The MFI was developed in 1995 to assess fatigue as a multidimensional concept. This questionnaire has been widely translated and used; however, its multidimensional structure has not been consistent across studies. In addition there is a need for short questionnaires in the clinical context. Therefore, we aim to develop a MFI-short version which is easy-to-use for the patient, quick for the clinician, with solid psychometric properties. We examined three MFI-20 validated versions and tested their structure, by leading a CFA on 422 cancer patients to evaluate their psychometric qualities (Study 1). Then, we conducted both EFA and CFA by deleting all reversed items. We assumed that 10 items were sufficient to conserve a good explanation of the data and a multidimensional structure (Study 2). Study 1 revealed that there is no consensus about the structure of the three MFI versions tested and CFA showed inadequate fit (RMSEA > .10; CFI < .9). Those three versions showed an inadequate fit with regard to expected fit indices. In study 2, the multidimensional structure was confirmed with only ten items. The best model gave a three-factor solution with a χ2 = 67.130, a RMSEA = .072 (95% CI [.048-.096]) and a CFI equal to .97. The MFI short-form maintains good psychometric quality. This new questionnaire is adapted to cancer populations, and may be useful for clinicians to screen their patients' fatigue. Further validations of this MFI-short form are warranted to confirm its psychometric properties in other populations.