Purpose: This study aimed to identify profiles of caregivers to cancer patients at higher risk of having at least one moderately or highly unmet supportive care need based on 1) relevant socio-demographic (e.g. age, gender) and clinical (e.g. type of cancer, phase of the cancer pathway) variables highlighted in the literature and easily identifiable in routine, and 2) caregivers' anxiety and depression symptoms. Method: Three hundred and sixty-four main caregivers completed a questionnaire assessing their supportive care needs (SCNS-P&C-F) and anxiety and depression symptoms (HADS) during the treatment or follow-up stage of patients with digestive, breast, or lung cancer. Decision trees were used to identify profiles of caregivers with the Conditional inference Tree (CTree) technique. Results: In our study, only the combination of three main variables was important to predict the risk of unmet supportive care needs of caregivers: anxiety and/or depression symptoms, the age of caregivers or patients, and the presence/absence of metastases. Emotional distress has the greatest impact, exceeding that of the socio-demographic and clinical variables considered in this study. Conclusions: This study shows the importance of considering a set of variables and their combinations rather than evaluating their effects separately. Routinely assessing the anxiety and depression symptoms of caregivers using the HADS could improve the screening of caregivers at higher risk of unmet supportive care needs based on socio-demographic and clinical variables only. This study provides recommendations on how to identify caregivers at risk of unmet needs, in the context of an inability to support all caregivers.