MOST BEHAVIORAL INTERVENTIONS designed to prevent STDs, including HIV, rely on behavioral outcomes to assess efficacy. Whereas many interventions are associated with change in some behavioral measures, primarily condom usage, 1–6 it is unknown if these improvements resulted in lowered infection rates. In contrast, two randomized controlled trials of behavioral interventions (Project RESPECT and Project SAFE [Sexual Awareness For Everyone]), using systematic screening at baseline and follow-up, have demonstrated efficacy in reducing STD rates. 6–8 Further analysis from Project Respect 9 found only weak associations between behaviors examined and incident infection. Preliminary analysis of behaviors in Project SAFE showed that certain behaviors were related both to reduced cumulative infection rates (0–12 months’ follow-up) and to the intervention. 7 A third intervention trial without baseline screening utilized chart review and endpoint LCR detection of chlamydia and gonorrhea; results demonstrated significant improvements in three variables related to number of sex acts and condom use but no differences in overall infection rates and no indication that improvements in behavior were associated with infection. 10 An earlier controlled, randomized trial 11 demonstrated an increase in condom use among men receiving intervention without a corresponding decrease in disease.
Correspondence between behavioral measures and infection and the utility of behavioral outcomes as surrogate measures of intervention efficacy have been the subject of considerable debate. This dialogue gained momentum in 1995 with data showing that STD clinic clients who reported that they always used condoms were as likely to become infected as those who said they never used them. 12 This raised questions about participant veracity and recall and brought into focus the complexity of behavioral measurement. 4,12–17 Many feel that no clear relationship exists between STD acquisition and certain discrete sexual behaviors, particularly various aspects of condom use. 4,9,16-20 The relationship between biologic and behavioral outcomes is exceedingly complex, dependent not only on widely reported behaviors such as consistency of condom use and frequency of sexual acts but also on correctness of condom use, type and number of partners, and infection status of partner(s). 4,7,9,16–19,21–23 In order to refine our understanding of this relationship, development of more precise behavioral measures is essential. 16,17,22
We continue this discussion by presenting a comprehensive analysis of behaviors responsible for reduced reinfection rates in Project SAFE during each follow-up interval, 0 to 6 months, 6 to 12 months, and 0 to 12 months. Because sexual behavior is highly complex, we constructed composite measures to more fully explicate the relationships between behavior and infection in our population. 22 Our goal is to show that infection is associated with behavior, if behavior is measured comprehensively, incorporating context.
This is an analysis of data collected as part of a controlled, randomized trial of a behavioral risk-reduction intervention for Mexican American and African American women. This study was approved by the institutional review boards at the University of Texas Health Science Center and the San Antonio Metropolitan Health District. After providing written consent, participants were screened/treated for STDs (all initially had a nonviral STD) and interviewed at baseline and at 6 and 12 months’ follow-up. Participants could also return to our research clinic for an unscheduled problem visit. At each visit a targeted physical examination was performed with collection of genital tract specimens for microbiologic testing, including Neisseria gonorrhoeae, Chlamydia trachomatis, and other lower genital tract pathogens. The primary test for gonorrhea and chlamydia (including off-site problem visits) was DNA-probe testing of endocervical samples (GenProbe PACE 2; GenProbe, San Diego, CA). Our infection measure was defined by new cases of chlamydia and/or gonorrhea. Study procedures are provided in greater detail in the original publication. 7
Highly trained interviewers collected very detailed information on sexual behaviors with up to five partners in the 3 months before each interview. We conducted in-depth analyses so that we could better understand our data; our goal was to understand and meaningfully measure which behaviors significantly contributed to infection. Selection of cutpoints for continuous variables was based on theoretical considerations and analysis of their distributions relative to the outcome of interest. All variables considered were part of our initial theoretical framework and were addressed in the intervention. Our holistic approach to variable construction was informed by 18 months of ethnographic data collection 24 and focused on examining interrelationships among various related constructs. We created composite measures when the addition of context clarified the relationship between behaviors and infection. Nurses’ notes supplemented clinical and behavioral variables. A detailed discussion of variable conceptualization follows.
Sex with untreated partner—unprotected sex with a partner before he was treated/completely treated.
Following treatment of their baseline infection, all participants were told to have their partner(s) treated and to avoid sex until all treatment was completed. At the 6-month interview, women were asked if they had sex with a man (the term “partner” was not used in interviews because it confused some participants) before he was completely treated. Possible responses were: “no,” “yes, with condoms,” “yes, without condoms,” “don't know if he was completely treated—used condoms,” and “don't know if he was completely treated—used no condoms.” Women were categorized as having sex with an untreated partner if they answered “yes” or “don't know” without using condoms or if they answered “no” but nurses’ notes indicated that the partner was never treated (all patients were asked). The values of this variable used in all logistic regression models are identical because the variable refers to the period following the baseline infection (taken from the 6-month questionnaire). Sex with an untreated partner continued to have predictive value at 12 months because some women stayed with the same partner or returned to the originally infected man and/or because this variable reflects a propensity for risk behavior.
