Although experimental studies have shown lipoprotein(a) antiangiogenic and antitumoral effects, the association of lipoprotein(a) levels with cancer in population studies remains elusive and poorly documented. The aim of this study was to analyse the relationship between lipoprotein(a) plasma levels and the incidence of cancer over 10 years of follow-up. Data from two French centres of the PRIME cohort were used, representing 5237 men aged 50–59 years and free from a history of cancer at baseline. Data on medical history, socioeconomic and lifestyle factors were obtained by questionnaire. Lipoprotein(a) plasma levels were analysed from fasting blood samples collected at baseline. The relationship between lipoprotein(a) levels and first incident cancer was studied using the multivariate Cox proportional hazards models for all-site and the main-site-specific cancers, adjusted for various potential confounders including age, centre, smoking status and alcohol consumption. During follow-up, 456 new cancers were identified. No significant association was found between lipoprotein(a) and the all-site or main-site-specific cancers (hazard ratios for quartiles 2–4 vs. 1, respectively: 1.24, 1.11, 1.29, P=0.23). However, a higher risk seemed to be observed for highest lipoprotein(a) levels in all sites, lung, colorectal or tobacco/alcohol-related cancers. For prostate cancer, the lowest risk was observed for the highest levels of lipoprotein(a) (P=0.12). In conclusion, no evident association was found between the lipoprotein(a) levels and the incidence of cancer. Nevertheless, a higher cancer risk seemed to be observed for the highest lipoprotein(a) levels. Further research focusing on the lipoprotein(a) qualitative structure, that is, apolipoprotein(a) polymorphism could help clarify this highly complex relation.