BACKGROUND Perioperative bleeding during total knee arthroplasty (TKA) is an ongoing problem for surgeons. Intravenous or topical application of tranexamic acid (TXA) can effectively stop bleeding, but there is still no uniform standard for the best method of administration and dose. METHODS From October 2016 to September 2018, 218 patients with unilateral primary knee osteoarthritis requiring knee replacement were enrolled and randomly divided into four groups. Group 1 (n = 55) received intra-articular injection (IAI) of TXA and peri-articular injection (PAI) of placebo, group 2 (n = 55) received IAI of placebo and PAI of TXA, group 3 (n = 51) received IAI of TXA and PAI of TXA, and group 4 (n = 57) received double placebo (IAI of placebo and PAI of placebo). The demographic characteristics, surgical indices, hematological indices, wound healing history, and thromboembolic events were investigated. RESULTS Eight patients were lost to follow-up and 210 patients were included in the analysis. The median TBLs in patients who received IAI of TXA and PAI of placebo and those who received IAI of placebo and PAI of TXA were 470.81 ml and 481.54 ml, respectively. These TBL levels were significantly higher compared to those in patients who received IAI of TXA and PAI of TXA (359.18 ml, P ≤ .001), but significantly lower compared to those in patients who received the double placebo (522.71 ml, P ≤ .001). Compared to other groups, more patients in the double placebo group needed a blood transfusion (P = .013). In the short-term, the double placebo group had higher VAS pain scores and less ROM after surgery (P = .011 and P = .001, respectively). In the long-term (6-month follow-up), there were no significant differences in ROM, VAS, DVT, PE, or wound-related complications. CONCLUSIONS The combined use of IAI and PAI of TXA can significantly reduce the TBL and the need for blood transfusion without delaying wound healing or increasing the risk of DVT and PE. In the short-term after surgery, this combined method reduces the pain VAS scores and improves the ROM; however, there are no long-term effects on VAS and ROM.