Mutual monogamy—one steady, faithful partner in the study interval.
This is a composite variable consisting of type of relationship (steady versus casual), perceived partner fidelity (woman's assessment), and number of partners (0–1 versus >1) in each study interval. Women were considered mutually monogamous cumulatively only if they had the same steady, faithful partner (or no sex partner) throughout the 12-month period.
Our conceptualization of “steady” requires explanation. Prior research experience with the target and similar populations taught us that the meaning of “steady relationship” varies among women. In order to refine this construct so that we were grouping women with similar relationships together, we asked participants to clarify if their relationship was steady all the time, most of the time, or off and on. Prior experience with this question indicated that off-and-on relationships were often high risk; for example, even faithful (while the relationship is “on”) male partners may have had sex with others during the “off” periods (study participants may not consider this unfaithful behavior because they were not steady at the time). Off-and-on relationships were consequently categorized as “casual”; steady all or most of the time was categorized as “steady.” Prior work also indicated that women living apart from their husbands because of marital problems were often at considerable risk because they would periodically have sex with their spouses (who often were having sex with others). Consequently, such relationships were categorized as casual.
Unsafe sex—never using condoms with one or more casual partners or both ≥5 unprotected acts in the past 3 months and incorrect or problematic condom use.
This dichotomous variable (unsafe versus safer sex) integrates relationship type, number of unprotected acts, and condom-usage problems. Relationships among risk and preventive behaviors may be systematic and perhaps causal, e.g., individuals who are less discriminating in their partner choice may compensate for their heightened risk by increasing their condom use. 4 We believe that many women make reasoned albeit not foolproof decisions about when to use condoms. Consequently, we designed “unsafe sex” to be partner-specific about use/never use of condoms. We reasoned that it was much higher risk to never use condoms with a casual than with a steady partner. Never using condoms with a casual partner indicates that the woman is not exercising caution. We also reasoned that incorrect condom usage adds to high-risk exposure among participants who elect to use condoms (regardless of relationship type), particularly if the woman is already at higher risk due to a high number (≥5 in 3 months) of unprotected acts. Five or more unprotected acts was selected as a cutpoint because it was reasonable and was significantly related to cumulative infection. We examined the various types of incorrect or problematic use and conceptualized it as broadly as possible to include breakage, slippage, other problems (e.g., became lodged inside the woman), and failure to always check that the penis is completely covered (speaks to vigilance).
Our prior ethnographic work indicated that some couples placed the condom “halfway” to retain some flesh contact; this often led to its being wedged inside the vagina. A woman who used condoms with her casual partner but not with her steady partner could be in the safer-sex category. Her riskier relationship status would be accounted for in another variable, mutual monogamy. There is some overlap between the variables unsafe sex and mutual monogamy in that both incorporate the variable “steady”; however, tests of collinearity at every time point indicated that the highest R2 was 0.259 at 0 to 6 months.
Rapid partner turnover—having a new sex partner within 3 months of another sex partner.
This composite variable combines acquisition of a new sex partner with the interval between partners. Dates of first and last sex with each man were used to establish time span. A man was considered an “existing partner” if he was listed as a partner in the immediately preceding interview. All others were considered “new,” including ex-partners from the more distant past (they nonetheless represented a new exposure). This method of differentiating “old” versus “new” is more precise than asking participants to classify their partners themselves. There is some overlap between the variables rapid partner turnover and mutual monogamy: many women with rapid partner turnover in a given interval had more than one partner during that interval and thus were not monogamous. However, tests of collinearity at every time point indicated that the highest R2 was 0.217 at 0 to 12 months. Because scoring this variable is dependent on data from a prior interview, there are no baseline values.
Douching after sex.
We asked women if they currently douched and if so, under which circumstance(s): after menses, after sexual intercourse, to keep clean in general, when they perceive an odor, to prevent disease, and to prevent pregnancy. Douching after sex was the best predictor of infection. Douching has been associated with an increased incidence of chlamydia, gonorrhea, and bacterial vaginosis. 25–27 Most researchers who examine the relationship between pelvic infection and douching do not separately consider douching after sex. For example, Scholes et al. 26 found a fourfold higher likelihood of chlamydia infection among women with any douching in the last 12 months versus those with none. Koumans and colleagues found that douching in the last 90 days was associated with chlamydia and gonorrhea. 25 Douching in general may diminish the vagina's natural defense mechanisms and thus facilitate infection by altering normal flora or causing inflammation. 28,29 Douching after sex may possibly have a more powerful effect because the force created by the douching solution theoretically could facilitate entry of pathogens present in the upper vagina into the endocervical canal.
