In Reply We appreciate the thoughtful critique made by Yuan and Niuregarding our meta-analysis¹ on parent-based sexual health interventions. They call attention to 3 types of heterogeneity (clinical, methodologic, and statistical), and they suggest we focused exclusively on statistical heterogeneity. Clinical heterogeneity refers to variability in clinical characteristics (eg, participant characteristics, types of interventions, and outcomes); methodologic heterogeneity refers to variability in trial design or quality; and statistical heterogeneity refers to variability in observed treatment effects.² We agree these forms of heterogeneity are important to consider, and we have in fact considered all 3 throughout the planning and evaluation of our study. We first considered clinical and methodologic heterogeneity in determining the scope of our meta-analysis and study selection criteria. We limited the scope to peer-reviewed randomized clinical trials of parent-based sexual health interventions that reported behavioral outcomes among US adolescents. Yet even within this sample restriction, we recognized there remained important clinical and methodologic heterogeneity and therefore reported study-level data on adolescent characteristics, parent characteristics, intervention design, and study settings in the study characteristics table (Table 1).¹ We also comment on issues related to heterogeneity (particularly in the assessment of outcomes) in the Discussion section. In analyses, we considered the importance of clinical heterogeneity in participant and intervention characteristics and found adolescent age, sex, and race/ethnicity, as well as intervention design and parent involvement, moderated the effects of interventions on condom use. It is true that we did not run statistical comparisons of potential moderators on outcomes where there was not statistical heterogeneity. There is not a clear consensus regarding the best approach to handling clinical/methodologic heterogeneity in the absence of significant statistical heterogeneity.² As Gagnier et al² recommend, we determined our approach to reporting and examining moderators a priori, and we agree with their assertion that there remains “a need to develop a comprehensive set of recommendations for how to perform valid, applicable, and appropriate investigations of clinical covariates.”^2(p12)

在答复中，我们赞赏Yuan和Niuregarding对基于父母的性健康干预措施进行的荟萃分析^1所进行的深思熟虑的批评。他们呼吁关注3种类型的异质性（临床，方法和统计），并且他们建议我们只关注统计异质性。临床异质性是指临床特征（例如参与者特征，干预类型和结果）的变异性；方法学异质性是指试验设计或质量的差异；统计异质性是指观察到的治疗效果的差异。²我们同意必须考虑这些形式的异质性，并且实际上我们在研究的整个计划和评估过程中都考虑了这3种情况。在确定我们的荟萃分析和研究选择标准的范围时，我们首先考虑了临床和方法学的异质性。我们的范围仅限于同行评审的基于父母的性健康干预措施的随机临床试验，该试验报告了美国青少年的行为结局。即使在此样本限制范围内，我们也认识到仍然存在重要的临床和方法异质性，因此在研究特征表中报告了有关青少年特征，父母特征，干预设计和研究设置的研究水平数据（表1）。^1个我们还将在“讨论”部分中评论与异质性有关的问题（尤其是在结果评估中）。在分析中，我们考虑了临床异质性在参与者和干预特征中的重要性，并发现青少年的年龄，性别，种族/民族以及干预设计和父母参与，减轻了干预对避孕套使用的影响。的确，我们没有对没有统计学异质性的结果进行潜在主持人的统计比较。在缺乏显着的统计异质性的情况下，关于处理临床/方法异质性的最佳方法尚无明确共识。²正如Gagnier等人²建议，我们确定了先验地报告和检查主持人的方法，并且我们同意他们的主张，即“仍然需要针对如何进行有效，适用和适当的临床协变量研究制定一套全面的建议。” ^{2 （p12）}