香京julia种子在线播放

This is a systematic review and NMA of preoperative carbohydrate intervention trials in adult patients undergoing elective surgery. The Preferred Reporting Items for Systematic Reviews (PRISMA) and Meta-analyses for RCTs were used to organize the reporting (13). The study protocol was registered (registration number: CRD42022312944) with the International Prospective Register of Systematic Reviews (PROSPERO) following the standard reporting method.

MEDLINE (PubMed), Web of Science, EMBASE, EBSCO, the Cochrane Central Register of Controlled Trials, and China National Knowledge Infrastructure (CNKI) were searched to identify eligible trials. We updated the literature search weekly, and the search was performed from database inception until 16 September 2022 (details are shown in Supplementary Table 1).

Eligible trials included the preoperative administration of at least 10 g carbohydrate loading (orally or intravenously) before 4 h of the surgery started, and with fasting, placebo, or water, undergoing any type of elective surgery in adults. Studies also included carbohydrate-based solutions containing other compounds (such as glutamine and whey protein). Patients with diabetes mellitus or those who were receiving emergency surgery were also excluded.

Two investigators independently screened articles by title, abstract, and full text using the inclusion criteria. The inclusion of a study was decided by consensus between the two investigators. When differences occurred, investigators consulted or discussed with a third one to solve them.

Five categories were used to classify the preoperative administration for the included RCTs:

The primary outcome was mean change from baseline to the end point (within the first 24 h after surgery) in insulin resistance, as measured by the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) method according to the following equation: HOMA-IR = [fasting insulin(μU/mL) × fasting glucose (mmol/L)]/22.5)]. Secondary outcomes were included: residual gastric volume (RGV) during the operation; insulin sensitivity (measured by the hyperinsulinemic glucose clamp method) within the first 24 h after surgery; fasting plasma glucose (FPG) within the first 24 h after surgery; fasting serum insulin (Fin) level within the first 24 h after surgery; the serum levels of C-reactive protein (CRP) within the first 24 h after surgery; postoperative pain, patients’ satisfaction, thirst, hunger, anxiety, postoperative nausea and vomit (PONV), fatigue, weakness (all measured on a visual analog scale [VAS]), and the occurrences of postoperative infection.

The following study characteristics were extracted for each eligible study: (1) trial information: the first author, study year, the study country, and trial name; (2) patient characteristics: sample size in each treatment, the type of surgery, and American Society of Anesthesiologists (ASA) grade; (3) intervention details: the type, total dose, administrate route, and timing of each treatment; (4) outcome measures: the primary or secondary outcomes including insulin resistance, RGV, insulin sensitivity, FPG, Fin level, the serum levels of CRP, pain, thirst, hunger, anxiety, nausea and vomit, fatigue, weakness within the first 24 h after surgery, and the occurrences of postoperative infection.

The quality of every eligible trial was assessed independently by two researchers based on the Cochrane risk of bias 2.0 tool in RCTs in a blind fashion (14), which contains five domains: randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Disagreements were discussed and resolved through consensus.

We estimated the effect sizes for group differences with respect to baseline changes. We used the imputation of correlation when standard deviations were not available for the mean change value, but were available for baseline and endpoint values (15). Arithmetic difference between baseline and end point was used when the study did not report mean change. Meta-analytic calculations were conducted using R Version 4.1.2 (RStudio, Boston, MA, USA) (16). We performed a Bayesian network model and all analyses were conducted using the “gemtc” package version 1.0-1 (17) and jagsUI packages version 1.5.2 (18). Network plot command of Stata version 16.0 (StataCorp, College Station, TX 77845, USA) was used to draw the comparison-adjusted funnel (19).

Mean difference (MD) was used to model continuous variables, whereas dichotomous outcomes were modeled using a binomial likelihood and logit link (20). The outcomes were converted to standard units. Additionally, missing standard deviations were calculated from standard errors, ranges, or interquartile ranges as described in the Cochrane Handbook (21). In this study, a NMA was conducted within a Bayesian framework to assess the relative effectiveness of preoperative carbohydrate loading for recovery after elective surgery.

The consistency model and the inconsistency model were used to analyze all outcomes, and the difference in deviance information criterion (DIC) and I² was used to compare the overall findings. If the difference in DIC between the two models was ≥ 5, we used the inconsistency model. Both a fixed-effect model and a random model were run for each result, and a lower DIC value indicated a greater model fit.

The Markov chain Monte Carlo (MCMC) algorithm was used to estimate the posterior densities of all unknown parameters in each model. It was based on simulations of 200,000 iterations in each of four chains and provided evidence for confirming the convergence of the models.

The trials we included were tested for consistency and inconsistency. We used the node splitting method to perform to compare the treatment effect direct and indirect comparisons of multiple interventions, and P > 0.05 was considered to indicate good consistency (22, 23).

Probability values were summarized and are reported as the surface under the cumulative ranking (SUCRA) curve. When the intervention was certain to be the worst, the SUCRA value would be 0, and when it was certain to be the best, the SUCRA value would be 1 (24).

