Based on various outcome measures (the total efficiency rate, VAS, WOMAC, LKS, ROM), different pooled data from 12 trials were used. The data were divided into stratified analyses according to different interventions of control groups.

Combination group versus acupuncture groupResult of the total efficiency rate

A total of seven studies [9, 19, 21, 23, 25,26,27] involving 441 KOA individuals compared the total efficiency rate of acupuncture combined with active exercise training and acupuncture for KOA. The results demonstrated that combined treatment was superior to acupuncture in the total efficiency rate [RR = 1.13, 95%CI (1.05, 1.22), I2 = 0%, P = 0.70] (Fig. 6).

Fig. 6

Forest plot of total efficiency rate in comparison with combination group versus acupuncture group

Result of the VAS

A total of seven studies [9, 19, 21, 23, 25,26,27] reported the VAS score in 441 KOA individuals. The results showed that the combined group was better at reducing pain than the acupuncture group [MD =  − 0.74, 95%CI (− 1.04, − 0.43), I2 = 68%, P < 0.05] (Fig. 7A). By exploring heterogeneity, we found the risk of bias in two trials [9, 27] was high. After removing the two trials (the duration of treatment in these two trials was not 4 weeks, while the other trials were all 4 weeks), sensitivity analysis showed that the overall effects did not change [MD =  − 0.72, 95%CI (− 0.88, − 0.56), I2 = 30%, P = 0.22] (Fig. 7B).

Fig. 7

Forest plot of VAS in comparison with combination group versus acupuncture group. (A) All studies. (B) After sensitivity analysis. (C) After subgroup analysis

Because of the heterogeneity of the VAS score, a subgroup analysis of active exercise training after acupuncture and active exercise training with needles showed that heterogeneity was reduced in both groups [MD =  − 0.55, 95%CI (− 0.84, − 0.25), I2 = 52%, P = 0.08], [MD =  − 1.15, 95%CI (− 1.46, − 0.83), I2 = 0%, P = 0.33] (Fig. 7C). The results of the subgroup analysis also suggested that exercise during acupuncture might be more effective than exercise after acupuncture in reducing pain.

Result of the WOMAC total score

A total of seven studies [9, 19, 21, 23, 25,26,27] reported the WOMAC total score in 441 KOA individuals. The results showed that the combined group was better at relieving knee symptoms and improving knee function than the acupuncture group [MD =  − 6.97, 95%CI (− 10.74, − 3.19), I2 = 76%, P < 0.05] (Fig. 8A). By analyzing the sources of WOMAC heterogeneity, we found that after excluding one trial [22] (duration of treatment for 5 weeks, frequency of treatment twice a week), heterogeneity was reduced[MD =  − 5.21, 95%CI (− 7.91, − 2.52), I2 = 52%, P = 0.06] (Fig. 8B).

Fig. 8

Forest plot of WOMAC total score in comparison with combination group versus acupuncture group. (A) All studies. (B) After sensitivity analysis. (C) After subgroup analysis

Due to the heterogeneity of the results, we also performed a subgroup analysis based on the time of exercise intervention (active exercise training after acupuncture or active exercise training with needles), which showed that knee function improved in both groups [MD =  − 7.09, 95%CI (− 12.16, − 2.01), I2 = 83%, P < 0.05], [MD =  − 7.27, 95%CI (− 11.35, − 3.28), I2 = 0%, P = 0.7] (Fig. 8C). The results of the subgroup analysis also suggested that exercise during acupuncture might be more effective in improving knee joint function than exercise after acupuncture.

Result of the WOMAC-dysfunction A total of three trials [9, 19, 27] reported WOMAC-dysfunction in 186 KOA individuals. The results showed that the dysfunction score of the combined group was significantly lower than that of the acupuncture group, indicating that the combined group could better improve the functional status of KOA patients[MD =  − 7.69, 95%CI (− 18.34, 2.96), I2 = 92%, P < 0.05] (Fig. 9A1).

Fig. 9

Forest plot of WOMAC-dysfunction (A), WOMAC-stiffness (B), WOMAC-pain (C) in comparison with combination group versus acupuncture group. (1) All studies. (2) After sensitivity analysis

By analyzing the sources of WOMAC-dysfunction heterogeneity, we performed a sensitivity analysis. We found that the heterogeneity decreased after excluding one trial [27] with the duration of treatment for 5 weeks and frequency of treatment twice a week [MD =  − 1.66, 95%CI (− 6.70, 3.38), I2 = 60%, P = 0.11] (Fig. 9A2).

Result of the WOMAC-stiffness A total of three trials [9, 19, 27] reported WOMAC-stiffness in 186 KOA individuals. The results showed that the combined group was better at relieving knee stiffness than the acupuncture group[MD =  − 1.08, 95%CI (− 2.19, 0.02), I2 = 87%, P < 0.05] (Fig. 9B1). By analyzing the sources of WOMAC-stiffness heterogeneity, we found that heterogeneity was reduced after excluding a three-arm study [19] [MD =  − 1.60, 95%CI (− 2.36, − 0.84), I2 = 60%, P = 0.11] (Fig. 9B2).

