Background Pink skin tumours are difficult to differentiate, clinically and dermoscopically. In previous studies, mainly focused on pigmented lesions, pattern analysis provided the best sensitivity and specificity values, as compared to other algorithms. These findings suggest that the global dermoscopic appearance, based on the evaluation of prevalent features, could represent a valuable and practical approach even when dealing with pink lesions.Objective In this study, we aimed to evaluate the diagnostic accuracy of a new dermoscopic approach for pink tumours based on the prevalent criterion, as compared to a standard diagnostic method (Menzies algorithm).Methods The databases of two referral centres were retrospectively evaluated to retrieve dermoscopic images of amelanotic/hypomelanotic skin lesions. Two experts in dermoscopy, blinded for the final diagnosis and for clinical and demographic information, evaluated separately dermoscopic pictures of 1000 lesions according to the Menzies score and to the prevalent criterion method.Results According to the high sensitivity model of the Menzies score, 129 (12.9%) lesions were considered as nonsuspicious (of which 16 were false negative) and 871 (87.1%) as suspicious (of which 212 were false positive), with 97.6% sensitivity and 34.8% specificity. According to the high specificity model, 370 (37%) lesions were evaluated as non-suspicious (of which 105 were false negative) and 630 (63%) as suspicious (of which 60 were false positive), with 84.4% sensitivity and 81.5% specificity. Concerning the prevalent criterion method, 316 (31.6%) lesions were evaluated as non-suspicious (of which 46 were false negative) and 684 (68.4) as suspicious (of which 55 were false positive), with 93.2% sensitivity and 83.1% specificity.Conclusions This study demonstrated that focusing on the prevalent dermoscopic features could allow to detect malignant pink tumours with similar sensitivity but higher specificity than using the conventional Menzies scoring system.These authors equally contributed to this work and share the first authorship.