Objective To investigate the best puncture site of peripherally inserted central catheter (PICC) for premature infants. Methods A total of 162 preterm infants who underwent PICC were divided into three groups according to their puncture sites, i.e. the elbow vein puncture group (n=66), axillary vein puncture group (n=39), and knee medial saphenous vein puncture group (n=57). The one-step success puncture rate, unplanned cupping rate, and catheter indwelling time before unplanned cupping of three groups were compared and main reason of unplanned cupping was analyzed. Results The one-step success puncture rates of elbow vein, axillary vein, and knee medial saphenous vein puncture groups were 92.42%, 94.87%, and 91.23%, respectively. The unplanned cupping rates of three groups were 18.18%, 10.26%, and 17.54%. The catheter indwelling time before unplanned cupping of three groups was (22.25±21.10)d, (28.5±15.42)d, and (22.2±4.06)d. The differences of three groups were not statistically significant (P>0.05). The main reason for unplanned cupping was tube plugging. Conclusion For PICC at different sites of premature infants, the one-step success puncture rate, unplanned cupping rate, and catheter indwelling time before unplanned cupping are similar. When the vein at commonly used puncture site is unable to perform PICC, the axillary vein and great saphenous vein can also be used for vein puncture and inserting PICC through the axillary vein shows certain advantages. The study provides more suitable sites for premature infants to establish central venous access through peripheral vein.