Length of hospital stay
We considered the LOS as our primary outcome, as in our opinion this may represent the overall effects of preoperative CHO-loading best, due to the potential beneficial effects on several important confounders such as better postoperative blood sugar control, faster gastro-intestinal recovery, lesser chance of nosocomial infections and also the reduction of hospital costs (41). Other safety outcomes were not assessed in this study.
We found that LOS of those patients who received CHO loading was significantly shorter by a mean of 1.7 [-3.04;-0.38] days when compared to preoperative fasting that should be considered as a clinically relevant difference. Although difficult to explain, but some authors believe that CHO-loading could result in a more balanced glucose homeostasis that is the main reason for reduced LOS (15, 21). However, our data did not provide satisfactory support to this hypothesis. More detailed assessment of glucose homeostasis should be considered in the future.
Liu et al., found that the loss of muscle mass was lower in the CHO-group, which was considered responsible for the faster recovery (18). Unfortunately, based on the available data on postoperative blood sugar and insulin levels, but also on CRP levels – as the only reported and poolable marker of postoperative inflammation – we found no significant differences between the groups.
Hospital LOS is influenced by multiple factors, such as adequate pain and fluid management, early mobilization, earlier start of oral food and fluid intake, etc., not analyzed in the current meta-analysis as data was either scarce or not reported at all (16, 18, 19, 20, 24). It is also important to note that there is a difference in the terms of ‘length of hospital stay’, ‘total lengths of hospital stay’ (18) and ‘readiness to discharge’ (26) that was also inconsistently defined in the articles. LOS may also be affected by local organizational issues such as timing of surgery after hospital admission and the timing of patient discharge. Therefore, we entirely agree with Cho et al (26), who suggested to use the term, “time to ready to discharge” in the future as an outcome measure instead of the actual LOS.
There was no significant difference in the hospital LOS when CHO-loading was compared to placebo. In fact, the pooled results were almost identical with a very narrow confidence interval (0.01 [-0.18;0.16]). Regarding the type of the placebo in general, plain water is one option (19, 30, 31, 38, 42, 43, 44), but in this case, patients could be able to identify the difference between the drug and placebo by its sweetness, hence blinding is questionable. An alternative option is the use of artificial sweeteners (27, 29, 33, 34, 45, 46, 47, 48, but in this case, one cannot exclude their metabolic effects (19). In our meta-analysis, the used placebos differed greatly in the studies. In some cases, the authors used plain water (31, 32), while Yuill et al. (37) used fluid with minerals that was not flavored either, so double-blindness cannot be guaranteed which is a biasing factor all in general. Nevertheless, in our opinion, this is the most suitable placebo for investigating the effects of carbohydrates in a placebo-controlled trial design. Although it remains difficult to give an exact explanation of the difference between the fasting and placebo comparisons, but one cannot exclude that it is not necessarily the CHO-loading but the fasting that is responsible for the worse outcome. A hypothesis that is worth further investigation in the future.
Postoperative blood sugar and insulin levels
Postoperative hypo- and hyperglycemia should be avoided due to several potential adverse effects both surgical a general that may delay recovery in general after surgery (2). In the current study we found that the postoperative metabolic status as indicated by serum glucose and insulin levels did not differ significantly between the groups and showed similar tendency in both comparisons to fasting or placebo.
When CHO-loading was compared to fasting, there was a minimal mean difference on postoperative day 1(-0.01 [-0.49; 0.46]) and on postoperative day 2 (0.21[-0.30;0.72]) between the groups in the blood glucose levels, but patients in the CHO-group had a tendency of lower serum insulin levels, but the difference did not reach statistical difference. When compared CHO-loading to placebo, the results were similar, with a tendency of favoring CHO-loading, but without statistical significance. However, we were only able to pool data from 3 studies for each comparison, with a high heterogeneity.
It is important to note that the used amount, composition, and timing of administration of CHO-loading fluids notably differed in the included studies. Another reason for heterogeneity is that several studies used different timepoints for measuring insulin and glucose levels. Some even recorded these parameters on postoperative day 3 (22) and 7 (22, 49). We believe that expecting such a long-term effect of a relatively low dose CHO-loading preoperatively do not really make sense. Even measuring blood sugar levels on day 1 and 2 is questionable, especially if patients are fed already on day one. As for future research, we believe that timing of postoperative glucose and insulin level measurements should be standardized. A possible approach could be one that was applied by Rizavnovic et al, where they measured glucose levels two hours prior surgery, six hours after surgery, and in the mornings of postoperative day 1 and 2 (35).
C-reactive protein
CRP is the most commonly used inflammatory biomarker that may also be used for assessing postoperative inflammatory response (41). In both comparisons (to fasting or placebo) there was a similar tendency of lower CRP levels in the CHO-group, but it did not reach statistical significance. How CHO loading could affect postoperative inflammatory response remains uncertain and detailed description of the current hypotheses are beyond the scope of this manuscript.
Strengths and limitation
The main strength of our analysis is that we strictly followed our protocol, which was registered in advance. Both univariate and multivariate analyses were performed to achieve objectivity. Rigorous methodology was applied, and our results were derived from randomized controlled trials only. However, because of the different methodology that was used in most studies, we were unable to pool all 55 identified studies that were eligible for data extraction, and only 26 (15-40) of them could be used for pooled analyses. The other 29 studies could not be included in any analysis (41-69) (Supplemetary Table).
High heterogeneity of data is another limiting factor. We worked with very heterogenous patient population, who were undergoing different types and durations of surgeries. The definition of fasting and placebo also differed to some extent among the studies. The administered carbohydrate fluids varied in contents and amount. The used randomized control trials measured glucose and insulin levels in inappropriate times. Also, the different measuring timepoints seriously limited our study.
The main message of this meta-analysis is that CHO-loading as compared to preoperative fasting resulted in significantly shorter length of hospital stay. However, when compared to placebo, this difference disappeared. Based on the non-significant findings in all other outcomes, explanation of these result remains difficult.
Implication for practice and research
Although our results suggest that preoperative fluid intake with or without CHO is superior to fasting, but the pathophysiology remains unclear. To find out the exact benefits and further details of this intervention renders the need for future research in which circumstances such as timing, dosing, type of fluid, etc. are much more standardized. More information is also needed on glucose homeostasis in order to enable us to determine whether the observed effects are related to the fluid (i.e.: water) intake per se, or its carbohydrate content. Regarding outcomes, we suggest the term “ready to discharge” instead of “length of hospital stay” as some confounding circumstances (i.er.: social background, bed availability, etc.), may affect the time of discharge. Finally, we recognized that postoperative complications are rarely used as an outcome that should also be included as important endpoints on future trials. We would like to highlight the importance of implementing scientific results into daily practice, because with this approach may allow faster recovery and lesser incidence of postoperative complications (70, 71).
As compared to the used placebos, such as clear water, water with artificial sweeteners and preformed glucose-free solutions with mineral, we could not find significant differences.