Latino adolescents are disproportionately affected by teenage pregnancy, sexually transmitted infections, and HIV, persistent sexual and reproductive health (SRH) disparities that represent a national public health concern. Despite progress nationally, Latina adolescents continue to exhibit above-average teenage pregnancy, birth, and repeat birth rates.1 Particularly concerning are the 17% increase in reportable sexually transmitted infections among 10- to 19-year-old Latino adolescents since 2014 and the 6% rise in new HIV diagnoses among 13- to 19-year-old Latino adolescents between 2016 and 2017 alone.2–4 Given these statistics, research is needed to strengthen the evidence on programs to reduce Latino adolescent SRH disparities.
In this issue of Pediatrics, Evans et al5 systematically reviewed and meta-analyzed the empirical evidence regarding the effectiveness of SRH interventions for Latino adolescents. Their findings suggested small to moderate but significant mean pooled intervention effects across all examined behavioral outcomes (abstinence, condom use, and number of sex partners). Building on this evidence, strengthening interventions for Latino adolescents is a priority. Important considerations for the interpretation of meta-analytic review evidence, including clinical, methodologic, and statistical heterogeneity,6 may yield additional insights into factors associated with intervention effectiveness.7,8 We examine these 3 types of heterogeneity in relation to Latino SRH interventions and draw implications for strengthening future interventions.
Intervention meta-analyses estimate weighted mean effect size across studies on a selected outcome. However, studies’ differing characteristics, such as study population or intervention content and delivery, may influence intervention effectiveness. This variability is defined as clinical heterogeneity.9 In interpreting meta-analyses’ results, consideration of clinical heterogeneity is important because variability in individual effects due to differences in clinical characteristics is not reflected in pooled effect sizes. The search strategy and inclusion criteria defined by Evans et al5 sought to minimize clinical heterogeneity, but residual clinical heterogeneity is inevitable in meta-analyses and often cannot be adjusted for.9 To illustrate, Evans et al5 highlighted that reporting on the foreign-born participant proportions was incomplete across studies, thereby excluding this clinical heterogeneity domain from formal moderation analyses. Therefore, developers of Latino SRH interventions should interpret meta-analysis findings with attention to clinical heterogeneity in domains that may moderate intervention effectiveness.10
Studies included in meta-analyses frequently vary in methodologic quality (eg, underlying theories and cultural relevance, measures of reliability and validity, attrition, etc). This variability, defined as methodologic heterogeneity, can result in differing degrees of validity across included studies and affect the strength of meta-analytical findings.11 To assess the impact of methodologic heterogeneity on their findings, Evans et al5 conducted sensitivity analyses with attention to several potential sources of bias, including random sequence generation, overall attrition, and selective reporting. However, other domains of potential methodologic heterogeneity, such as refusal bias, differential attrition, or information bias, remained unaccounted for. For example, of studies with response data, all 5 that observed significant effects for condom use reported refusal rates <20%, whereas the remainder that did not find significant effects reported average refusal rates close to 60%, illustrating the importance of considering the methodologically best evidence versus all evidence.12
In meta-analyses, statistical heterogeneity pertains to variability in observed effect sizes across evaluated studies. Statistical heterogeneity can arise from chance, clinical factors, methodologic factors, or a combination thereof.13 Common measures of statistical heterogeneity include the Q and the I2 statistics: Q reflects the statistical significance of heterogeneity, and I2 yields the percentage of variability in effects that cannot be explained by chance alone.14 Variability in observed effects beyond that expected by chance can be attributed to clinical or methodologic heterogeneity. Evans et al5 formally tested for statistical heterogeneity using the Q and I2 statistics, observing no significant statistical heterogeneity for abstinence and number of partners but reporting significant statistical heterogeneity for condom use. However, the Q statistic weakly detects true heterogeneity when sample sizes are small15 and may have been underpowered to detect meaningful statistical heterogeneity for abstinence and number of partners. Interestingly, alternative methods for evaluating statistical heterogeneity are emerging in the literature.15,16 Notably, for condom use, Evans et al5 formally tested for moderation by multiple clinical and methodologic factors and identified one important moderator of effect sizes: cultural tailoring of interventions.
Persistent SRH disparities among Latino adolescents remain a national public health priority. Evans et al5 meta-analyzed interventions for Latino adolescents, finding significant pooled overall effects for behavioral outcomes, including abstinence, condom use, and number of sex partners. Nevertheless, strengthening the effects of future SRH interventions for Latino adolescents is needed. Meta-analyses are integral to strengthening intervention science. Attention to clinical, methodologic, and statistical heterogeneity across studies can yield insights into factors associated with bolstering intervention effectiveness. Cultural tailoring to increase the effectiveness of condom interventions for Latino adolescents is one such intervention effect modifier. Future research to strengthen the understanding of other clinical and methodologic factors enhancing the effectiveness of Latino SRH interventions is warranted.
Opinions expressed in these commentaries are those of the authors and not necessarily those of the American Academy of Pediatrics or its Committees.
FUNDING: Supported by the William T Grant Foundation (grant 189030) and the National Institutes of Health (NIH) through funding for the Center for Drug Use and HIV Research (P30DA011041). The funding bodies did not influence the content or opinions expressed in this article. Funded by the National Institutes of Health (NIH).
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2019-3572.
sexual and reproductive health
POTENTIAL CONFLICT OF INTEREST: Dr Guilamo-Ramos discloses the receipt of grants and personal fees from ViiV Healthcare outside the submitted work. He serves as a member of the US Presidential Advisory Council on HIV/AIDS and as the Vice Chair of the Board of Directors of the Latino Commission on AIDS. Mr Hidalgo and Dr Keene have indicated they have no potential conflicts of interest to disclose.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.