Behavioral Sexual Risk-Reduction Counseling in Primary Care to Prevent Sexually Transmitted Infections: A Systematic Review for the U.S. Preventive Services Task Force

The Centers for Disease Control and Prevention estimates that approximately 20 million new cases of sexually transmitted infections (STIs) occur each year in the United States, half of which are among persons aged 15 to 24 years (1). In 2003 and 2004, 38% of sexually active female adolescents aged 14 to 19 years had an STI (2). In 2010, the inflation-adjusted annual direct medical costs of STIs were estimated to be $16.9 billion in the United States (3). In 2008, the U.S. Preventive Services Task Force (USPSTF) recommended high-intensity behavioral counseling interventions for sexually active adolescents and in adults at increased risk for STIs (B recommendation). The evidence was insufficient, however, to assess the balance of benefits and harms of behavioral counseling to prevent STIs in nonsexually active adolescents and in adults not at increased risk for STIs (I statement). This systematic review updates the previous review that formed the basis of the 2008 recommendation. We developed an analytic framework (Appendix Figure 1) with 4 key questions (Appendix Table 1) that address counseling's effects on patient health outcomes (key question 1), behavioral outcomes (key question 2), other positive outcomes (key question 3), and harms of counseling (key question 4). Appendix Figure 1. Analytic framework. STI = sexually transmitted infection. Appendix Table 1. Key Questions Methods The full report describes our methods in detail (4). Data Sources and Searches To identify the cumulative body of literature, we examined all studies included in the previous USPSTF review and searched MEDLINE, PubMed, Cochrane Central Register of Controlled Trials, and CINAHL from 1 January 2007 through 4 November 2013 to identify relevant articles published since the previous review (5). We also searched the bibliographies of relevant reviews and Web sites of governmental agencies and professional organizations, and we consulted with outside experts. Between 4 November 2013 and this publication, we actively monitored published literature for potentially important new trials directly relevant to the key questions in this systematic review; none were located. Study Selection Two investigators independently reviewed abstracts and relevant full-text articles against prespecified inclusion criteria. We included trials evaluating counseling interventions targeting risky sexual behaviors to prevent STIs in adults and adolescents. We excluded studies limited to persons with HIV (or populations with very high prevalence of HIV [>10% in the study sample]), inmates and parolees, and persons in inpatient or residential settings because results limited to these groups may not be applicable to general primary care populations. We required that included interventions be conducted in, or participants be recruited from, primary care or other outpatient clinical settings, including reproductive health clinics, STI clinics, and mental health clinics. We included English-language trials conducted in very high human development countries according to the World Health Organization (6). We accepted the following comparators as control groups: usual care, attention control, minimal intervention (<15 minutes of intervention contact), wait list, or no intervention. We included trials reporting 1 or more of the following at 3 months or later after baseline: patient health outcomes (STI incidence and morbidity or mortality related to STIs), sexual behavioral outcomes (for example, condom use or number of sexual partners), and harms of the intervention (for example, care avoidance). Data Extraction and Quality Assessment Two investigators independently assessed the methodological quality of each study using USPSTF criteria (7). Studies were rated as good, fair, or poor quality. Good-quality studies had adequate randomization procedures, allocation concealment, blinding of outcome assessors, reliable outcome measures (for example, at least standard laboratory procedures or efforts to minimize demand characteristics for self-reported outcomes), similar groups at baseline and follow-up, low attrition, acceptable statistical methods, and adequate adherence to the intervention. Fair-quality trials met some but not all of these criteria. Poor-quality studies had a serious flaw (for example, attrition >40%, differential attrition >20% between groups, or substantial baseline differences between groups) or multiple important limitations that would invalidate the study findings. We excluded all poor-quality studies. We resolved disagreements through discussion and, if necessary, consultation with a third investigator. One investigator abstracted data from all included studies into a standard evidence table. A second investigator checked the data for accuracy. Data Synthesis and Analysis We created summary tables for each key question that included trial characteristics and summaries