Globally, young people are at elevated risk of STIs, HIV and unintended pregnancy. Notable gender and racial disparities exist. In the United States, for example, 2013 chlamydia rates were more than four times as high among 15–19-year-old females as among males of the same age, and the rate among black females was five times the rate among white females in that age-group.¹ Worldwide, in 2013, among adolescents aged 15–19, two-thirds of new HIV infections were among females.² In some countries, the disparity is even greater; for example, HIV prevalence among young people aged 15–24 in South Africa is 14% for females and 4% for males.³ In addition, the consequences of unintended pregnancy, along with the associated risks of childbearing and the responsibilities of child care, fall disproportionately on females.

Along with efforts to deliver clinical services, reduce structural vulnerability^4–6 and foster protective social norms, a key strategy for improving adolescent sexual health outcomes has been group- and curriculum-based sexuality and HIV education. Indeed, in response to the call of international agreements such as the 1994 International Conference on Population and Development (ICPD),^7–10 international agencies continue to prioritize comprehensive HIV and sexuality education,^11–13 and many countries are undertaking national efforts or have adopted policies for education to help prevent adolescent pregnancy and HIV transmission.^14–16

Sexuality education curricula may be delivered in schools, community settings or clinics as a stand-alone program or as a component of a multifaceted intervention, such as a young women’s financial literacy program. These programs go by various names, including "family life education," "AIDS education" or "health education," to name a few. The term "comprehensive sexuality education" has evolved historically and continues to be used elastically. It has often been used to describe curricula of any duration that provide complete, medically accurate content, including information about contraception and condoms, regardless of whether topics such as gender, rights, equality, diversity and power are addressed. In general, the term has been used to distinguish such curricula from abstinence-only approaches.^17–19

Despite extensive investments in and evaluations of sexuality and HIV education for young people, questions of effectiveness persist. Indeed, many researchers note that significant room for program improvement remains,^20–26 and a number of reviews have sought to tease out characteristics common to effective curriculum-based interventions.^25,27,28 Consensus has been reached about several of such characteristics—i.e., the benefits of comprehensive versus abstinence-only content, and of participatory, skills-building teaching approaches—but overall, the literature raises a number of questions. "What is success?" and "What exactly works?" remain matters of ongoing debate. At least two articles have critiqued some of the reviews on methodological grounds.^26,29 Others noted that whether a program is deemed successful or not may depend on whether an evaluation measures a behavioral outcome or a biological or health variable. Because reported sexual behavior does not always correlate with health outcomes, and because of issues regarding reporting by intervention participants—such as the validity of self-reports of sexual behavior and the potential for social desirability bias—one review concluded that "trials with reported sexual behaviors as their outcome are insufficient."^{30(p. S11)} Indeed, many reviews recommend the use of a higher bar, biological outcomes, as a more reliable, objective measure of program efficacy.^{20,22,27,30–34} Of course, tracking biological and health outcomes requires large sample sizes and substantial resources. Thus, the use of adolescents’ self-reported sexual behavior change remains the only practical option for many studies. The results of such studies can still be instructive, but impact data are preferable for generating lessons about evidence-based programs and identifying key program characteristics.

In addition, researchers have noted the difficulty of identifying key characteristics that are consistent across studies.^24,35 For example, Chin and colleagues found that no moderator variables—dosage (average number of program hours), setting (school or community), focus (HIV and STIs, pregnancy, or both), facilitator (adult, peer or both), number of components (single versus multiple) and targeting (tailoring of materials to participants)—were consistently associated with effectiveness or lack of effectiveness in their meta-analysis of 66 studies.¹⁷ Johnson and colleagues, in their review, found that interventions were more successful when they delivered more intensive content—for example, through more sessions, more condom skills training or more motivational training—though they note that "finer grained analyses of intervention content may yield better explanation of efficacy."^{28(p. 82)}

This point is well taken, as only characteristics that are looked for will be found and proven or disproven as consequential. In recent years, drawing largely from the ICPD, emerging evidence, and field experiences of feminist and nongovernmental organizations in developing countries, international agencies, donor programs and researchers have increasingly highlighted gender as a topic integral to comprehensive sexuality education.^19,36–40 Although this approach is gaining currency on the ground in some settings, meaningful attention to gender is still far from the norm.^41,42 Programs have lagged in integrating a gender or power perspective into comprehensive sexuality education because there is a lack of clarity about what a gender or power perspective means, and especially, about how to implement such an approach clearly enough and with enough detail for both the educator and the learner. This article explores whether the inclusion of content focusing on gender and power matters for program efficacy and the ways in which effective curriculum-based programs have addressed gender and power.

