Introduction

Overnight fasting has long been recommended for elective surgery to ensure that the patient has an empty stomach before anesthesia, which reduces the risk of aspiration [1–6]. However, questions have been recently raised as to whether such fasting is appropriate, especially regarding the fact that overnight fasting depletes the carbohydrate reserves and changes the metabolism, such as the endocrine response (increased glucagon, depleted insulin) and metabolic response (glycogen breakdown, protein breakdown, lipolysis) [4, 6]. Endogenous insulin resistance (IR) is a central feature of the postoperative metabolic response to surgical injury, and might delay the patient recovery after surgery, which could be associated with infectious complications and morbidity [5, 7]. In addition, overnight fasting is uncomfortable (including preoperative thirst, mouth dryness, hunger and anxiety) for surgery patients [2–4]. Therefore, comprehensive perioperative care programs (such as the ERAS protocol) based on the best-available practices in elective surgery were constructed to minimize the metabolic stress of the operation and to support the function of the vital organs [8]. These perioperative care programs have been shown to result in remarkable improvements in both the recovery after surgery and the development of complications [8–10]. In these programs, preoperative carbohydrate (CHO) consumption is an important component that supports both aims, and therefore, is one of the key treatments in the ERAS protocol [8, 11, 12].

Studies have demonstrated that the metabolic responses, including the IR, were significantly attenuated or abolished by the administration of CHO 2 h before elective open general or gastrointestinal surgery, total hip replacement, and thyroidectomy, primarily owing to a decreased reduction in peripheral glucose disposal and increased glucose oxidation rates [2, 13–20]. Preoperative administration of oral CHO could significantly reduce the length of the hospital stay and recovery time as compared with fasting or consumption of a placebo (flavored water) [19, 21]. CHO neither delayed gastric emptying nor affected the gastric acidity [2, 4, 22], and so it was considered to be safe for elective surgical patients [4, 14, 23, 24]. In addition, CHO has been shown to improve preoperative well-being, the postoperative stress hormone response, and prevent surgery-induced immunodepression in patients scheduled for elective general or gastrointestinal surgery [14, 15, 22, 25]. However, some studies showed that preoperative CHO did not shorten the length of hospital stay after major general or gastrointestinal surgery, or improve the IR after elective colorectal, upper-gastrointestinal or coronary artery bypass surgery [18, 25–27].

A traditional review [28] showed that preoperative oral CHO was associated with a shorter length of hospital stay, but another systematic review [29] indicated that statistical significance was not reached for the length of the hospital stay, although trends were seen. The data from these reviews were either outdated or insufficient for health care users. Therefore, we performed a systematic review to evaluate the efficacy and adverse events associated with preoperative CHO for surgical patients.

Methods

We performed this systematic review of the available literature in accordance with the PRISMA guidelines for the conduct of meta-analyses of randomized controlled trials (RCTs).

Search strategy

Systematic, comprehensive literature searches were conducted of PubMed, the Cochrane Library, EMBASE, ISI Web of Knowledge, China Journal Full-text Database, Chinese Biomedical Database, Chinese Scientific Journals Full-text Database, and CMA digital periodicals. All searches were conducted in May 2010, and updated in September 2010, without language or publication status restrictions. The search terms included “preoperative,” “carbohydrates,” “CHO,” “glucose,” “fasting,” and “surgery”. If possible, subject heading terms, such as Medical Subject Headings terms, were added in all searches. Reference lists from relevant review articles were hand-searched. A search of the ClinicalTrials.gov website was also done to identify RCTs, which had been completed but not published. Requests for original data were made by contacting the authors or principal investigators of the studies. All searches were conducted independently by two reviewers (Lun Li and Tiantian Sun); differences were checked by the other investigator and resolved by discussion.