of results and qualitatively examined the range of results and potential associations with effect size. We stratified our analyses on the basis of age (adolescents vs. adults, including age-based subgroup analyses when reported [810]) and estimated intervention intensity: high (>2 hours of intervention contact), moderate (0.5 to 2 hours of intervention contact), and low (brief single session or <0.5 hour of intervention contact). These cut points were selected to correspond with a typical, single brief session that would be feasible in a primary care office (low intensity); a longer single session or 2 to 3 brief sessions that may be feasible in selected primary care settings (medium intensity); and what would probably require multiple nonbrief sessions, usually necessitating specialized and trained staff that could be referred from primary care (high intensity). We categorized populations on the basis of STI risk. Low/mix referred to a mix of sexually active and presexually active participants (for adolescents only). General referred to sexually active adults with no further risk factors and not in a setting with increased risk (for adults only). Increased referred to participants with increased risk based on sociodemographics (sexually active teenagers, low-income inner-city residents, racial/ethnic subgroups with higher STI prevalence, men who have sex with men [MSM], and mentally ill or disabled persons), sexual history (for example, persons reporting high-risk behaviors), or setting (for example, STI clinics). The prior STI category was limited to persons with a current or recent STI at baseline. Additional potential moderators or mediators that we examined in exploratory qualitative analysis include characteristics of the interventions (degree of cultural tailoring, group vs. individual format, condom negotiation or other communication training as an intervention component, counselor characteristics, setting, type of control group, or number of sessions) and population (sex, sexual orientation, socioeconomic status, mental health issues, or history of abuse). We did random-effects meta-analyses for STI incidence using the DerSimonianLaird method (11). We analyzed odds ratios (ORs) because they were the most commonly reported outcome, which allowed us to include the largest number of studies in the meta-analysis. We ran sensitivity analyses using the profile likelihood method because some of our pooled estimates were derived from a small number of trials (12). Results were very similar, and all statistically significant results remained statistically significant with the profile likelihood method. Results shown on forest plots are from the DerSimonianLaird analyses. Statistical heterogeneity was assessed using the I 2 statistic (13). We used Stata, version 11.2 (StataCorp), for all meta-analyses. Role of Funding Source The Agency for Healthcare Research and Quality (AHRQ) funded this review under a contract to support the work of the USPSTF. Members of the USPSTF and an AHRQ medical officer assisted in defining this review's scope. The AHRQ staff provided oversight for the project and assisted in the external review of the companion draft evidence synthesis. Although approval from AHRQ was required before submission of the manuscript, the authors are solely responsible for its content and the decision to submit it for publication. Results Thirty-one trials (810, 1441), reported in 57 publications (810, 1467), were selected from our review of 3241 abstracts and 218 full-text articles (Appendix Figure 2). Of the 31 included trials (n= 70324), 16 were newly published and not included in the previous review (n= 56110). Most evidence comes from trials in women and nonwhite or minority populations. Most trials targeted high-risk groups based on demographic characteristics, high-risk behaviors, or presence of a recent STI. Study details (including target populations) are presented in Appendix Tables 2 and 3 for adolescents and adults, respectively, and Supplements 1 and 2. Appendix Figure 2. Summary of evidence search and selection. CE = comparative effectiveness; KQ = key question. Appendix Table 2. Summary of Included Studies: Adolescents Appendix Table 3. Summary of Included Studies: Adults Supplement 1. Design and baseline population characteristics of included studies targeting adolescents Supplement 2. Design and baseline population characteristics of included studies targeting adults Although the interventions were very heterogeneous, there were some shared components. All interventions sought to minimize high-risk sexual behaviors (for example, unprotected sexual intercourse or multiple partners) and maximize protective behaviors (for example, condom use). Interventions provided basic information about STIs and commonly included risk assessment, hands-on skill training in condom use, problem solving, decision making, goal-setting, and communication surrounding condom use and safe sex. The depth with which topics were covered varied. Some interventions included additional components,