Rationale

A strong theoretical base supports attention to gender and power in comprehensive sexuality education. Connell’s theory of gender and power, for example, provides a rich theoretical underpinning of the social structures that characterize the gendered relationship between males and females.⁴³ Wingood and DiClemente⁴⁴ have extended this theory, highlighting how economic factors, relationships of unequal power and gender norms manifest in exposures and risk factors that increase women’s risk of HIV. Pulerwitz and colleagues⁴⁵ operationalized the concept by developing and validating the Sexual Relationship Power Scale, a tool used to measure relationship power in HIV and STI research.

In addition, a large body of empirical evidence indicates that gender and power matter for sexual and reproductive health behavior and outcomes. This evidence is consistent across three interrelated domains: gender norms (including masculinity, femininity and equality), power in sexual relationships, and intimate partner violence. Harmful gender norms have been correlated with a number of adverse sexual and reproductive health outcomes and risk behaviors, even after other variables have been controlled for. For example, studies have found that individuals who adhere to harmful gender attitudes are significantly less likely than those who do not to use contraceptives or condoms.^46–51 Also, compared with women and female adolescents’ reports of more equitable relationships, reports of low power in sexual relationships have been independently correlated with negative sexual and reproductive health outcomes, including higher rates of STIs and HIV infection.^52–54 And women and female adolescents who have experienced intimate partner violence are significantly more likely than those who have not to have a host of adverse outcomes—from low rates of condom use^55–57 to higher rates of pregnancy^58–61 and STIs or HIV infection.^{52,54–56,62,63}

Thus, a strong rationale exists for why attention to gender and power has the potential to improve the outcomes of curriculum-based sexuality and HIV education programs for young people. But does inclusion of these topics in a group- and curriculum-based context make a difference? We could identify no rigorous studies that examined the efficacy of sexuality education programs in terms of whether they included content on gender and power. The current review was undertaken to determine what existing evaluations of interventions for young people might suggest. Mindful of researchers’ concerns regarding the strength of evidence for effectiveness and their recommendation to look at actual health outcomes (as opposed to self-reports of behavior change), the review included rigorous evaluations of group- and curriculum-based sexuality and HIV education that assessed health outcomes— specifically, pregnancy, childbearing, HIV or other STIs— to compare programs that included attention to gender and power with those that did not. This study also seeks to provide initial insight into which characteristics of gender and power programs appear to influence effectiveness.

METHODS

The electronic databases searched were PubMed, ERIC, Cochrane Central Register of Controlled Trials and Eldis. The following search terms were used: "evaluation," "outcome," "impact" or "effect;" "program" or "intervention"; "HIV," "AIDS," "STI," "STD," "sexually transmitted infection," "sexually transmitted disease" or "pregnancy;" and "adolescent," "adolescence," "youth," "young people" or "teen." The reference sections in 36 reviews, meta-analyses and systematic reviews of interventions aiming to decrease sexual risk, including three Cochrane Reviews, were hand searched, as were the Centers for Disease Control and Prevention’s Compendium of Evidence-Based HIV Prevention Interventions Web site and the Office of Adolescent Health’s Teen Pregnancy Prevention Resource Center Web site.^{17,20–25,28–36,64–83}

Studies were included if they were evaluations of behavior-change interventions to prevent HIV, STIs or unintended pregnancy that were group- and curriculum-based, or were multicomponent interventions in which one of the main components entailed participants meeting in a group and following a curriculum; exclusively or predominantly assessed effects on adolescents aged 19 or younger; were published between 1990 and 2012; used rigorous designs, such as randomized controlled trials or quasi-experimental studies that adjusted for baseline differences; had a minimum sample size of 100; and measured the effect of the intervention on health outcomes—i.e., STIs, HIV, pregnancy or childbearing.