Inclusion criteria and study selection

Randomized controlled trials were eligible for inclusion if they evaluated the preoperative administration of CHO for surgical patients. The comparisons we evaluated were CHO versus overnight fasting, CHO versus placebo (flavored water), and CHO versus i.v. infusion of glucose. Papers that reported at least one outcome that we were interested in, such as the changes in blood glucose and insulin levels (at the induction of anesthesia, at the end of surgery, on the first day after the operation), the IR index (IRI), the insulin sensitivity index (ISI), gastric pH or volume, the length of hospital or ICU stay, preoperative well-being (anxiety, hunger, thirsty, nausea, or dryness of mouth), postoperative vomiting, and aspiration during surgery, were included. Three reviewers (Zehao Wang, Jinhui Tian, and Kang Yi) independently assessed the potential citations for inclusion, and disagreements were resolved by a fourth reviewer (Kehu Yang).

Data abstraction

Data were abstracted and entered into an Excel database by three authors (Lun Li, Tiantian Sun, and Xiangji Ying). The following fields were extracted: country, patient characteristics (age, sex, etc.), and treatment protocols (details of intervention and comparison(s), sample size, etc.), and outcomes and measured effects. The outcomes were extracted preferentially by the intention-to-treat method. Any disagreements were resolved by a fourth reviewer (Kehu Yang). If data were lacking in the article, we contacted the first author or the corresponding author for further information. If they did not respond within 4 weeks, we extracted as much information as possible from the paper.

Quality and evidence level assessment

The methodological qualities and evidence levels were evaluated by three independent reviewers (Lun Li, Tiantian Sun, and Zehao Wang), and differences were resolved by consultation with a fourth reviewer (Kehu Yang). The following items were assessed according to the Cochrane handbook 5.0 recommended standards [30]: randomization, blinding, concealed allocation, baseline comparability, subject loss to follow-up, intention-to-treat analysis, selective reporting, incomplete outcome data, and other biases. To evaluate the methodological quality for the outcomes evaluated, we used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach [31, 32], which specifies four levels: high, moderate, low, and very low quality evidence.

Data analysis and subgroup analysis

The data were analyzed using the Review Manager Version 5.0 software program. For dichotomous outcomes, the results were expressed as the risk ratio (RR) with the 95 % confidence interval (CI). If there were continuous data, the mean difference (MD), or the standardized mean difference (SMD) if different scales were used, was used to assess the effects of treatment. The data were pooled using the fixed-effects model, but the random-effects model was also considered to ensure the robustness of the model. The percentage of variability across trials attributable to heterogeneity beyond chance was estimated with the I 2 statistic, which was deemed significant for p < 0.05 or for I 2 > 50 %. In case of significant heterogeneity, the results of the random-effects model were noted. The subgroup analyses were also conducted, for example, using different controls (CHO vs. overnight fasting, CHO vs. placebo, CHO vs. i.v. infusion of glucose) without pooling the overall effects. In addition, the subgroup analyses of different surgeries were conducted, such as for patients undergoing laparoscopic cholecystectomy, colorectal surgery, total hip replacement, cardiac surgery, etc.

Results

Search results

After a comprehensive search, we found 2,605 citations. We excluded 486 duplicates and 1,981 citations based on screening the titles and abstracts. After screening full text articles, we excluded 31 citations (not relevant to CHO), 36 citations (not relevant to surgery), 26 citations (not RCTs), 17 citations (traditional review), and 2 which did not have the outcomes we were interested in investigating [3, 4]. Finally, 22 trials (26 citations) [2, 14, 15, 18, 19, 21, 22, 24–27, 33–47] were included for the analyses (Fig. 1).

Fig. 1
figure 1

A flow diagram of the studies

Full size image

Characteristics of included trials

The 22 trials that we analyzed were from Sweden (n = 6), China (n = 4), Denmark (n = 2), the UK (n = 2), Finland (n = 2), the Czech Republic (n = 2), Croatia (n = 1), Germany (n = 1), Turkey (n = 1), and the Netherlands (n = 1). They were about colorectal surgery (n = 7), cholecystectomy (n = 3), orthopedic surgery (n = 5), cardiac surgery (n = 2), and two or more surgery types (n = 5). These studies examined CHO versus placebo (n = 13), CHO versus overnight fasting (n = 15), and CHO versus glucose (i.v.) (n = 3). The timings of CHO administration were 4 h before anesthesia (n = 1), the evening before surgery, and 2–3 h before anesthesia (n = 13), and 2–3 h before anesthesia (n = 8). The number of patients in these trials ranged from 12 to 252. Other information about the trials is presented in Table 1.

