A Meta-Analysis of the Efficacy of Behavioral Interventions to Reduce Risky Sexual Behavior and Decrease Sexually Transmitted Infections in Latinas Living in the United States

Abstract

The objective of this meta-analysis was to determine the effect of behavioral interventions in reducing risky sexual behavior and incident sexually transmitted infections (STI) among Latina women living in the United States. Studies were found by systematically searching the MEDLINE, EMBASE, and PsychInfo databases without language restriction. Two independent reviewers screened abstracts and full texts of articles to find randomized control trials testing the effects of behavioral interventions aimed at changing risky sexual behavior among Latinas. Articles were selected using prespecified inclusion criteria. Two independent reviewers extracted data from the included trials in duplicate using a standardized data extraction form. Six randomized control trials met the inclusion criteria for a total of 2,909 participants. Using random effects models with inverse variance weighting, we found a protective effect of the behavioral intervention on reported risky sexual behavior (odds ratio = 0.52; 95% confidence interval = 0.42, 0.64) and on incident nonviral STI (odds ratio = 0.65; 95% confidence interval = 0.46, 0.93). Behavioral interventions targeted toward Latina populations are effective in reducing risky sexual behaviors and incident STI and should be considered by policymakers as a potential tool for HIV/STI prevention in this population.

Latina women represent a high-risk subgroup for incident HIV and sexually transmitted infections (STIs). In the United States, Latinas are more than four times more likely to become infected with HIV than White women (Centers for Disease Control and Prevention [CDC], 2012). The high prevalence of STIs in this population further increases the risk of HIV transmission (CDC, 2011; Peragallo, Gonzalez-Guarda, McCabe, & Cianelli, 2012). Factors that increase the risk of HIV/STI among Latinas in the United States include low socioeconomic status, intimate partner violence, lack of knowledge of HIV, substance abuse, high reported concurrency among partners, and lack of awareness of their partner’s status (Adimora, Schoenbach, & Doherty, 2007; Campbell et al., 2008; CDC, 2013b; El-Sadr, Mayer, & Hodder, 2010; Hodder et al., 2010; Peragallo et al., 2012; Sareen, Pagura, & Grant, 2009). Cultural values and gender norms among Latinos, including those living the United States, have been shown to place Latinas at further risk of HIV/STI through a lack of empowerment, difficulty in negotiating condom use, and delayed testing and diagnosis (Cianelli, Ferrer, & McElmurry, 2008; Davila, 2005; Davila, Bonilla, Gonzalez-Ramirez, & Villarruel, 2007; Gonzalez-Guarda, Peragallo, Urrutia, Vasquez, & Mitrani, 2008; Hernandez, 2011; Moreno, 2007; Peragallo, 1996; Sowell, Holtz, & Velasquez, 2008).

Despite the potential for disease prevention, Latinas have been shown to be reluctant to discuss condom use with their partners (Deren, Shedlin, Decena, & Mino, 2005). Behavioral interventions aimed at increasing condom use among Latinas, therefore, have the potential for a substantial decrease in risk of HIV/STI acquisition.

The proposed mechanism of these behavioral interventions is primarily through culturally targeted interventions aimed at increasing partner communication, self-efficacy of condom use, HIV-related knowledge, knowledge of perceived barriers, and perceived HIV risk (Peragallo et al., 2012). There are several existing behavioral interventions designed for women, often Latina women in particular, that aim to increase the self-efficacy of condom use in an effort to decrease incident HIV/STIs (CDC, 2013a).

Meta-analyses on the effects of behavioral interventions on increasing condom use found that, in general, behavioral interventions increased condom use and decreased incident cases of HIV/STIs (Mize, Robinson, Bockting, & Scheltema, 2002; Scott-Sheldon, Huedo-Medina, Warren, Johnson, & Carey, 2011). Scott-Sheldon et al. (2011) and Mize et al. (2002) found that behavioral interventions seemed to be more effective in Latinos than other ethnic subgroups; however, only Mize et al. examined the effects of interventions targeted at women, including Latinas. Additionally, other reviews have found that behavioral interventions are more effective among Latinos if specific cultural needs are addressed in the intervention (Albarracin, Albarracin, & Durantini, 2008; Cardoza, Documet, Fryer, Gold, & Butler, 2012; Russell, Alexander, & Corbo, 2000).

