Supplemental perioperative intravenous crystalloids for postoperative nausea and vomiting
Abstract
Background
Postoperative nausea and vomiting (PONV) is a common complication following general anaesthesia. It may be associated with patient dissatisfaction, increased costs of treatment, and unintended admission to hospital.
Supplemental intravenous crystalloid administration in the perioperative period may be a simple intervention to prevent PONV.
Objectives
To assess whether supplemental intravenous crystalloid administration prevents PONV in patients undergoing surgical procedures under general anaesthesia.
Search methods
We searched the Cochrane Central Register of Controlled Trials (CENTRAL; 2018, Issue 7), MEDLINE (1946 to August 2018), Embase (1947 to August 2018), and the Cumulative Index of Nursing and Allied Health Literature (CINAHL; 1971 to August 2018). We searched clinical trials registers for ongoing or unpublished completed studies (August 2018), handsearched conference proceedings of anaesthesiology societies, as published in three major journals (British Journal of Anaesthesia, European Journal of Anaesthesiology, and Anesthesiology; August 2018), and conducted backward and forward citation searching of relevant articles.
Selection criteria
We included randomized controlled trials of participants older than six months undergoing surgical procedures under general anaesthesia and given supplemental perioperative intravenous crystalloids, defined as a volume larger than that received by a comparator group, to prevent PONV.
Data collection and analysis
We used the standard methodological procedures described by Cochrane.
Main results
We included 41 studies (4224 participants). Participants underwent ambulatory or short length of stay surgical procedures, and were predominantly American Society of Anesthesiology (ASA) class I or II. There is one study awaiting classification and three ongoing studies. All studies took place in surgical centres, and were conducted in geographically diverse settings. Risk of bias was generally unclear across all domains.
Supplemental intravenous crystalloid administration probably reduces the cumulative risk of postoperative nausea (PON) (risk ratio (RR) 0.62, 95% confidence interval (CI) 0.51 to 0.75; 18 studies; 1766 participants; moderate‐certainty evidence). When the postoperative period was divided into early (first six hours postoperatively) and late (at the time point closest to or including 24 hours postoperatively) time points, the intervention reduced the risk of early PON (RR 0.67, 95% CI 0.58 to 0.78; 20 studies; 2310 participants; moderate‐certainty evidence) and late PON (RR 0.47, 95% CI 0.32 to 0.69; 17 studies; 1682 participants; moderate‐certainty evidence).
Supplemental intravenous crystalloid administration probably reduces the risk of postoperative vomiting (POV) (RR 0.50, 95% CI 0.40 to 0.63; 20 studies; 1970 participants; moderate‐certainty evidence). The intervention specifically reduced both early POV (RR 0.56, 95% CI 0.41 to 0.76; 19 studies; 1998 participants; moderate‐certainty evidence) and late POV (RR 0.48, 95% CI 0.29 to 0.79; 15 studies; 1403 participants; moderate‐certainty evidence).
Supplemental intravenous crystalloid administration probably reduces the need for pharmacologic treatment of PONV (RR 0.62, 95% CI 0.51 to 0.76; 23 studies; 2416 participants; moderate‐certainty evidence).
The effect of supplemental intravenous crystalloid administration on the risk of unplanned postoperative admission to hospital is unclear (RR 1.05, 95% CI 0.77 to 1.43; 3 studies; 235 participants; low‐certainty evidence).
No studies reported serious adverse events that may occur following supplemental perioperative intravenous crystalloid administration (i.e. admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death).
Authors' conclusions
There is moderate‐certainty evidence that supplemental perioperative intravenous crystalloid administration reduces PON and POV, in ASA class I to II patients receiving general anaesthesia for ambulatory or short length of stay surgical procedures. The intervention probably also reduces the risk of pharmacologic treatment for PONV. The effect of the intervention on the risk of unintended postoperative admission to hospital is unclear. The risk of serious adverse events resulting from supplemental perioperative intravenous crystalloid administration is unknown as no studies reported this outcome. The one study awaiting classification may alter the conclusions of the review once assessed.
PICOs
Plain language summary
Extra intravenous fluid given during surgery to prevent nausea and vomiting
Review question
This review looks at whether giving extra intravenous fluid to people during general anaesthesia prevents nausea and vomiting after their surgery is done.
Background
Nausea and vomiting is a common complication after having general anaesthetic for surgery. About 30% of people suffer from nausea and vomiting after surgery, even after receiving medication intended to prevent it.
During surgery, a patient receives salt‐containing fluid through an intravenous drip and the amount of fluid given may affect how they feel afterwards. Some complications, like nausea and vomiting, may be reduced after getting extra intravenous fluid during surgery. Some complications, like shortness of breath, may be worse with extra fluid.
Search date
The search was up‐to‐date as of August 2018.
Study characteristics
We looked at studies where people had general anaesthesia for surgery, and received larger or smaller amounts of intravenous fluid, and were later checked to see if they developed nausea and vomiting after their surgeries were done. We found 41 studies, with 4224 participants analysed in our review.
Key results
Our review suggests that giving people extra intravenous fluid during surgery under general anaesthesia probably decreases the risk of having either nausea or vomiting after surgery, and probably reduces the need for medication to treat nausea.
It is unclear how giving extra intravenous fluid affects the risk of unexpectedly needing hospital admission after minor surgery. No studies looked at whether extra intravenous fluid makes other complications worse.
Certainty of the evidence
There are two reasons why the conclusions of this review may not be exactly correct. First, many of the studies were not designed perfectly. Second, the studies did not agree on exactly how helpful the extra intravenous fluids were for preventing nausea and vomiting. Most studies did find it at least somewhat helpful.
Authors' conclusions
Summary of findings
| Supplemental IV crystalloid compared to comparator IV crystalloid volume for preventing postoperative nausea and vomiting. | |||||
| Patient or population: participants aged 6 months or older undergoing surgical procedures under general anaesthesia | |||||
| Outcome | Illustrative comparative risks* (95% CI) | Relative effect | № of participants | Certainty of the evidence | |
| Assumed risk | Corresponding risk | ||||
| Risk with comparator IV crystalloid | Risk with supplemental IV crystalloid | ||||
| Risk of PON, defined as the presence of subjective nausea, reported dichotomously or based on a study‐defined dichotomous threshold on a continuous scale such as a VAS | Cumulative events, explicitly reported for the entire study period | RR 0.62 | 1766 | ⊕⊕⊕⊝ | |
| 482 per 1000 | 183 fewer per 1000 (120 to 236 fewer) | ||||
| Early events, occurring in the first 6 hours postoperatively | RR 0.67 (0.58 to 0.78) | 2310 (20 RCTs) | ⊕⊕⊕⊝ | ||
| 307 per 1000 | 101 fewer per 1000 (68 to 129 fewer) | ||||
| Late events, occurring at the time point closest to or including 24 hours postoperatively | RR 0.47 (0.32 to 0.69) | 1682 (17 RCTs) | ⊕⊕⊕⊝ | ||
| 187 per 1000 | 99 fewer per 1000 (58 to 127 fewer) | ||||
| Risk of POV, reported dichotomously by any discrete episodes of vomiting | Cumulative events, explicitly reported for the entire study period | RR 0.50 | 1970 | ⊕⊕⊕⊝ | |
| 295 per 1000 | 147 fewer per 1000 (109 to 177 fewer) | ||||
| Early events, occurring in the first 6 hours postoperatively | RR 0.56 (0.41 to 0.76) | 1998 (19 RCTs) | ⊕⊕⊕⊝ | ||
| 106 per 1000 | 47 fewer per 1000 (25 to 63 fewer) | ||||
| Late events, occurring at the time point closest to or including 24 hours postoperatively | RR 0.48 (0.29 to 0.79) | 1403 (15 RCTs) | ⊕⊕⊕⊝ | ||
| 68 per 1000 | 35 fewer per 1000 (14 to 48 fewer) | ||||
| Risk of requiring pharmacologic treatment for PONV, reported dichotomously as the use of any medication intended to treat nausea or vomiting during the postoperative period | Cumulative events, explicitly reported for the entire study period | RR 0.62 | 2416 | ⊕⊕⊕⊝ | |
| 284 per 1000 | 108 fewer per 1000 (68 to 139 fewer) | ||||
| Risk of unintended postoperative admission to hospital, reported dichotomously as admission to an inpatient unit of a participant after an intended ambulatory surgical procedure | Cumulative events, explicitly reported for the entire study period | RR 1.05 | 235 | ⊕⊕⊝⊝ | |
| 288 per 1000 | 14 more per 1000 (66 fewer to 124 more) | ||||
| Risk of suffering a serious adverse event, reported dichotomously as the occurrence of any of: admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death | Cumulative events, explicitly reported for the entire study period | ‐ | This outcome was not reported for included trials | ‐ | |
| ‐ | ‐ | ||||
| * For all outcomes, the assumed and corresponding risks (and their 95% CI) are based on the proportion of events in the comparator and intervention groups, respectively. CI: confidence interval; PON: postoperative nausea;POV: postoperative vomiting; RR: risk ratio; VAS: visual analogue scale | |||||
| GRADE Working Group grades of evidence | |||||
| 1Downgraded one level due to risk of publication bias, following inspection of funnel plot. 2Downgraded two levels due to imprecision and inconsistency. | |||||
Background
Description of the condition
Postoperative nausea and vomiting (PONV) is a common and dreaded complication following anaesthesia. In the absence of risk factors, the baseline risk of PONV is 10%. The presence of female gender, history of motion sickness or PONV, non‐smoking status, and the use of postoperative opioids increase PONV risk to as high as 79% (Apfel 1999). Even with the administration of prophylactic antiemetic medications, the risk of PONV can still be approximately 30% (Habib 2006; Watcha 1992). Complications after surgery lead to patient dissatisfaction and it has been shown that patients rank PONV a highly undesirable complication, wishing to avoid it even more than postoperative pain (Eberhart 2002; Macario 1999; Myles 2000). PONV is so distressing that patients are willing to pay out of pocket to prevent its occurrence (Gan 2001).
