Dorien Bangma

68 | CHAPTER 4 or the Measuring Independent Living in the Elderly Study (MILES; Lassen-Greene et al., 2017; Okonkwo et al., 2009, 2006). However, despite the same study cohort has been used, there is insufficient evidence that the same sample has been used in these studies and, therefore, these studies are treated and described as separate studies in the content and meta-analysis. Two studies focused on people living with PD (Pirogovsky et al., 2013, 2014). However, these studies did use the same sample and are, therefore, considered and described as one study within the present systematic review. Of longitudinal studies, baseline data (if applicable) are used for cross-sectional comparisons. However, one research group performed a follow-up at one-year (Martin, Griffith, et al., 2008; Triebel et al., 2009), two-years (Clark et al., 2014; Gerstenecker et al., 2016) and six-years (Martin et al., 2019) of the same sample and, therefore, only the baseline data of people living with AD and people living with MCI as described in Martin et al. (2019) was used to evaluate cross-sectional data. In the content analysis, group differences were considered significant when alpha < .05, independent of the used alpha level in the original study. Two studies, however, did not describe the p-values of their group comparisons and used a more conservative alpha level to evaluate significance (Marson et al., 2000; Tracy et al., 2017). Potential significant results using a less conservative p-value could, therefore, not be determined for these studies and included in this systematic review. Correlations were interpreted as weak when r = .30, moderate when r = .50 and strong when r = .70 (Fritz et al., 2012). Meta-analysis. In addition, a meta-analysis was performed for each NDD group separately when more than one of the included studies focused on a particular group. In the meta-analyses, the total scores (i.e., sum of domains or items) of the FDM tests were used. Therefore, studies that did not report total scores (Arcara et al., 2019; Czaja et al., 2017; Gerstenecker et al., 2018; Giannouli & Tsolaki, 2014; Marson et al., 2000; Okonkwo et al., 2006, 2009) and studies that did not provide means and/or standard deviations for total scores (Mahurin et al., 1991; Marson et al., 2009; Sheppard et al., 2017) were excluded from the meta- analysis. Furthermore, two studies did not include a healthy control group (Gerstenecker et al., 2016; Loewenstein et al., 1995) and were, therefore, not used in the meta-analysis. Regarding longitudinal studies, only the baseline results were used for the meta-analysis. Furthermore, only the most recent published study was used in the meta-analysis of the two PD studies that presumably used the same sample (Pirogovsky et al., 2013, 2014). In total, thirty-one studies were included in the meta-analysis (Table 4.1; Figure 4.1). When studies included more than one subgroup (e.g., MCI converters and MCI non-converters) and/or used more than one FDM test or total score, a combined fixed effect size was calculated (Borenstein et al., 2009). Using Comprehensive Meta-Analysis software version 2.2.064 the pooled mean effect size (method: random) was calculated for each group. Hedges’ g was considered small when g = .20, medium when g = .50 and large when g = .80 (Fritz et al., 2012). Funnel plot asymmetry was tested with a regression method to evaluate publication bias (Egger et al., 1997). Finally, heterogeneity was evaluated by calculating the I 2 value. I 2 values of 25%, 50% and 75% were interpreted as low, moderate and high heterogeneity, respectively (Higgins et al., 2003).