<b><i>Background/Aims:</i></b> Fatigue is recognised as a common and burdensome symptom among dialysis patients. A growing body of research is devoted to understanding fatigue in advanced kidney disease, yet its measurement is challenging within this context. Our aim was to evaluate the factor structure underlying the multidimensional fatigue inventory (MFI-20) and to examine its associations with clinical factors and mood. <b><i>Methods:</i></b> Data was evaluated for confirmatory factor analysis (CFA) from the screening phase of a multicentre randomised placebo-controlled trial of sertraline in haemodialysis (HD) patients. Four hundred seventy patients completed the MFI-20, which purports to measure 5 components of fatigue (general fatigue, mental fatigue, physical fatigue, reduced motivation and reduced activity). CFA models were evaluated in MPlus 7.3 using the robust maximum likelihood (MLR) estimation. <b><i>Results:</i></b> The evaluation of the original 5 factors revealed low internal reliability for the general factor and reduced activity, and high intercorrelations between all sum scores. CFA revealed poor model fit for the original 5-factor MFI-20 model (confirmatory fit index = 0.738; Tucker-Lewis index = 0.689; root mean squared error of approximation = 0.101). Alternative models, including 1, 3 and bi-factor models all demonstrated poor fit to the data. No reliable factor model was confirmed prohibiting the examination of factors associated with fatigue. <b><i>Conclusions:</i></b> We were not able to confirm the factor structure of the MFI-20 in a large sample of HD patients. Certain items may lack suitable face validity in this context.