T he Centers for Disease Control and Prevention estimates that approximately 20 million new cases of sexually transmitted infections (STIs) occur each year in the United States, half of which are among persons aged 15 to 24 years (1). In 2003 and 2004, 38% of sexually active female adolescents aged 14 to 19 years had an STI (2). In 2010, the inflation-adjusted annual direct medical costs of STIs were estimated to be $16.9 billion in the United States (3).
In 2008, the U.S. Preventive Services Task Force (USPSTF) recommended high-intensity behavioral counseling interventions for sexually active adolescents and in adults at increased risk for STIs (B recommendation). The evidence was insufficient, however, to assess the balance of benefits and harms of behavioral counseling to prevent STIs in nonsexually active adolescents and in adults not at increased risk for STIs (I statement). This systematic re-view updates the previous review that formed the basis of the 2008 recommendation. We developed an analytic framework (Appendix Figure 1, available at www.annals .org) with 4 key questions (Appendix Table 1, available at www.annals.org) that address counseling's effects on patient health outcomes (key question 1), behavioral outcomes (key question 2), other positive outcomes (key question 3), and harms of counseling (key question 4).

METHODS
The full report describes our methods in detail (4).

Data Sources and Searches
To identify the cumulative body of literature, we examined all studies included in the previous USPSTF review and searched MEDLINE, PubMed, Cochrane Central Register of Controlled Trials, and CINAHL from 1 January 2007 through 4 November 2013 to identify relevant articles published since the previous review (5). We also searched the bibliographies of relevant reviews and Web sites of governmental agencies and professional organizations, and we consulted with outside experts. Between 4 November 2013 and this publication, we actively monitored published literature for potentially important new trials directly relevant to the key questions in this systematic review; none were located.

Study Selection
Two investigators independently reviewed abstracts and relevant full-text articles against prespecified inclusion criteria. We included trials evaluating counseling interventions targeting risky sexual behaviors to prevent STIs in adults and adolescents. We excluded studies limited to persons with HIV (or populations with very high prevalence of HIV [Ͼ10% in the study sample]), inmates and parolees, and persons in inpatient or residential settings because results limited to these groups may not be applicable to general primary care populations.
We required that included interventions be conducted in, or participants be recruited from, primary care or other outpatient clinical settings, including reproductive health clinics, STI clinics, and mental health clinics. We included English-language trials conducted in "very high" human development countries according to the World Health Organization (6). We accepted the following comparators as control groups: usual care, attention control, minimal intervention (Ͻ15 minutes of intervention contact), wait list, or no intervention. We included trials reporting 1 or more of the following at 3 months or later after baseline: patient health outcomes (STI incidence and morbidity or mortality related to STIs), sexual behavioral outcomes (for example, condom use or number of sexual partners), and harms of the intervention (for example, care avoidance).

Data Extraction and Quality Assessment
Two investigators independently assessed the methodological quality of each study using USPSTF criteria (7). Studies were rated as good, fair, or poor quality. Goodquality studies had adequate randomization procedures, allocation concealment, blinding of outcome assessors, reliable outcome measures (for example, at least standard laboratory procedures or efforts to minimize demand characteristics for self-reported outcomes), similar groups at baseline and follow-up, low attrition, acceptable statistical methods, and adequate adherence to the intervention. Fairquality trials met some but not all of these criteria. Poorquality studies had a serious flaw (for example, attrition Ͼ40%, differential attrition Ͼ20% between groups, or substantial baseline differences between groups) or multiple important limitations that would invalidate the study findings. We excluded all poor-quality studies. We resolved disagreements through discussion and, if necessary, consultation with a third investigator. One investigator abstracted data from all included studies into a standard evidence table. A second investigator checked the data for accuracy.