Chi-square analysis was used to assess group differences in follow-up behaviors when baseline values were not available. When they were available, logistic regression analysis (controlling for baseline values) was utilized. Chi-square analysis was also used to determine group differences in direction of change from baseline values, i.e., the percentage of women changing from low- to high-risk behavior and vice versa. Multiple logistic regression, controlling for age <19 years (a strong independent predictor of infection), was also used to determine which behaviors affected by intervention were responsible for group differences in infection. Results are based on intention-to-treat group assignments. Behavioral analysis presented here is based on the subset of 477 women with complete data from both follow-up visits.
Baseline sample characteristics of the 477 women are similar to those previously described for the entire sample (n = 549). 7 Low levels of income and education characterize the population; ages range between 14 and 45 years, with 70% under 24 years. Infection rates for this cohort, by study and control groups, respectively, are 12.0% versus 16.7% at 0 to 6 months, 8.8% versus 16.7% at 6 to 12 months, and 17.7% versus 25.9% at 0 to 12 months. These rates are comparable to those previously reported for the entire study population. 7
The proportions of study and control group participants engaging in the high-risk variants of five behavioral variables are provided in Table 1. Results are presented for the follow-up intervals 0 to 6 months and 6 to 12 months and cumulatively for the entire study period, 0 to 12 months. Cumulative results reflect worst-case values, i.e., women were assigned to the lower-risk category cumulatively only if they were at low risk in that measure at both 6 and 12 months. There were 42 women at 0 to 6 months, 23 at 6 to 12 months, and 15 at 0 to 12 months who abstained from sex; data from these women are incorporated into the lower-risk variant of each behavior. Relationships between the behaviors and infection, as measured by the adjusted odds ratios, did not change substantially when women who abstained from sex were excluded from analysis.
Probability of infection during follow-up is affected by two shifts in behavior: change from initially lower-risk activities or relationships to higher-risk ones, as well as shifts from higher-risk to lower-risk behaviors. The proportions of women demonstrating these changes are provided for variables with applicable baseline values (mutual monogamy, unsafe sex, and douching after sex) in Table 2.
Sex With Untreated Partner
Significantly more study than control women followed instructions to refrain from unprotected sex with an untreated or incompletely treated partner after their initial infection (Table 1). Since there were no comparable baseline values, this variable is not included in Table 2.
At baseline there were no significant group differences in mutual monogamy (past 6 months). Both groups showed a drop in nonmonogamous unions at the follow-ups at 0 to 6 and 6 to 12 months (Table 1). However, significantly more study than control women were in mutually monogamous relationships during these intervals, and only the study group showed cumulative change. The major thrust of intervention-associated change was in the shift from nonmutually monogamous unions to monogamous ones (Table 2). This was accomplished by convincing an existing partner to change or by dropping an unfaithful partner and either acquiring a new, faithful man or abstaining. The intervention was also associated with minimizing shifts from initially low-risk to higher-risk relationships, especially at the 6-month follow-up.
Simple measures of condom use—specifically, mean percentage of coital acts that were unprotected, mean number of unprotected acts, and 100% condom use—were not significantly related to infection at any time period (Pearson r values for these variables measured cumulatively were −0.04, −0.01, and 0, respectively). A cutpoint of <5 unprotected acts in the last 3 months yielded a more useful measure cumulatively (either as an additive model for two 3-month periods [P = 0.002]7 or as a worst-case analysis [P = 0.001] used here). The variable of <5 unprotected acts, however, was not a significant predictor of infection at 0 to 6 or 6 to 12 months because it was confounded by the high-risk status of participants whose sexual activity consisted of a few episodes of unprotected casual sex. Over time (represented by the cumulative measure), many of these women shifted to ≥5 unprotected acts, lessening confounding. “Unsafe sex” overcame the limitations of the <5 unprotected acts cutpoint by incorporating context. During cumulative follow-up, 172 women practiced “unsafe sex”: of these, 48.8% never used condoms with one or more casual partners, 34.9% had ≥5 unprotected acts/problematic use, and 16.3% had both behaviors.
