The purpose of this article is to evaluate the influence of the definition of hyperprogressive disease (HPD) on the incidence of the phenomenon, the patient’s characteristics and the correlation with survival outcomes on a same NSCLC cohort. We retrospectively studied a cohort of 406 patients with advanced NSCLC treated with PD-1/PD-L1 inhibitors from November 10, 2012 to April 5, 2017 in eight French institutions. Measurable lesions were defined as per the RECIST 1.1 criteria on at least 2 CT-scans before the initiation of immunotherapy and 1 CT-scan during treatment, allowing the calculation of dynamic indexes like Tumour Growth Rate (TGR) and Tumour Growth Kinetics (TGK). We observe that the different definitions result in incidences of the HPD phenomenon varying moderately. The concordance between these different definitions (using the Jaccard similarity index) vary widely. For every definition, patients experiencing HPD are associated with poorer survival outcomes but only two of these definitions are significantly correlated with Overall Survival (p < 0.05). Our study suggests that the previous six definitions of HPD do not represent the same tumoral behaviour. A new definition derived from our study appears to be more relevant to describe the characteristics expected with hyperprogression (increase of the tumour kinetics and poor survival). Additional studies on larger groups of patients with more diverse characteristics would be necessary to confirm the accuracy of these results and definition.