Defining meaningful postoperative change in epilepsy surgery patients: Measuring the unmeasurable?

Abstract

Two hundred and ninety epilepsy surgery patients completed the Hospital Anxiety and Depression Scale and were assessed on a list learning task preoperatively and 1 year postoperatively. Deterioration and improvement in verbal memory were determined using reliable change indices (RCIs) at 80 and 90% confidence limits. Almost one third of patients (27%) experienced a deterioration in verbal learning. The number of left temporal lobectomy patients who had deteriorated outnumbered the right temporal lobectomy patients by 2:1. Significant improvements in verbal learning were seen in 21% of the right temporal lobectomy group and 10% of the left temporal lobectomy group. Patients who were seizure-free postoperatively were not more likely to experience a postoperative deterioration or improvement in memory than those who continued to experience seizures. No significant relationships were found between subjective ratings of postoperative memory function and objective indices of change. Reliable, objective indices of postoperative deterioration in memory function may bear little relation to the patient’s subjective experience. This should be considered when statistical predictions are used as the basis of preoperative counseling.

Introduction

Clinicians are increasingly aware of the utility of multivariate statistical procedures that allow the generation of models that predict postoperative cognitive change in epilepsy surgery patients. The earliest models integrated baseline neuropsychological measures with demographic data to produce models that predict postoperative change in memory function [1]. Later studies have included data from the sodium amytal test and structural MRI to predict postoperative decline [2], [3], [4], [5].

The combination of large patient cohorts with a rich source of preoperative data on the function and structure of the brain allows for increasingly sophisticated predictive models. These models are frequently presented as an objective basis on which to base preoperative counseling and can be useful in the design and implementation of preemptive rehabilitation strategies where a significant postoperative decline is predicted. However, the clinical utility of these models depends heavily on the definition of change used.

There has recently been a shift away from simple measures of change based on the subtraction of raw preoperative data from postoperative scores to the use of reliable change indices (RCIs) and standardized regression-based (SRB) methodologies [6], [7], [8], [9]. RCIs combine the standard deviation and reliability characteristics of a test to provide a statistically robust determination of meaningful change in test–retest situations. RCIs have now been published for a number of standard neuropsychological tests [6], [7], [10], [11] and are widely employed in epilepsy surgery programs across the world to assess postoperative change. Similarly, standardized regression based methodologies create models of change based on change in unoperated groups. These models predict follow-up scores based on baseline performance, age, education, and gender. The difference between the observed and predicted follow-up scores is standardized using the standard error of the estimate.

Although these methods are undoubtedly statistically superior to simple subtraction scores, it does not appear that these definitions of change are any more closely allied with the subjective experience of postoperative change than traditional measures [12].

There is a large body of literature that suggests that there is little correlation between traditional neuropsychological test scores and measures of change and patients’ complaints of memory problems [12], [13], [14], [15], [16], [17], [18], [19], [20]. These studies suggest that subjective memory complaints in both postsurgical and unoperated groups bear little relation to formal neuropsychological test scores and may be more closely related to levels of anxiety, depression, and seizure frequency, although the relationship is by no means straightforward and may be mediated by intelligence and other cognitive factors [16]. Sawrie et al. [12] found that measures of subjective and objective change (SRB) on the California Verbal Learning Task were not significantly correlated in a sample of 45 surgical patients. Consistent with previous studies [16], [19], [20], they found that postoperative measures of anxiety and depression were significantly correlated with self reported memory decline 1 year after surgery.

The aim of this study was to further examine the relationship between three measures of postoperative change on a list-learning task—(1) RCIs at an 80% confidence interval (RCI80); (2) RCIs at a 90% confidence interval (RCI90); (3) ±1 SD—and measures of mood, seizure status and subjective memory complaints recorded 1 year postoperatively in a cohort of temporal lobe epilepsy surgery patients.