Although not usually life‐threatening, PONV may lead to complications commonly associated with vomiting, including dehydration, electrolyte imbalance, and aspiration of gastric contents. In some surgical cases, PONV has also led to: wound complications, oesophageal rupture, subcutaneous emphysema, pneumomediastinum, and bilateral pneumothoraces (Atallah 2004; Bremner 1993; Schumann 1999; Temes 1999; Thompson 1978). PONV is among the most frequently observed complications in the postanaesthesia care unit (PACU), and its presence correlates strongly with delayed discharge from the PACU, unanticipated admission following ambulatory surgery, and increased costs (Gold 1989; Parra‐Sanchez 2012; Wetchler 1992). Therefore, PONV leads not only to patient dissatisfaction, but also increased costs related to length of hospital stay.
PONV hinders patient mobilization and delays resumption of oral intake of food, fluids, and medications. Therefore its prevention is typically included in early recovery after surgery programmes (Mortensen 2014). Preventing PONV positively impacts patient satisfaction, surgical outcomes, and resource utilization.
There are numerous prophylactic treatments for PONV. For instance, ondansetron 4 mg intravenously (0.1 mg/kg in children) is a commonly used pharmacologic antiemetic. Dexamethasone 4 mg to 10 mg (0.1 mg/kg in children) has also been demonstrated to have antiemetic properties, and is safe to use in a surgical population (De Oliviera 2013; Polderman 2018). Other medications shown to prevent PONV in meta‐analysis include tropisetron, dolasetron, cyclizine, granisetron, and droperidol. Droperidol is rarely used due to its association with QT prolongation, and related US Food and Drug Administration (FDA) black box warning (Guy 1991; McCormick 2002). Pharmacologic interventions are often used in combination (Apfel 2004), and multimodal prophylaxis is recommended in patients predicted to be at high risk of PONV (Gan 2014). An upcoming Cochrane Review will examine the use of pharmacologic prophylaxis for PONV (Weibel 2017).
Anaesthetic technique also influences the risk of PONV. The use of volatile anaesthetics increases the risk of PONV, while the use of regional anaesthesia, or total intravenous anaesthesia with propofol, is comparatively protective against nausea and vomiting (Borgeat 2003; Scuderi 2000). Administration of intravenous dextrose‐containing solutions may also prevent PONV (Dabubondoc 2013)
There are non‐pharmacologic approaches to PONV prevention as well. Acupuncture, specifically acustimulation of the P6 acupoint, reduces PONV by 30% when used in combination with ondansetron 4 mg intravenous in comparison to ondansetron alone (Lee 2015). Additionally, inhalation of isopropyl alcohol vapours has been demonstrated to reduce requirements for rescue anti‐emetic medications, when compared to saline placebo (Hines 2018).
Description of the intervention
This Cochrane systematic review examined the effect of supplemental perioperative intravenous crystalloid administration on PONV.
Intravenous crystalloids are widely administered before, during, and after procedures requiring general anaesthesia. They are inexpensive and have relatively few adverse effects. A prior systematic review has suggested that supplemental intravenous crystalloids may be effective in preventing PONV (Apfel 2012). These authors noted that intravenous administration of 15 mL to 30 mL/kg may generally be regarded as a substantial supplemental volume, whereas 0 mL to 3 mL/kg might be considered "restrictive". However, studies of supplemental perioperative intravenous crystalloids were noted to vary widely on the specific volumes administered.
How the intervention might work
Investigation of the effect of perioperative intravenous crystalloid administration on PONV was initially motivated by the results of observational studies suggesting that perioperative volume status influenced postoperative complication rates (Dawson 1980; Fahy 1969; Shires 1961). This work showed that PONV was among the most prevalent events after surgery and motivated subsequent inquiry into the relationship between perioperative volume resuscitation and PONV (Keane 1986; Spencer 1988; Yogendran 1995).
Multiple reviews have explained the complex physiology of nausea and vomiting (Blackburn 2015; Borison 1953; Palazzo 1984). Briefly, the vomiting centre, located in the lateral reticular formation of the medulla, co‐ordinates efferent activity to the respiratory, gastrointestinal, and abdominal musculature to produce vomiting. This centre receives afferent stimuli from a variety of sites: the pharynx, gastrointestinal tract chemo‐ and stretch receptors, the brain (including vestibular information from cranial nerve VIII), aortic baroreceptors, and the chemoreceptor trigger zone. The chemoreceptor trigger zone is a neural centre physiologically outside of the blood‐brain barrier, which provides afferent information to the vomiting centre in response to noxious stimuli in the blood.
Patients typically present for surgery with a fluid deficit secondary to fasting, bleeding, bowel preparation, and other causes of dehydration. Preoperative orthostatic hypotension is associated with PONV, and preoperative volume expansion reduces intraoperative gut hypoperfusion. It has been proposed that brainstem, vestibular, and intestinal hypoperfusion, with concomitant ischaemia, may mediate nausea and vomiting (Gan 1997; Pusch 2002a; Pusch 2002b). Supplemental intravenous crystalloids could serve to mitigate this effect; however, no proven explanation for the putative role of volume status in this model exists.
A prior Cochrane Review found that buffered intravenous solutions were not superior to non‐buffered intravenous solutions for preventing postoperative vomiting (Bampoe 2017).
Why it is important to do this review
Despite evidence‐based, multimodal prophylactic regimens, PONV remains a prevalent clinical problem (Gan 2007). The use of pharmacologic agents alone reduces the risk of PONV but increases the risk of side effects (Alkaissi 2004). Intravenous crystalloids are an attractive treatment modality because they are relatively inexpensive and have few side effects. Many different intravenous fluid interventions have been tested in a wide variety of surgical and anaesthetic contexts. Not surprisingly, results have been conflicting. Previous reviewers have suggested the presence of reporting bias in the early literature (Apfel 2012). Moreover, there is a lack of consensus regarding the volume of supplemental intravenous crystalloid required for PONV prophylaxis, or the ideal timing of its administration.
We conducted this systematic review to consolidate knowledge from the existing literature, so that perioperative clinicians may be provided a comprehensive assessment of the influence of intravenous crystalloid supplementation on the risk of PONV.
Objectives
To assess whether supplemental intravenous crystalloid administration reduces PONV in patients undergoing surgical procedures under general anaesthesia.
Methods
Criteria for considering studies for this review
Types of studies
We included all randomized controlled trials (RCTs) that evaluated the effect of supplemental perioperative intravenous crystalloid administration for the prevention of PONV.
We did not exclude any study based on language of publication or publication status.
Types of participants
We included participants older than six months, undergoing any type of surgical procedure performed under general anaesthesia. For subgroup and sensitivity analyses, we defined children as six months to 17 years, and adults as 18 years or older.
Types of interventions
We included studies that examined supplemental perioperative intravenous crystalloid administration. Given the lack of agreement in the literature on specific volumes administered, we defined the intervention as an intravenous crystalloid volume larger than that received by a comparator group. The comparator is defined as an intravenous crystalloid volume smaller than that received by an intervention group, and we also included studies in which the comparator received no supplemental perioperative intravenous crystalloid. We included studies regardless of the timing of administration, including preoperative, intraoperative, postoperative, or a combination of these. Timing of administration was classified by the point at which administration was initiated.
We also included studies that administered dextrose‐containing crystalloids, but since intravenous dextrose may independently reduce PONV (Dabubondoc 2013), we conducted sensitivity analyses to ensure that the inclusion of these studies did not influence our overall meta‐analyses.
We excluded non‐intravenous routes of crystalloid administration (i.e. oral).
We excluded studies that compared only supplemental intravenous colloids to a comparator. However, we included studies including both colloids and crystalloids, as long as they had an intervention group receiving only supplemental crystalloid, in a volume greater than that received by a comparator group that also received only crystalloid.