Programs of any length could be evaluated, and programs were not required to include all of the criteria that have been outlined in various standards for sexuality education, as long as they were not abstinence-only. The programs included in this review are typical of the diverse scope of non–abstinence-only programs that operate in much of the world.

Studies were excluded if they were conducted among special populations (such as drug users, men who have sex with men or commercial sex workers).

Criteria were established for classifying curricula as addressing gender—gender norms, gender equality, and harmful or biased practices and behavior driven by gender—and power inequalities in intimate relationships. Specifically, curricula had to go beyond the conventional content on resisting sexual advances (refusal skills) to include at least one explicit lesson, topic or activity covering an aspect of gender or power in sexual relationships—for example, how harmful notions of masculinity and femininity affect behaviors, are perpetuated and can be transformed; rights and coercion; gender inequality in society; unequal power in intimate relationships; fostering young women’s empowerment; or gender and power dynamics of condom use.

The classification of an intervention as addressing gender and power was first determined by assessing the description provided in the primary article, and when available, related articles. If this was insufficient, the curriculum or curriculum summaries were obtained; in some instances, the authors were contacted for details on program content. Notes describing the way that gender and power were addressed in the intervention were taken as needed. This content review was conducted by three researchers other than the author and was blind, i.e., information on the results of the program was not provided to the researchers.

Also, so that other potential influences on outcome could be considered, additional information was extracted about each program and study, including the study design, the theoretical basis and general description of the intervention, the duration of the intervention, and the pedagogical approach. The independent effect of each intervention on health outcomes—pregnancy, childbearing, HIV or STIs—was recorded separately.

RESULTS

Of 8,230 citations identified, 7,614 were excluded after examination of the title. Of the remaining 616 citations, 316 were excluded after examination of the abstract, and 300 articles were reviewed in full. Twenty-seven articles (22 studies) met all inclusion criteria.^84–110 Of the 22 studies included in this review, 14 were conducted in the United States,^{84,85,88–90,93,95–97,99,100,104,105,107} six in low- or middle-income countries^{86,91,92,94,101,108} and two in high-income countries other than the United States.^98,109 Fifteen were randomized controlled trials,^{84–86,88–91,93,94,96,98,100,101,105,109} and seven were longitudinal cohort studies with controls.^{92,95,97,99,104,107,108} Sample sizes ranged from 148¹⁰⁵ to more than 9,000 participants.¹⁰¹ Seven studies enrolled females only,^{85,89,90,93,99,104,105} 15 included both females and males^{84, 86,88,91,92,94–98,100,101,107–109} and none enrolled males only. Ten of the included studies were conducted in schools,^{84,88,91,92,95–98,107,109} five in clinic settings,^{89,90,93,104,105} four in community settings,^{94,99,100,108} two in multiple settings^86,101 and one at a Marine recruit training base (Appendix Table 1).⁸⁵

Program Efficacy

About half of the programs (10) demonstrated significant decreases in pregnancy,^84,86,89,100 childbearing,⁹¹ STIs,^{89,90,93,94,105} or STIs and pregnancy combined;⁸⁵ just over half (12) failed to show a significant, independent effect on any of these outcomes.^{88,92,95–99,101,104,107–109} Of the 10 effective programs, five enrolled both females and males.^{84,86,91,94,100} In one study, effects were the same for females and males (both sexes had a reduction in HSV-2 incidence, and the incidence of HIV infection and pregnancy involvement were unchanged for both).⁹⁴ Another study found differential effects by sex: decreased pregnancy among females, but no change in males’ reports of causing a pregnancy.¹⁰⁰ In the remaining three studies, pregnancy or childbearing information was gathered or reported for females only,^84,86,91 although for one of these, a subsequent study found that the program’s effect on reducing pregnancy did not differ for females and males.¹¹¹

Study and program dimensions that may help explain why some comprehensive sexuality education programs were effective and others were not are examined below. Characteristics of the research designs, different aspects of interventions, and finally the gender and power content of the curricula are examined.

