How reliable is the classification of cognitive impairment across different criteria in early and late stages of multiple sclerosis?

BACKGROUND

Prevalence rates of cognitive impairment (CI) in multiple sclerosis (MS) vary between 40% and 80%. Differences in classification criteria for CI may explain this variance.

OBJECTIVE

This study reviewed and compared classification criteria for CI in patients with early and late MS.

METHODS

The paper consists of two parts: a systematic review of published classification criteria and the presentation of new data. Criteria were reviewed in respect to percentage of abnormal parameters and cut-offs concerning standard deviations. Thereafter, criteria were applied to cognitive data of 25 patients with early MS (duration ≤ 2 y), 52 matched patients with late MS (≥ 12 y), and 75 matched controls. The test battery assessed alertness, divided attention, mental flexibility, verbal and visual learning, memory, and visuospatial abilities.

RESULTS

Seventy classification criteria were revealed and grouped into 20 distinct approaches that can be subdivided into three basic classification strategies. Most commonly, CI was defined as performing 1.5 SD or 2 SD below the normative mean in 18-30% of test parameters (n=42). Other criteria utilized cognitive domains (n=6), composite indices (n=8), or combinations of cut-offs and strategies. The stringency of the criteria was correlated with the prevalence rate of CI (r=-.43) and disease duration (r=.48). In the new data, a substantial effect of classification criteria was found with a prevalence rate ranging from 0 to 68% in early and 4 to 81% in late MS. Increased rates of CI in patients vs. controls were found following 18 out of 20 criteria in the sample of late MS. In early MS, an increased rate of CI was only found following a liberal 1.5 SD cut-off criterion. Inter-rater reliability between all criteria was moderate. However, between criteria of comparable stringency the inter-rater reliability was found to be strong.

CONCLUSION

Classification based on different published criteria is not fully comparable and criteria need to be better homogenized.