Data Synthesis and Analysis
We created summary tables for each key question that included trial characteristics and summaries of results and qualitatively examined the range of results and potential associations with effect size. We stratified our analyses on the basis of age (adolescents vs. adults, including age-based subgroup analyses when reported [8 -10]) and estimated intervention intensity: high (Ͼ2 hours of intervention contact), moderate (0.5 to 2 hours of intervention contact), and low (brief single session or Ͻ0.5 hour of intervention contact). These cut points were selected to correspond with a typical, single brief session that would be feasible in a primary care office (low intensity); a longer single session or 2 to 3 brief sessions that may be feasible in selected primary care settings (medium intensity); and what would probably require multiple nonbrief sessions, usually necessitating specialized and trained staff that could be referred from primary care (high intensity). We categorized populations on the basis of STI "risk." "Low/mix" referred to a mix of sexually active and pre-sexually active participants (for adolescents only). "General" referred to sexually active adults with no further risk factors and not in a setting with increased risk (for adults only). "Increased" referred to participants with increased risk based on sociodemographics (sexually active teenagers, low-income inner-city residents, racial/ethnic subgroups with higher STI prevalence, men who have sex with men [MSM], and mentally ill or disabled persons), sexual history (for example, persons reporting high-risk behaviors), or setting (for example, STI clinics). The "prior STI" category was limited to persons with a current or recent STI at baseline. Additional potential moderators or mediators that we examined in exploratory qualitative analysis include characteristics of the interventions (degree of cultural tailoring, group vs. individual format, condom negotiation or other communication training as an intervention component, counselor characteristics, setting, type of control group, or number of sessions) and population (sex, sexual orientation, socioeconomic status, mental health issues, or history of abuse).
We did random-effects meta-analyses for STI incidence using the DerSimonian-Laird method (11). We analyzed odds ratios (ORs) because they were the most commonly reported outcome, which allowed us to include the largest number of studies in the meta-analysis. We ran sensitivity analyses using the profile likelihood method because some of our pooled estimates were derived from a small number of trials (12). Results were very similar, and all statistically significant results remained statistically significant with the profile likelihood method. Results shown on forest plots are from the DerSimonian-Laird analyses. Statistical heterogeneity was assessed using the I 2 statistic (13). We used Stata, version 11.2 (StataCorp), for all metaanalyses.

Role of Funding Source
The Agency for Healthcare Research and Quality (AHRQ) funded this review under a contract to support the work of the USPSTF. Members of the USPSTF and an AHRQ medical officer assisted in defining this review's scope. The AHRQ staff provided oversight for the project and assisted in the external review of the companion draft evidence synthesis. Although approval from AHRQ was required before submission of the manuscript, the authors are solely responsible for its content and the decision to submit it for publication.

RESULTS
Thirty-one trials (8 -10, 14 -41), reported in 57 publications (8 -10, 14 -67), were selected from our review of 3241 abstracts and 218 full-text articles (Appendix Figure  2, available at www.annals.org). Of the 31 included trials (n ϭ 70 324), 16 were newly published and not included in the previous review (n ϭ 56 110). Most evidence comes from trials in women and nonwhite or minority populations. Most trials targeted high-risk groups based on demographic characteristics, high-risk behaviors, or presence of a recent STI. Study details (including target populations) are presented in Appendix Tables 2 and 3 (available at www .annals.org) for adolescents and adults, respectively, and Supplements 1 and 2 (available at www.annals.org).
Although the interventions were very heterogeneous, there were some shared components. All interventions sought to minimize high-risk sexual behaviors (for example, unprotected sexual intercourse or multiple partners) and maximize protective behaviors (for example, condom use). Interventions provided basic information about STIs and commonly included risk assessment, hands-on skill training in condom use, problem solving, decision making, goal-setting, and communication surrounding condom use and safe sex. The depth with which topics were covered varied. Some interventions included additional components, such as HIV testing and contraceptive counseling. Many interventions were culturally tailored to a target group, usually based on age, gender, and ethnicity.
The interventions included 1 to 13 sessions, which ranged from mail-, computer-, or video-only interventions to up to 17 hours of face-to-face contact. We categorized 16 of the intervention groups as high intensity, 10 as moderate intensity, and 9 as low intensity. Most of the highintensity interventions involved group sessions with extensive educational and behavior change components. Most moderate-intensity interventions involved 1 or 2 individual meetings for a total of 45 to 60 minutes of contact, although several involved group meetings. Most lowintensity interventions involved brief individual meetings with a counselor or primary care provider or were limited to print, computer-based, or video-based materials. Almost all (k ϭ 28) trials were done in the United States. The most common settings were primary care (k ϭ 15) (10, 14 -18, 20, 22, 25-27, 29, 34, 36, 40) and STI clinics (k ϭ 8) (9,23,28,30,35,38,39,41).
The STI results were generally based on laboratory tests for bacterial infections, most commonly gonorrhea and chlamydia. Because studies provided treatment of baseline infections, bacterial infections at follow-up were considered new infections. For studies that included viral infection outcomes, only infections after baseline assess- Review Behavioral Counseling to Prevent STIs ment were counted in the results. Most trials collected their own samples at follow-up assessment, and many supplemented their testing with patient medical records to identify STIs that occurred between assessments. A few relied on only medical records or patient self-report for STI results.

