Relationships between neuropsychological variables

The relationships between neuropsychological domains are complex and an aim of this thesis was to investigate these relationships, if possible from the literature. A clear pattern of three outcome groups emerged across outcome domains: there was a group who remained stable (the largest proportion); a group who improved; and a group who deteriorated. This common pattern raises the possibility that the same young people would have fallen into each of these outcome groups across outcome domains, i.e. those who deteriorated in language may also have deteriorated in IQ and mood. Unfortunately, the review was unable to investigate this question, as individual level results were generally presented separately for each outcome within papers, without reference to

participant characteristics, and many of the studies only reported one outcome domain.

Furthermore, within-subject comparisons to investigate the relationships between outcome domains were rarely performed by studies. Additionally, study authors did not always adequately consider other causal explanations for their findings other than the effect of resection. For example, most studies that reported psychological wellbeing outcomes referenced a biological account of mental health outcome after TLE surgery, due to the disruption of mesiotemporal networks. However, any paediatric surgery can be a traumatic event in the life of a child which may have long-term effects (Lerwick, 2013), so studies comparing outcomes of temporal lobe epilepsy surgery with outcomes

134 of other surgeries are warranted. In sum, further appropriately designed studies are required to find out the effects of epilepsy on children’s lives and the mechanisms for these effects, so that young people, families and clinicians can make a fully informed decision about the risks and benefits of proceeding with surgery. The logic model in Figure 4.1 is presented as a tentative framework for further exploration of the surgical, child and outcome variables that may shape long-term outcomes to surgery, so that future research can explore these relationships. Logic models present

relationships between inputs and outputs in a systematic and visual way (Kellogg WK Foundation, 2004) and they have been used in a number of health-related systematic reviews (Anderson et al., 2011). Logic models can be useful for planning analyses (Anderson et al., 2011) and the model in Figure 4.1 may be helpful for shaping further investigations of the relationship between variables. Figure 4.1 shows variables that were investigated in the studies, displayed in bold text, and also variables that are recommended as a focus for further research, not in bold.

Ultimately, this review was unable to construct an account in the narrative synthesis of how the intervention works and for whom, in terms of the neuropsychological outcomes of TLE surgery, which is the stated aim of narrative synthesis according to the Cochrane collaboration (Ryan, 2013). However, its findings on the methodological and reporting quality were perhaps more significant. The recommendations arising from these findings are outlined in Section 4.4.

135 Figure 4.1 Logic model for understanding potential relationships between outcomes of paediatric epilepsy surgery, for further investigation.

136 4.2 Are limitations in reporting and study design biasing conclusions drawn about the efficacy of temporal lobe epilepsy surgery in children?

As discussed in Section 3.3, all of the studies reporting neuropsychological outcomes of temporal lobe epilepsy surgery in children carried significant risk of bias. Some bias may have been caused by features of the study design, some by the quality of reporting. Key limitations in study design were as follows:

 Nearly all studies were retrospective, which may have biased results as data were not collected at standardised time points, and because the sample may be biased because only those who receive follow-up are included, and data for those lost-to-follow-up may be missing.

 Study methods did not adequately control for other variables that may explain intervention effects. As Hermann et al. (2002) and Strauss et al. (2000) note, studies measuring

cognitive outcome of paediatric epilepsy surgery rarely include a control group of age- matched children with comparable epilepsies, who undergo neuropsychological assessment at the same time-points as the surgical group. Therefore, it is not possible to disentangle the effects of surgery from expected cognitive development during that time without surgery. This review, like others (Sherman et al., 2011; Tellez-Zenteno et al., 2007), found that these types of medical control group are rare, so methods were unable to distinguish outcomes of surgery from outcomes of AEDs as usual.

 The results of participants with a range of resection types, ages, seizure durations and follow-up durations, so the outcomes associated with each of these characteristics were hidden.

 For some outcome domains, there was a lack of pre-surgical assessment, so the reported outcome data may have been due to pre-existing differences between young people.  Studies generally lacked enough participants for the sample to be considered representative

or for statistical tests to yield meaningful results.