There were no baseline group differences in unsafe sex, and the proportion of women engaging in higher-risk sex decreased in both groups at the follow-ups of 0 to 6 and 6 to 12 months (Table 1). However, significantly more control than study women engaged in unsafe sex at these times. Also, only study women showed a cumulative decrease during the 12-month follow-up period. The major thrust of intervention-associated change was to prevent a shift from safer to unsafe sex, especially at 6 to 12 and 0 to 12 months, where close to twice the proportion of controls as study women adopted unsafe sexual practices (Table 2). The intervention was also associated with shifts from higher- to lower-risk sex. Of interest, of the 331 women with any condom use during the 12-month study period, 42.6% (37.4% study, 48.4% control;P = 0.04) experienced some incorrect or problematic use; their cumulative infection rate was 30.5%, compared with 16.8% among condom users with no problems.
Douching After Sex
At baseline, significantly more study than control women douched after sex (Table 1). Although both groups showed significant decreases in this behavior throughout follow-up, study women made the most dramatic changes. The major thrust of intervention-associated change was a shift from douching to not douching after sex; twice as many study women as controls stopped this risk behavior (Table 2).
Rapid Partner Turnover
Acquisition of a new partner measured simply as “yes” or “no” or as number of new partners did not sufficiently reflect risk status. The 252 women who retained existing partner(s) during the 0 to 12–month period had a cumulative infection rate of 16.7%; of these, 233 with one partner had a rate of 15.5% and 19 with 2 or more partners (all were concurrent relationships) had an infection rate of 31.5%. Fifteen women had no sex partner. The remaining 210 women had one or more new partners; their infection rate was 29.1%. Seventy of these women waited at least 3 months between last sex with the former partner and first sex with the “new” man, whereas the other 140 did not; their respective infection rates were 8.6% and 39.3%. Acquiring a new partner in and of itself, therefore, was not associated with infection; the abstention interval between partners (taking time to meet a new man, waiting to have sex, or both) appeared to be the critical factor.
Abstaining at least 3 months between sexual partners may be a proxy for selectivity about men; however, it also limits concurrent relationships and the number of partners that can be accrued and may also affect number of unprotected acts in a given interval. To assess the relative impact of the waiting period independent of total number of partners, concurrent relationships, and number of unprotected acts, we compared these variables in a multiple logistic regression analysis using the subsample of women who had new partners (n = 210). Among the 70 women who waited ≥3 months to resume sexual activity, 33 had 1, 31 had 2, and 6 had 3 partners; having >1 partner did not impact infection rates. Among the remaining 140 women, 15 had 1, 40 had 2, and 85 had ≥3 partners; respective infection rates were 33.3%, 40.0%, and 40.0%. Among the 125 women with >1 partner, 90 had concurrent relationships (infection rate = 43.3%) and 35 did not (infection rate = 31.4%). Although there were no subgroup differences in coital frequency, more abstaining than nonabstaining women had less than five unprotected acts (41.4% versus 19.3%;P < 0.001). Logistic regression results indicated that nonabstainers were much more likely to be infected than those who waited (adjusted odds ratio = 4.4;P = 0.006). Neither multiple partners (0–1 versus 2 versus ≥3) nor concurrent relationships were significantly related to infection in this model. Unprotected acts had a significant effect but did not appreciably reduce the impact of the waiting period on infection.
As shown in Table 1, there were no group differences in rapid partner turnover at 0 to 6 months; approximately 20% to 23% of both groups acquired a new sex partner within 3 months of having sex with another man. However, during the second 6-month period, only 10.4% of study women versus 22.8% of controls did so (P < 0.001). Cumulative group differences were not significant (P = 0.15) because of 0 to 6–month values. Direction of change is not provided because there are no relevant baseline values.
Table 3 provides data on the relationship between infection and behavior, including results of multiple logistic regression analyses for all time periods. Unprotected sex with an untreated/incompletely treated partner had by far the strongest association with infection, both at 0 to 6 months (adjusted odds ratio = 8.1) and cumulatively (adjusted odds ratio = 5.6). Even at 6 to 12 months, women with this risk factor were 3.7 times more likely to have a new infection than those who did not. Mutual monogamy was significantly associated with infection at 6 to 12 months and 0 to 12 months. Even at 0 to 6 months, when its significance level was reduced (P = 0.06), its adjusted odds ratio was 2.1. Unsafe sex was significant across all time periods, but particularly at 0 to 6 months (adjusted odds ratio = 2.9). Rapid partner turnover was significantly related to infection at all time points, but particularly at 6 to 12 months (adjusted odds ratio = 4.7). Douching after sex was significantly associated with infection at 0 to 6 months but not at 6 to 12 months; it was significant at the 0.054 level cumulatively. These behaviors considered jointly accounted for the intervention effect; the term for group assignment was not significant when forced into these models.