Since these meta-analyses were published, several randomized control trials of behavioral interventions to increase condom use among Latina women have been published (Fisher et al., 2011; Peragallo et al., 2012; Wingood et al., 2011). Although prior meta-analyses have found that interventions appear to be efficacious among Latinos, no meta-analysis has focused specifically on the high-risk subgroup of Latina women. Given that recent studies focusing on this population have not been included in prior meta-analyses, this meta-analysis expands on previous work and provides a detailed examination of the literature and interventions targeted at Latinas. The purpose of this study was to investigate the efficacy of behavioral interventions aimed at promoting condom use and decreasing risky sexual behavior among Latinas in the United States through a meta-analysis of randomized controlled trials. The efficacy of behavioral interventions in reducing STI incidence was also assessed as a secondary outcome.

Method

Data Sources and Search Strategy

A comprehensive search of the literature using the MEDLINE, PsychInfo, and EMBASE databases (January 1950 through July 2013) and the National Institutes of Health (NIH) clinical trial registry (clinicaltrials.gov) was performed. Each database was searched using relevant medical subject headings and all fields keyword expansions of “behavioral intervention” AND “trial” AND (“condom use” OR “Sexually transmitted infections”) using search limits of “human” and “female.” The complete MEDLINE search can be found in the supplementary material (available online at http://heb.sagepub.com/supplemental). The search was not restricted by language, although all articles that met the inclusion criteria were published in English. Citations for conference abstracts, dissertations, and other potentially relevant literature were included in the search. References from retrieved articles and previous meta-analyses were checked for additional studies that met the inclusion criteria.

Study Selection

Two study investigators (MA, CG) independently screened 1,066 titles and abstracts and selected 90 articles for full-text review. Any disagreement between the investigators was resolved by consensus through a discussion of the article in question. The full text of the remaining 90 articles was reviewed by investigators (JH, LW), and disagreement was resolved by a third investigator (CG). Articles that met the following inclusion criteria were included in the meta-analysis: (a) intervention targeted to women, (b) randomized control trial, (c) intervention was behavioral in nature, (d) reported an outcome of condom use and/or STI acquisition, (e) follow-up ≥3 months, (f) the control group received standard of care or minimal intervention, (g) the study population was at least 65% Latina or results were stratified by race/ethnicity, (h) study was conducted in the U.S. noninstitutionalized population, and (i) study reported effect size and variance or the information needed to calculate summary measures. The inclusion criterion of 65% or more Latina in the study population cutoff was selected to be consistent with a previous meta-analysis on behavioral interventions in Latinos (Herbst et al., 2007). Figure 1 details the study selection process.

Figure 1.

Selection of studies for inclusion in the meta-analysis.

Data Collection

Data were extracted from the six articles that met the inclusion criteria by two investigators (DR, LW) independently and in duplicate using a standardized data collection form (attached in supplemental materials). The reviewers abstracted data on intervention type, duration, and content. The reviews also obtained data necessary to assess the quality of each trial, including randomization, blinding, clarity of eligibility criteria, adherence to the intervention, loss to follow-up, method used to obtain self-reported data, use of intention-to-treat analysis, and length of the recall period.

Reviewers collected data on the outcomes of interest: condom use, risky sex, and incident STI. They also recorded the study definition for each outcome and, if relevant, whether the effect measures were adjusted and unadjusted for confounders. Potential modifiers and mediators, including study population, baseline STI risk, mean age, socioeconomic status of participants, and length of the intervention were also collected. Any disagreement between the two investigators was resolved by a third reviewer (MA).