Study Design and Efficacy

All studies employed rigorous designs: randomized controlled trials or longitudinal cohort studies with controls. Two-thirds (10 out of 15) of the programs that were evaluated with a randomized controlled trial achieved significant reductions in pregnancy,^84,86,100 childbearing,⁹¹ STIs,^{89,90,93,94,105} or STIs and pregnancy combined.⁸⁵ None of the programs evaluated with a longitudinal cohort study with controls demonstrated a significant independent effect on any of these indicators.^{92,95,97,99,104,107,108}

Sample size may affect the likelihood of detecting an effect. Six studies had sample sizes of between 100 and 500 participants, two (33%) of which showed significantly decreased pregnancy or STI rates^100,105 and four (67%) of which did not.^92,97,99,104 Of the six studies with sample sizes of between 501 and 1,000 adolescents, four (67%) decreased pregnancy or STI rates^84,89,90,93 and two did not.^88,95 Ten studies had more than 1,000 participants in their sample, with four (40%) demonstrating a significant decrease in pregnancy, childbearing, STIs, or STIs and pregnancy combined,^85,86,91,94 and six having no significant effect on these outcomes.^{96,98,101,107–109} Thus, while programs in the smallest sample size category were least likely to report an impact on health outcomes, some programs evaluated under even this scenario showed decreased pregnancy or STI rates. Larger sample sizes did not guarantee detection of a significant effect—only half of the studies with more than 500 participants showed significant decreases in pregnancy, childbearing or STIs.

Postintervention follow-up was six months or less in five studies; one evaluation showed a significant decrease in pregnancy risk⁸⁴ and four had no effect on pregnancy or STIs.^{92,104,107,108} Nine of the 17 studies (53%) that followed participants for one year or longer found significantly decreased rates of pregnancy, childbearing or STIs,^{85,86,89–91,93,94,100,105} and eight of the 17 did not.^{88,95–99,101,109} For the outcomes examined, longer term follow-up (a year or more) appears more likely to detect a beneficial impact than follow-up at six months or less.

Eight of the studies were published in the 1990s, and 14 were published in 2000 or later. One of the eight studies from the 1990s showed significantly decreased pregnancy rates,⁸⁴ whereas the other seven studies had no effect on pregnancy or STI outcomes.^{92,95,96,98,99,104,107} Among the 14 studies published in 2000 or later, nine (64%) had an effect on childbearing, pregnancy, STIs, or pregnancy and STIs,^{85,86,89–91,93,94,100,105} and five did not.^{88,97,101,108,109} Programs evaluated more recently, i.e., 2000 or later, appear more likely to demonstrate beneficial health impacts than the earlier studies.

Pedagogy, Theory, Duration and Efficacy

As noted above, one program characteristic that has been fairly consistently correlated with effective programs is interactive, learner-centered and skills-based teaching approaches. Of the 22 programs, all but one explicitly described using interactive, participatory, learner-centered or critical thinking pedagogy. Thus, while good pedagogy may be a prerequisite for positive results, alone it is not what distinguished effective from ineffective programs.

Another characteristic cited as important for efficacy is that the program be based on a theory of behavior change (such as theory of reasoned action, health belief model, social cognitive theory, etc.) or have a clear model for how the program will lead to behavior change.⁷⁹ This characteristic had no explanatory power in this set of studies. Most—20 out of 22 programs—described their theory or model of behavior change. Of those that specified a theory of behavior change, half (10) significantly decreased pregnancy or STIs^{84–86,89–91,93,94,100,105} and half (10) did not.^{88,95–99,101,104,107,109} This finding is consistent with a recent review of HIV prevention programs for young people in Sub-Saharan Africa that found no difference in the effectiveness of programs that were based on theory and those that were not.¹¹²

Nor did intervention duration appear to differentiate effective from ineffective programs. Both sets of studies—programs that significantly decreased adverse health outcomes and programs that did not—comprised interventions with a similar range in duration: 1–2 single-session interventions, 5–8 of intermediate duration (2–17 sessions, totaling 5–50 hours), and three that ran for at least one school year.

Program Setting, Multiple Components and Efficacy

Evaluations were conducted in multiple settings, including schools, clinics, community-based organizations, a Marine recruit training base or a mix of these. Out of the 10 evaluations conducted in schools, two (20%) found a significant independent reduction in pregnancy or childbearing,^84,91 and eight had no effect on pregnancy or STI outcomes.^{88,92,95–98,107,109} Four out of five programs implemented in clinics (80%) significantly decreased STIs;^89,90,93,105 one of the clinic-based programs had no effect.¹⁰⁴ The four community-based programs were as likely to have had an effect on health outcomes as not: Two significantly reduced pregnancy or STI rates^94,100 and two did not.^99,108 The program implemented in a Marine recruit training base significantly decreased pregnancy and STIs (as a combined variable).⁸⁵ Of the two programs that were implemented in multiple settings, one had a significant, positive effect on pregnancy,⁸⁶ and the other had no effect on health outcomes.¹⁰¹ Schools appeared to be more challenging settings, whereas clinic-based programs were more likely to have a significant effect.

