We conducted a systematic review of the past five decades of published data on cognitive and behavioural outcomes of individuals with an extra sex chromosome. The data were obtained primarily through published articles, although authors were contacted for clarification where appropriate. The online Web of Knowledge database was used to identify English-language articles, using the search terms (XXX OR XXY OR XYY OR sex chromosome (trisomy OR anomaly OR abnormality OR aneuploidy) OR Klinefelter) AND (cognit* OR language OR behav* OR brain OR neuro*). The search was confined to articles in databases for genetics, neuroscience, behavioural science, and psychiatry. Further articles were traced through PubMed and from reference lists in identified papers.
Because letter combinations such as XXX are used with various meanings in other disciplines, the search criteria generated many irrelevant papers, but these were readily identified, leaving 702 papers with a focus on chromosome abnormalities. As shown in , we excluded articles describing animal studies, those dealing with conditions other than SCTs, and those with no cognitive or neurodevelopmental content, leaving a pool of 424 publications to evaluate. Of these, 113 did not include new empirical data, and 153 were case reports.
Procedure for selection of articles on sex chromosome trisomies (SCTs) for inclusion in the review.
As we aimed to characterize the phenotype in as unbiased a way as possible, we next excluded articles with major ascertainment bias (n=64), that is, studies of individuals recruited through institutions for cognitive or psychiatric disorders, as well as those who were identified via patient support groups or through postnatal chromosome testing. Our rationale was that it was not usually possible to tell why a chromosome test had been conducted, or why an individual had joined a support group. Thus, we erred on the side of caution and excluded those who may have come to attention because of cognitive or behavioural difficulties. However, we did not exclude males with XXY who were identified as adults because of infertility or endocrinological concerns, because there is no reason to suppose that they were cognitively atypical. Studies of a sample of adults identified via population chromosome screening of tall males were also included.
Having identified 94 studies without major ascertainment bias, we next excluded studies that did not meet the following methodological inclusion criteria: (1) power of at least 0.69 to detect an effect size (Cohen’s d) of at least 1.0 on a one-tailed test, computed using G*Power 3;8
for categorical data, this effect size corresponded to an odds ratio of 6.11;9
and (2) sufficient information provided to evaluate power and other key methodological points. To avoid excluding most studies, the power criterion had to be set lower than is typically used; in effect it meant that studies were excluded unless the total sample (one SCT group plus any comparison groups) was at least 20. Where no comparison group of typically developing individuals had been assessed, the study was included if the sample size for a given karyotype was >10 and if well standardized measures had been used, allowing an estimation of performance relative to a comparable normative sample. In addition, to boost information on XYY cases from an unbiased source, two studies including more than 12 prenatally ascertained cases of at least one karyotype were included,5,10
despite lack of normal comparison groups or standardized measures.
Overall, 46 articles met our criteria (see Tables SI
published online), but these were reports of data from only 12 independent samples. Most samples came from state- or countrywide screens of newborn infants initiated in the 1950s to 1960s, of which five – based in Boston and Denver in the USA, Edinburgh in the UK (two surveys), and Toronto in Canada – had sample sizes for at least one SCT that met our power criterion. Studies are grouped by cohort in Table SI
. Of the 46 articles, 35 are descriptions of data from cases identified neonatally. The remaining articles comprise (1) four reports on a group of males with XXY and XYY identified in Copenhagen, Denmark, by screening of adult males from the general population who were 184cm tall or more; (2) five studies from four cohorts of males with XXY that met our power criterion where, although the diagnosis was made postnatally, this was prompted by concerns about physical development or fertility in adulthood, rather than cognitive or behavioural difficulties; and (3) the two aforementioned uncontrolled studies of children identified on prenatal screening. Where it appeared that there was duplication or overlap in data reported in different papers, only the publication with the largest sample is considered.
When reviewing results from this literature, a p value of <0.05 was regarded as statistically significant. Where relevant, we report effect sizes, to provide a standardized measure of difference between two groups. This can be more informative than significance levels, especially when considering the clinical importance of deficits.