Data Analysis

Multiple measures of condom use were reported in the included studies necessitating the combination of related condom-use measures for meta-analysis. The primary outcome of interest was designated risky sexual behavior and included the following outcomes: never used condoms, inconsistent condom use, and noncompliance (sex with an untreated or incompletely treated partner after treatment for an STI). Respondents reporting any of these outcomes were coded as having risky sexual behavior in a dichotomous risky versus nonrisky sexual behavior variable. The secondary outcome of interest, incident STI, was created from the following reported measures: incident nonviral STI (Shain et al., 2004), incident chlamydia infection (Peragallo et al., 2012), and incident chlamydia or gonorrhea infection (Shain et al., 1999).

In studies where outcomes were measured at more than one follow-up time point (Peragallo et al., 2005; Peragallo et al., 2012; Shain et al., 1999; Shain et al., 2004; Wingood et al., 2011), the longest follow-up time was selected for the primary outcome analysis because it was felt to best represent sustained intervention effect. One study (Wingood et al., 2011) also reported multiple recall periods for behavioral outcomes at each of the follow-up time points. The shortest recall period was used for the analysis in an effort to select the measure with the least misclassification possible.

A sensitivity analysis based on follow-up time was planned a priori to see whether there was evidence of sustained effects of the intervention. Quality parameters were investigated through a posteriori sensitivity analyses of studies using blinding and by analyzing studies using intention-to-treat analysis to see if different effect sizes were estimated in these groupings of presumably higher quality studies. Additional a posteriori sensitivity analyses were conducted to examine substitution of alternate risky sexual behavior measures as well as to recall periods and year of publication, in order to determine whether outcome definition heterogeneity or extended recall time changed the results. A sensitivity analysis by study retention was considered but rejected because of the narrow range of participant retention in selected studies.

Estimates are reported as odds ratios (ORs) and as Cohen’s d standardized mean differences. The OR was selected for its ease in interpretability; however, because of heterogeneity in the outcome variables, Cohen’s d standardized mean differences are also presented. Wingood et al. (2011) reported adjusted OR in addition to unadjusted estimates. The adjusted OR was used for the meta-analysis. The OR for risky sex was either reported directly in the article or calculated using aggregate data on the number of participants in each group who reported the risky sex outcome at longest follow-up. The standard error was calculated using reported confidence intervals or from the same data used to calculate OR. Cohen’s d was calculated either from rate data or was converted from the OR, when rate data were not reported in the article (Sanchez-Meca, Marin-Martinez, & Chacon-Moscoso, 2003).

The natural log of the OR and standard error were used to pool the estimates using fixed effects and DerSermonian and Laird random effects models with inverse variance weighting. Heterogeneity was assessed using the Cochrane’s Q test and the I-squared statistic. A cut point of p < .10 was used to determine if significant herterogeneity was present. The random effects models are presented because the results are a more conservative estimate of the standard error.

The following subgroup analyses were preplanned: (a) interventions targeted specifically for Latinas and (b) baseline HIV/STI status preintervention. These subgroups were chosen because we hypothesize that interventions targeted to Latinas will more effective in this population than general interventions for women. We also believe that women with a previous STI at baseline are a higher risk population than women without an STI and therefore the effects of the intervention may differ between these two risk groups.

Publication bias was assessed using a funnel plot and the Begg and Egger tests with p-values cut point set at .10. This cut point was selected to minimize the potential for Type II error when assessing publication bias. Study quality was assessed using a checklist adapted to fit the needs of this meta-analysis (Jadad, Moher, & Klassen, 1998; Verhagen et al., 1998). All analyses were performed using Stata 11.2 (StataCorp, College Station, TX).

Results

The results from the search strategy and study selection are detailed in Figure 1. A total of six studies met the eligibility criteria for inclusion (Peragallo et al., 2005; Peragallo et al., 2012; Shain et al., 1999; Shain et al., 2004; Suarez-Al-Adam, Raffaelli, & O’Leary, 2000; Wingood et al., 2011). One study was found by searching the references of articles selected for full-text review (Suarez-Al-Adam et al., 2000).