Multicomponent interventions have been hypothesized to be more effective than single-component interventions. Of the 22 studies in this review, 14 were single-component sexuality education interventions. Eight were multicomponent, that is, they included at least one other type of program element, such as service learning,^84,88 community awareness raising,^86,101 health services or vouchers for services,^{86,90,99–101,108} or activities or support in nonhealth areas (e.g., jobs, academics, art, sports),^99,100 in addition to curriculum-based sexuality education. Half (four out of eight) of the multicomponent interventions demonstrated a significant decrease in pregnancy or STI rates,^84,86,90,100 and almost half (six out of 14) of the single-component interventions demonstrated such an effect.^{85,89,91,93,94,105} This did not vary by type of component. For example, of the two programs that had a service-learning component in addition to group- and curriculum-based sexuality education, one had a significant reduction in pregnancy rates,⁸⁴ and the other had no effect on health outcomes.⁸⁸ Thus, as Chin and colleagues¹⁷ found in their meta-analysis, multicomponent interventions were not found to be associated with a greater likelihood of effect than single-component interventions among these studies.

Gender and Power Content and Efficacy

Disaggregating the evaluated programs by gender and power content found that 10 curricula included attention to issues of gender or power,^{84,86,89–91,94,99–101,105} and 12 did not.^{85,88,92,93,95–98,104,107–109} The two groups of curricula—those that included gender or power and those that did not—were similar in most other program aspects analyzed. Table 1 shows roughly similar breakdowns by location, female-only vs. mixed sex, sample size, last follow-up survey, whether participatory and learner-centered teaching methods were used, and whether the program was theory based. Dimensions in which programs that included attention to gender or power appeared to differ from other programs were setting, number of components and some study design aspects.

The inclusion of gender and power content exerted a powerful effect on program outcomes. Among the 10 programs that addressed gender and power, eight (80%) led to significant decreases (Table 2) in at least one of the health outcomes (pregnancy, childbearing or STIs).^{84,86,89–91,94,100,105} In contrast, among the 12 programs that did not address gender and power, only two (17%) significantly reduced rates of pregnancy or STIs.^85,93

Other Possible Factors

Because study design characteristics and the setting of the intervention also may have led to a greater or lesser likelihood of detecting or leading to an impact, the question is whether the association between gender and power content and program efficacy still holds when considered in relation to these other characteristics.

As noted above, randomized controlled trials were far more likely to detect significant reductions in STIs, pregnancy or childbearing than were longitudinal cohort designs. If we look at the gender and power content of only the 15 programs evaluated with a randomized controlled trial design, 89% (8 out of 9) of the programs that addressed gender or power had a beneficial effect, compared with 33% (2 out of 6) of those that did not (Figure 1). Larger sample size also tended to help detect an effect. Yet if only the 16 evaluations with sample sizes of greater than 500 are considered, 86% (6 out of 7) of the programs with a gender and power component led to significant reductions in STIs or pregnancy, compared with 11% (1 out of 9) without such a component. A similar pattern is found for length of follow-up: Of the 17 studies that had a postintervention follow-up of one year or longer, 78% (7 out of 9) of the programs that addressed gender or power reduced adverse health outcomes, compared with 25% (2 out of 8) of those that did not. And, when only the 14 studies published since 2000 were considered, 88% (7 out of 8) of the programs that addressed gender or power were found to be effective in decreasing STIs or pregnancy, compared with 33% (2 out of 6) of programs that did not.

In terms of setting, only two of 10 school-based programs brought about a significant decrease in pregnancy, childbearing or STIs. These were also the only two school-based programs that addressed gender or power. Clinic-based programs were far more likely to reduce adverse health outcomes than programs implemented in other settings, with four out of five clinic-based programs proving effective. All three clinic-based programs that addressed gender or power had a positive effect, whereas only one of the other two programs had an effect.

