Cancer diagnoses which are not confirmed by pathology are often under-registered in cancer registries compared to pathology-confirmed diagnoses. It is unknown how many patients have a non pathology-confirmed cancer diagnosis, and whether their characteristics and survival differ from patients with a pathology-confirmed diagnosis. Participants from the prospective population-based Rotterdam Study were followed between 1989 and 2013 for the diagnosis of cancer. Cancer diagnoses were classified into pathology-confirmed versus non pathology-confirmed (i.e., based on imaging or tumour markers). We compared participant characteristics and the distribution of cancers at different sites. Furthermore, we investigated differences in overall survival using survival curves adjusted for age and sex. During a median (interquartile range) follow-up of 10.7 (6.3-15.9) years, 2698 out of 14,024 participants were diagnosed with cancer, of which 316 diagnoses (11.7%) were non pathology-confirmed. Participants with non pathology-confirmed diagnoses were older, more often women, and had a lower education. Most frequently non pathology-confirmed cancer sites included central nervous system (66.7%), hepato-pancreato-biliary (44.5%), and unknown primary origin (31.2%). Survival of participants with non pathology-confirmed diagnoses after 1 year was lower compared to survival of participants with pathology-confirmed diagnoses (32.6% vs. 63.4%; risk difference of 30.8% [95% CI 25.2%; 36.2%]). Pathological confirmation of cancer is related to participant characteristics and cancer site. Furthermore, participants with non pathology-confirmed diagnoses have worse survival than participants with pathology-confirmed diagnoses. Missing data on non pathology-confirmed diagnoses may result in underestimation of cancer incidence and in an overestimation of survival in cancer registries, and may introduce bias in aetiological research.