Adolescents
Incidence of STIs was reduced in all 8 comparisons targeting adolescents (n ϭ 3407) (Figure 1), although results were not statistically significant in 2 trials (10,15). Pooled results showed a 62% reduction in the odds of contracting an STI with high-intensity counseling after 12 months (DerSimonian-Laird OR, 0.38 [95% CI, 0.24 to 0.60]; I 2 ϭ 65%; profile likelihood OR, 0.38 [CI, 0.23 to 0.62]; I 2 ϭ 55%; k ϭ 5). The CIs suggest that a reduction of 40% or more in the odds of incident STIs with these interventions is likely. When reported, STI rates at follow-up ranged from 13% to 40% in the control groups of trials with high-intensity interventions compared with 5% to 24% in the intervention groups. The 2 moderateintensity intervention groups resulted in reductions of 33% to 47% in the odds of having an STI, only 1 of which was statistically significant. The only low-intensity intervention trial (n ϭ 219) involved a single brief contact with the primary care provider plus a video and print materials for  persons aged 12 to 15 years, most of whom were reportedly not sexually active. The young age of the participants and the reliance of the trial on self-report of STI symptoms rather than biological confirmation may have contributed to the low number of STIs at follow-up and statistically nonsignificant group differences (15). Six trials (n ϭ 3030) reported sexual behavior outcomes, and most found beneficial effects for some behavioral outcomes (8, 14 -18). Measures of behavior change were very heterogeneous. Measures of condom use or unprotected sex were the most commonly reported behavioral outcomes.
Most trials were limited to sexually active African American and Latina girls; only 1 trial included sexually active male and female teenagers (9). Only 1 of the trials in adolescents was rated as good quality. Several had attrition greater than 15% (up to 34% in 1 trial), and many did not describe their allocation concealment procedures. Four were conducted in, or adolescents were recruited from, primary care settings. Although all 4 of these trials reported reductions in the odds of contracting an STI of 33% or more, not all effects were statistically significant (10,15,17,18).
Heterogeneity was high (I 2 ϭ 65%) in the analysis of high-intensity interventions because of the very large effect size for the outcome of chlamydia infection incidence in 1 study (18). However, I 2 was reduced to 0% in sensitivity analyses when infections of gonorrhea and trichomonas, instead of chlamydia, were analyzed, with only minor attenuation of pooled effect size (OR changed from 0.38 to 0.48). This study did not report a composite outcome of any STI, and we chose, a priori, chlamydia infection as our primary outcome because it has the highest prevalence of the 3 STIs examined.

Adults
Nineteen trials (n ϭ 61 909) reported STI outcomes in adult populations, and 7 of these were conducted in, or participants were recruited from, primary care (10,20,22,29,34,36,40). All trials were included in the metaanalysis except 3 that did not provide necessary data on any of the commonly reported STIs (20,24,35).
High-intensity interventions resulted in a 30% reduction in the odds of contracting an STI (DerSimonian-Laird OR, 0.70 [CI, 0.56 to 0.87]; I 2 ϭ 23%; profile likelihood OR, 0.71 [CI, 0.55 to 0.86]; I 2 ϭ 6%; k ϭ 9) (Figure 2). Upper CIs indicate high probability of at least a 13% to 14% reduction in STIs with high-intensity interventions. The proportion of persons with an STI at follow-up in the high-intensity intervention groups ranged from 2% to 63% compared with 5% to 68% in the control groups. Three of the high-intensity trials were done in primary care settings with ORs ranging from 0.48 (CI, 0.24 to 0.97) to 0.82 (CI, 0.46 to 1.45) (29,36,40).
The pooled effects for low-and moderate-intensity trials did not show a reduction in the odds of contracting an STI (Figure 2), and most trials did not report group differences, including the 3 trials that could not be included in the meta-analysis. Some of the low-and moderateintensity trials were effective, however.
Most (9 of 12) of the high-intensity interventions with behavioral outcomes reported beneficial results for at least 1 behavioral outcome. All 4 of the high-intensity trials reporting ORs found increases of 24% to 42% in the odds of condom use (Appendix Figure 3, available at www .annals.org). The moderate-intensity interventions had mixed findings, with ORs for condom use and unprotected sex outcomes ranging from 0.98 to 2.2 and CIs crossing 1.0 for the smaller ORs. Most of the low-intensity interventions showed no group differences in behavioral outcomes; although some ORs were large (up to 5.2), CIs were generally very wide and all crossed 1.0. Specific measures of sexual behavior were reported inconsistently, and some trials reported many interrelated behavioral outcomes, which raised concern about opportunistic reporting and elevated type II error rates.
Many of the adult trials were limited to African American and Latina women; however, several studies included men and women and 1 focused on African American men. A few studies included very narrow subpopulations (for example, psychiatric patients or women with genital warts). Only 6 of the adult trials were rated as good quality. Along with concerns about selective reporting of behavioral outcomes, common concerns in the fair-quality trials included high attrition (15% to 40%) and lack of information about randomization and allocation concealment.