Study Characteristics

Study characteristics are reported in Table 1. The total number of participants was 2,909, and the number of participants in the individual studies ranged from 60 to 775. The longest follow-up postintervention ranged from 3 months to 2 years. The studies by Peragallo et al. used the same intervention in different populations (Peragallo et al., 2005; Peragallo et al., 2011) and the studies by Shain et al. used the same intervention; however, an additional enhanced group was added as a separate intervention group in the 2004 study (Shain et al., 2004). The enhanced intervention group was given the same intervention as the standard group plus extra optional sessions. Participants in the standard and enhanced intervention group were combined and compared with the control group for this analysis.

Table 1.

Study Characteristics.

Study	Year	N	Age (Mean ± SD)	Latina (%)	SES (%)	Education (Mean ± SD)	Study Population	Intervention Name	Intervention Description	Length of Intervention	Control Group	Outcome	Maximum Follow-Up
Peragallo	2005	657	―	100	70.5%a	48.2%b	Community; Sexually active without condom use	SEPA	“Culturally tailored sessions on understanding their bodies, HIV/AIDS and STDs, condoms (myths and use), negotiating safer sex practices, violence prevention, and partner communication,” “Skills training . . . role-playing . . . consistent with Latinas’ values and beliefs”	3+ sessions	HIV test counseling	Condom use: always (95% to 100%) versus inconsistent (0% to 95%)	6 months
Peragallo	2012	548	38.5 ± 8.5	100	69.5%c	13.4 ±3.45	Community; Sexually active	SEPA	“Culturally specific and theoretically based group intervention for Hispanic women . . . hands on activities, role playing . . . skills demonstration, homework to build self-efficacy and direct provision of information”	5 sessions	One session intervention at month 12	Any condom use; Chlamydia infection	12 months
Shain	1999	775	21.6 ± 0.43	69	$254.5 ±$l7.7e	10.8 ±0.11	Clinic; Previous STI infection	Project SAFE	“A sex and culture-specific behavioral intervention . . . adapted [the] AIDS Risk Reduction Model . . . lively and open discussions, games, videotape watching, behavior modeling and role play. Learning was personalized”	3 sessions	Standard STI counseling	Compliance; Five or more unprotected sexual acts; Chlamydia or gonorrhea infection	24 months
Shain	2004	617	21.0 ± 0.61	76	$302.9 ±$23.7e	10.7 ±0.18	Clinic; Previous STI infection	Project SAFE 2	Same intervention as described in Shain, 1999	3 sessions	Standard STI counseling	Compliance; Nonviral STI	12 months
Suarez	2000	60	34.5	100	―	46.0%b	Community; Sexually active without condom use	―	“Based on social cognitive theory . . . providing information and building skills to promote safer sex ... 3 weeks”	7 sessions	Health-promotion also lasting 3 weeks	Safer sex (using condoms for all sex acts or abstaining from	6 months
Wingood	2011	252	30.3 ± 6.9	100	54.8%d	43.8%b	Community; Sexually active	AMIGAS	“A culturally congruent HIV prevention intervention for Latin women . . . [focused on] healthy relationships, cultural and gender pride . . . video testimonials of Latina women who were living with HIV . . . explore how experiences affect risk . . . role-playing”	4 sessions	1 -session health intervention	Never used condoms; Consistent condom use (use of condoms for all sexual acts)	6 months

Note. Suarez et al. did not report the SD for age. SES = socioeconomic status; STD = sexually transmitted disease; STI = sexually transmitted infection.

^aAt or below 100% of the poverty line.

^bPercentage less than high school.

^cMonthly income <$2,000.

^dOn public assistance.

^eMean monthly per capita income.

Although the specific interventions varied between studies, all of the interventions focused on increasing awareness of HIV/STI transmission and teaching skills in negotiating condom use. The number of sessions per intervention ranged from three to seven. The interventions of the included trials were designed to be culturally appropriate for the Latina participants with the exception of the Suarez-Al-Adam et al. (2000) study. A sensitivity analysis was performed without the Suarez-Al-Adam et al. study, and the results of the meta-analysis remained unchanged. The Shain et al. (1999, 2004) trials were conducted in public clinics with participants that were diagnosed with a nonviral STI at baseline. The study populations of the other included trials were recruited from communities. The differences in the community-based versus clinic-based populations were the basis for our subgroup analysis.

