Only two radiofrequency techniques (thermal and cooled) had sucient data for the follow-up visit at 6 months for meta-analysis. The patients had signicant improvements in pain at this follow-up for the two radiofrequency techniques compared with the baseline level (random-eects model: 4 trials, MD = 3.69, 95 % CI = 2.814.57, P < 0.00001 for thermal radiofrequency; randomeects model: 3 trials, MD = 3.57, 95 % CI = 2.624.52, P < 0.00001 for cooled radiofrequency) (Fig. 10). The ecacy of the two techniques was similar (P = 0.85). Only two radiofrequency techniques (thermal and cooled) had sucient data at the 12-month follow-up for metaanalysis. The patients had signicant improvements in pain at this follow-up for the two radiofrequency techniques compared with the baseline level (random-eects model: 5 trials, MD = 3.38, 95 % CI = 2.444.32, P < 0.00001 for thermal radiofrequency; random-effe

53, 95 % CI = 2.644.42, P < 0.00001 for cooled radiofrequency) (Fig. 11). The ecacy of the two techniques was similar (P = 0.82).

3.5.3. Methodological uncertainty To assess the methodological uncertainty, we investigated if the results changed in the analyses or comparisons using xed-eects model or random-eects model (Table S2). There were changes in one comparison. For treating LFJ, the three techniques for the follow-up visit at 6 months reached borderline statistical signicance (P = 0.06) using the random eects model but become signicantly dierent (P < 0.00001) when using the xed-eects model. Other results estimated by random-eects and xed-eect models appeared similar.

3.6. Complications and adverse eects No serious complications were reported after receiving treatment using the three radiofrequency techniques [25,27,28,30,31,33,38], and only minor complications were reported, such as pain, hemorrhage, and infection [19,21,23,29,34,39]. 4. Discussion A comparison of the ecacy among three radiofrequency techniques in the treatment of low back pain has not been well investigated. We endeavored to compare the ecacy of the three techniques using a meta-analysis. All three techniques showed signicant improvements in LFJ or SIJ pain for up to 12 months when compared with the baseline level. Previous meta-analysis studies also concluded that thermal radiofrequency has signicant reductions in LFJ pain up to 12 months , and cooled radiofrequency can signicantly relieve pain in patients with SIJ pain . For treating LFJ or SIJ pain, the three techniques did not show signicant dierences in pain relief for the follow-