Types of outcome measures
The following outcomes were subject to meta‐analysis.
Primary outcomes
-
Risk of PON, defined as the presence of subjective nausea, reported dichotomously or based on a study‐defined dichotomous threshold on a continuous scale such as a visual analogue scale (VAS).
-
Risk of POV, reported dichotomously by any discrete episodes of vomiting.
Secondary outcomes
-
Risk of requiring pharmacologic treatment for PONV, reported dichotomously as the use of any medication intended to treat nausea or vomiting during the postoperative period.
-
Risk of unintended postoperative admission to hospital, reported dichotomously as admission of a participant to an inpatient unit after an intended ambulatory surgical procedure.
-
Risk of suffering a serious adverse event, reported dichotomously as the occurrence of any of: admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death.
The risk of PONV was reported in a minority of studies. PONV was also inconsistently defined across studies, casting doubt on the meaningfulness of analysing this nebulous outcome. As such, we opted to focus on analysis of PON and POV, which study investigators defined in a more consistent manner.
For the risk of PON, when continuous data were reported (e.g. using a visual analogue scale), we analysed these separately from dichotomous data in order to better characterize the magnitude of effect.
We measured POV dichotomously, based on the presence or absence of vomiting during the postoperative period. Some studies presented retching, the production of emetogenic movements without the expulsion of gastric contents, on its own or with vomiting. We combined retching and vomiting data when it would clearly not cause a unit of analysis error.
Although our analyses focused on the risk of these outcomes over the cumulative study period, when the data were available we also analysed the risk of these outcomes occurring at different postoperative time points postoperatively (i.e. early, late). In accordance with prior reviews on this topic, the early postoperative period was defined as the highest incidence of PON or POV within six hours after surgery, while the late postoperative period was defined as the time period reporting PON or POV nearest to 24 hours after surgery (Apfel 2012).
Search methods for identification of studies
Electronic searches
We identified RCTs through literature searching with systematic and sensitive search strategies as outlined in Chapter 6.4 of the Cochrane Handbook of Systematic Reviews of Interventions (Lefebvre 2011). We did not apply restrictions to language or publication status.
We searched the following databases for relevant trials (August 2018).
-
Cochrane Central Register of Controlled Trials (CENTRAL; 2018, Issue 7).
-
MEDLINE (1946 to August 2018).
-
Embase (1947 to August 2018).
-
Cumulative Index of Nursing and Allied Health (CINAHL; 1971 to August 2018).
We developed the initial search strategy using MEDLINE, identifying relevant index terms and the keywords to cover the concepts of the perioperative period, nausea and vomiting, intravenous administration, and crystalloid fluids. This search strategy was then adapted to the other electronic databases. All search strategies can be found in Appendix 1.
We scanned the following trials registries for ongoing and unpublished trials (August 2018).
-
The World Health Organization International Clinical Trials Registry Platform (WHO ICTRP).
-
ClinicalTrials.gov (clinicaltrials.gov).
Searching other resources
We scanned the reference lists and citations of included trials and any relevant systematic reviews identified for further references to additional trials. When necessary we attempted to contact trial authors for additional information.
We screened conference proceedings of anaesthesiology societies, published in three major anaesthesiology journals, from the two preceding years: British Journal of Anaesthesia, European Journal of Anaesthesiology, and Anesthesiology. This search was completed on 4 August 2018.
Data collection and analysis
Selection of studies
We merged search results using Covidence, and removed duplicated records. Two review authors (KJ, MW) read titles and abstracts and removed obviously irrelevant reports. We resolved discrepancies in the title and abstract screening by discussion with two other review authors (RG, SB). Two authors (KJ, MW) retrieved the full text of the potentially relevant reports. Four authors (KJ, MW, RG, SB) examined the full‐text reports to determine which met the eligibility criteria, and made final decisions on study inclusion. We recorded the number of studies retrieved at each stage and reported this information using a PRISMA flow diagram (Moher 2009), where we reported brief details of closely related papers excluded from this review (Figure 1).
Study flow diagram.
Data extraction and management
Two review authors (KJ, MW) independently read the included studies and extracted the following data using a Cochrane template data extraction form (Appendix 2).
-
Participants: total number of participants randomized to each group.
-
Interventions: details of intervention and comparison (including type of intravenous crystalloid, crystalloid volumes used, and timing of intravenous crystalloid administration).
-
Outcomes: study outcomes as measured and reported by study authors (to include types of assessment measures, and time of measurement).
-
Outcome data: results of outcome measures.
We resolved discrepancies by discussion with two other review authors (RG, SB). After agreement, one study author (KJ) entered study data and information for evaluation of the risk of bias into Review Manager 5 (Review Manager 2014). We attempted to contact study authors to obtain additional information when required.
Assessment of risk of bias in included studies
Two review authors (KJ, MW), independently assessed the retained studies with the Cochrane 'Risk of bias' tool (Higgins 2011; Review Manager 2014). We resolved disagreements by discussion with the assistance of a third review author (RG). As is standard, we considered the following methodological criteria:
-
random sequence generation (selection bias);
-
allocation concealment (selection bias);
-
blinding of participants and personnel (performance bias);
-
blinding of outcome assessment (detection bias);
-
incompleteness of outcome data (attrition bias);
-
selective outcome reporting (reporting bias);
-
other sources of bias.
We considered randomization adequate when generated by a computer or random number table algorithm. We considered concealment adequate if the study prevented participant recruiters, investigators, and participants from knowing the allocation of each subsequent study participant (e.g. central randomization by a third party, or the use of sequential, opaque, sealed envelopes). We considered blinding adequate if measures were clearly described that would reasonable prevent participants and personnel from being aware of group allocation (e.g. intervention completed while patient was anaesthetized, using a concealed intravenous crystalloid container and pump operated by a third party not otherwise responsible for patient care). We considered outcome data adequate if all dropouts or withdrawals were accounted for, or if the number of dropouts was small (< 20%) and similar for both interventions. We considered trials as having a low risk of reporting bias if each measurement stated in the methods section was included in the results. We considered non‐intention‐to‐treat as selective reporting.
Measures of treatment effect
Using Review Manager 5 (Review Manager 2014), we presented the results for dichotomous data as risk ratios (RRs) with 95% confidence intervals (CIs) and for continuous data as mean differences (MDs) with 95% CIs. We planned to adapt data presented with different scales as standardized mean differences (SMDs) and 95% CIs. For SMDs, we considered 0.2 a small effect, 0.5 a moderate effect and 0.8 a large effect.
Unit of analysis issues
For studies containing greater than two groups, we merged the data from the similar groups when they were equivalent according to the criteria of our protocol (Jewer 2016). When this was not feasible, we entered the data separately and divided the comparator equally.
If we had identified cluster‐randomized trials, we would have meta‐analysed standard errors and effect estimates using the generic inverse‐variance method in Review Manager 5 (Review Manager 2014)
Dealing with missing data
When data were missing, we attempted to contact the corresponding author. We did not treat medians and means as equivalent. When possible, we calculated missing statistics from other quoted statistics. When participant dropout was encountered, we used an intention‐to‐treat analysis. We explored the effect of missing data using 'best‐case' and 'worst‐case' scenario sensitivity analyses.
Assessment of heterogeneity
We considered clinical heterogeneity during evaluation of the manuscripts, prior to pooling the results. We quantified statistical heterogeneity by calculating the I2 statistic, judging the amount of heterogeneity as low (I2 < 40%), moderate (I2 = 40% to 75%), or high (I2 > 75%) (Guyatt 2011; Higgins 2003).
Assessment of reporting biases
Publication bias is introduced when medical journals are more likely to report studies favouring one treatment than they are to report studies favouring another treatment. In intervention studies, manuscripts may be more likely to be published if they demonstrate efficacy of an intervention over a placebo or control arm. As greater than 10 studies contributed to each of our primary outcomes, we used visual funnel plot analysis in our assessment of reporting bias (Duval 2000). We planned to use the classical fail‐safe number had this not been the case (< 10 studies).
For each outcome, we constructed a funnel plot in Review Manager 5 (Review Manager 2014), with the standard error or precisions (1/standard error) on the y‐axis and the logarithm of the odds ratio on the x‐axis. When no publication bias or small study effects were present, the graph had the shape of an inverted funnel. A vertical line at the logarithm of the effect size found (log odds ratio) would divide the studies such that they are evenly distributed on each side of the line. This provided an estimation of the putative publication bias‐free effect size.
Data synthesis
We analysed data with Review Manager 5 using random‐effects models for all comparisons, given the anticipated moderate to high amount of heterogeneity across studies (Higgins 2003; Review Manager 2014). Random‐effects models give the same results as fixed‐effect models in the absence of statistical heterogeneity. When there is statistical heterogeneity, random‐effects models widen the confidence interval, thus decreasing the chance of finding an effect when there is none. They may increase the weight of smaller studies.