Of interest, douching after sex did not appear to be a proxy for failure to use condoms, i.e., substitution of douching for condom use. Of the 462 sexually active women in the study, those who douched after sex were no more likely to have ≥5 unprotected acts in the last 3 months than those who did not (77.1% versus 78.1%). Furthermore, the douching effect did not necessarily appear to be a proxy for having risky partners. Although significantly more women in nonmutually monogamous unions than in monogamous ones douched after sex (19% versus 9.6%;P < 0.006), douching was associated with higher infection rates in both groups: 22.2% versus 7.0% among the mutually monogamous and 42.3% versus 28.8% among the nonmutually monogamous.
Mutually monogamous and nonmonogamous groups were further examined. Over the 12-month study period, 203 mutually monogamous women had a cumulative infection rate of 8.4%. Only marginal reductions in infection rates resulted when they had safer sex, had low partner turnover, or did not douche (infection rates reduced to 7.7%, 7.7%, and 7.0%, respectively). On the other hand, sex with an untreated partner had considerable impact. Overall, 42.6% of study women and 33.3% of control women were both monogamous and avoided unprotected sex with an incompletely treated man (n = 182). Their infection rate was 4.4%. Fewer than 8% of these participants were high risk in other behaviors. Because of this and their already low infection rate, the addition of further variables (unsafe sex, rapid partner turnover, or douching) had minimal impact. The lowest infection rate achieved by any subset defined by these variables was 3.6% for the 168 monogamous women who neither had sex with an untreated partner nor douched after sex.
The 274 nonmonogamous women had a cumulative infection rate of 31.4%. Rates were reduced most by adding safer sex (23.0%; n = 122) or low partner turnover (23.1%; n = 143). Avoidance of unprotected sex with an untreated partner further reduced these rates to 16.4% (n = 104) and 16.8% (n = 119), respectively. The lowest cumulative infection rate attained by nonmonogamous women was 11.9% for the 59 women who avoided sex with an untreated partner and had both safer sex and low partner turnover. On the other hand, the 42 nonmonogamous women who had sex with an untreated/incompletely treated partner had an infection rate of 57.1%. The STD incidence ratio based only on monogamy and sex with an untreated partner was 13.0. Those nonmonogamous women who had sex with an untreated partner and additionally had rapid partner turnover had a cumulative infection rate of 61.1%; however, there were only 18 participants in this subset. In the nonmonogamous group, 47% of study women versus 29.6% of controls (P = 0.003) avoided sex with an untreated partner and experienced safer sex throughout the study. Forty-eight and a half percent of study women versus 38.7% of controls (P = 0.10) avoided sex with an untreated partner and had low partner turnover during this time. The latter differences were stronger at 6 to 12 months (65.2% versus 45.6%;P < 0.007), when the study group made great strides in reducing partner turnover.
One measure of how strongly the behavioral variables are jointly related to infection is the percentage of participants whose infection status is correctly predicted or classified (“infected” or “not infected”) by the multiple logistic regression model. We used the five behavioral variables described above along with age in models to predict infection during the first and second 6 months and at 0 to 12 months; we then calculated the cutpoint for the logistic function that balanced false-positives and false-negatives. During 0 to 6 months, the behavioral model correctly predicted infection status for 82.4% of participants overall; 76.5% of the infected women and 83.4% of the uninfected were correctly classified. The 6 to 12–month model correctly predicted infection for 74.4% overall, with 70% of the infected and 75.1% of the uninfected correctly classified. The 0 to 12–month model correctly predicted infection status for 75.3% of the women, with 71.8% of the infected and 76.2% of the uninfected correctly classified. (We also ran the cumulative regression model substituting “any sex with an untreated/incompletely treated partner” for “unprotected sex” due to the latter's potential overlap with “unsafe sex,” but this substitution slightly decreased predictive value.) The complex behavior variables defined in this study when used jointly demonstrated very strong ability to predict STD infection for both the infected and uninfected subsets.
Study results indicated that reduced infection rates associated with Project SAFE 7 were attributable to change in five modifiable behaviors. Further analysis showed that interventions work not only by reducing high-risk behaviors but also by maintaining initially low-risk behaviors that might otherwise tend to become higher risk with time. Over the 12-month study period, the shift among control-group women to high-risk behavior in the three variables for which we had baseline data was not trivial. A delayed study-group response was also found for one behavior, “rapid partner turnover,” which was likely at least partly responsible for our earlier observation 7 that reductions in infection rates were greater at 12 months (49%) than at 6 months (34%). It may have taken time for study women to become more selective in their partner acquisition(s). This disparity in speed of behavior modification illustrates the importance of examining intervention effects over time: changes in various behaviors do not necessarily occur simultaneously or at the same rate in different individuals.
