If patients in whom test scores were,21 only in two domains clustering together in a single PCA factor were reclassified as ‘not impaired’, 43% of our study population could be considered as cognitively impaired, a prevalence lower than previously reported. The first component from our PCA un-rotated model showed similar loadings on all tests suggesting similar contribution on the overall variance and therefore, may support the appropriateness of using a simple average of test scores to summarise cognitive function. However,Nestorone the GPT showed a somewhat smaller contribution reflecting its lower correlation with the other tests, and it could be suspected that the cognitive domain it measures has less weight in a simple average. As an alternative approach to measure cognitive function we averaged the two parts of the HVLT-R and CTT to calculate an NPZ-3 score. Here, we observed a greater level of agreement between the two definition of NCI as 36% and 32% of study participants showed NPZ-3,21 and scores,21 in two or three tests, respectively. Our study population was large, homogenous, effectively treated and derived from a multi-centre study which confers some strength to our findings. However,Icaritin our testing battery was brief and by no means comprehensive. Using a larger battery may generate different prevalence of NCI even in the same population, perhaps increasing sensitivity. However, depending of the level of correlation between tests, including a larger number of observations may increase the probability of $2 abnormal results. Simioni et al reported a very high prevalence of NCI in effectively suppressed patients using a very large battery and adjusting their analysis for multiple comparisons but not for correlation between the 39 scores used. On the other hand, a number of short batteries exploring each cognitive domain with a single test have demonstrated good sensitivity and specificity compared to a comprehensive test battery. Prevalence of NCI is highly dependent not only on the definition used, but also on normative datasets utilised for analysis since a number of socio-demographic and cultural factors might impact performance on neuropsychological tests.