This variability suggests that approaches based on targeting high-prevalence communities could be more effective than attempts to target high-risk individuals. Both core definitions showed similarly high levels of outmigration. Because migration has been shown to influence the equilibrium prevalence of various STDs, 5 a more precise quantification of it could lead to more accurate modeling of the spread of STDs. It is unclear from our results which core definition is more relevant to STD spread, because a slightly larger amount of partner concurrency in one core was offset by a slightly lower amount of condom use by the other.
Early age at first sexual experience was strongly associated with core group membership, much more clearly than past or current socioeconomic factors. Early age at first sex is already an established marker for later high-risk sexual behaviors such as increased STD rates and higher numbers of partners among women. 16 We have shown that early age at first sex is also associated with prevalence and incidence of herpes simplex virus type 2 in this population in males and females. 17 In contrast, sociocultural definitions based on socioeconomic factors are of little application in this population, because the associations are too weak and also change with age.
Further exploration of a modern trajectory in sexual activity with age and education is warranted. Men and women who left school early were more likely to be in the core group at age 18, but not at ages 21 and 26. Hence, migration out of the core group between ages 18 and 21 was disproportionately of those with less education. Those with higher education enter the core group by age 21 and remain in it proportionately to other education groups. This trajectory should be explored in larger samples.