A systematic review and meta-analysis of suicidality in autistic and possibly autistic people without co-occurring intellectual disability.

The studies included people without intellectual disability only, so these figures say nothing about autistic people with a co-occurring intellectual disability, who were deliberately excluded because the self-report measures used are harder to apply to them. The individual studies disagreed enormously with each other (I² around 96%), and when the authors worked out the range in which a new study's true prevalence would fall, it was very wide: for suicidal ideation it ran from below zero (the lower bound was −3.3%) to 72.7%. The tools for measuring suicidality varied and most were not designed or validated for autistic people. Studies generally did not separate passive thoughts of being better off dead from active thoughts of killing oneself, and timeframes ranged from the past week to lifetime, so the pooled figures carry real uncertainty. Only 4 of the 37 studies assessed were at low risk of bias; 28 were unclear and 5 high, mostly because sample sizes were not justified and participation rates were not reported, so the pooled estimate rests on studies of limited quality. Nearly all the pooled studies were run in high-income countries (91.7%), yet about 75% of suicides happen in low- and middle-income countries, so these rates may not travel. The review required studies to distinguish suicide attempts from self-harm without suicidal intent, which is defensible but means some relevant literature may have been missed. Agreement between the two screeners at title and abstract stage was only fair, so some eligible studies may have been screened out early.

The only funding statement in the text is the Economic and Social Research Council, a UK public research funder. No industry funding and no individual author competing interests are reported in the supplied text.

Do not treat 34.2% as a precise figure to quote to a patient or a service. The studies disagreed so much that the range in which a new study's true prevalence would fall ran from below zero to 72.7% for suicidal ideation, and only 4 of 37 studies were rated at low risk of bias. The subgroup differences, such as lower rates in Asia, are associations between studies rather than proof that those factors cause the difference, and the authors themselves suggest under-reporting in some countries may explain the pattern rather than genuinely lower risk. This paper is also not evidence that any screening or prevention programme works.