There was little heterogeneity among the potential modifiers and mediators of intervention efficacy, which, in combination with the small sample size, limited the ability to detect which factors may be important to the success of the intervention.

Pooled Effects

Individually, each study found a protective effect of the behavioral intervention on risky sex. However, the point estimates were varied, and some studies did not have statistically significant findings. The pooled effect showed little heterogeneity (I-squared = 0.0%) and the pooled OR and 95% confidence interval using random effects modeling was 0.52 (0.42, 0.64) for all included trials (Figure 2). Both subgroups, STI at baseline and no STI at baseline, had statistically significant protective effects against risky sex (Figure 2). The STI at baseline group had a slightly more protective effect than the community-based group with an OR of 0.46 (0.33, 0.63) and 0.57 (0.44, 0.76), respectively. The pooled Cohen’s d was 0.36 (0.25, 0.47), indicating a moderate difference in risky sexual behavior between groups.

Figure 2.

The effect of behavioral interventions on risky sex.

Three studies measured nonviral STI incidence during the follow-up period (Peragallo et al., 2012; Shain et al., 1999; Shain et al., 2004). The two Shain et al. studies were appropriately powered to detect a difference in STI incidence between the intervention and control groups. The Peragallo et al. study from 2011 also assessed incident STI; however, they were underpowered to detect an effect of the behavioral intervention on STI acquisition. The pooled random effects model showed a protective effect with an OR of 0.65 (0.46, 0.93) (Figure 3). There was some heterogeneity (I-squared = 23.5% [0.0% to 65.0%]), which informed the decision to use the random effects model. The pooled Cohen’s d was 0.24 (0.02, 0.47), indicating a small difference between STI acquisition between the intervention and control group.

Figure 3.

Behavioral interventions on incident sexually transmitted infections.

Quality of Studies and Risk of Bias

Among the included studies, all had adequate randomization methods, clearly stated eligibility criteria, and descriptions of losses and exclusions. Although studies had relatively high retention (ranging from 69.1% to 91.2%), only two used intention-to-treat analysis. Three studies (50%) did not use any blinding of participants or assessors. Quality measures for each study are described in Table 2.

Table 2.

Quality Assessment of Studies.

Study	Blinding	Adequate Randomization	Intention to Treat Analysis	Eligibility Criteria Clearly Specified	Power Calculations Described	Longest Follow-Up Time (Months)	Recall for Self-Reported Outcomes (Months)	Study Retention (%)	Description of Losses and Exclusions
Peragallo et al. (2005)	No	Yes	No	Yes	Yes	6	3	69.1	Yes
Peragallo et al. (2012)	No	Yes	Yes	Yes	Yes	12	3	69.5	Yes
Shain et al. (1999)	Single	Yes	No	Yes	No	12	12	89.0a	Yes
Shain et al. (2004)	Single	Yes	No	Yes	No	24	12	91.2a	Yes
Suarez-Al-Adam et al. (2000)	No	Yes	No	Yes	No	3	1	76.6	Yes
Wingood et al. (2011)	Double	Yes	Yes	Yes	No	6	1.3	88.9	Yes

^aAnalysis on behavioral outcome used only participants with high compliance.

Sensitivity analyses were performed to assess the robustness of our findings with respect to outcome variable selection and relevant quality measures (Table 3). None of the sensitivity analyses had results that varied significantly from the pooled effect measure.

Table 3.

Sensitivity Analyses.

	Number of Studies	Pooled OR (95% CI)	Heterogeneity: p Value of Q Statistic
Intention to treat analysis	2	0.54 (0.38, 0.77)	.682
Only follow-up ≥1 year	3	0.49 (0.38, 0.64)	.721
Studies published in the past 10 years	4	0.55 (0.43, 0.69)	.616
Interventions designed for Latinas	5	0.53 (0.43, 0.65)	.706
Shain et al. (1999) outcome: ≥5 unprotected sex acts	6	0.55 (0.45, 0.68)	.779
Wingood: 90-day recall	6	0.53 (0.43, 0.65)	.769
Wingood: Inconsistent condom use, 30-day recall	6	0.50 (0.41, 0.62)	.449
Wingood: Inconsistent condom use, 90-day recall	6	0.47 (0.36, 0.63)	.160
Pooled	6	0.52 (0.42, 0.64)	.761

Note. OR = odds ratio; CI = confidence interval.

