According to the search strategy, 5130 articles were screened (Pubmed 1368, Cochrane Library 33, Web of Science 1513, Embase 1212, Clinical trial platform 1, CNKI 276, VIP 423, Wanfang data 304). A total of 2562 duplicate articles were removed, leaving 2568. After screening titles and abstracts, a total of 2539 articles were excluded, and the remaining 29 articles were re-screened. After full-text screening of 29 articles, 14 articles were excluded because full-text was not available, there were no outcome indicators, or there were not enough extractable data. Ultimately, our study included 15 articles [3, 4, 16,17,18,19,20,21,22,23,24,25,26,27,28] involving a total of 755 patients, 386 patients in the PCNL group and 369 patients in the FURL group. A flowchart of the selection process is shown in Fig. 1.

Flow chart of articles retrieval

The studies included were published between 2015 and 2023 and their main characteristics and clinical results are summarized in Table 1. The methodological assessment of NOS ranged from 6 to 8, indicating that all studies in our meta-analysis were of high quality (Table 2).

Twelve articles [4, 16,17,18,19, 21,22,23,24,25,26,27] reported operating time, and the results of heterogeneity test of articles were (P = 0.000, I2 = 93.72%), indicating that the heterogeneity of included articles was significant. Random-effects model was applied for meta-analysis. The results showed that there was no statistically significant difference in operating time between PCNL group and FURL group: SMD = 0.079, 95%CI:(−0.43, 0.589), Z = 0.306, P = 0.760 (Fig. 2). Publication bias test including Egger’s test P = 0.180, Begg’s test P = 0.205, indicating no publication bias. The combined effect value adjusted by Trim’s method is 95%CI:(−0.922,0.207). The above conclusion still holds true. Sensitivity analysis results indicated that the results for pooled effects were generally robust. Funnel plots of publication bias test and forest plot of sensitivity analysis were showed in Figure S1.

Forest plot showing pooled SMD with 95% CI of operation time

Eight articles [4, 17, 18, 21,22,23, 26, 27] reported intraoperative blood loss, and the results of heterogeneity test were (P = 0.000, I2 = 93.07%), indicating that the heterogeneity of the included articles was significant. Random-effects model was applied for meta-analysis. The results showed that there was a statistically significant difference in intraoperative blood loss between PCNL group and FURL group: SMD = 1.713, 95%CI:(0.858, 2.568), Z = 3.928, P = 0.000 (Fig. 3). A statistically significant difference may indicate that intraoperative blood loss was shorter in the FURL group than in the PCNL group. Publication bias test including Egger’s test P = 0.084, Begg’s test P = 0.083, indicating no publication bias. We utilized Trim’s method to adjust combined effect value. The combined effect value adjusted by Trim’s method is 95%CI:(0.319,2.243). The above conclusion still holds true. Sensitivity analysis results indicated that the results for pooled effects were generally robust. Funnel plots of publication bias test and forest plot of sensitivity analysis were showed in Figure S2.

Forest plot showing pooled SMD with 95% CI of blood loss

Thirteen articles [3, 4, 16,17,18,19, 21,22,23,24,25,26,27] reported hospital stay, and the results of heterogeneity test were (P = 0.000, I2 = 94.12%), indicating that the heterogeneity of the included articles was significant. Random-effects model was applied for meta-analysis. The results showed that there was a statistically significant difference in postoperative hospital stay between PCNL group and FURL group: SMD = 2.611, 95%CI: (1.726, 3.496), Z = 5.784, P = 0.000 (Fig. 4). A statistically significant difference may indicate that postoperative hospital stay was shorter in the FURL group than in the PCNL group. Publication bias test including Egger’s test P = 0.008, Begg’s test P = 0.020, indicating potential publication bias. We utilized Trim’s method to adjust combined effect value. The combined effect value adjusted by Trim’s method is 95%CI:(0.389,2.387). The above conclusion still holds true. Sensitivity analysis results indicated that the results for pooled effects were generally robust. Funnel plots of publication bias test and forest plot of sensitivity analysis were showed in Figure S3.

Forest plot showing pooled SMD with 95% CI of hospital stay

Thirteen articles [3, 4, 16, 17, 19,20,21,22,23,24,25,26, 28] reported the incidence of postoperative complications in patients, and the results of literature heterogeneity test were (P = 0.220, I2 = 22.06%), indicating that the heterogeneity of included articles was low, and fixed-effects model was applied for meta-analysis. The results showed that [OR = 1.793,95%CI: (1.139, 2.822), Z = 1.586, P = 0.113] (Fig. 5A). We found that the significance of OR was inconsistent with the significance of P values, which may be due to the relatively large heterogeneity of a study. Publication bias test including Egger’s test P = 0.577, Begg’s test P = 0.714, indicating no publication bias. The combined effect value adjusted by Trim’s method is 95%CI:(0.952,2.602). So we ended up using Trim's method corrected results. Therefore, the results showed that there was no statistically significant difference in complication rate between PCNL group and FURL group. Sensitivity analysis results indicated that the results for pooled effects were not robust. The sensitivity analysis results also confirm the pooled effect we have previously obtained (Fig. 5B). Funnel plots of publication bias test were showed in Figure S4.

Forest plot showing pooled OR with 95% CI of complication rate. A Forest plot of complication rate; B Forest plot of sensitive analysis

Eleven articles [3, 4, 16, 19, 21, 22, 24,25,26,27,28] reported the stone-free rate in patients, and the results of literature heterogeneity test were (P = 0.011, I2 = 56.39%), indicating that the heterogeneity of the included articles was significant. Random-effects model was applied for meta-analysis. The results showed that there was no statistically significant difference in stone-free rate between PCNL and FURL groups: OR = 1.339, 95%CI: (0.576, 3.112), Z = 0.678, P = 0.497 (Fig. 6). Publication bias test including Egger’s test P = 0.576, Begg’s test P = 0.815, indicating no publication bias. The combined effect value adjusted by Trim’s method is 95%CI:(0.522,2.670). The above conclusion still holds true. Sensitivity analysis results indicated that the results for pooled effects were generally robust. Funnel plots of publication bias test and forest plot of sensitivity analysis were showed in Figure S5.

Forest plot showing pooled OR with 95% CI of stone-free rate

Five articles [3, 16, 19, 20, 25] reported the symptom-free rate in patients, and the results of literature heterogeneity test were (P = 0.162, I2 = 45.09%), indicating that the heterogeneity of included articles was low, and fixed-effects model was applied for meta-analysis. The results showed that [OR = 3.826,95%CI: (1.124, 13.026), Z = 0.966, P = 0.334] (Fig. 7A). We found that the significance of OR was inconsistent with the significance of P values, which may be due to the relatively large heterogeneity of a study. Publication bias test including Egger’s test P = 0.648, Begg’s test P = 0.602, indicating no publication bias. The combined effect value adjusted by Trim’s method is 95%CI:(0.561,10.238). So we ended up using Trim's method corrected results. Therefore, the results showed that there was no statistically significant difference in symptom-free rate between PCNL group and FURL group. Sensitivity analysis results indicated that the results for pooled effects were not robust. The sensitivity analysis results also confirm the pooled effect we have previously obtained (Fig. 7B). Funnel plots of publication bias test were showed in Figure S6.

Forest plot showing pooled OR with 95% CI of symptom-free rate. A Forest plot of symptom-free rate; B Forest plot of sensitive analysis