 Follow-up durations were inappropriately short for the variables being measured, such as intellectual functioning and psychological functioning, which are likely to change over the long term as the young person develops.

It should be noted that the conclusions of simple uncontrolled case studies are not necessarily biased, if they do not make claims of generalisability or causation, and if they have a well-defined cohort and assessment procedures and analyses used are adequately reported, as defined by

137 STROBE (Appendix D). However, the studies were not concordant with STROBE and key

limitations in reporting quality were as follows:

 Individual participant data was not presented, so it was not possible to disaggregate

outcome information for participants with different resections, ages, or other characteristics. For some studies this included reporting results for adults and children combined together.  Key demographic variables, such as age and side of surgery, were not reported, meaning

that the results of the participants cannot be generalised.

 The nature of the intervention or assessments undertaken were not adequately reported.  A thorough history of other interventions, including AEDs was not presented. For example,

many children remain on antiepileptic drugs after surgery and it is difficult to disentangle whether outcomes are the result of surgical intervention, medication, or even maturation of the child’s brain with the passage of time (Schmidt, Baumgartner, & Loscher, 2004). As other reviews have noted (Tellez-Zenteno, Dhar, Hernandez-Ronquillo, & Wiebe, 2007) AED use is sometimes inadequately reported in paediatric epilepsy surgery literature; authors may neglect to mention how many AEDs were tried, and for how long, before surgical candidacy of participants was assessed, and whether participants remain on AEDs after surgery.

Therefore, the review was able to answer this research question, with the finding that limitations in reporting and study design are likely to bias conclusions drawn about the efficacy of temporal lobe surgery for neuropsychological functioning in children.

Despite these limitations to methodological and reporting quality of the studies included in this review, centres should be commended for publishing their studies and it should be appreciated that not all studies had the expressed aim of assessing the efficacy of temporal lobe epilepsy surgery, rather most study aims were simply to present the findings from their recent case series of epilepsy surgery patients. It is positive that surgical centres communicate their outcomes, for the purpose of quality assurance, service improvement and to highlight training needs. It should also be noted that many surgical centres were not represented in the included studies, suggesting that many centres do not routinely publish their outcomes, as noted by Shastin et al. (2015), or that their published outcomes did not meet the criteria of this review, perhaps not presenting results for temporal surgery separately, or not presenting neuropsychological outcome. The reporting of routine outcome data represents a valuable opportunity for practice-based evidence to complement the results of trials, which have strong internal validity but may lack external validity (Barkham et al., 2010). This fusion of evidence-based medicine and practice-based evidence approaches allows the

138 best available evidence to be collected from rigorous research conducted in routine clinical settings (Barkham & Margison, 2007).

However, the conclusions of the included published case series often overstated their findings, and implied a causal link between outcome at follow-up and surgery, which could not safely be inferred from their results due to the method used. Although most studies documented their limitations and many used tentative language when drawing their conclusions, these papers may then be cited as supporting evidence by further studies, without reference to their methodological limitations (e.g. Spencer et al., 2008), and used to support stronger conclusions. This may contribute to the emergence of a consensus that has a spurious certainty that is not based on evidence at the ground level of research.

The review raised epistemological questions about the nature of what is considered as evidence. Despite many of the studies having high risk of bias, it should be noted that they all passed through the peer review process, and, as shown in Section 3.3.7, the average number of identified biases per paper from each journal did not correlate with journal’s impact factor. This finding is important as it casts doubt upon the assumption made by some clinicians that impact factor can be seen as a

measure of journal quality (Saha et al., 2003), at least for the literature concerning neuropsychological outcomes of paediatric epilepsy surgery. The implications of the methodological limitations in this literature may not be fully appreciated and may not be

communicated to the family when they are deciding whether or not to proceed with temporal lobe surgery. Clinical implications and recommendations for study design, reporting and new ways to develop high quality evidence are presented in Section 4.4.