Quality of life measures in epilepsy

Sample.

Data were gathered from a national, multicenter randomized controlled trial of vigabatrin add-on therapy. All enrollees were taking phenytoin or carbamazepine monotherapy at study entry. Of the 192 subjects who were randomized, 142 completed the study, including follow-up questionnaires and interviews at a 28-week follow-up. Mean age of the 142 subjects was 38.2 years (range 18.8 to 66.8 years), and 48% were men. Over 85% of the subjects were white, 12.2% were African American, and 1.4% were Asian American. Predominant seizure types represented included 51.6% with complex partial seizures, 26.2% with tonic clonic seizures, and 21.7% with both complex partial and tonic-clonic seizures. Seizure etiology was unknown in 51.7% and related to trauma in 17% of the patients. Mean seizure frequency at baseline was 3.6 seizures per month. Subsets of these 142 subjects that met different criteria for “changed” and “unchanged” were included in the analyses presented here.

Criteria for change.

We evaluated two primary criteria for positive change. The most stringent criterion for classifying a subset of subjects as changed based on changes in seizure frequency was achieving seizure freedom—that is, a 100% reduction in complex partial and tonic-clonic seizure frequency from the 12 weeks before baseline to the 12-week interval before the 28-week follow-up. Participants were classified as unchanged if they had less than a 50% change in seizure frequency. (Fewer than 10% of subjects in the analysis group had a 50% or more increase in seizure frequency; thus, there were not enough people in the subgroups to evaluate responsiveness based on negative change.)

We used a two-category or greater improvement in patients’ self-ratings of their overall condition as another criterion of positive change. Subjects were asked at baseline and again at the 28-week follow-up to categorize their overall quality of life at that time (Appendix). Participants who had no change in this rating were classified as unchanged for this criterion.

HRQOL measures.

Two generic and two epilepsy-targeted measures were evaluated (table 1).

Generic measures.

The SF-36 consists of 36 items that measure eight dimensions of health: physical functioning, social functioning, role limitations due to physical problems, role limitations due to emotional problems, emotional well-being, energy/fatigue, general health perceptions, and pain.22-23 These scales can be summarized into two underlying dimensions: physical and mental health summary scores. The SF-12 is a 12-item subset of the SF-36 that provides accurate estimates of the SF-36 mental and physical health composite summary scores.24

In addition to the SF-36 scoring method adduced by Ware and colleagues,24 the 36 items contained in the SF-36 were estimated using the RAND-36 Health Status Inventory (HSI) scoring procedure, which is based on item response theory (IRT).25 Rather than assume that each item in a scale contributes equal information and adding items together, IRT scoring models estimate the association between a respondent’s underlying level on a characteristic and the probability of a particular item response.26 Consequently, IRT models can yield more accurate estimates of latent traits than do classic test theory scoring methods.25,26 The RAND-36 HSI includes a mental health composite, a physical health composite, and a global health composite.22

Epilepsy-targeted measures.

Quality of Life in Epilepsy (QOLIE)-89 is an epilepsy-targeted measure that includes the SF-36 as a generic core.6 Four domain scores (epilepsy-targeted, mental health, physical health, and cognitive distress) and an overall QOLIE-89 score have been derived. Previously reported analyses of the QOLIE-89 provide support for its reliability and construct validity.8 A short form of the measure, the QOLIE-31, yields an overall score.7,9 Construction of the QOLIE-31 was based on expert consensus regarding dimensions of HRQOL issues covered in the QOLIE-89 thought to be the most important for people with epilepsy, including seizure worry, emotional well-being, energy/fatigue, medication effects, work–driving–social limitations, and cognitive distress.

Data collection.

HRQOL measures, self-ratings of overall condition, and seizure data used in these analyses were collected at baseline and at a 28-week follow-up. Seizure frequency data for classifying subjects as changed or unchanged were determined based on the 12-week interval before randomization into the clinical trial (baseline) and the 12-week interval before the 28-week follow-up after baseline. HRQOL data were returned directly to UCLA investigators; seizure and other clinical data were returned to investigators at Hoechst Marion Roussel and subsequently transferred to UCLA investigators. Participating institutional review boards approved all procedures for recruitment and interactions with subjects.

Analyses.

Responsiveness indices included effect size, standardized response mean, Guyatt responsiveness statistic, and F statistics comparing change scores between the changed and unchanged groups.27 All indices were calculated for each of the HRQOL measures with the two primary external change criteria discussed above: a 100% reduction in seizure frequency and a two-category or greater improvement in patients’ self-ratings of their overall condition. Cohen’s effect size criteria of “small,” “medium,” and “large” were applied to observed effect sizes for each change criterion and each HRQOL measure.28 The median rank of each HRQOL measure across responsiveness indices was determined. These rankings provide comparative assessments of the degree of responsiveness of each measure.

Classification of subjects by external criteria.

Of the 142 subjects, 22 (15%) became seizure free and were thus classified as changed according to that criterion. Sixty-one (43%) had less than a 50% increase or decrease in seizure frequency and composed the unchanged group. For the patients’ self-ratings of their overall condition, 27 (21%) of the 142 subjects had ratings that were at least two categories better at follow-up compared with baseline. Fifty-one (40%) had no change in their global rating and comprised the unchanged group for this criterion.

Comparisons of responsiveness (tables 2 and 3⇓).

The QOLIE-31 yielded responsiveness indices comparable to the QOLIE-89, despite having one-third the number of items. Effect sizes for the overall QOLIE-31 score were 0.70 for the criterion of 100% reduction in seizure frequency and 0.74 for the overall condition criterion, corresponding to a medium-to-large effect.28 Effect sizes for the overall QOLIE-89 score for these criteria were 0.68 and 0.63. For the QOLIE-89 epilepsy-targeted domain, corresponding effect sizes were 0.66 and 0.80. Using the seizure freedom criterion, the QOLIE-89 physical health composite had a higher effect size than the QOLIE-89 mental health composite (0.65 versus 0.57). For the QOLIE-89, effect sizes were medium and large for all domains and summary scores except for cognitive distress, where effect sizes were small.

In general, both epilepsy-targeted measures had larger responsiveness indices than the SF-36 and SF-12. However, the responsiveness indices of the RAND-36 HSI mental health and global health composites were comparable to those of the epilepsy-targeted measures for the criterion of change in seizure frequency. The traditional scoring of the SF-36 revealed less sensitivity to change, with effect sizes for the mental health summary score of 0.58 and only 0.38 for the physical health summary score. SF-12 responsiveness indices were even smaller than those of the SF-36.

In general, other responsiveness indices (standardized response mean, Guyatt statistic, and F-ratio) yielded similar findings for generic versus epilepsy-targeted measures. p Values associated with F-statistics were all significant except for those of the SF-12 composites for the change in patients’ self-ratings of their overall condition, and for the SF-36 physical health composite for the change in seizure frequency criterion.