Publication bias was assessed using a funnel plot and the Begg and Egger’s tests. The funnel plot (Figure 4) shows little evidence of publication bias. Similarly, the Begg and Egger’s tests had nonsignificant p values, .707 and .403, respectively. Although the statistical findings were nonsignificant, the Begg and Egger’s tests are underpowered to detect publication bias since only six studies are included in the meta-analysis.

Figure 4.

Funnel plot.

Discussion

This meta-analysis aimed to determine the impact of behavioral interventions targeted at reducing sexual risk factors and STIs among Latina women living in the United States. We found that women in the behavioral intervention groups were less likely to engage in risky sexual behavior and less likely to acquire a nonbacterial STI than those in the standard of care groups.

Our results are consistent with findings from other meta-analyses on behavioral interventions to reduce risky sexual behaviors that included studies intervening on both genders and multiple ethnic groups in the United States (Albarracin et al., 2008; Cardoza et al., 2012; Crepaz et al., 2007; Crepaz et al., 2009; Herbst et al., 2007; Huedo-Medina et al., 2010; Mize et al., 2002; Neumann et al., 2002; Noar, 2008; Russell et al., 2000; Scott-Sheldon et al., 2011). Our findings are consistent with a 2007 systematic review and meta-analysis of behavioral interventions to reduce HIV risk behaviors of Hispanics by Herbst et al. (2007). Herbst et al. reported a 31% reduced odds of acquiring an incident STI and a 25% reduced odds of displaying risky sexual behavior, which is similar to results obtained in this meta-analysis. Another more recent meta-analysis from 2011 demonstrated the efficacy of behavioral interventions to increase condom use and reduce STIs within a wide range of populations (Scott-Sheldon et al., 2011), and it further suggested that interventions were more successful in HIV reduction in samples with a greater proportion of Latino participants. These findings are concordant with Mize et al.’s 2002 meta-analysis that found that interventions to increase risk reduction behaviors were more effective among Latina samples compared with other ethnicities.

Latina women have been shown to face unique cultural challenges in negotiating condom use with their partner (Cianelli et al., 2008; Gonzalez-Guarda et al., 2008; Hernandez, 2011; Moreno, 2007; Peragallo, 1996; Sowell et al., 2008), which reinforces the importance of interventions that address these culture- and gender-specific issues. Five of the six trials included in this meta-analysis used ethnographic data from Latinas to inform their interventions so that they addressed the appropriate cultural context within which sexual decision making takes place (Peragallo et al., 2005; Peragallo et al., 2012; Shain et al., 1999; Shain et al., 2004; Wingood et al., 2011). When we restricted our analysis to the five studies using interventions targeting Latinas, the resulting protective OR was very similar to our overall results.

Prior research has indicated that interventions are more successful in lowering incident STIs in participants with a baseline STI diagnosis but are not more successful in improving sexual risk behaviors in participants with a baseline STI diagnosis (Scott-Sheldon et al., 2011). Our subgroup analysis by STI status at baseline showed that effect sizes and confidence intervals for risky sexual behavior were quite similar between the subgroups, suggesting that there was not a difference in effect sizes, which is consistent with existing literature. We did not perform a statistical test for significance between subgroups by STI status at baseline for the outcome of risky sexual behavior because of the small number of studies in this subgroup analysis. Since only three studies reported STI outcomes, we were not able to examine a similar relationship for the STI outcome.

We did not find a dose–response effect with respect to the number of sessions for each intervention; however, we were likely underpowered to detect a dose–response relationship. The number of sessions ranged from three to seven, and there does not appear to be an increased protective effect from the longer interventions. Determining the optimal length and most efficacious content of interventions was out of the scope of this meta-analysis, but it is an important area of research before any widespread implementation of behavioral interventions.