To investigate the source of heterogeneity, meta-regression was used to explore and account for the heterogeneity with the risk of bias, the category of surgery, and the blinding of these studies’ designs.

The planned sensitivity analyses of the outcomes were conducted to evaluate the robustness of the model. First, in addition to the Bayesian random effect network, sensitivity analyses were performed using a fixed-effect network. Second, the transitivity assumption was tested by splitting the “water or placebo” group within the network. Third, all analyses were repeated after excluding high-risk trials and data from imputation methods. In addition, for the primary outcome, we planned to add subgroup analyses conducted for different surgical categories, and a comparison-adjusted funnel plot was used to assess the presence of small-study effects bias.

The Confidence in Network Meta-Analysis (CINeMA) methodological framework and application were used to evaluate confidence in NMA effect estimates for all outcomes and treatment comparisons (25, 26).

A total of 9411 records were retrieved, of which 58 articles (N = 4936 patients) fulfilled the eligibility criteria and were included in the meta-analysis, the retrieval process is shown in Figure 1. A total of five interventions were included in this meta-analysis: oral low-dose carbohydrate (10–50 g) loading, oral high-dose carbohydrate (more than 50 g) loading, carbohydrate by intravenous perfusion (Carbohydrate, iv), placebo/water, and fasting. Detailed trial and patient characteristics are shown in Table 1.

The overall quality of RCTs included in the network was high and moderate. The risk of bias of 58 studies included in the meta-analysis is shown in Figure 2 (details of the risk of bias 2.0 assessment in each trial are shown in Supplementary Figure 1). According to the risk of bias 2.0 tool of Cochrane Collaboration, 25 (43%) studies were high-quality across all domains and 12 RCTs (21%) were at high risk of bias.

The primary outcome of this study was postoperative insulin resistance, and it was measured by the homeostasis model assessment-insulin resistance (HOMA-IR) method. The network plot for the primary outcome is shown in Figure 3. Each circle represented an intervention, and the area of each circle was proportional to the number of patients for which the intervention was accepted and indicated the sample size, and the width of the line was proportional to the number of trials that directly compared the two interventions.

Twenty-four studies with 1,390 patients reported insulin resistance during the study period (27–51). Both interventions associated insulin resistance compared with placebo/water, with MD ranging from –3.25 (95%CrI, –5.27 to –1.24) for administrated oral low-dose carbohydrate to –2.57 (95%CrI, –4.33 to –0.78) for oral high-dose carbohydrate loading before surgery. The subgroup analysis based on the category of surgery revealed that the association of oral low-dose carbohydrate compared to placebo/water would correlate with insulin resistance (MD, –4.37 [95%CrI, –8.42 to –0.47]) for patients undergoing major abdominal surgery. Figure 4 shows the results. The result of CineMA represents the confidence in this estimate was low (Supplementary Table 2).

Among all trials included, oral low-dose carbohydrate loading had the highest probability of being the best intervention (SUCRA value of 0.74 compared with other interventions). The corresponding results of SUCRA values are shown in Figure 5. Inconsistency analysis calculated by the node split method showed no significant difference between direct and indirect evidence of this network model, with P-value ranging from 0.05 to 0.32 (Supplementary Table 3). The result of the network meta-regression shows that the covariates we included may not affect the value of insulin resistance (Supplementary Table 4).

After excluding studies with a high risk of bias and data of trials with imputation methods (network plot is shown in Supplementary Figure 2), there was an association of oral low-dose carbohydrate loading (MD, –1.29 [95%CrI, –2.26 to –0.27]) with insulin resistance for postoperative patients compared with placebo/water remained. Oral low-dose carbohydrate loading (MD, –1.17 [95% CrI, –1.88 to –0.43]) administration was associated with insulin resistance compared with fasting. The subgroup analysis showed that when patients undergoing major abdominal surgery, administrated oral low-dose carbohydrates before surgery was associated with insulin resistance (MD, –1.35 [95% CrI, –2.64 to –0.01]) compared with fasting. Figure 6 shows the forest plot results. And the SUCRA followed a similar pattern, with oral low-dose carbohydrates having the highest probability of being the best intervention when compared with other interventions; the SUCRA value is 0.88 (Supplementary Table 5).

A comparison-adjusted funnel plot for postoperative insulin resistance implies the presence of publication bias between the carbohydrate groups and controls (Supplementary Figure 3).

Supplementary Figure 4 represents network plots for each secondary outcome. The confidence in these estimates was generally moderate to very low (Supplementary Table 6).

The latest practice guidelines for preoperative fasting recommend that clear liquids may be ingested for up 2 h before an operation; however, it reported less thirst and hunger for fasting time of 2–4 h compared to more than 4 h of fasting, however, it reported equivocal findings for RGV, blood glucose values, hunger, and thirst of nutritional or carbohydrate drinks at 2–4 h relative to more than 4 h of fasting (1).