Table 1 Characteristics of the included trials
Full size table

Quality assessment

Of these 22 trials, all trials mentioned randomization, 16 trials mentioned blinding, and 12 trials mentioned concealed allocation, but only 10 studies reported the details of randomization, concealed allocation, and blinding. Other information regarding these trials is presented in Table 2.

Table 2 Quality assessment of the included studies
Full size table

Meta-analysis and systematic review (Supplementary Table 1)

The changes in the insulin level

Greater decreases in the insulin level at the induction of anesthesia in the CHO group were observed than in the placebo group (SMD −0.55, 95 % CI −0.86, −0.24), but statistically significant differences in the changes in the insulin levels at the induction of anesthesia were not observed between the CHO group and the overnight fasting group (SMD −0.53, 95 % CI −1.66, 0.59, I 2 = 94 %), or between the CHO group and the i.v. glucose infusion group (SMD −0.29, 95 % CI −0.91, 0.33).

Significant differences in the changes in the insulin levels at the end of surgery were not observed between the CHO and overnight fasting groups (SMD 0.91, 95 % CI −1.55, 3.37, I 2 = 98 %), between the CHO and placebo groups (SMD −0.52, 95 % CI −1.31, 0.28, I 2 = 78 %), or between the CHO and i.v. glucose infusion groups (SMD 2.00, 95 % CI −0.95, 4.96, I 2 = 98 %).

Significantly higher increases in the insulin levels on the first day after surgery in the CHO group were found than those in the overnight fasting group (SMD 0.82, 95 % CI 0.49, 1.16) and the i.v. infusion of glucose group (SMD 0.65, 95 % CI 0.31, 0.98), but no significant difference was observed between the CHO and placebo groups (SMD 0.25, 95 % CI −0.73, 1.23, I 2 = 91 %).

The changes in the glucose level

There were no statistically significant differences in the changes in the glucose levels at the induction of anesthesia in these three comparisons: CHO versus overnight fasting (SMD −0.01, 95 % CI −0.64, 0.63, I 2 = 81 %), CHO versus placebo (SMD −0.27, 95 % CI −0.70, 0.16, I 2 = 78 %), and CHO versus i.v. glucose infusion (SMD 0.22, 95 % CI −0.12, 0.55). Significant differences were also not observed in the changes in the glucose levels at the end of surgery between the CHO and overnight fasting groups (SMD −0.71, 95 % CI −2.44, 1.03, I 2 = 95 %), and between the CHO and i.v. glucose infusion groups (SMD −2.23, 95 % CI −4.51, 0.05, I 2 = 97 %). However, a smaller increase in the glucose level at the end of surgery in the CHO group was seen than that in the placebo group (SMD −1.47, 95 % CI −2.15, −0.79, I 2 = 18 %).

Greater increases in the glucose levels on the first day after surgery were observed in the CHO group than those in the overnight fasting group (SMD 0.77, 95 % CI 0.43, 1.10) and the i.v. infusion of glucose group (SMD 0.79, 95 % CI 0.50, 1.08, I 2 = 0 %), but there was no statistically significant difference in the glucose levels on the first day after the surgery between the CHO and placebo groups (SMD 0.04, 95 % CI −0.28, 0.36).

Postoperative ISI and IRI

A significant difference in the change in the ISI was seen between the CHO and placebo groups (SMD 1.06, 95 % CI 0.32, 1.81), but there were no significant differences in the other two comparisons: CHO versus overnight fasting (SMD 0.34, 95 % CI −0.73, 1.40, I 2 = 93 %), CHO versus i.v. glucose infusion (SMD −0.46, 95 % CI −1.14, 0.21, I 2 = 72 %). There were also no significant differences in the changes in the postoperative IRI between the CHO and overnight fasting groups (SMD −1.02, 95 % CI −2.60, 0.56, I 2 = 86 %) or between the CHO and placebo groups (SMD −0.61, 95 % CI −1.96, 0.75, I 2 = 90 %).