Subgroup analysis and investigation of heterogeneity
In the event of moderate (I2 = 40% to 75%) or high (I2 > 75%) statistical heterogeneity, we started with visual inspection of the forest plots, then proceeded with the following a priori subgroup analyses:
-
volume of supplemental intravenous crystalloid administered (control: intervention volume ratio of less than 1:3 or greater than 1:3);
-
timing of supplemental intravenous crystalloid administration (preoperative, intraoperative, or postoperative);
-
age (6 months to 17 years, 18 years or older).
For outcomes that have a moderate or high level of heterogeneity after subgroup analyses, the results of the subgroup analyses are only presented in a narrative manner.
For outcomes involving multiple studies with paediatric participants, the subgroup results for paediatric participants are also specifically reported, to elucidate this important source of clinical heterogeneity, and to provide specific guidance for clinicians working with this specific patient population.
Sensitivity analysis
We performed sensitivity analyses for outcomes involving studies that used dextrose‐containing fluids, as this is an intervention that independently reduces the risk of PONV (Dabubondoc 2013). The volume of supplemental intravenous crystalloid administered varied in each study, therefore we conducted sensitivity analyses to determine the effect of including studies that infused larger absolute volumes of supplemental intravenous crystalloid to their respective comparator groups (i.e. 10 mL/kg or more). We also sought to assess the influence of studies at relatively higher risk of bias. For each outcome involving studies with one or fewer domains at high or unclear risk of bias, we performed a sensitivity analysis using only those studies with low risk of bias.
'Summary of findings' table and GRADE
The GRADE approach appraises the certainty of a body of evidence based on the extent to which one can be confident that an estimate of effect or association reflects the item being assessed. The certainty of a body of evidence takes into consideration within‐study risk of bias (methodological quality), the directness of the evidence, the heterogeneity of the data, the precision of effect estimates, and the risk of publication bias. We used the principles of the GRADE system (Guyatt 2008; Santesso 2016), to provide an overall assessment of the certainty of the body of evidence associated with each outcome.
We used GRADEpro software to create summary of findings Table for the main comparison (GRADEpro GDT).
Results
Description of studies
See: Included studies; Excluded studies; Studies awaiting classification; and Ongoing studies.
Results of the search
We identified 1210 records from database searches, plus 21 records from forward and backward citation searches, grey literature searches, and clinical trials registry searches. After excluding duplicates, we scrutinized the titles and abstracts of 888 records. From these, we assessed 58 full reports for eligibility, of which we excluded 13. Details of excluded studies are in the table Characteristics of excluded studies.
Included studies
We included 41 studies in the review (Ali 2003; Amireh 2009; Ashok 2017; Behdad 2011; Bennett 1999; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Egeli 2004; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Hashish 2007; Heidari 2012; Heshmati 2004; Holte 2004; Ismail 2017; Keane 1986; Lambert 2009; Lee 2009; Magner 2004; Maharaj 2005; McCaul 2003; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Ooi 1992; Paganelli 2008; Sharma 2010; Shin 2007; Singh 2013; Soleimani 2018; Spencer 1988; Yilmaz 2014; Yogendran 1995; Yoon 2008).
Study selection is detailed in Figure 1.
One of the included studies was a completed, unpublished RCT with data available on ClinicalTrials.gov (Yilmaz 2014). Five studies were published in a non‐English language: two in Farsi (Behdad 2011; Najafianaraki 2010), two in Korean (Shin 2007, Yoon 2008), and one in Portuguese (Paganelli 2008). These studies were translated for interpretation and included in the meta‐analysis.
Three of the included studies did not report data in sufficient detail to use in any analysis (Bennett 1999; Egeli 2004; Singh 2013). We were unable to obtain further details from the authors.
For further details, see Characteristics of included studies.
Participants
The 41 RCTs included in the review reported data from 4224 participants.
Thirty‐four studies exclusively enrolled adult participants. One study enrolled predominantly adults but also had participants as young as 12 years old (Shin 2007). Six studies enrolled paediatric participants, encompassing various age ranges: three to seven years (Ashok 2017), four to 18 years (Egeli 2004), one to 12 years (Elgueta 2013; Goodarzi 2006), six to 12 years (Heshmati 2004), and two to 15 years (Yilmaz 2014).
Thirty‐two studies included participants classified as ASA I or II (Ali 2003; Amireh 2009; Ashok 2017; Bennett 1999; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Hashish 2007; Heidari 2012; Holte 2004; Ismail 2017; Keane 1986; Lambert 2009; Lee 2009; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Ooi 1992; Sharma 2010; Shin 2007; Soleimani 2018; Spencer 1988; Yilmaz 2014; Yoon 2008). Three studies included participants classified as ASA I to III (Maharaj 2005; Paganelli 2008; Yogendran 1995), and two studies included participants classified as ASA I only (Behdad 2011; Heshmati 2004; Magner 2004). Four studies did not report the ASA classification of their participants (Dagher 2009; Egeli 2004; McCaul 2003; Singh 2013).
One study specifically described selecting for participants at high risk for PONV (Bhukal 2012). The other studies were inconsistent in their reporting of baseline risk factors for PONV.
All studies enrolled participants undergoing surgery with general anaesthesia. One study also included participants receiving deep sedation (Bennett 1999).
There were a variety of elective surgeries performed in these studies, performed on an ambulatory basis or with a short length of stay (i.e. one day). Among studies that solely focused on one type of surgical procedure, seven studies focused on laparoscopic gynaecologic surgery (Chauhan 2013; Hashish 2007; Lambert 2009; Magner 2004; Maharaj 2005; McCaul 2003; Monti 1999), six studies focused on laparoscopic cholecystectomy (Amireh 2009; Holte 2004; Ismail 2017; Lee 2009; Paganelli 2008; Sharma 2010), six studies focused on otorhinolaryngologic procedures (Behdad 2011; Dagher 2009; Egeli 2004; Elgueta 2013; Heshmati 2004; Yilmaz 2014), two studies focused on unspecified laparoscopic surgeries (Cook 1990; Murshed 2012), two studies focused on therapeutic abortion (Elhakim 1998; Ooi 1992), one study focused on strabismus repair (Goodarzi 2006), one study focused on open cholecystectomy (Chaudhary 2008), one study focused on dental extractions (Bennett 1999), one study focused on orthopaedic surgery (Heidari 2012), one study focused on cervical cerclage (Najafianaraki 2010), and one study focused on breast cancer surgery (Soleimani 2018). Twelve studies involved a heterogeneous mix of surgical procedures, typically of an abdominal or gynaecologic nature (Ali 2003; Ashok 2017; Bhukal 2012; Chohedri 2006; Gwak 2007; Keane 1986; Onyando 2014; Shin 2007; Singh 2013; Spencer 1988; Yogendran 1995; Yoon 2008).
Settings
All studies took place in surgical centres. Twenty‐one studies took place in Asia, eight in Europe, six in North America, four in Africa, and two in South America. Six studies were completed in each of India and Iran. Five studies were completed in each of the UK and the USA. The remaining studies originated from other countries (i.e. Bangladesh, Brazil, Canada, Chile, Denmark, Egypt, Ireland, Jordan, Kenya, Lebanon, South Korea, and Turkey).
Interventions
Twenty‐nine of the 41 included studies used Ringer's lactate as their intervention supplemental crystalloid (Ali 2003; Amireh 2009; Ashok 2017; Behdad 2011; Chaudhary 2008; Chauhan 2013; Cook 1990; Dagher 2009; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Hashish 2007; Heidari 2012; Heshmati 2004; Holte 2004; Ismail 2017; Lambert 2009; Lee 2009; Magner 2004; Maharaj 2005; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Sharma 2010; Singh 2013; Spencer 1988; Yoon 2008). Five studies used normal saline (Bennett 1999; Bhukal 2012; Chohedri 2006; Paganelli 2008; Yilmaz 2014). One study used both Ringer's lactate and normal saline (Soleimani 2018). One study used an acetate‐containing balanced crystalloid solution called Plasmalyte (Yogendran 1995).
One study used 5% dextrose in saline (Ooi 1992), one study used 5% dextrose in Ringer's lactate (Egeli 2004), two studies used a combination of 5% dextrose in water and Ringer's lactate (Keane 1986; McCaul 2003). One study had a study arm using Ringer's lactate and a study arm using 5% dextrose in Ringer's lactate (Cook 1990). Finally, one study had a study arm using Ringer's lactate and a study arm using 5% dextrose in water (Shin 2007); however, the latter arm was not included in analyses as 5% dextrose in water is not a crystalloid solution. Three studies used more than one type of intravenous crystalloid solution (Cook 1990; McCaul 2003; Soleimani 2018).