It makes good theoretical sense that the variable “unprotected sex with an untreated/incompletely treated partner” had the greatest cumulative impact on reinfection: it likely constituted a direct conduit to pathogens. Douching after sex was the least powerful explanatory variable. It is not clear whether douching had a direct impact or if women were more likely to douche after sex when they suspected that their partner was at high risk, mistakenly thinking it would wash away pathogens. It is interesting that douching after sex was associated with increased infection risk even in mutually monogamous women, possibly serving as a marker for questionable partner fidelity.
Paths to reduced infection rates differed for mutually monogamous and nonmonogamous women. Monogamous women benefitted most from avoiding unprotected sex with an incompletely treated/untreated partner, thereby preventing repeated reinoculation. Little additional benefit was gained from practicing other lower-risk behaviors: the nature of mutual monogamy either limits opportunity to engage in higher-risk practices (e.g., rapid partner turnover) or reduces the impact of safer behaviors (e.g., condom use, particularly when both partners are fully treated). Whereas a woman may have little control over her partner's behavior, she usually (except in cases of domestic violence) has control over discontinuing the relationship; more study group than control group women either convinced an existing partner to become faithful or ended the relationship. For women with a steady partner, a good strategy to avoid reinfection would be (1) to be vigilant about partner treatment, (2) to strengthen the relationship and insist on fidelity, and (3) to be prepared to end the relationship if the partner is unwilling to change. However, consistent, correct condom use is very important at the beginning of a relationship to allow time to more carefully assess partner fidelity.
Nonmonogamous women achieved significant reductions in infection by avoiding sex with an untreated partner and practicing either low partner turnover or safer sex. Nonmonogamous women who did not practice these safer behaviors had infection rates of over 50%. The infection rate of the few nonmonogamous women who practiced all three behaviors was still much higher than that of monogamous participants who simply avoided unprotected sex with untreated partners.
What is especially important in the current study is that study group women, monogamous and nonmonogamous, were significantly more likely than control counterparts to adopt strategies that substantially reduced infection rates. Monogamous study group women were more likely than controls to avoid unprotected sex with an incompletely treated partner. Nonmonogamous study women were more likely than controls to avoid sex with untreated partners and either practice safer sex or, by the 6 to 12–month follow-up, have low partner turnover.
Whereas mutual monogamy is the safest approach, not all women realistically can or want to achieve it; consequently, it is important for interventions to provide women with a range of risk-reduction options. Interventions should realistically and meaningfully deal with the needs, lifestyles, and culture of its target population. Women for whom mutual monogamy is not a realistic option must be given other choices, including vigilance regarding partner treatment, consistent and correct condom use, and improving partner selectivity.
Because of the complex interplay of behavioral variables and the many opportunities for measurement error and misclassification, other randomized controlled trials have not reported strong correspondence between behavior and infection. 9,10 We attribute our results to deliberate attempts to minimize misclassification and incorporate context into variable conceptualization. 22 This requires very detailed data collection. For example, acquiring a new partner was considered in light of the abstention interval between partners (≥3 months). In so doing, it was clear that the risk, at least in our data set, was not in obtaining a new partner per se but in acquiring one too quickly. The interval between partners is also being studied by others and is referred to as the GAP. 30–32
If the objective of a given intervention is to encourage a specific behavior, e.g., consistent condom use, it should seek to maintain existing protective behaviors while changing risky ones. Appropriate evaluation endpoints would be behavioral self-reports. However, if the goal is to reduce infection, much more is required both of intervention design and its evaluation. Many behaviors need to be addressed; participants must realize that change in only one behavior is often insufficient to avoid infection. They need to see “the bigger picture” so that they do not offset decreased risk in one behavior by increased risk in another.
Assessing the impact of behavior change on infection is complex. For example, simple condom-use measures do not provide sufficient insight into the mechanisms by which these behaviors translate into disease: their impact depends on context and, as argued by others, 4,9,19,33 behavior change is dynamic. Individuals who adopt a lower-risk behavior such as consistent condom use may simultaneously become more likely to choose riskier partners. Insight into the complex relationship between infection and behavior can be gained by considering several behaviors concurrently.
