In total, 1443 relevant studies were identified during the initial examination. After eliminating duplicate studies, 1234 studies were obtained. After screening the titles and abstracts, 51 studies were subjected to full-text screening, with 10 studies finally included [14,15,16,17,18,19,20,21,22,23], as shown in Fig. 1. The basic features of the included studies are presented in Table 1. Two studies were published in English [14, 15] and eight in Chinese [16,17,18,19,20,21,22,23], and all were randomised controlled trials with a total sample size of 5131 patients. Types of surgery included laparoscopic surgery, thoracoscopic surgery, and knee arthroscopy. Figures 2 and 3 present the bias risk assessment. The Cochrane Bias Risk tool reported a potentially ambiguous and high risk of bias, mainly in implementing blinding, given the challenges in double blinding for such procedural trials.

The four included studies [14, 15, 17, 22] showed no heterogeneity (I2 = 0%, p = 0.77); therefore, a fixed-effects model was used for meta-analysis. According to the results, there was no significant difference in the incidence of postoperative vomiting between the intervention group and the control group (OR = 0.81; 95%CI [0.58–1.12]; p = 0.20) (Fig. 4). The experimental and control groups had a sufficient sample size, with both including 664 patients; therefore, the possibility that early drinking water would not increase the incidence of vomiting in patients could be supported to a certain extent. However, owing to the small number of included studies assessing this outcome and the limited involvement of surgical types, further studies are needed.

Primary outcome: vomiting. The forest plot of pooled data on changes in the incidence of vomiting in the experimental and control groups using a fixed-effects model

Four studies [20,21,22,23] mentioned aspiration indicators. Only two studies [20, 21] were included, given that the other two studies [22, 23] did not include aspiration. I2 = 0% was used as the fixed-effects model for the meta-analysis. There was no significant difference in the postoperative aspiration rate between the intervention and the control groups (OR = 0.78; 95%CI [0.45–1.37]; p = 0.40), as shown in Fig. 5.

Primary outcome: aspiration. The forest plot of pooled data on changes in the incidence of aspiration in the experimental and control groups using a fixed-effects model. CI, confidence interval

The five included studies [14, 15, 17, 20, 22] had low inter-study heterogeneity (I2 = 27%, p = 0.24); therefore, the fixed-effects model was applied for the meta-analysis. Compared with the control group, the incidence of postoperative nausea (OR) in the intervention group was 0.89 (95%CI [0.69- 1.15]; p = 0.38), although the difference was not statistically significant (Fig. 6).

Secondary outcome: nausea. Forest plot of pooled data for nausea across studies with comparator group using a fixed-effects model. CI, confidence interval

Statistical heterogeneity among the seven included studies [14,15,16,17,18, 21, 23] was large (I2 = 99%; p < 0.01), with an MD of -9.44 (95%CI [-12.04- -6.83], p < 0.01) for thirst scores in the intervention group when compared with those of the control group. Subgroup analysis was performed according to the different numerical scoring criteria. The results of the subgroup analysis with a score from 0 to 10 showed that early drinking significantly improved thirst (MD = -2.94; 95%CI [-5.43- -0.45]; p = 0.02; I2 = 99%), while the results of the other subgroup with a score from 0 to 100 also revealed that early drinking significantly improved thirst (MD = -23.38; 95%CI [-32.06- -14.71]; p < 0.01; I2 = 94%), as shown in Fig. 7. Sensitivity analyses were performed for each subgroup, excluding the studies individually. No sources of heterogeneity were detected; this could be attributed to the individualisation of numbers expressing the degree of thirst for each patient according to their own comprehension, resulting in a wide variation in scores.

Secondary outcome: degree of thirst. Forest plot of pooled data for the degree of thirst across studies with comparator group using a random-effects model. CI, confidence interval; SD, standard deviation

Five studies [16, 19, 20, 22, 23] reported the effect of early water intake on anal exhaust time in postoperative patients. The heterogeneity between the studies was high (I2 = 95%, p < 0.01); therefore, a random-effects model was used. Patients in the intervention group had a shorter anal exhaust time than patients in the control group, with a statistically significant difference (MD = -5.48; 95%CI [-7.74- -3.22]; p < 0.01), as shown in Fig. 8. Sensitivity analysis was performed, with the effect estimates held constant, indicating the robustness of the pooled results.

Secondary outcome: anal exhaust time. Forest plot of pooled data on anal exhaust time across studies with comparator group using a random-effects model

Three studies [16, 20, 23] included postoperative patient defaecation time as an outcome indicator; there was high heterogeneity between these studies (I2 = 59%; p = 0.09); therefore, a random-effects model was employed. Patients in the intervention group had significantly shorter postoperative defaecation times than those in the control group (MD = -6.34, 95%CI [-8.90- -3.79], p < 0.01), as shown in Fig. 9. A sensitivity analysis was conducted, where the effect estimates were kept constant, revealing the reliability and stability of the combined outcomes.

Secondary outcome: first defecation time. Forest plot of pooled data on first defecation time across studies with comparator group using a random-effects model

The TSA results of both primary outcomes indicated that the curve was not crossed the traditional and TSA thresholds, and the cumulative sample size was not reached the expected value, requiring further research, as shown in Supplementary Figs. 1 and 2.

We assessed the quality of the outcome using the GRADE evaluation. The quality of the studies that were included was generally low, which led to a downgrading of the risk of bias to “serious”. Moreover, the I2 values for both the degree of thirst, anal exhaust time and first defecation time displayed a significant level of inconsistency, resulting in a downgrade of the inconsistency rating to “serious”. In the GRADE evaluation, except for the quality assessment of degree of thirst, anal exhaust time and first defecation time, the quality of other outcomes was moderate. Supplementary Table 1 summarized the overall results of the GRADE assessment.