The further miniaturization of liquid-phase microextraction (LPME) systems has important significance and major challenges for microscale sample analysis. Herein, we developed a rapid and flexible droplet-droplet microfluidic microextraction approach to perform nanoliter-scale miniaturized sample pretreatment, by combining droplet-based microfluidics, robotic liquid handling, and LPME techniques. Differing from the previous microextraction methods, both the extractant and sample volumes were decreased from the microliter scale or even milliliter scale to the nanoliter scale. We utilized the ability of a liquid-handling robot to manipulate nanoliter-scale droplets and micrometer-scale positioning to overcome the scaling effect difficulties in performing liquid-liquid extraction of nanoliter-volume samples in microsystems. Two microextraction modes, droplet-in-droplet microfluidic microextraction and droplet-on-droplet microfluidic microextraction, were developed according to the different solubility properties of the extractants. Various factors affecting the microextraction process were investigated, including the extraction time, recovery method of the extractant droplet, static and dynamic extraction mode, and cross-contamination. To demonstrate the validity and adaptability of the pretreatment and analysis of droplet samples with complex matrices, the present microextraction system coupled with MALDI-TOF mass spectrometry (MS) detection was applied to the quantitative determination of 7-ethyl-10-hydroxylcamptothecin (SN-38), an active metabolite of the anticancer drug irinotecan, in 800-nL droplets containing HepG2 cells. A linear relationship (y = 0.0305x + 0.376, R2 = 0.984) was obtained in the range of 4-100 ng/mL, with the limits of detection and quantitation being 2.2 and 4.5 ng/mL for SN-38, respectively.