Supplemental crystalloid administration started before induction of anaesthesia (preoperatively) in 24 studies (Ali 2003; Amireh 2009; Bennett 1999; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Hashish 2007; Heidari 2012; Holte 2004; Lambert 2009; Lee 2009; Magner 2004; Maharaj 2005; Monti 1999; Murshed 2012; Onyando 2014; Ooi 1992; Sharma 2010; Shin 2007; Singh 2013; Yilmaz 2014; Yogendran 1995), after induction of anaesthesia (intraoperatively) in 15 studies (Ashok 2017; Behdad 2011; Bhukal 2012; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Heshmati 2004; Ismail 2017; Keane 1986; McCaul 2003; Najafianaraki 2010; Paganelli 2008; Spencer 1988; Yoon 2008), and after emergence from anaesthesia (postoperatively) in one study (Egeli 2004). One study started supplemental crystalloid administration preoperatively for one study arm, and intraoperatively in another study arm (Soleimani 2018).
Generally, intervention groups were administered a volume of intravenous supplemental crystalloid of at least 10 mL/kg. There were a minority of studies where the comparator groups were administered a volume of intravenous supplemental crystalloid comparable to this volume (Ashok 2017; Chauhan 2013; Dagher 2009; Goodarzi 2006; Hashish 2007; Holte 2004; Ismail 2017; Magner 2004; Paganelli 2008; Sharma 2010; Yilmaz 2014).
Details of the intervention are presented in Table 1.
| Study | Timing | Comparator group | Intervention group(s) |
| Preoperative | RL 2 mL/kg | RL 15 mL/kg | |
| Preoperative | No crystalloid bolus | RL 10 mL/kg | |
| Intraoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Intraoperative | RL 4 mL/kg | RL 10 mL/kg | |
| Intraoperative | RL 20 mL/kg | ||
| Preoperative | NS 1 to 2 mL/kg | NS 15 mL/kg | |
| Intraoperative | NS 4 mL/kg | NS 10 mL/kg | |
| Preoperative | RL 2 mL/kg | RL 12 mL/kg | |
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Preoperative | NS 2 mL/kg | NS 20 mL/kg | |
| Preoperative | No crystalloid bolus | RL 20 mL/kg | |
| Preoperative | RL 20 mL/kg with 1 g/kg dextrose | ||
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Postoperative | No crystalloid bolus | D5RL 60 to 120 mL/hour | |
| Intraoperative | RL 10 mL/kg/hour | RL 30 mL/kg/hour | |
| Intraoperative | No crystalloid bolus | RL 1000 mL | |
| Intraoperative | RL 10 mL/kg/hour | RL 30 mL/kg/hour | |
| Intraoperative | RL 6 mL/kg/hour | RL 18 mL/kg/hour | |
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Preoperative | No crystalloid bolus | RL 10 mL/kg | |
| Intraoperative | No crystalloid bolus | RL 4 mL/kg/hour | |
| Preoperative | RL 15 mL/kg | RL 40 mL/kg | |
| Intraoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Mixed | No crystalloid bolus | Intraoperative RL 1000 mL then postoperative 1000 mL D5W | |
| Preoperative | No crystalloid bolus | RL 900 to 1000 mL | |
| Preoperative | RL 5 mL/kg/hour | RL 30 mL/kg/hour | |
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Preoperative | RL 2 mL/kg per hour fasted | RL 3 mL/kg per house fasted | |
| Intraoperative | No crystalloid bolus | RL 1.5 mL/kg per hour fasted | |
| Intraoperative | D5RL 1.5 mL/kg per hour fasted | ||
| Preoperative | No crystalloid bolus | RL 1000 mL | |
| Preoperative | RL 1.5 mL/kg per hour fasted | RL 15 mL/kg | |
| Preoperative | RL 2 mL/kg per hour fasted | RL 2 mL/kg per hour fasted then RL 10 mL/kg | |
| Preoperative | No crystalloid bolus | RL "maintenance" rate per hour fasted (maximum 1000 mL) | |
| Preoperative | No crystalloid bolus | 20 mL/kg 4% dextrose/0.18% saline solution | |
| Intraoperative | NS 10 mL/kg/hour | NS 1000 mL bolus then 10 mL/kg/hour | |
| Preoperative | RL 10 mL/kg | RL 20 mL/kg | |
| Preoperative | RL 30 mL/kg | ||
| Preoperative | RL 2 mL/kg | RL 20 mL/kg | |
| Preoperative | No crystalloid bolus | RL 30 mL/kg | |
| Preoperative | NS 1.5 mL/kg/hour | NS 1.5 mL/kg/hour then RL 5 mL/kg | |
| Intraoperative | NS 1.5 mL/kg/hour then RL 5 mL/kg | ||
| Intraoperative | No crystalloid bolus | RL 1000 mL | |
| Intraoperative | NS 10 mL/kg/hour | NS 20 mL/kg/hour | |
| Preoperative | Plasmalyte 2 mL/kg | Plasmalyte 20 mL/kg | |
| Intraoperative | RL 2 mL/kg | RL 18 mL/kg |
D5RL: dextrose 5% in Ringer's Lactate; D5W: dextrose 5% in water, NS: normal saline, RL: Ringer's Lactate
Comparators
In all studies, participants in the comparator group received a smaller volume of perioperative crystalloid than did participants in the intervention group, or they received no perioperative crystalloid.
In 26 studies, both the comparator group and the intervention group received the same type of intravenous crystalloid (Ali 2003; Ashok 2017; Behdad 2011; Bennett 1999; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Dagher 2009; Elgueta 2013; Goodarzi 2006; Gwak 2007; Hashish 2007; Holte 2004; Ismail 2017; Lee 2009; Magner 2004; Maharaj 2005; Murshed 2012; Najafianaraki 2010; Paganelli 2008; Sharma 2010; Shin 2007; Yilmaz 2014; Yogendran 1995; Yoon 2008). In 14 studies, the comparator group did not receive any perioperative intravenous crystalloid bolus (Amireh 2009; Cook 1990; Egeli 2004; Elhakim 1998; Heidari 2012; Heshmati 2004; Keane 1986; Lambert 2009; McCaul 2003; Monti 1999; Onyando 2014; Ooi 1992; Singh 2013; Spencer 1988).
Funding sources
Six studies disclosed a funding source. Of these, five studies cited an academic source of funding, such as a hospital or university department (Ashok 2017; Chauhan 2013; Holte 2004; Maharaj 2005; Yilmaz 2014), while one study disclosed funding from an industry source (Spencer 1988, Baxter Health Care). Two studies stated that they had no funding (Bhukal 2012; Soleimani 2018). For the remaining studies, the source of funding was unclear.
Excluded studies
We excluded 13 studies for not meeting the inclusion criteria (Abraham‐Nording 2012; Alnema 2011; Apfel 2012; Brandstrup 2003; Cuthbertson 2011; Dabubondoc 2013; Gaiser 2002; Heidari 2011; Holte 2007a; Holte 2007b; Lei 2017; Mintz 2004; Yavuz 2014). Five were not focused on PONV (Abraham‐Nording 2012; Brandstrup 2003; Cuthbertson 2011; Holte 2007a; Holte 2007b), five were not randomized controlled trials (Alnema 2011; Apfel 2012; Lei 2017; Mintz 2004; Yavuz 2014), one studied participants undergoing neuraxial anaesthetic (Gaiser 2002), and two did not have an intervention group receiving intravenous crystalloids (Dabubondoc 2013; Heidari 2011).
See Characteristics of excluded studies.
Studies awaiting classification
We identified one study in a clinical trial register that was terminated; we will await publication of study results before assessing eligibility (Laws 2003).
For further details, see Characteristics of studies awaiting classification
Ongoing studies
We identified three ongoing studies (NCT03141645; NCT03142464; NCT03485443).
One study aims to investigate preoperative intravenous fluid administration in participants 18 years or older, undergoing laparoscopic cholecystectomy (NCT03141645). They will compare participants receiving preoperative intravenous fluid administration against two groups: one that receives intraoperative ondansetron, and one that receives neither preoperative intravenous fluid nor intraoperative ondansetron. The primary outcome is PONV within the first postoperative 24 hours. The study hypothesis is that participants receiving preoperative intravenous fluid administration and patients receiving intraoperative ondansetron will have a similar reduction in risk of PONV, compared with the comparator group receiving neither.
One study aims to examine postoperative intravenous fluid administration in participants 18 years or older, undergoing laparoscopic cholecystectomy (NCT03142464). The study compares a restrictive fluid administration strategy against their usual practice of postoperative fluid administration. The primary outcome is renal function, reflected by serum creatinine, while nausea rated on a visual analogue scale (VAS) is a secondary outcome measure.