Length of hospital stay and ICU stay

The length of the hospital stay in the CHO group tended to be longer than that in i.v. infusion glucose group (SMD 0.45, 95 % CI 0.12, 0.78), but there were no significant differences between the CHO and overnight fasting groups (SMD −0.06, 95 % CI −0.49, 0.37, I 2 = 65 %) or between the CHO and placebo groups (SMD −0.32, 95 % CI −0.81, 0.17, I 2 = 77 %).

There were also not any statistically significant differences in the length of ICU stay between the CHO and overnight fasting groups (SMD −0.15, 95 % CI −0.44, 0.14, I 2 = 7 %) or between the CHO and placebo groups (SMD 0.22, 95 % CI −0.14, 0.59).

Postoperative gastric pH and residue volume

The gastric pH was not significantly different between the CHO and overnight fasting groups (SMD 0.01, 95 % CI −0.35, 0.36, I 2 = 56 %) or between the CHO and placebo groups (SMD −0.08, 95 % CI −0.37, 0.20, I 2 = 12 %).

There were no significant differences in the gastric residual volume between the CHO and overnight fasting groups (SMD −0.11, 95 % CI −0.36, 0.15, I 2 = 19 %) or between the CHO and placebo groups (SMD −0.03, 95 % CI −0.30, 0.24, I 2 = 4 %).

Postoperative vomiting and aspiration during surgery

There were no significant differences in the incidence of postoperative vomiting between the CHO and overnight fasting groups (RR 0.90, 95 % CI 0.47, 1.72) or between the CHO and placebo groups (RR 1.31, 95 % CI 0.23, 7.45). Six trials [14, 15, 18, 22, 38, 41, 42, 45] mentioned aspiration, and none of them reported any incidents of either vomiting or aspiration (CHO vs. overnight fasting, CHO vs. placebo).

Preoperative well-being

The results regarding preoperative hunger and anxiety were inconsistent for the comparisons between the CHO and overnight fasting groups and between the CHO and placebo groups. Less thirst was observed in the CHO group than that in the overnight fasting group. However, the results about thirst between the CHO group and placebo group were contradictory. No statistically significant differences were found in the rates of nausea and dry mouth between the CHO group and the overnight fasting group or between the CHO group and the placebo group (Supplementary Table 2).

Results of the subgroup analysis of patients undergoing different surgeries

Colorectal surgery (Supplementary Table 3)

Seven RCTs [15, 21, 22, 35, 36, 39, 41–43, 47] supplied enough data to evaluate the effect of CHO on patients undergoing colorectal surgery. In the comparison of CHO with overnight fasting, statistically significant differences were only observed in the insulin level on the first day after surgery (SMD 0.82, 95 % CI 0.49, 1.16), the glucose level on the first day after surgery (SMD 0.77, 95 % CI 0.43, 1.10), the ISI (SMD −0.38, 95 % CI −0.65, −0.11, I 2 = 94 %), and the IRI (SMD −1.82, 95 % CI −2.67, −0.98). However, there were no significant differences in any other outcomes (insulin and glucose levels at the end of surgery, the length of hospital stay, gastric volume, or gastric pH). Less hunger, anxiety, and thirst were observed in both RCTs [15, 39, 41, 42], but preoperative nausea was not reduced [15, 41, 42].

In the comparison of CHO with placebo, a smaller increase in the insulin level (SMD −1.35, 95 % CI −2.12, −0.57) and glucose level (SMD −1.06, 95 % CI −1.81, −0.31) at the end of surgery, a smaller decrease in the ISI (SMD 1.06, 95 % CI 0.32, 1.81) and a smaller increase in the IRI (SMD −1.34, 95 % CI −2.12, −0.57) were seen. However, there were no significant differences in any other outcomes (the length of hospital stay, gastric volume, and gastric pH). CHO could not reduce the preoperative anxiety and nausea in one RCT [43]; however, the results for preoperative thirst were inconsistent in the two RCTs [15, 41–43].

