Genome-wide association study (GWAS) was performed for 16 agronomic traits including nitrogen use efficiency (NUE) and yield-related components using a panel of 190 mainly japonica rice varieties and a set of 38 390 single nucleotide polymorphism (SNP) markers. This panel was evaluated under rainfed upland conditions in Madagascar in two consecutive cropping seasons with two contrasted nitrogen input levels. Using another set of five grain traits, we identified previously known genes (GW5, GS3, Awn1 and Glabrous1), thus validating the pertinence and accuracy of our datasets for GWAS. A total of 369 significant associations were detected between SNPs and agronomic traits, gathered into 46 distinct haplotype groups and 28 isolated markers. Few association signals were identified for the complex quantitative trait NUE, however, larger number of quantitative trait loci (QTLs) were detected for its component traits, with 10 and 2 association signals for nitrogen utilization efficiency and nitrogen uptake efficiency, respectively. Several detected association signals co-localized with genes involved in nitrogen transport or nitrogen remobilization within 100 kb. The present study thus confirmed the potential of GWAS to identify candidate genes and new loci associated with agronomic traits. However, because of the quantitative and complex nature of NUE-related traits, GWAS might have not captured a large number of QTLs with limited effects.