TY - JOUR T1 - Sodium valproate: who knew what and when? Cumulative meta-analysis gives extra insights JF - BMJ Evidence-Based Medicine JO - BMJ EBM SP - 127 LP - 129 DO - 10.1136/bmjebm-2018-111068 VL - 24 IS - 4 AU - Carl Heneghan AU - Jeffrey K Aronson Y1 - 2019/08/01 UR - http://ebm.bmj.com/content/24/4/127.abstract N2 - Sodium valproate is licensed in the EU for treating generalised, partial or other forms of epilepsy. It has also been used to treat bipolar disorder and to prevent migraine. In February of this year, the European Medicines Agency recommended that sodium valproate should not be used during pregnancy unless no other effective treatment is available, and that it must not be used in women able to have children, unless the conditions of a pregnancy prevention programme are met.1 These measures to protect women and their children are welcome, but we argue that they should have been instituted several years ago, as the evidence was clear as far back as 1990 that there were risks of congenital malformations in women exposed to valproate.In 1992, Antman and colleagues used cumulative meta-analysis to show that expert recommendations often lagged behind pooled estimates of clinical trials.2 In a cumulative meta-analysis, the combined result is summarised by adding each new clinical trial result to the previous. This can be a useful technique for assessing both benefits and harms, allowing interpretation of what was known and when it emerged.In 2015, Tanoshima et al performed a systematic review of 59 cohort studies published between 1983 and 2014, using cumulative meta-analysis. We used their analysis (see figure 1 to interpret signals of valproate-associated major congenital malformations and to determine when they emerged.3 In 1990, Hunter and Allen published an assessment of the risk of congenital malformations among … ER -