One study aims to evaluate the effect of intraoperative hydration on postoperative vomiting in paediatric patients undergoing otorhinolaryngological surgery (NCT03485443). They will compare participants receiving an intervention of normal saline at a rate of 30 mL/kg/hour during the intraoperative period with a comparison group receiving normal saline at a rate of 10 mL/kg/hour during the intraoperative period. The primary outcomes assessed will be postoperative nausea and vomiting in the PACU. Rescue antiemetic administration will also be documented, as will intensity of postoperative pain.
For further details, see Characteristics of ongoing studies.
Risk of bias in included studies
We assessed the risk of bias in the included studies in terms of allocation sequence generation, blinding, incomplete reporting of outcome data, and selective reporting. Risk of bias was generally low to moderate across all included studies, but eight studies were at high risk of bias (Bennett 1999; Egeli 2004; Keane 1986; Lambert 2009; McCaul 2003; Monti 1999; Soleimani 2018; Yogendran 1995). Three studies were at low risk of bias across all domains (Amireh 2009; Holte 2004; Ismail 2017), while nine studies had one domain at unclear risk of bias but were otherwise at low risk of bias (Ali 2003; Ashok 2017; Bhukal 2012; Chauhan 2013; Elgueta 2013; Gwak 2007; Magner 2004; Maharaj 2005; Murshed 2012). Outcomes that included data from studies at higher risk were subject to further sensitivity analyses to assess the influence of these studies on results.
For details, see Figure 2, Figure 3, and Characteristics of included studies
'Risk of bias' graph.
'Risk of bias' summary.
Allocation
All studies were RCTs. Twenty‐one studies provided adequate information to determine that the randomization process was prospective and unpredictable (Ali 2003; Amireh 2009; Ashok 2017; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Dagher 2009; Elgueta 2013; Goodarzi 2006; Gwak 2007; Heidari 2012; Holte 2004; Ismail 2017; Magner 2004; Maharaj 2005; McCaul 2003; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Soleimani 2018). In the remaining studies, the randomization process was not adequately described to exclude randomization bias.
Allocation concealment was adequate if study personnel were unaware of the allocation of each subsequent study participant (i.e. using sequentially numbered, opaque, sealed envelopes or centralized third‐party allocation). Five studies sufficiently described effective methods to conceal group allocations from study personnel (Amireh 2009; Gwak 2007; Holte 2004; Ismail 2017; Najafianaraki 2010). The remaining studies provided insufficient details on their allocation process.
Blinding
Participants and personnel were adequately blinded in 19 studies (Ali 2003; Amireh 2009; Ashok 2017; Bhukal 2012; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Elgueta 2013; Elhakim 1998; Holte 2004; Ismail 2017; Magner 2004; Maharaj 2005; Murshed 2012; Onyando 2014; Shin 2007; Yilmaz 2014; Yogendran 1995). In one study supplemental intravenous crystalloid was administered over 24 hours postoperatively, so blinding of participants and care staff would have been very difficult (Egeli 2004). In one study, outcome assessors were blinded, but patients and anaesthesia personnel were not described as blinded during the preoperative and intraoperative periods, respectively (Soleimani 2018). In one study, blinding was not mentioned at all (Monti 1999). In the remaining studies, blinding of either participants or personnel was not adequately described to rule out a lack of blinding.
As all outcomes were evaluated in the postoperative period, adequate blinding of the outcome assessor was theoretically possible in all studies where fluid administration occurred preoperatively or intraoperatively. Twenty‐nine studies stated that the outcome assessor was blinded to study group (Ali 2003; Amireh 2009; Ashok 2017; Behdad 2011; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Heshmati 2004; Holte 2004; Ismail 2017; Lambert 2009; Lee 2009; Magner 2004; Maharaj 2005; McCaul 2003; Murshed 2012; Onyando 2014; Ooi 1992; Soleimani 2018; Yilmaz 2014; Yogendran 1995; Yoon 2008). In one study, fluid administration took place postoperatively and outcome assessors were not blinded (Egeli 2004). In one study, participants received the intervention preoperatively while awake and later completed a self‐administered questionnaire (Bennett 1999). In one study, blinding was not mentioned at all (Monti 1999). Outcome assessor blinding was not adequately described in the remaining studies, therefore they had an unclear risk of bias.
Incomplete outcome data
We judged six studies to have unclear risk of attrition bias: for not reporting participant counts (Singh 2013); for not providing reasons for exclusions (Cook 1990); for inadequately reporting results to facilitate evaluation of attrition (Monti 1999); for having > 20% attrition with balanced withdrawals amongst study groups (Shin 2007); for having > 10% attrition with balanced withdrawals (Bennett 1999); and for < 10% attrition with balanced withdrawals (Yoon 2008). Two studies had high risk of attrition bias, where approximately 10% of participants were excluded or lost to follow‐up without indication of their allocation (McCaul 2003; Yogendran 1995). The remaining studies reported no participant losses during the trial, or only a small number of participants that was unlikely to substantially affect study results.
Selective reporting
Five studies were at high risk of reporting bias: four for not completing an intention‐to‐treat analysis for excluded participants (Bennett 1999; Lambert 2009; McCaul 2003; Yogendran 1995), and one for failing to report results of a planned outcome, vomiting (Keane 1986). Two studies had an unclear risk of bias, as they did not complete an intention‐to‐treat analysis but were only missing two participants each (Cook 1990; Dagher 2009). One paper failed to report several planned outcomes, but these were not pertinent to this review, so the paper was placed at unclear risk of reporting bias (Spencer 1988). One study did not report their data in adequate detail to rule out reporting bias, and was at unclear risk (Singh 2013). The remaining papers were at low risk of reporting bias.
Other potential sources of bias
In one study, a statistically significant difference in operative time existed between the intervention and the comparator despite prospective randomization and allocation concealment, but it was unclear whether this would bias other study results (Onyando 2014).
Effects of interventions
See summary of findings Table for the main comparison.
Primary outcomes
1. Risk of PON, defined as the presence of subjective nausea, reported dichotomously or based on a study‐defined dichotomous threshold on a continuous scale such as a VAS
Studies reporting risk of PON
Thirty‐two studies (3268 participants) assessed PON (Ali 2003; Amireh 2009; Behdad 2011; Bennett 1999; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Elhakim 1998; Gwak 2007; Hashish 2007; Heshmati 2004; Ismail 2017; Keane 1986; Lambert 2009; Magner 2004; Maharaj 2005; McCaul 2003; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Ooi 1992; Paganelli 2008; Sharma 2010; Shin 2007; Singh 2013; Soleimani 2018; Spencer 1988; Yogendran 1995; Yoon 2008). Three of these studies reported data in insufficient detail to be used in our meta‐analysis of PON, and we were unable to obtain further details from the authors (Bennett 1999; Bhukal 2012; Singh 2013).
Most studies reported PON data dichotomously (Ali 2003; Amireh 2009; Behdad 2011; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Gwak 2007; Hashish 2007; Heshmati 2004; Ismail 2017; Keane 1986; Lambert 2009; Magner 2004; Maharaj 2005; McCaul 2003; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Ooi 1992; Paganelli 2008; Sharma 2010; Shin 2007; Spencer 1988; Yogendran 1995; Yoon 2008). In studies presenting continuous data, several grading scales were used: five‐point Likert scale (Bennett 1999; Bhukal 2012), 10 cm or 100 mm VAS (Elhakim 1998; Sharma 2010), 0 to 10 verbal grading scale (Maharaj 2005), and ordinal grading scale (Magner 2004). In some studies, investigators assessed nausea on a continuous scale and converted this measurement to a dichotomous value using a threshold level, for instance 1 on a 0 to 10 verbal scale or 50 mm on a 100 mm VAS (Ali 2003; Amireh 2009; Chaudhary 2008; Gwak 2007; Maharaj 2005; Onyando 2014). One study reported both dichotomous and continuous data for PON (Maharaj 2005).
Risk of PON, when cumulative events were explicitly reported for the entire study period
Eighteen studies (1766 participants) reported dichotomous data for risk of PON during the cumulative study period (Ali 2003; Amireh 2009; Behdad 2011; Chauhan 2013; Dagher 2009; Gwak 2007; Heshmati 2004; Ismail 2017; Lambert 2009; Magner 2004; Maharaj 2005; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Ooi 1992; Sharma 2010; Yoon 2008). Supplemental intravenous crystalloid decreased risk of PON during the cumulative study period (risk ratio (RR) 0.62, 95% confidence interval (CI) 0.51 to 0.75; Analysis 1.1; Figure 4). This outcome had moderate statistical heterogeneity (I2 = 57%) that could not be reduced by subgroup analyses for: the relative amount of crystalloid administered, timing of crystalloid administration, or age (i.e. paediatric participants). We rated the certainty of this evidence using GRADE as moderate, having been downgraded due to risk of publication bias, as indicated by inspection of a funnel plot generated from included study data.
Forest plot of comparison: 1 Supplemental IV crystalloid administration for preventing PONV versus control, outcome: 1.5 Risk of overall PON (when cumulative nausea events were explicitly reported for the entire study period), as measured by the presence of subjective nausea, reported dichotomously or based on a study‐defined dichotomous threshold on a continuous scale such as a VAS.