This NMA represents the most comprehensive analysis of currently available data regarding preoperative carbohydrate loading for patients undergoing elective surgery. We combined direct and indirect evidence from 58 trials comparing four different intervention arms in 4,936 patients undergoing elective surgery. The study that included sufficient numbers of patients to prove a potential association in clinical outcomes was of patients undergoing elective surgery, and it included the most patients available in the current literature. To maintain the homogeneity of interventions, our research divided the dose of carbohydrate loading into low dose (10–50 g) and high dose (>50 g). Our main findings indicate that among patients undergoing elective surgery, preoperative low-dose carbohydrate loading has been found to be associated with insulin resistance and postoperative infection rates.

Three published meta-analyses explored the influence of low-carbohydrate loading on postoperative outcomes (2, 8, 12). However, reports of the effects of carbohydrate loading on insulin sensitivity remain inconsistent. Smith et al. (8) conducted that no significant association was between carbohydrate loading and insulin resistance An earlier NMA of 43 RCTs found that only high-dose carbohydrate administration resulted in a statistically significant associated with insulin resistance compared with fasting, and water or placebo, but with wide confidence intervals so the results are not credible (12). A recent meta-analysis has investigated that compared with fasting, preoperative administration of carbohydrate associated with insulin resistance (2). In our study, we found that oral carbohydrate loading was associated with insulin resistance compared with placebo or water, and the association was still observed in an analysis of excluded high-risk trials and data for the imputation methods. A separate subgroup analysis based on the surgical categories identified the true effect of low-dose carbohydrate loading on insulin resistance, especially those undergoing major abdominal surgery that would otherwise be confounded by other surgical categories. This effect might be due to the preoperative carbohydrate loading, which stimulates an endogenous insulin release and switches off the overnight fasting metabolic state, toward anabolism (63). It should be mentioned that the confidence effect estimate is low or very low, and the significant heterogeneity among studies (different categories of surgery, different types of carbohydrates, and different populations); therefore, the result regarding the effect of carbohydrate loading on insulin resistance must be interpreted with caution.

The present meta-analysis found that oral high-dose carbohydrate (>50 g) was more effective in postoperative outcomes than relative to low-dose carbohydrate, and there is no dose–response relationship between carbohydrate and postoperative outcomes. This may be related to the fact that there is less data available in the network for low-dose carbohydrate comparisons, so some results have wider confidence intervals than in high-dose comparisons.

The gold standard of insulin sensitivity is measured by the hyperinsulinemic–euglycemic clamp method in humans (76). However, we found a small number of studies (n = 7) for this outcome, which could be due to the fact that it is a time-consuming, labor-intensive, and invasive procedure. The multiple-treatments meta-analysis found no evidence that carbohydrate loading was more or less effective in reducing insulin sensitivity compared with placebo/water or fasting. Therefore, more randomized controlled trials need to be included in future analyses to further confirm this outcome.

A recent meta-analysis has investigated that compared with fasting, preoperative administration of carbohydrates decreased patients’ thirst, hungry, and pain (2). Meanwhile, in our study, there was no difference in postoperative patients’ comfort between the administration of preoperative carbohydrates and control groups, and no other significant differences were found in any of the other secondary outcomes. However, some of these results had wide confidence intervals, indicating that data availability is limited. Future well-designed randomized studies will need to examine the biochemical effects and recovery of preoperative carbohydrate loading in elective surgery.

This review has some strengths: First, a comprehensive search was conducted to identify eligible trials; independent study selection, data extraction, and risk of bias assessment were performed by two reviewers; and the CINeMA was used to assess confidence in the NMA results. Second, we also conducted a network meta-regression to evaluate which variables might influence the postoperative outcomes. This review used a Bayesian framework to overcame the tendency of the frequentist approach to be unstable in parameter estimation and obtain biased results (77). Third, we tested different model assumptions to verify the reliability of outcomes in this NMA. Fourth, a NMA is performed to analyze the effect of preoperative carbohydrate loading on various postoperative recovery indicators among elective surgery patients, compensating for the lack of direct comparison between them.

This study has several limitations. First, the results of this meta-analysis are highly dependent on the quality of the trials included. According to the CINeMA results, the evaluation of the credibility of results was from moderate to very low, and there was large uncertainty regarding all the estimates. Second, although 58 RCTs were retrieved, only 21 trials reported postoperative low-dose carbohydrate administration in the network, two studies reported preoperative carbohydrate by intravenous perfusion, and there were relatively few direct comparisons. Third, this may, however, be a type II error (false-negative findings), as only a few trials are available to assess postoperative outcome indicators in many second outcomes. Fourth, small trials tend to report larger beneficial effects than large trials; however, only three trials in our review included more than 100 patients per arm, which may introduce bias due to small-study effects (78). Fifth, the SUCRA value was used to estimate a ranking probability of comparative effectiveness between the different interventions. Sixth, many trials, lack good design, resulting in combining different types of carbohydrates into one group and placebo and water into one group for the main analysis. Finally, double-blinding was not applied in many trials designs included, which may affect the results, but this is also difficult to resolve because fasting and drinking are easily known by the participants, and subsequent experiments need to be further refined.