Overall, gender or power content remains a consistently important characteristic of effective programs, even when other variables are considered.

Are These Results Due to Chance?

Another question that arises is whether the results of this gender analysis could reflect chance. Indeed, a critique²⁶ that has been made of lists of "programs that work" is that programs demonstrating just a single, positive effect—which may indicate that a result was achieved by chance—are typically classified as "evidence-based, effective" programs. To examine whether the findings regarding the importance of addressing gender and power may have been due to chance, the evaluations categorized as effective were examined for evidence of other positive effects, such as increased reports of condom use, decreased number of partners, improved self-efficacy and reductions in intimate partner violence. Among the eight programs that addressed gender and power and demonstrated significant decreases in pregnancy or STIs, all eight also found significant, independent beneficial effects on reported behavior, attitudes, or other desirable health or social outcomes. Most reported several additional positive effects. In contrast, of the two programs that did not address gender or power that found a significant effect on any health outcome (i.e., decreases in STIs, pregnancy, or STIs and pregnancy as a combined outcome), only one of them also had positive effects on reported behaviors, knowledge and self-efficacy;⁹³ the other program had no effect on any self-reported risk behavior (multiple partners, casual partners or inconsistent condom use).⁸⁵ Nine out of 10 programs that found reductions in pregnancies or STIs also demonstrated decreases in other risk factors, and it is thus highly unlikely that their effect on health outcomes was by chance. The one "effective" program that considered STIs and pregnancy as a combined outcome but did not show a decline in other risk behaviors did not address gender or power.

How Do Successful Programs Approach Gender and Power?

Finally, this review sought to identify the specific qualities of a gender and power program that may contribute to positive results. Despite the small number of programs, some common characteristics emerged. In addition to the interactive and learner-centered pedagogical approaches noted above, these elements included:

•Explicit attention to gender or power in relationships. This approach includes providing teachers with specific content, activities and vocabulary to explore gender stereotypes and power inequalities in intimate relationships. Some also provide explicit instructions for handling subtle, and not so subtle, sexual or homophobic harassment. A notable contrast is the SHARE program.¹⁰⁹ Although the authors of the study thoughtfully consider how power and gender norms relate to sexual behavior,^113,114 the curriculum itself does not provide explicit activities or tools for teachers to engage learners in these topics. The evaluation found no effect on pregnancy or on reported behaviors.

•Fostering critical thinking about how gender norms or power manifest and operate. Depending on the local context, this element may include critically examining and analyzing images of females in visual media and music,^89,90,100 harmful practices such as early marriage,⁹⁴ power disparities in relationships caused by economic or age differ- ences,^91,94 or how some of the differences in the ways males and females express their sexuality are the result of gender stereotypes.⁸⁴

•Fostering personal reflection. Participants are given opportunities to reflect on how the contextual factors of gender and power relate to their own life, sexual relationships or health. The Teen Outreach Program, for example, asks participants to think about how messages about gender affect their relationships, sexual and otherwise.⁸⁴ SIHLE, HORIZONS, Stepping Stones, the Children’s Aid Society– Carrera program and Project Safe explore how power operates in relationships and why it makes it difficult to protect one’s health.^{89,90,94,100,105} The ways programs foster such reflection varies. Some use personal writing exercises, another asks participants to think about their own current and past relationships while playing a game about relationship types and situations, and others provide short case studies and facilitate discussions about how power inequality and gendered sexual scripts influence condom use. Many programs also address sexual coercion and intimate partner violence.^84,89,90,94 In contrast, MEMA kwa Vijana’s multiyear curriculum, which did not decrease STIs or pregnancy, has a single session on gender but focuses on the equal abilities of males and females, rather than taking the next step to help participants reflect on how gender norms and stereotypes affect relationships, power, sexual and reproductive health, or HIV. Indeed, in a process evaluation of the program and reflection on barriers to and facilitators of change, the authors conclude that the intervention did not sufficiently address systemic social or structural factors, including gender.¹¹⁵