One study (30 participants) in this analysis used a dextrose‐containing solution in the intervention group (Ooi 1992), and a sensitivity analysis found that inclusion of this study did not substantially affect the RR or statistical heterogeneity. The inclusion of studies where comparator group participants received at least 10 mL/kg of supplemental intravenous crystalloid did not substantially affect the RR. We performed a sensitivity analysis involving only studies at low risk of bias (Ali 2003; Amireh 2009; Chauhan 2013; Gwak 2007; Ismail 2017; Maharaj 2005; Murshed 2012), and this did not substantially affect the RR.
Risk of PON during specific time points (i.e. early and late postoperative period)
Twenty studies (2310 participants) reported dichotomous data on risk of PON in the early postoperative period (Ali 2003; Amireh 2009; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Gwak 2007; Hashish 2007; Ismail 2017; Keane 1986; Magner 2004; Maharaj 2005; McCaul 2003; Murshed 2012; Onyando 2014; Paganelli 2008; Shin 2007; Spencer 1988; Yogendran 1995; Yoon 2008); of these, three studies (410 participants) used a dextrose‐containing solution (Cook 1990; Keane 1986; McCaul 2003). Seventeen studies (1682 participants) reported dichotomous data on risk of PON in the late postoperative period (Ali 2003; Amireh 2009; Cook 1990; Dagher 2009; Gwak 2007; Hashish 2007; Ismail 2017; Magner 2004; Maharaj 2005; McCaul 2003; Murshed 2012; Onyando 2014; Paganelli 2008; Shin 2007; Spencer 1988; Yogendran 1995; Yoon 2008); of these, two studies (98 participants) used a dextrose‐containing solution (Cook 1990; McCaul 2003).
Supplemental intravenous crystalloids decreased the risk of early PON (RR 0.67, 95% CI 0.58 to 0.78; Analysis 1.2). Heterogeneity was low (I2 = 9%). Supplemental intravenous crystalloids decreased the risk of late PON (RR 0.47, 95% CI 0.32 to 0.69; Analysis 1.3; Figure 5). Heterogeneity was low (I2 = 38%). We rated the certainty of evidence using GRADE as moderate for both early and late time points, having been downgraded due to risk of publication bias, as indicated by inspection of a funnel plot generated from included study data.
Forest plot of comparison: 1 Supplemental IV crystalloid administration for preventing PONV versus control, outcome: 1.9 Risk of pharmacologic treatment for PONV.
A sensitivity analysis found that inclusion of dextrose‐containing solutions did not substantially affect the RR or statistical heterogeneity for risk of early PON. For risk of late PON, removing dextrose‐containing solutions increased statistical heterogeneity (I2 = 47%), but did not substantially affect the RR. The inclusion of studies where comparator group participants received at least 10 mL/kg of supplemental intravenous crystalloid did not substantially affect the RR for risk of early PON or late PON. We performed sensitivity analyses involving only studies at low risk of bias for early PON (Ali 2003; Amireh 2009; Chauhan 2013; Elgueta 2013; Gwak 2007; Ismail 2017; Magner 2004; Maharaj 2005; Murshed 2012) and late PON (Ali 2003; Amireh 2009; Elgueta 2013; Gwak 2007; Ismail 2017; Magner 2004; Maharaj 2005; Murshed 2012), but the RR was not substantially affected in either case.
Risk of PON, when reported using continuous data
Five studies (415 participants) reported continuous data for early PON (Chaudhary 2008; Elhakim 1998; Maharaj 2005; Sharma 2010; Soleimani 2018). One study reported both dichotomous and continuous data for early and late PON, and was accordingly included in analyses of PON as a dichotomous outcome, as well as analyses of PON as a continuous outcome (Maharaj 2005).
Supplemental intravenous crystalloids decreased the severity of early PON on a 100 mm VAS (mean difference (MD) ‐16.38, 95% CI ‐21.81 to ‐10.96; Analysis 1.4). Statistical heterogeneity was moderate (I2 = 47%), but there were insufficient studies to conduct planned subgroup analyses.
Five studies (415 participants) reported continuous data assessing late PON (Chaudhary 2008; Elhakim 1998; Maharaj 2005; Sharma 2010; Soleimani 2018). On a 100 mm VAS, supplemental intravenous crystalloids decreased the severity of nausea (MD ‐9.62, 95% CI ‐14.91 to ‐4.32; Analysis 1.5). Statistical heterogeneity was high (I2 = 71%), but there were insufficient studies to conduct planned subgroup analyses.
There were insufficient studies to conduct sensitivity analyses for dextrose‐containing solutions, for comparator group volume infused, or for risk of bias.
2. Risk of POV, reported dichotomously by any discrete episodes of vomiting
Studies reporting risk of POV
Thirty‐one studies (3105 participants) evaluated POV (Ali 2003; Amireh 2009; Ashok 2017; Behdad 2011; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Elgueta 2013; Elhakim 1998; Gwak 2007; Hashish 2007; Heidari 2012; Heshmati 2004; Ismail 2017; Magner 2004; Maharaj 2005; McCaul 2003; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Paganelli 2008; Sharma 2010; Shin 2007; Singh 2013; Spencer 1988; Yilmaz 2014; Yoon 2008); however, one study did not report sufficiently detailed data to be included our analyses for risk of POV (Singh 2013).
Four studies (500 participants) reported POV in paediatric participants, aged 6 months to 18 years (Ashok 2017; Elgueta 2013; Heshmati 2004; Yilmaz 2014).
Risk of POV, when cumulative events were explicitly reported for the entire study period
Twenty studies (1970 participants) provided overall data for POV across all time points (Ali 2003; Amireh 2009; Ashok 2017; Behdad 2011; Bhukal 2012; Chaudhary 2008; Dagher 2009; Elgueta 2013; Gwak 2007; Heidari 2012; Heshmati 2004; Ismail 2017; Magner 2004; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Sharma 2010; Yilmaz 2014; Yoon 2008); of these, four studies (500 participants) included paediatric participants (Ashok 2017; Elgueta 2013; Heshmati 2004; Yilmaz 2014). Supplemental intravenous crystalloids decreased the cumulative risk of POV over the entire study period (RR 0.50, 95% CI 0.40 to 0.63; Analysis 1.6; Figure 6). Heterogeneity was low (I2 = 31%). We rated the certainty of this evidence using GRADE as moderate, having been downgraded due to risk of publication bias, as indicated by inspection of a funnel plot generated from included study data.
Forest plot of comparison: 1 Supplemental IV crystalloid administration for preventing PONV versus control, outcome: 1.6 Risk of cumulative POV.
For paediatric participants, supplemental intravenous crystalloid administration also reduced the cumulative risk of POV over the entire study period, but to a lesser degree than for adults (RR 0.69, 95% CI 0.57 to 0.85; I2 = 0%).
There were insufficient studies to conduct sensitivity analyses for dextrose‐containing solutions. The inclusion of studies where comparator group participants received at least 10 mL/kg of supplemental intravenous crystalloid did not substantially affect the RR. A sensitivity analysis involving only studies at low risk of bias did not substantially affect the RR (Ali 2003; Amireh 2009; Ashok 2017; Bhukal 2012; Elgueta 2013; Gwak 2007; Ismail 2017; Magner 2004; Murshed 2012).
Risk of POV during specific time points (i.e. early and late postoperative period)
We analysed 19 studies (1998 participants) for early POV (Ali 2003; Amireh 2009; Chauhan 2013; Chohedri 2006; Cook 1990; Dagher 2009; Elhakim 1998; Gwak 2007; Hashish 2007; Ismail 2017; Magner 2004; Maharaj 2005; McCaul 2003; Murshed 2012; Onyando 2014; Paganelli 2008; Shin 2007; Spencer 1988; Yoon 2008); of these, two studies (98 participants) used dextrose‐containing solutions (Cook 1990; McCaul 2003). Fifteen studies (1403 participants) assessed late POV (Ali 2003; Amireh 2009; Cook 1990; Dagher 2009; Elhakim 1998; Gwak 2007; Hashish 2007; Ismail 2017; Magner 2004; McCaul 2003; Murshed 2012; Onyando 2014; Paganelli 2008; Shin 2007; Yoon 2008); of these, two studies (98 participants) used dextrose‐containing solutions (Cook 1990; McCaul 2003).
Supplemental intravenous crystalloids decreased early POV (RR 0.56, 95% CI 0.41 to 0.76; Analysis 1.7). Statistical heterogeneity was low (I2 = 0%). Supplemental intravenous crystalloids also decreased postoperative vomiting late POV (RR 0.48, 95% CI 0.29 to 0.79; Analysis 1.8). Statistical heterogeneity was low (I2 = 0%). We rated the certainty of this evidence using GRADE as moderate for both time points, having been downgraded due to risk of publication bias, as indicated by inspection of a funnel plot generated by included study data.