The comparison by Richens and colleagues 19 of condom and seat belt use is especially instructive in this regard. They cite research showing no noticeable difference in road deaths among European countries who have and have not passed seat belt laws. Researchers 19 have hypothesized risk-compensation: drivers wearing seat belts may have felt safer and therefore were more likely to speed and take other driving risks, which may have caused the number of road casualties to remain at their original level. There is debate among safety experts concerning the possibility that risk-reduction interventions may be undermined by such compensatory changes. The authors suggest that this type of behavioral adaptation may apply to condom promotion interventions, i.e., failure of increased condom usage to reduce infection rates may reflect compensatory shifts to increased casual partners, more partners overall, more sexual activity, or higher rates of partner change.
The “gold standard” for STD prevention interventions remains biologic outcome. Although our study showed strong convergent validity between behavioral constructs and biologic endpoints, the need for the latter has not decreased. We agree with Schachter 20 that biomarkers along with behavioral outcomes should be utilized to evaluate efficacy before such interventions are implemented (urine-based testing makes this more easily achieved). Use of biologic endpoints is necessary not only to demonstrate intervention efficacy in reducing disease but also to deepen our understanding of which behaviors, under which circumstances, increase the probability of infection.
Additional studies are needed on the relationship between infection and behavior, particularly development of a validated behavioral-risk index consisting of age and several behaviors/circumstances. Such a tool, if easily administered, would be especially valuable in clinical practice to identify high-risk women who might benefit from increased screening, treatment, and counseling. Work on such an index for our study population is in progress. 23
Although the purpose of this article is not to address the value of incident bacterial STDs as surrogates for HIV infection, we feel it is important to state our concurrence with recent statements. 18,20 While acknowledging that bacterial STDs are not a perfect model for HIV infection, Schachter 20 believes that an intervention that reduced incident STDs is encouraging and should be further explored, whereas one that did not do so “should certainly raise a red flag about its likely success in reducing HIV incidence.” The position statement issued by the NIMH/APPC Workgroup on Behavioral and Biologic Outcomes in HIV/STD Studies addressed problems in using STD incidence as a surrogate for HIV data. However, it recognized that the absence of a simple relationship between the two “does not refute the utility of using STD data to demonstrate that sexual transmission of a pathogen has been interrupted by an intervention.”18
The intervention evaluated here was strongly associated with reductions in both risk behaviors and infection. Moreover, behavioral risk status at follow-up was highly associated with infection status at follow-up. Efficacy against HIV infection could not be evaluated because of its low incidence in our target population. 7 However, we believe that this intervention may also help prevent infection with HIV because resulting reductions in risky behaviors appear to have disrupted the sexual transmission of other pathogens. Relevance to HIV prevention should not, however, detract from the importance of reducing the enormous burden of other STDs on women and children.
1. Kelly JA, Murphy DA, Washington CD, et al. The effects of HIV/AIDS intervention groups for high-risk women in urban clinics. Am J Public Health 1994; 84: 1918–1922.
2. DiClemente RJ, Wingood GM. A randomized controlled trial of an HIV sexual risk-reduction intervention for young African-American women. JAMA 1995; 274: 1271–1276.
3. St. Lawrence JS, Brasfield TL, Jefferson KW, Alleyne E, O'Bannon RE, Shirley A. Cognitive-behavioral intervention to reduce African American adolescents’ risk for HIV infection. J Consult Clin Psychol 1995; 63: 221–237.
4. Aral SO, Peterman TA. Measuring outcomes of behavioural interventions for STD/HIV prevention. Int J STD AIDS 1996; 7: 30–38.
5. Jemmott JB, Jemmott LS, Fong GT. Abstinence and safer sex HIV risk-reduction interventions for African Am adolescents: a randomized controlled trial. JAMA 1998; 279: 1529–1536.
6. Stephenson JM, Imrie J, Sutton SR. Rigorous trials of sexual behaviour interventions in STD/HIV prevention: what can we learn from them? AIDS 2000; 14: S115–S124.
7. Shain RN, Piper JM, Newton ER, et al. A randomized, controlled trial of a behavioral intervention to prevent sexually transmitted disease among minority women. N Engl J Med 1999; 340: 93–100.
8. Kamb ML, Fishbein M, Douglas JM, et al. Efficacy of risk-reduction counseling to prevent human immunodeficiency virus and sexually transmitted diseases: a randomized controlled trial. Project RESPECT Study Group. JAMA 1998; 280: 1161–1167.
9. Peterman TA, Lin LS, Newman DR, et al. Does measured behavior reflect STD risk? An analysis of data from a randomized controlled behavioral intervention study. Project RESPECT Study Group. Sex Transm Dis 2000; 27: 446–451.
10. The NIMH Multisite HIV Prevention Trial. Reducing HIV sexual risk behavior. The National Institute of Mental Health (NIMH) Multisite HIV Prevention Trial Group. Science 1998; 280: 1889–1894.
