SEXUAL ORIENTATION AND ADOLESCENT SUBSTANCE USE: QUESTIONABLE CORRELATIONS

Abstract

In April's issue of Addiction, Marshal et al.[1] report that the average odds of substance use among lesbian, gay and bisexual youth were 190% higher than for heterosexual youth. Their meta-analysis included 19 studies, of which 13 were based upon general population samples. Consequently, rates of bi- or homosexuals were generally very low (i.e. <5%). Marshal et al. discuss various limitations of the studies in their analysis, and point out that the majority of them did not have large enough cell sizes to explore differences between subgroups of bi- or homosexual youth. However, low prevalence rates may represent a more serious methodological limitation than low absolute numbers, but this important issue is not addressed by the researchers. Under-reporting is often assumed to be a major problem in measuring behaviours that are socially defined as undesirable, but this may not necessarily be the case [2,3]. Over-reporting may, in fact, be a greater threat to the validity of survey data on such phenomena, given that their true prevalence is low. Same-sex sexual orientation is a case in point: because the vast majority of adolescents are most probably heterosexual many of them will, by definition, have the possibility to report falsely that they are bi- or homosexual. Even a low proportion of such response errors can have a significant effect, simply because the group of heterosexuals is so large. Due to over-reporting, one may thus assume that a sizeable fraction of those who report same-sex sexual orientation (or other uncommon feelings or behaviours) are, in fact, false positives. As a consequence, rates may be inflated and observed correlates may deviate from the true pattern of correlates. When analysing correlations between various uncommon phenomena, the problems may accumulate. Marshal et al.[1] reported that the effect sizes of same-sex sexual orientation were generally larger for rare forms of substance use than for drinking and marijuana use, but it goes without saying that this finding remains questionable. In relation to this, it should be noted that some respondents may present themselves intentionally in a negative way [3,4], and that ‘faking bad’ is likely to be more prevalent in adolescent than in adult samples [3]. Over-reporting has been approached by asking the respondents whether they have used a fictitious drug. Adolescents with false positive answers to such a question were excluded from the analyses in a study by Pape & Storvoll [3], and as a consequence rates and socio-demographic correlates of reported use drugs such as heroin and LSD changed markedly. The authors thus concluded that ‘researchers should consider carefully whether it is defensible to analyze survey data in order to obtain knowledge about rare behaviors’ (p. 55). Correspondingly, Skog [2] pointed out that one should be very cautious when interpreting survey results on uncommon phenomena. However, many researchers appear to be unaware of the fact that false positives may seriously deteriorate the validity of measures on such phenomena. Marshal et al.'s [1] meta-analysis, as well as many of the studies on which it is based, seem to illustrate the point. None.