In the comparison of CHO with i.v. infusion glucose, a greater increase in the insulin levels at the end of surgery (SMD 3.51, 95 % CI 2.99, 4.03) and on the first day after surgery (SMD 0.65, 95 % CI 0.31, 0.98) and a longer hospital stay (SMD 0.45, 95 % CI 0.12, 0.78) were observed. However, no significant differences in the other outcomes (insulin and glucose levels at the induction of anesthesia, glucose level at the end of surgery, ISI) were seen.

Laparoscopic cholecystectomy (Supplementary Table 4)

Three RCTs [22, 33, 37] supplied enough data about laparoscopic cholecystectomy for an analysis. However, there were no significant differences in the insulin level at the end of surgery, the IRI, gastric volume, gastric pH, or vomiting in the comparison of CHO with overnight fasting, in the gastric volume, gastric pH and vomiting in the comparison of CHO with placebo, or in the glucose level at the end of surgery in the comparison of CHO with i.v. infusion glucose. Two RCTs [33, 45] reported that CHO could not reduce the preoperative hunger any better than overnight fasting, and could not decrease nausea better than the placebo.

Total hip replacement (Supplementary Table 5)

Three RCTs [19, 44, 46] supplied enough data for an analysis of subjects undergoing total hip replacement. A smaller increase in the glucose level at the end of surgery (SMD −1.98, 95 % CI −2.92, −1.03) was observed in the comparison of CHO with placebo. However, there were not any statistically significant differences in the other outcomes (insulin level at the end of surgery or glucose levels at the induction of anesthesia).

Cardiac surgery

Two RCTs [27, 38] supplied enough data about cardiac surgery for an analysis. However, there were no significant differences in the glucose levels at the induction of anesthesia, or the length of the hospital and ICU stays when CHO was compared with overnight fasting or placebo, nor were there any differences in the incidence of vomiting when CHO was compared with overnight fasting. There were no significant differences in preoperative hunger, nausea, or mouth dryness in the comparisons of CHO with placebo or overnight fasting. Less thirst was observed in the comparison of CHO with overnight fasting [38].

Evidence level assessment

We evaluated the quality of the evidence of the outcomes except for the preoperative well-being. The quality of evidence for aspiration was low, as some important methodological items of the included studies were inadequate, and the sample size of the included studies was small. In the remaining 34 outcomes, only 4 outcomes achieved a moderate evidence level; 13 outcomes had very low quality of the evidence, and the remaining 18 outcomes had a low quality of evidence (Supplementary Table 1).

Discussion

Evidence summary

The present meta-analysis of 13 trials comparing CHO with placebo showed that greater decreases in the insulin level at the induction of anesthesia and smaller increases in the glucose level at the end of surgery were observed in the CHO group than those in the placebo group. Fewer decreases in the postoperative ISI were observed in the CHO group than that in the placebo group. This indicated that CHO may attenuate the immediate postoperative IR as compared to the placebo. However, the number of included studies was too small to draw a definitive conclusion. In addition, the quality of the evidence for most outcomes was relatively low, so further research is likely to have an impact on our confidence in the estimate of the effects of CHO.

The meta-analysis of 15 trials comparing CHO with overnight fasting showed that CHO increased the insulin and glucose levels on the first day after surgery more than the levels observed in the overnight fasting group. However, our results were inconsistent with the results of a primary study [39], as the baseline insulin levels between the CHO and overnight fasting groups differed. In addition, only one study with 149 patients was evaluated for this outcome, and the quality of evidence for these two outcomes was low. As a result, the differences from the available evidence were not affirmative.

The meta-analysis of the three trials of CHO as compared to i.v. glucose infusion showed that there were increased insulin and glucose levels on the first day after surgery and a longer hospital stay in the CHO group than those in the i.v. glucose infusion group. However, the baseline insulin levels differed between the CHO and i.v. glucose infusion groups in the previously published study [39]. Moreover, the numbers of included studies were small for these two outcomes, and the quality of the evidence for these two outcomes was low, so we were not confident about the results.

With regard to the patients’ preoperative well-being, CHO could reduce preoperative thirst as compared to overnight fasting, but not as compared to placebo. However, CHO could not decrease the incidence of preoperative nausea or mouth dryness as compared to either overnight fasting or placebo. In terms of the preoperative hunger and anxiety, the results of the included studies were inconsistent. However, the data about the preoperative patient well-being, from the included studies, were not pooled to produce a more conclusive result, as the data from all relevant studies were not presented in a format that allowed pooling. Nevertheless, the available evidence suggested that the preoperative well-being might be improved after CHO intake to at least some degree.