Result of the WOMAC-pain A total of three trials [9, 19, 27] reported WOMAC-pain in 186 KOA individuals. The results showed that the combined group was better at relieving pain than the acupuncture group[MD =  − 1.08, 95%CI (− 2.57, 0.40), I2 = 58%, P = 0.09] (Fig. 9C1). By analyzing the sources of WOMAC-pain heterogeneity, we found that heterogeneity was reduced after excluding a three-arm study [19] [MD =  − 1.95, 95%CI (− 3.39, − 0.51), I2 = 0%, P = 0.72] (Fig. 9C2).

Result of the ROM

A total of two studies [23, 26] reported the ROM in 119 KOA individuals. The results showed that the combined group was better at improving joint range of motion than the acupuncture group [MD = 6.25, 95%CI (2.37, 10.04), I2 = 0%, P = 0.71] (Fig. 10).

Fig. 10

Forest plot of ROM in comparison with combination group versus acupuncture group

Combination group versus exercise or western medicine or exercise medicine groupResult of the total efficiency rate

A total of five studies [19, 20, 22, 24, 28] involving 321 KOA individuals compared the total efficiency rate of acupuncture combined with active exercise training with exercise or western medicine or exercise medicine on KOA. The results demonstrated that combination group was superior to control group in the total efficiency rate [RR = 1.31, 95% CI (1.18, 1.47), I2 = 48%, P = 0.10] (Fig. 11).

Fig. 11

Forest plot of total efficiency rate in comparison with combination group versus exercise or western medicine or exercise medicine group

Result of the VAS

A total of five studies [19, 20, 22, 24, 28] reported the VAS in 321 KOA individuals. The results showed that the combined group was better at reducing pain than the control group [MD = 1.42, 95% CI (− 1.85, − 1.00), I2 = 65%, P = 0.02] (Fig. 12A). By exploring heterogeneity, we found that the heterogeneity decreased after excluding a three-arm trial [19] with only exercise and no other treatment[MD =  − 1.24, 95%CI (− 1.57, − 0.91), I2 = 0%, P = 0.41] (Fig. 12B).

Fig. 12

Forest plot of VAS in comparison with combination group versus exercise or western medicine or exercise medicine group. (A) All studies. (B) After sensitivity analysis. (C) After subgroup analysis

Due to the heterogeneity of the results, we further performed a subgroup analysis of the duration of treatment. The results showed that heterogeneity was reduced in both groups [MD =  − 1.07, 95%CI (− 1.67, − 0.46), I2 = 46%, P = 0.17], [MD =  − 1.75, 95%CI (− 2.03, − 1.48), I2 = 10%, P = 0.33] (Fig. 12C). The results of the subgroup analysis also indicated that the longer the treatment period, the more pain reduction in KOA individuals.

Result of the WOMAC total score

A total of four studies [19, 20, 22, 24] reported the WOMAC total score in 257 KOA individuals. The results showed that the combined group was better at relieving knee symptoms and improving knee function than the control group [MD =  − 7.05, 95%CI (− 11.43, − 2.66), I2 = 86%, P < 0.05] (Fig. 13A). By analyzing the sources of WOMAC heterogeneity, we found that after excluding one trial [24] with exercise during acupuncture and treatment duration of 20 days, heterogeneity was reduced [MD =  − 5.36, 95%CI (− 9.26, − 1.46), I2 = 62%, P = 0.07] (Fig. 13B).

Fig. 13

Forest plot of WOMAC total score in comparison with combination group versus exercise or western medicine or exercise medicine group. (A) All studies. (B) After sensitivity analysis. (C) After subgroup analysis

Due to the heterogeneity of the results, we also performed a subgroup analysis based on the type of exercise (isometric strength training + isotonic strength training or not), which showed that knee function improved in both groups [MD =  − 3.41, 95%CI (− 6.49, − 0.32), I2 = 0%, P = 0.32], [MD =  − 10.44, 95%CI (− 12.81, − 8.07), I2 = 40%, P = 0.20] (Fig. 13C).

Result of the WOMAC-dysfunction A total of two trials [19, 22] reported WOMAC-dysfunction in 137 KOA individuals. The results showed that the dysfunction score of the combined group was significantly lower than that of the control group, indicating that the combined group could better improve the functional status of KOA individuals [MD =  − 5.34, 95%CI (− 7.81, − 2.87), I2 = 20%, P = 0.26] (Fig. 14A).

Fig. 14

Forest plot of WOMAC-dysfunction (A), WOMAC-stiffness (B), WOMAC-pain (C) in comparison with combination group versus exercise or western medicine or exercise medicine group

Result of the WOMAC-stiffness A total of two trials [19, 22] reported WOMAC-stiffness in 137 KOA individuals. The results showed that the combined group was better at relieving knee stiffness than the control group [MD =  − 0.39, 95%CI (− 0.73, − 0.06), I2 = 0%, P = 0.47] (Fig. 14B).

Result of the WOMAC-pain A total of two trials [19, 22] reported WOMAC-pain in 137 KOA individuals. The results showed that the combined group was better at relieving pain than the control group [MD =  − 1.43, 95%CI (− 2.13, 0.73), I2 = 0%, P = 0.50] (Fig. 14C).