OBJECTIVES: To validate the short version of the QLQ-C30 obtained for patients with non-small-cell lung cancer in patients with head and neck, prostate, breast or cervix cancer. METHODS: We analysed data of 636 patients distributed: 237 diagnosed with head and neck cancer, 146 diagnosed with breast cancer, 140 diagnosed with cervix cancer and 113 diagnosed with prostate cancer. The analysis followed a 4-step approach. First, we conducted a Mokken nonparametric item response analysis to ascertain the QLQ-C30 dimensionality and separate several scale if appropriate. Second, we conducted a parametric Samejima’s graded response model (GRM) to assess the item characteristics and information for each scale. Third, we did a confirmatory factor analysis (CFA) to test the scales unidimensionality and to obtain standardised factor loadings to suggest a reduced version of the QLQ. Finally, we assessed the discriminative validity of the reduced version by using receiver-operator curve (ROC) analysis. RESULTS: Mokken analysis of the QLQ-C30 resulted in a unidimensional scale, with an overall scalability defined a medium scale. The unconstrained GRM showed that most items presented appropriate difficulty and discrimination parameters. The CFA supported an underlying unidimensional latent structure for the whole QLQ-C30 (CFI = 0.98; RMSEA = 0.05) with modification indexes pointing to important redundancy of information. The selection of items with standardized factor loadings > 0.70 lead to a 6-item QLQ that showed good discriminative validity against independent criteria of quality of life (ROC area = 0.76; 95% CI = 0.72 to 0.80) as compared with the values for the whole scale (ROC area = 0.70; 95% CI = 0.66 to 0.74). CONCLUSIONS: The EORTC reduced scale was validated in this study; it presents good psychometric properties and includes a unidimensional structure of patient-perceived quality of life.