LAUSR

With great interest we have read the meta-analysis by Ng et al. [1] on ‘Effects of surgery on neurocognitive function in patients with glioma’ which was recently published in this journal. The authors aimed to identify the tests that are commonly used for neuropsychological assessment in glioma patients. Furthermore, they aimed to quantify changes in cognitive function in glioma patients on the short and long term after surgery. Based on 11 studies that were included in the meta-analysis, it was concluded that surgery in glioma patients confers a benefit for performance on all cognitive domains that were assessed, except for executive functions which deteriorated after surgery. In addition, the authors underscore the need for more prospective and longitudinal studies assessing the effects of surgery on cognitive functioning in patients with glioma, and recommend using a standardized neuropsychological assessment battery in doing so. Although this meta-analysis draws attention to the important fact that surgery in glioma patients may be accompanied by changes in cognitive functioning, a few concerns are at issue. The authors chose to exclude studies that reported neuropsychological data as z scores only. However, we are already familiar with at least nine studies published in recent years (but before the publishing date of the meta-analysis) that described cognitive changes in this patient group, yet, reported results in terms of z scores and were therefore excluded, e.g., [2]. Instead of excluding studies, the authors could have summed these standardized (z) scores across studies weighting them appropriately using the sample sizes, which is a fairly common approach in meta-analysis [3]. This would have provided a more comprehensive and reliable overview regarding cognitive changes in glioma patients. Moreover, the excluded studies reported at best no significant declines after surgery, instead of a ‘beneficial effect of surgery on cognitive functioning’ as was concluded from the meta-analysis. This conclusion may cause a serious underestimation of cognitive dysfunction in patients after glioma surgery by clinicians. In addition, it is not only important to address which neuropsychological tests are used to assess cognitive functioning (as the authors indeed do), but it also an absolute prerequisite to address how changes in performance on these tests are defined and presented in the included studies. When considering changes in cognitive performance over time, very common phenomena related to repeated neuropsychological assessment should be taken into account, such as improved performance due to practice effects, that might result in overestimations of improvement, or in underestimations of decline. In the meta-analysis, details should have been presented on how changes in cognitive performance over different studies have been evaluated—as this varied widely over the included studies, from using raw difference scores (subtracting preoperative from postoperative scores) to taking into account practice effects and imperfect test–retest reliabilities (e.g., by the use of reliable change indices). As the authors did not take these fundamental neuropsychological issues into account this has most likely resulted in an underestimation of cognitive dysfunction in glioma patients undergoing surgery in the meta-analysis. Given the different approaches that are taken by studies to determine on changes in cognitive functioning, we understand the complexity of this with regard to the statistical procedures, however, these caveats should at least have been given full attention in the discussion section. * Sophie J. M. Rijnen [email protected]