Background: Semantic-specific impairment on verbal fluency tasks (VFTs) is a recognised marker of Alzheimer’s disease. However, most research has been conducted in high-income countries (HICs), with limited data from Sub-Saharan Africa. In a two-stage pilot and validation study, VFTs were examined in relation to demographic variables and cognitive impairment in Tanzania.
Methods: In a community pilot study, older adults (n=131) from the Hai district, Kilimanjaro completed a neuropsychological battery including the IDEA screen and five VFTs: two semantic (animals, market items), two phonemic (words beginning with C and B), one action. Suspected cognitive impairment was determined according to IDEA score cut-offs (≤7). An additional validation cohort including cognitively healthy participants (n=71), patients with mild cognitive impairment (n=4) or probable Alzheimer’s dementia (n=16), were recruited through memory clinics in Dodoma and Moshi. In each study, VFT z-scores were derived from the mean and standard deviation of cognitively healthy participants (IDEA>7) recruited from the same region. Phonemic-semantic discrepancy scores were calculated as phonemic minus semantic z-scores.
Results: Across cohorts, participants performed better on semantic than phonemic or action fluency (p<0.001). All VFTs correlated with education, frailty and total IDEA scores in both cohorts (p<0.015). Age correlated with all VFTs in the community sample (p<0.005) and with animal, market and letter B fluency (p<0.02) in the validation cohort. IDEA scores correlated with letter B-animal fluency discrepancy scores in both cohorts (p≤0.030) and letter C-animal fluency discrepancy scores in the validation cohort (p=0.005). In both cohorts, all raw VFT scores correlated with delayed memory and object naming (p<0.05). Delayed memory additionally correlated with letter C-animal fluency discrepancies in the validation cohort (p=0.007). All VFTs effectively classified both clinically determined and suspected cognitive impairment in community and clinical cohorts (AUROC≥0.7,p≤0.006). Phonemic-animal discrepancies additionally classified clinically determined cognitive impairment (AUROC>0.7,p≤0.005).
Conclusion: In both East African cohorts, VFT performance reflected similar demographic and cognitive patterns as reported in HICs, despite lower absolute scores. Additionally, phonemic-animal discrepancies were an effective marker of clinically relevant cognitive impairment. The present study therefore validates the inclusion of VFTs, including phonemic-semantic discrepancies, in neuropsychological assessments for early cognitive impairment in Sub-Saharan Africa.