Characteristics Influencing the Effectiveness of the Interventions Population Characteristics
Most of the included trials were done in fairly narrow populations known to have high STI prevalence. Many of the trials targeting African American and/or Latina women were effective in reducing STI incidence. Some trials analyzed subgroups to examine whether their intervention was effective in particular subpopulations, such as smaller age groups, men and women separately, MSM separately from exclusively heterosexual men, persons with and without a history of STIs, and persons with and without a history of substance abuse. Age group was the most common subpopulation difference tested. All 3 trials that reported results separately for adolescents and adults found group differences for adolescents but not adults in at least 1 active intervention group (8 -10). Other than the greater likelihood of benefit in adolescents, no clear evidence suggested that interventions were more or less likely to be effective for any important subpopulation. Some subpopulations, however, were poorly represented, such as MSM and American Indian and Alaska Native persons. Subgroup results were usually consistent with overall study results.

Intervention Characteristics
Intervention intensity was the only characteristic that clearly influenced outcomes in these trials. On the basis of qualitative synthesis, we found no clear relationship between the effect size and degree of cultural tailoring, group versus individual format, condom negotiation training, other communication training, counselor characteristics, setting, type of control group, or number of sessions. We could not isolate effects of these features, however, because they were not evenly distributed across the spectrum of intervention intensity or population risk.

Harms of Sexual Risk-Reduction Counseling
Three trials explicitly reported on adverse events (n ϭ 6837) and found no harms related to the counseling interventions (35,36,41). We found no statistically significant paradoxical increase in the overall incidence of STIs among any of the studies. A subgroup analysis in 1 trial, however, showed a statistically significant deleterious effect on STI incidence in MSM, with 12.5% of control and 18.7% of intervention participants having an STI at follow-up (adjusted relative risk, 1.41 [98.3% CI, 1.05 to 1.90]) (41). The intervention involved a 25-minute session in an STI clinic before taking a rapid HIV test and a brief follow-up intervention after receiving the results. One other trial testing the brief video-based intervention in STI clinics provided subgroup results for MSM and did not see a deleterious effect (28).
No consistent evidence was found that interventions increased sexual activity in adolescents. Although 1 trial reported a short-term increase in the proportion of youth who were sexually active in the previous 3 months (15), another reported a decrease in this proportion (14). Other trials found no differences in frequency or number of partners (8,17,68).

