Anne Fleur Kortekaas-Rijlaarsdam

33 MPH AND ACADEMIC PERFORMANCE: META-ANALYSIS 2 differences in comorbidity), all meta-analytic effect sizes were calculated using a random effects model. The I 2 statistic was used to assess heterogeneity of effect sizes, where values of 25%, 50% and 75% indicate low, moderate and high heterogeneity, respectively (Higgins, Thompson, Deeks, & Altman, 2003). Rosenthal’s fail-safe n was calculated to determine the number of studies with a null effect necessary to cancel out significant effect sizes, where fail-safe n values >5 k + 10 were considered robust and k refers to the number of samples on which the relevant effect size was calculated (Rosenthal, 1995). Further, Egger funnel plot asymmetry was used to assess publication bias (Egger, Smith, Schneider, &Minder, 1997). Associations between effect size and sample size were investigated to assess the possibility that studies with small samples and large effect sizes were more easily published than studies reporting non-significant findings. All tests of significance were two-sided with α = .05. Risk of bias was estimated for each study based on Cochrane guidelines (Higgins & Green, 2011). R E SU LT S A combined total of 1777 children from 34 different studies was included in the meta- analyses. Another 425 children from seven studies were included in the qualitative synthesis because results were either reported in figures only or exact values were not reported, including six studies on math performance and three studies on spelling accuracy. Table E2.1 provides an overview of which studies qualified for meta-analysis and gives a narrative description of the results of those studies that did not qualify. Meta-analyses were conducted for math accuracy (29 studies, N=1528) and math productivity (17 studies, N=912). For reading, meta-analyses were conducted for reading accuracy (nine studies, N=207) and number of items attempted (five studies, N=100). Most studies (88.2%) used a placebo-controlled crossover design. Four studies (11.8%) used a between-subject design. In 73.5% of the studies, medication dosage was clinically titrated on symptom improvement before start of the trial. In the other 26.5% of the studies, dosages were fixed. When multiple dosages were used in randomized order, we included results from the dosage showing greatest effects on academic outcomes to optimize MPH-efficacy (please see Table E2.1 for details). While all studies predominantly involved primary school children, one study also included children from middle school (aged 12-16, 16.5%).