This study had several limitations. First, patients may have
improved cognitive performance only because of repetitive practice
of the test, especially memory tests and matrices. However,
the study was designed to minimize the ‘learning effect’ by
1 month intervals and full education and practices of the subjects
to be more familiar with the tests and perform at their best from
baseline. The second limitation was a loss to follow-up, which may
be related to selection bias. However, follow-up rates were similar
in both groups, as was improvement in depression and anxiety
symptoms. Third, the investigator was not blind to the participant’s
group, so there might be some possible confounding effects
implementing neurocognitive tests throughout the period. Finally,
possible interaction between the antidepressant and allowed
benzodiazepines is another limitation of our study. But the dosage
of them was kept unchanged during the study.