•Valuing oneself and recognizing one’s own power. Acknowledging one’s power to effect change in one’s own life, relationship or community is another consistently recurring theme in the successful gender and power programs. For example, the Teen Outreach Program engages participants in community service,⁸⁴ and the Children’s Aid Society–Carrera program is undergirded by a belief in participants’ "pure potential."¹⁰⁰ Some programs for females aim to foster gender pride.^89,90,105 Many emphasize young women’s power, strength, self-respect and agency.^{89–91,94,105} In settings where racialized social structures may affect one’s sense of self-worth, some programs interweave gender pride with ethnic or racial pride. For example, in the United States, SIHLE fosters young women’s pride in being black and female, and includes reading and analyzing poetry by famous black female writers.⁸⁹

These qualities—addressing gender and power explicitly, using participatory and learner-centered teaching approaches, fostering both critical thinking and personal reflection about how these concepts affect one’s own life and relationships, and valuing one’s own potential as an individual and as a change agent—overlap and reinforce each other, helping learners to apply the content to their own sexual and reproductive lives.

DISCUSSION

The nearly opposite outcomes of programs that address gender and power and programs that do not is striking. This finding is consistent with theory, as well as with the body of evidence that links gender, power and intimate partner violence with sexual and reproductive health outcomes, including HIV. It echoes the increasingly frequent call to address the multiple contextual factors that shape adolescent sexual behavior.^{30,66,67,116,117} Indeed, reviews of adolescent sexual risk reduction programs in South Africa by Harrison and colleagues looked beyond individual-level pathways and concluded that addressing contextual factors such as gender and poverty was important for success.²² Findings are also consistent with reviews of more diverse program types—i.e., reviews that included different kinds of interventions, not just those that were focused on adolescents and were group- and curriculum-based. These reviews have found that programs that address gender or power have positive effects on sexual and reproductive health—including knowledge, attitudes, reported behavior change and health outcomes.^118,119

A main limitation of this review, as with all reviews, is the possibility of missing eligible studies. While resources precluded perusal of additional databases such as Psyc-INFO and CINAHL, multiple databases were searched thoroughly and the references from 36 reviews and meta-analyses (most of which were themselves the product of broad searches) were hand searched. It is therefore likely that most, if not all, eligible studies were identified. Several other common limitations to systematic reviews were addressed: By setting a high bar for the measurement of effectiveness—the achievement of positive health outcomes—this review avoids the pitfall regarding the reliability of self-reported attitudes and behavior change. The drawback to ensuring a sample of studies with more reliable results, however, is that using a higher bar may in some way have biased the types of interventions included. Possible explanatory variables such as program duration, study design, pedagogical approach or inclusion of additional program components do not seem to be confounding the effects observed. Finally, the possibility of including chance findings was reduced by examining effects on the range of indicators that evaluations measured.

This review provides strong evidence that content on gender and power in intimate relationships should be considered a key characteristic of effective sexuality and HIV education. Many of the programs that addressed gender and power and significantly reduced pregnancy and STIs shared the following elements: They addressed gender and power explicitly, used participatory and learner-centered teaching approaches, facilitated critical thinking about gender and power in participants’ society, fostered personal reflection about how these concepts affect one’s own life and relationships, and helped participants value their own potential as individuals and as change agents.

Rigorous impact evaluations that compare models with and without a strong gender and power approach and a control would of course further add to this discussion. However, until such a body of work emerges, the theory, the antecedent evidence and the evaluation evidence reviewed here, all of which point to the import of gender and power, provide a powerful and persuasive argument for a shift toward programs that place attention on gender and power—or what may be termed an "empowerment" approach to sexuality education.¹²⁰

There are also implications for research. This review strongly concurs with the recommendations of others to use biological outcomes as a measure of efficacy when applicable and feasible, and to provide greater detail in reporting of interventions and control conditions.^{20,22,27,30–34} To advance further understanding of the potential impact for young people of sexuality and HIV education that addresses gender and power, a wider range of indicators must be included in evaluation studies. These include explanatory and outcome variables that measure gender attitudes, agency, power in relationships, critical thinking skills, intimate partner violence, advocacy or civic participation, school environment and safety, schooling outcomes, and school connectedness. Doing so will allow us to better understand the pathways through which an intervention operates, as well as expand the vision for what sexuality and HIV education can help achieve. Specifically, it can increase the chances that young people will have relationships characterized by equality, respect and nonviolence, and incur the benefits that such characteristics bring for a host of other outcomes.