No aspiration was observed in any of the studies that reported this outcome. The incidence of postoperative vomiting was not significantly different between the CHO and overnight fasting groups, or between the CHO and placebo groups, which meant that CHO was safe based on the available evidence. However, the sample sizes of the included studies were not large enough, and the quality of evidence was low, so caution must be exercised when CHO is used.

Among the outcomes we evaluated, CHO did not show better benefits in patients undergoing laparoscopic cholecystectomy as compared to overnight fasting, placebo, or i.v. infusion of glucose, or in patients undergoing cardiac surgery as compared to overnight fasting or placebo (except that less thirst was observed in the comparison of CHO with overnight fasting). A smaller increase in the glucose levels at the end of surgery was observed in the comparison of CHO with placebo in patients undergoing total hip replacement. For those undergoing colorectal surgery, a greater increase in the insulin and glucose levels on the first day after surgery and the glucose levels at the induction of anesthesia, as well as less IRI, hunger, anxiety, and thirst were observed in the comparison of CHO with overnight fasting. A smaller increase in the insulin and glucose levels at the end of surgery, a smaller decrease in the ISI, and a lower increase in the IRI were seen in the comparison of CHO with placebo. A greater increase in the insulin levels at the end of surgery and on the first day after surgery, and a longer hospital stay, were observed in the comparison of CHO with i.v. infusion glucose.

Overall, the 15 trials comparing CHO with overnight fasting showed that CHO increased the insulin and glucose levels on the first day after surgery as compared to the overnight fasting group, and with less preoperative thirst. Thirteen trials comparing CHO with placebo showed that greater decreases in the insulin level at the induction of anesthesia and a smaller increase in the glucose level at the end of surgery were found in the CHO group as compared to the placebo group. Fewer decreases in the postoperative ISI were observed in the CHO group than in the placebo group. In addition, CHO was associated with more increased insulin and glucose levels on the first day after surgery and a longer hospital stay than were noted in the i.v. glucose infusion group. No aspiration was observed in any of the included studies.

With regard to the different types of surgeries, the effects of CHO were different. CHO seemed to decrease the IRI as compared to overnight fasting, as well as increasing the ISI and decreasing the IRI as compared to placebo treatment in patients undergoing colorectal surgery. No significant effects were observed for any of the other surgeries evaluated. However, the number of studies included for the analysis of each type of surgery was limited, so we could not draw definitive conclusions.

In addition, the heterogeneity across the included studies was significant. There were likely two main reasons for this: first, the administration dose or timing of CHO was not the same across the included studies. There were two different administration methods: once or twice before surgery. The administration timings of the two doses of CHO protocol were the evening before surgery (first) and 2–4 h before anesthesia (second), while the administration time for the single CHO protocol was 2–4 h before anesthesia. These different numbers and times of CHO could be one source of the high heterogeneity. Second, many different types of surgery were performed. The types of surgery included laparoscopic cholecystectomy, colorectal surgery, total hip replacement, cardiac surgery, etc. Although we conducted the subgroup analyses and calculated the changes in the outcomes after the intake of CHO between two groups, heterogeneity still existed. As a result, the different surgeries could be another source of the high heterogeneity. Therefore, there was a high degree of inconsistency in the results, and when our results are applied to clinical practice, additional attention should be paid, and patients should be treated cautiously.

Study quality and evidence levels

The available evidence showed that inadequate random allocation or concealed allocation would result in overestimates of the effect [48]. However, 10 of the 22 included trials reported the details of randomization, concealed allocation, and blinding. It is well known that the subjective outcomes or patient-reported outcomes need to be blinded to patients, as a lack of blinding in randomized trials can be associated with more exaggerated estimated intervention effects [30, 49, 50]. However, the assessment of the blinding of these nine studies [2, 14, 15, 22, 33, 34, 38, 39, 41–43, 45], which reported preoperative well-being, was not satisfactory. Six of the nine studies mentioned blinding, but only five of the nine were blinded to patients. The overall quality of these trials focusing on CHO for elective surgery were poor, and the pooled results of these trials need to be cautiously interpreted when applying them to clinical practice.

