Practice effects distort translational validity estimates for a Neurocognitive Battery

With the globalization of biomedical research and the advent of "precision medicine," there is increased need for translation of neuropsychological tests, such as computerized batteries that can be incorporated in large-scale genomic studies. Estimates of translational validity are obtained by administering the test in the original and the translated versions to bilingual individuals. We investigated the translation of a neuropsychological battery from English to Arabic and how practice effects influence translational validity estimates.

The Penn computerized neurocognitive battery (Penn CNB) includes tests that were validated with functional neuroimaging and provides measures of accuracy and speed of performance in several cognitive domains. To develop an Arabic version of the CNB, the English version was translated into Arabic, then back translated and revised. The Arabic and the original English versions were administered in a randomized crossover design to bilingual participants (N = 22).

Performance varied by cognitive domain, but generally improved at the second session regardless of the language of the initial test. When performance on the English and Arabic version was compared, significant positive correlations were detected for accuracy in 8/13 cognitive domains and for speed in 4/13 domains (r = .02 to .97). When the practice estimates using linear models were incorporated, the translational validity estimates improved substantially (accuracy, r = .50-.96, speed, r = .63-.92, all correlations, p = .05 or better).

While crossover designs control for order effects on average performance, practice effects, regardless of language, still need to be removed to obtain estimates of translational validity. When practice effect is controlled for, the Arabic and English versions of the Penn-CNB are well correlated, and the Arabic version is suitable for use in research.