29 Cognitive Impact of SRS vs. WBRT: Systematic Review & Meta-analysis Data extraction and analysis The follow-up time points were converted to units of ‘months after radiotherapy’. To aid comparability across studies and following the classification used in previous studies, time points were clustered: short-term follow-up 1 to 4 months after radiotherapy, mid-term follow-up 5 to 8 months after radiotherapy, and longterm follow-up 9 to 15 months after radiotherapy. Baseline measurements always refer to the assessment before start of radiotherapy treatment. Additionally, neuropsychological tests were attributed to cognitive constructs in a data-driven classification, based on the subdivision as reported in the majority of the included studies (Supplementary Materials). Data was collected from text, tables and figures from the manuscripts and then tabulated. Missing data points were excluded from analyses and changes in sample size due to attrition were considered. For metaanalysis of the incidence of cognitive decline compared to baseline performance, we used the inverse variance method in a DerSimonian-Laird random effects model. For individual studies Clopper-Pearson confidence intervals were calculated. Heterogeneity between studies was assessed using Cochran’s Q test and the I2 statistic. Statistical analyses were performed with R 3.5.1 open-source software with the ‘meta’ package (http://www.R-project.org). Data quality A critical appraisal of the included studies was performed to assess data quality as reported in the manuscripts, for which a checklist consisting of 7 criteria was constructed (shown in Table 1). One point was awarded if the criterion was met and zero points if not, or if it was unclear based on the available information. A maximum score of 7 points could be obtained. A score between 5 and 7 indicates good to high quality, 3 and 4 medium quality and scores below 2 indicate low quality. RESULTS Study inclusion The initial search yielded 867 unique manuscripts. After applying the in- and exclusion criteria, 20 manuscripts reporting on 14 original datasets were included in this review (Figure 1). The majority of these studies were rated a good to high quality (Table 2). The one study rated as low quality was excluded from further analysis.24 Study and baseline patient characteristics and the main conclusions of the selected papers are shown in Table 3 and Table 4 respectively. Patient numbers varied considerably across studies with a median sample size of 81 (range: 20-208) and 35 (range: 7-111) at baseline for the WBRT and SRS studies, respectively. In total, 751 WBRT patients were included and 300 SRS patients. Since data on the incidence 2
RkJQdWJsaXNoZXIy MTk4NDMw