The number of patients included for each outcome was small, and inconsistencies existed across trials, so a high level of evidence was not produced by these trials. As a result, it is necessary to be prudent about using the results from these studies to recommend new strategies for clinical practice.

Study limitations

Our present study aimed to overview the efficacy and side effects of CHO for elective surgery, so we searched all RCTs that compared CHO with overnight fasting, placebo or i.v. glucose. This is the first article about this topic that performed a comprehensive search of all of the relevant medical electronic databases and ClinicalTrials.gov, and that tracked reference lists from review articles, and where the authors of the original articles were contacted. Three of our authors independently screened all of the trials, extracted data, and assessed the methodological quality and quality of evidence in order to avoid selection and performance biases.

Nevertheless, this paper still has several limitations. First, we just searched Chinese and English databases, while the databases in other languages were not searched, so studies that were published in other languages could not be included. Second, we did not consider the administration time, dose and amount of CHO, although these may influence gastric emptying. Therefore, the heterogeneity of the pooled results was high for each outcome. Third, a possible criticism of this systematic review is that it was based on trials of small sample sizes, and almost none of them reported sample calculations. Therefore, in the future, a sample calculation should be conducted before beginning a RCT to make it possible to achieve a sufficiently powerful analysis.

Implications for practice and research

A traditional review, which was published in 2009 [9], summarized the present understanding of the mechanisms underlying the positive clinical effects and finally stated that preoperative CHO loading was recommended before a major surgery. Despite the fact that preoperative CHO has some positive effects and has even been recommended in several countries, preoperative overnight fasting is still the standard of care in many hospitals [14]. This is because most patients and clinicians are afraid of the increased risk of aspiration. The results from our meta-analysis showed that no incidents of aspiration were observed in the six reports [14, 15, 18, 22, 38, 41, 42, 45] that mentioned aspiration. This is consistent with the results of a Cochrane systematic review [23], which showed that a shortened fluid fast was not associated with an increased risk of aspiration as compared with the standard “nil by mouth from midnight” fasting policy. However, the quality of the evidence for this outcome was low, as some important methodological items were inadequate, and the sample size was too small to accurately judge whether aspiration will occur when the sample size is increased or when the method is applied to more patients. Preoperative oral consumption of 800 ml of a carbohydrate-rich drink the evening before surgery, and 400 ml 2 h before anesthesia for elective surgery was recommended in the UK [51], and also in Germany [52], and in Scandinavia [53] during the past several years, but preoperative oral CHO for elective surgery did not appear to provide any major benefit based on the results of our present systematic review. Our study demonstrated that CHO might attenuate the postoperative IR as compared to placebo, and for colorectal surgery, the postoperative IR was decreased for the CHO group as compared with patients who were treated following overnight fasting or the administration of a placebo. However, the recommendations for applying these results in practice should be made cautiously, as poor methodological quality and a low quality of evidence were generally present for these studies. In addition, clinicians should keep in mind that preoperative oral CHO could not be used in patients with a hiatus hernia, known upper gastrointestinal tumor, slow gastric emptying, or diabetes [12].

The number of patients included in these trials was very small. Although a difference in treatment can be seen with a small sample (but with sufficient power analysis), we believe that samples should be large enough to detect possible differences, reasonable enough to be feasible, and small enough to detect efficient therapies [37, 54]. The methodological quality of the included trials was not good enough, which may have introduced some potential bias into the pooled results. Therefore, further RCTs should be designed using more rigorous methods to avoid potential biases and make the evidence more reliable.

Conclusion

Carbohydrates may be safe and attenuate postoperative insulin resistance better than placebo. However, the quality of the trials included in this analysis was poor, and the evidence levels for most outcomes were low. Therefore, caution should be exercised when applying these results for clinical practice. In addition, further, better-designed RCTs should be performed with rigorous methods to avoid potential biases and to make the evidence more reliable.