DISCUSSION
Consistent with the evidence considered in 2008, we found that high-intensity (Ͼ2 hours) interventions were likely to reduce the rate of STIs in both adults and sexually active adolescents. Condom use also increased with highintensity interventions, particularly in the short term. Some moderate-and low-intensity interventions were beneficial but less likely to show improvement over usual care. A summary of the evidence is shown in the Table. Although a few more trials included men than in the previous review, most still targeted adolescent and young adult African American and Latina women and generalizability to other populations is unclear.
Although we could not identify specific components that were associated with treatment benefit, interventions that were successful generally provided most or all of the following: information about STIs, such as prevalence, transmission, and details on how to reduce the risk for transmission; help in identifying personal risk for STIs; training in common behavior change processes, such as problem solving, decision making, and goal-setting; training in communication surrounding condom use and safe sex; and hands-on practice with condoms. Many successful interventions were also specifically tailored to the gender and race/ethnicity of the participants. These types of interventions are generally not feasible in a typical primary care visit, but they could be feasible in an integrated care setting that included a behavior specialist who was trained to provide sexual risk-reduction counseling. Materials for many of the included interventions are freely available from the authors or online.
We found no consistent evidence suggesting that sexual risk-reduction counseling is harmful for adults or adolescents. The 2 trials in young, mostly pre-sexually active adolescents that reported the proportion of participants engaging in any sexual activity had contradictory results, and sparse reporting precludes us from drawing conclusions. A review of community-based, comprehensive interventions for sexual risk reduction, however, found that similar interventions reduced sexual activity (69).
On the basis of pooled results, 11 high-risk adolescents (95% CI, 9 to 18) would have to receive high-intensity interventions to prevent 1 STI, which assumes a baseline cumulative incidence of 15% over 1 year. Trial data, however, were limited primarily to sexually active urban African American and Latina girls, and generalizability to other sexually active adolescents is unknown. In high-risk adult populations, 25 adults (CI, 17 to 59) would need highintensity counseling to prevent 1 STI in a setting with an annual cumulative STI incidence of 15%, again based largely on African American and Latina women. This would prevent 41 cases of STIs per 1000 adults. Based on real-world estimates of STI prevalence and patient volume (70), approximately 1300 STIs could potentially be prevented in a large county health department with widespread adoption of high-intensity counseling.
The effects of primary care-based counseling on sexual risk reduction may also potentiate the effects of other types of community-level interventions. For example, the likelihood of benefit from condom distribution programs is enhanced with additional individual, small-group, or community interventions on STI prevention (71). Thus, STI prevention may be enhanced if people hear risk-reduction messages from multiple sources, multiple times. Even relatively modest effects may contribute to clinically important effects in communities in which messages from other sources are also frequently encountered.
One of the main limitations of this report is that it includes relatively little information on populations other than high-risk African American and Latina women. Although this is an important population, the effects in men (particularly MSM and adolescent men) and women of other races or ethnicities are not as well-understood, yet they still experience substantial health burden from STIs.  ters for Disease Control and Prevention Community Guide found community-based individual, group, and community-level interventions to be effective in reducing the risk for STIs in MSM (72). Another limitation of our review is that intervention intensity was difficult to ascertain. We estimated the intervention time, but trials did not always provide details of contact time, and the chosen cut points of 30 minutes and 2 hours were somewhat arbitrary. Although greater contact time improved the likelihood of reducing STIs, the minimum time necessary for benefit was not clear. Further, the quality and intensity of counseling provided in usual care influence trial results; when usual care is extremely minimal, a relatively brief intervention might improve it enough to show a benefit.
We also excluded studies with high HIV prevalence, which we defined as more than 10% of the sample. Some STI clinics may have HIV prevalence in this range and serve populations similar to those in some included studies. Nevertheless, most of the studies we excluded for this reason were limited to only persons with HIV or their partners, which clearly represents a distinct subpopulation. In contrast, one of the studies in our review specifically described their patient population as having high HIV seroprevalence but was included because HIV prevalence in the study sample was lower than 10% (31).
We also did not consider important strategies of STI risk reduction that cannot be implemented in health care settings or that go beyond risk-reduction counseling (for example, STI or HIV testing, partner notification, schoolbased programs, and condom distribution programs). Other USPSTF reviews address STI testing (73), and the Centers for Disease Control and Prevention maintains a regularly updated compendium of evidence-based individual, group, and community-level interventions on risk reduction and associated dissemination materials, which target a wide range of populations and risk-reduction strategies (74).
More data are needed in mixed-sex populations and broadly applicable interventions that could be implemented in primary care. In addition, the effective low-and moderate-intensity interventions in adults should be replicated (28 -30, 38). Uses of interactive mobile, Web-based, or other automated expert systems have been only minimally assessed for STI risk reduction in primary care.
High-intensity interventions conducted in primary care or similar health care settings can reduce sexually transmitted infections and risky sexual behavior in adolescents and in adults who are at high risk for STIs, and they are unlikely to be harmful.