MicroRNAs (miRNAs) have emerged as promising cancer biomarkers. However, exploiting their informative potential requires careful optimization of their detection. Here, we compared the efficiency of commonly used RNA extraction kits in miRNA recovery from cells, plasma and urine/plasma-derived exosomes, using single-gene RT-qPCR and miRNA profiling. We used increasing amounts of starting material to investigate the impact of the input material size on miRNA extraction. We showed that miRNA recovery was largely influenced by the isolation method and by the amount of input material. In particular, the miRCURY™ kit provided highly pure RNA. However, its columns poorly recovered miRNAs from limiting amounts of cells and plasma, and rapidly saturated by large RNA species and plasma components, thus impeding miRNA recovery from high input amounts. Overall, the miRNeasy® kit permitted a better miRNA detection despite a less pure extracted RNA. Nevertheless, some miRNAs were preferentially or exclusively isolated by either of the methods. Trizol® LS resulted in very low purity RNA which affected RT-qPCR efficiency. In general, miRCURY™ biofluids kit efficiently extracted miRNAs from plasma. A careful selection of the RNA isolation method and the consideration of the type and size of input material are highly recommended to avoid biased results.