Our study findings should be interpreted in light of limitations of the underlying literature. The individual studies may suffer from social desirability bias. It is possible that women assigned to the intervention group feel more obligated to report a decrease in risky sexual behavior than participants in the control group. This differential misclassification would bias the results away from the null, creating an exaggerated effect measure. Social desirability bias was minimized in the Wingood et al. study through the use of audio-computer assisted self-interviewing (ACASI; Wingood et al., 2011). ACASI has been shown to elicit more accurate responses from participants, decreasing social desirability bias (Kissinger et al., 1999). Although the studies included in this meta-analysis likely suffer from social desirability bias to some extent, the biological data on incident STI provided an objective measure that supports the primary finding that behavioral interventions reduce risky sex behaviors. We found a decrease in incident STI among those in the intervention groups. Presumably, this means that these women increased condom use and decreased other risky sexual practices, which provides evidence that, despite potential social desirability bias, behavioral interventions decrease risky sexual behavior.

We originally aimed to assess increase in condom use as our primary outcome but the wide variability in the definition of condom use made it challenging to create a consistent condom use variable for analysis. Similarly, there was variability in the length of follow-up of the studies. The follow-up period ranged from 3 months to 2 years. Assessing the long-term sustainability of the behavioral intervention is vital to predicting the effects of the intervention on HIV/STI prevention. A sensitivity analysis found a similar protective effect of behavioral interventions when using only follow-up greater than or equal to 1 year; therefore, we can conclude that the effects of the intervention appear to be sustainable for at least 1 year following the completion of the intervention.

Of note, only two studies correctly used intention-to-treat analysis for assessment of the behavioral intervention. The other studies analyzed only the women who attended all of the intervention sessions, which may result in an overestimation of the protective effect of the intervention.

Finally, the use of an odds ratio as opposed to a risk ratio likely overestimates the protective effect of the behavioral interventions. Since the protective effect was large and the upper bounds of the confidence interval did not approach the null, there would still likely be a significant protective effect of behavioral interventions had we been able to use a risk ratio as our effect estimate.

Despite these limitations, this is the first meta-analysis, to our knowledge, that evaluated the effect of behavioral interventions aimed at reducing risky sex behaviors in the Latina population in the United States. Latinos currently represent 17% of the U.S. population, and it is estimated that their population will increase to nearly one third of the U.S. population by 2060 (U.S. Census Bureau, 2012). Given that HIV/STIs disproportionately affect Latinas and that their population is expected to grow significantly, it is imperative that evidence-based efforts be made to decrease the risk of HIV/STI in this population. Additionally, this population is important to study as they have unique challenges in negotiating condom use and decreasing their risk of HIV/STI acquisition. Our findings support the implementation of culturally specific behavioral interventions aimed at reducing the risk of HIV/STI transmission in Latina populations.

The effect measures from the individual studies were relatively similar, which contributed to the low heterogeneity among included studies. Despite the low heterogeneity, we chose to use random effects modeling and still found highly statistically significant findings. We performed numerous sensitivity analyses, none of which showed a substantial change in effect estimate or confidence interval width compared with the pooled estimate. This speaks of the robustness of our findings with respect to the selection of the outcome variables.

Conclusions

We conclude that behavioral interventions targeted to address the challenges of preventing STIs in Latina populations in the United States successfully reduce risky sexual behaviors and incidence STIs. The protective effects have been shown to persist for as long as 2 years following the intervention and are of near equal magnitude regardless of whether the intervention is implemented in women with STI at baseline or community-based populations, indicating that these interventions may be useful in a variety of settings. Prevention of HIV/STI in Latina populations is important, especially as this population continues to grow in the United States. The SEPA intervention from the Peragallo et al. (2005, 2012) studies and the AMIGAS intervention from the Wingood et al. (2011) study are endorsed by the Centers for Disease Control and Prevention and the Miami-Dade County Health Department, respectively, and are available options for public health practitioners to implement. These behavioral interventions, among others, were efficacious in reducing the odds of risky sexual behavior and STI acquisition among Latinas.