11. Boyer CB, Barrett DC, Peterman TA, Bolan G. Sexually transmitted disease (STD) and HIV risk in heterosexual adults attending a public STD clinic: evaluation of a randomized controlled behavioral risk-reduction intervention trial. AIDS 1997; 11: 359–367.
12. Zenilman JM, Weisman CS, Rompalo AM, et al. Condom use to prevent incident STDs: the validity of self-reported condom use. Sex Transm Dis 1995; 22: 15–21.
13. Weir SS, Feldblum PJ. Condom use to prevent incident STDs. Sex Transm Dis 1996; 23:76–77; discussion 79–82.
14. Galavotti C, Cabral R, Beeker C. Condom use to prevent incident STDs. Sex Transm Dis 1996; 23:77–79; discussion 79–82.
15. Turner CF, Miller HG. Zenilman's anomaly reconsidered: fallible reports, ceteris paribus, and other hypotheses. Sex Transm Dis 1997; 24: 522–527.
16. Fishbein M, Jarvis B. Failure to find a behavioral surrogate for STD incidence–what does it really mean? Sex Transm Dis 2000; 27: 452–455.
17. Fishbein M, Pequegnat W. Evaluating AIDS prevention interventions using behavioral and biological outcome measures. Sex Transm Dis 2000; 27: 101–110.
18. Pequegnat W, Fishbein M, Celentano D, et al. NIMH/APPC workgroup on behavioral and biological outcomes in HIV/STD prevention studies: a position statement. Sex Transm Dis 2000; 27: 127–132.
19. Richens J, Imrie J, Copas A. Condoms and seat belts: the parallels and the lessons. Lancet 2000; 355: 400–403.
20. Schachter J. Biologic versus behavioral endpoints–the duet continues. Sex Transm Dis 2000; 27: 456–457.
21. Shain RN, Piper JM, Newton ER, et al. A controlled randomized trial of a risk-reduction intervention: behaviors contributing to reduced infection rates at 12 months’ follow-up. Presented at the Meeting of the International Society for Sexually Transmitted Disease Research, Seville, Spain, October 19–22, 1997.
22. Shain RN, Perdue S, Piper JM, et al. Sexual risk-reduction behaviors amenable to intervention: the importance of context. Presented at the Meeting of the International Society for Sexually Transmitted Disease Research, Denver, Colorado, July 11–14, 1999.
23. Perdue ST, Shain RN, Piper JM, et al. Risk-reduction index for behaviors amenable to intervention. Presented at the Meeting of the International Society for Sexually Transmitted Disease Research, Denver, Colorado, July 11–14, 1999.
24. Ramos R, Shain RN, Johnson L. “Men I mess with don't have anything to do with AIDS”: using ethno-theory to understand sexual risk perception. Sociol Q 1995; 36: 483–504.
25. Koumans EH, Black C, Carter SA, et al. Complex relationships among douching, BV, and STDs. Presented at the 2000 National STD Prevention Conference: Untapped Opportunities Connecting Science with Solutions. Milwaukee, WI, December 4–7, 2000.
26. Scholes D, Stergachis A, Ichikawa LE, Heidrich FE, Holmes KK, Stamm WE. Vaginal douching as a risk factor for cervical Chlamydia trachomatis infection. Obstet Gynecol 1998; 91: 993–997.
27. Sweet R. Gynecologic conditions and bacterial vaginosis: implications for the non-pregnant patient. Infect Dis Obstet Gynecol 2000; 8: 184–190.
28. Onderdonk AB, Delaney ML, Hinkson PL, DuBois AM. Quantitative and qualitative effects of douche preparations on vaginal microflora. Obstet Gynecol 1992; 80: 333–338.
29. Eschenbach D, Patton D, Aura J, Thwin S. Effect of vaginal douching on vaginal physiology. Infect Dis Obstet Gynecol 2000; 8: 201–202.
30. Aral S, Adimora A, Manhart L, Gorbach P, Koumans E, Kraut JR. Temporal ordering of sex partnerships: concurrency and the “gap” as factors fueling or limiting the spread of STIs. Presented at the 2000 National STD Prevention Conference: Untapped Opportunities Connecting Science with Solutions. Milwaukee, WI, December 4–7, 2000.
31. Aral S. Sexual behavior in the US. Presented at STIs at the Millenium. Baltimore, MD, May 3–7, 2000.
32. Kraut JR, Aral SO. Sexually transmitted disease and the “GAP.” Presented at STIs at the Millennium. Baltimore, MD, May 3–7, 2000.
33. Aral SO, Peterman TA. Do we know the effectiveness of behavioural interventions? Lancet 1998; 351: 33–36.