A sensitivity analysis found that inclusion of dextrose‐containing solutions did not substantially affect the RR or statistical heterogeneity for risk of either early or late POV. The inclusion of studies where comparator group participants received at least 10 mL/kg of supplemental intravenous crystalloid did not substantially affect the RR for risk of early POV or late POV. A sensitivity analysis of studies at low risk of bias did not substantially affect the RR for risk of early POV (Ali 2003; Amireh 2009; Chauhan 2013; Gwak 2007; Ismail 2017; Magner 2004; Maharaj 2005; Murshed 2012) or for late POV (Ali 2003; Amireh 2009; Gwak 2007; Ismail 2017; Magner 2004; Murshed 2012).
Secondary outcomes
1. Risk of requiring pharmacologic treatment for PONV
Twenty‐three studies (2416 participants) measured the use of postoperative antiemetic medications (Ali 2003; Ashok 2017; Behdad 2011; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Cook 1990; Dagher 2009; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Holte 2004; Ismail 2017; Magner 2004; Maharaj 2005; Monti 1999; Murshed 2012; Onyando 2014; Sharma 2010; Soleimani 2018; Yogendran 1995; Yoon 2008); of these, one study (38 participants) used a dextrose‐containing solution (Cook 1990), and three studies (350 participants) examined pharmacologic treatment of PONV in paediatric participants, ranging from 1 to 12 years old (Ashok 2017; Elgueta 2013; Goodarzi 2006). One study measured the use of postoperative antiemetic medications but did not provide sufficient data to analyse this outcome (Singh 2013).
Supplemental intravenous crystalloids decreased the risk of requiring pharmacologic treatment of PONV (RR 0.62, 95% CI 0.51 to 0.76; Analysis 1.9; Figure 5). We rated the certainty of this evidence using GRADE as moderate, having been downgraded due to risk of publication bias, as indicated by inspection of a funnel plot generated from included study data.
This outcome had moderate statistical heterogeneity (I2 = 40%). The RR was not affected by any of our planned subgroup analyses: timing of fluid administration, relative volume of supplemental intravenous crystalloid administered, or age. A sensitivity analysis found that inclusion of dextrose‐containing solutions did not substantially affect the RR or statistical heterogeneity.
For paediatric participants, supplemental intravenous crystalloid administration did not appear to reduce the risk of requiring pharmacologic treatment of PONV (RR 0.81, 95% CI 0.50 to 1.30; I2 = 0%).
The inclusion of studies where comparator group participants received at least 10 mL/kg of supplemental intravenous crystalloid did not substantially affect the RR. A sensitivity analysis of studies at low risk of bias did not substantially affect the RR (Ali 2003; Ashok 2017; Bhukal 2012; Chauhan 2013; Elgueta 2013; Gwak 2007; Holte 2004; Ismail 2017; Magner 2004; Maharaj 2005; Murshed 2012).
2. Risk of unintended postoperative admission to hospital
Three studies (235 participants) quantified the rate of unplanned admission to hospital after ambulatory surgery (Ali 2003; Cook 1990; Maharaj 2005); of these, one study (38 participants) used a dextrose‐containing solution
Supplemental intravenous crystalloid administration did not affect this outcome (RR 1.05, 95% CI 0.77 to 1.43; Analysis 1.10). Heterogeneity was low (I² = 0%). We rated the certainty of this evidence using GRADE as low, having been downgraded due to imprecision and inconsistency of the results of included studies.
There were insufficient studies to carry out planned sensitivity analyses for dextrose‐containing solutions, comparator group volume infused, or for studies at low risk of bias.
3. Risk of suffering a serious adverse event (any of: admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death)
We found no information about this outcome in the included studies.
Discussion
Summary of main results
We included 41 trials with a total of 4224 participants in this meta‐analysis. Combination of the results of these studies showed that supplemental perioperative intravenous crystalloid administration probably reduces the risk of PON in the overall postoperative period (RR 0.62, 95% CI 0.51 to 0.75; moderate‐certainty evidence), and specifically during the early (RR 0.67, 95% CI 0.58 to 0.78; moderate‐certainty evidence) and late (RR 0.47, 95% CI 0.32 to 0.69; moderate‐certainty evidence) time points. Supplemental perioperative intravenous crystalloid administration probably reduces the risk of POV in the overall postoperative period (RR 0.50, 95% CI 0.40 to 0.63; moderate‐certainty evidence), as well as during early (RR 0.56, 95% CI 0.41 to 0.76; moderate‐certainty evidence) and late (RR 0.48, 95% CI 0.29 to 0.79; moderate‐certainty evidence) time points. The certainty of the evidence for all PON and POV outcomes, as assessed using GRADE, is rated as moderate. Supplemental perioperative intravenous crystalloid administration probably reduces the risk for treatment with antiemetic rescue medication (RR 0.62, 95% CI 0.51 to 0.76; moderate‐certainty evidence). The effect of the intervention on the risk of unintended postoperative hospital admission after ambulatory surgery is unclear (RR 1.05, 95% CI 0.77 to 1.43; 3 studies; 235 participants; low‐certainty evidence). No studies reported serious adverse events with this intervention (i.e. admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death).
Overall completeness and applicability of evidence
The majority of trials enrolled only ASA I to II patients, for ambulatory or short length of stay procedures (i.e. one day) (Ali 2003; Amireh 2009; Ashok 2017; Behdad 2011; Bennett 1999; Bhukal 2012; Chaudhary 2008; Chauhan 2013; Chohedri 2006; Cook 1990; Elgueta 2013; Elhakim 1998; Goodarzi 2006; Gwak 2007; Hashish 2007; Heidari 2012; Holte 2004; Ismail 2017; Keane 1986; Lambert 2009; Lee 2009; Monti 1999; Murshed 2012; Najafianaraki 2010; Onyando 2014; Ooi 1992; Sharma 2010; Shin 2007; Soleimani 2018; Spencer 1988; Yilmaz 2014; Yoon 2008). Otherwise, there was significant diversity amongst the included studies. Participants' baseline risk of PONV likely varied between studies, but this information was insufficiently reported to specifically analyse. Participants underwent a wide range of surgical procedures. Anaesthetic technique was varied, including induction and maintenance agents, use of muscle relaxants and reversal agents, intraoperative opioid administration, and pharmacologic PONV prophylaxis. Trials took place in a number of countries across the developed, emerging, and developing world. Although this variation likely introduced heterogeneity into the results, it also suggests that conclusions are generalizable to a sizeable scope of ambulatory surgical populations.
We found that PONV was very inconsistently defined across studies, so we had to focus on the related and more precisely defined outcomes of PON and POV. Most trials included in this review reported on one of our primary outcomes (i.e. risk of PON, risk of POV, or risk of PONV). There were few studies reporting continuous data for risk of PON; although far fewer studies and patients were pooled, we were able to assess the effect of supplemental intravenous crystalloid administration on PON severity. Nonetheless, this presents an area for further research.
Very few studies examined potential harms that patients may experience from vigorous volume administration. For instance, no studies examined the risk of serious adverse events (i.e. admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death), and no studies examined PACU length of stay. This is clearly a deficiency in the existing literature.
Due to differences in the way that studies defined the volume of supplemental intravenous crystalloid that was administered to patients, it was not possible to compare absolute volume administered across studies. Where applicable, we conducted subgroup analyses of supplemental intravenous crystalloid volume administered relative to comparator and intervention groups, and this was not found to be influential. Moreover, we conducted sensitivity analyses omitting studies where comparator groups received a volume of intravenous supplemental crystalloid comparable to most studies' intervention groups (10 mL/kg or more), and this appeared to have negligible influence on the effect of the intervention. However, more work may be required to elucidate an optimal dosing for this intervention.
Despite these limitations, there is sufficient data to suggest that supplemental intravenous crystalloid administration may be helpful to reduce the risk of PONV. The varied settings do provide a degree of generalizability, albeit in an ambulatory setting with generally healthy patients. These results may not be easily generalized to more comorbid patients, or more extensive surgical cases where hospital length of stay is expected to exceed one or two days.
Quality of the evidence
The vast majority of studies reported a consistent direction of effect, with overlap of confidence intervals, and pooled participant numbers exceeded optimal effect size calculations, so we rated down no primary outcome for imprecision. Assessment of population, interventions, and outcomes of all included studies discovered no risk of indirectness. We completed a thorough grey literature search.
However, for both risk of PON and risk of POV, inspection of funnel plots strongly suggested the risk of publication bias so we downgraded the evidence strength of these outcomes to moderate.
For the outcome of risk of requiring pharmacologic treatment of PONV, there were similar concerns, as well as inclusion of relatively more studies at risk of bias, accounting for > 10% of participants in the analysis (Monti 1999; Soleimani 2018; Yogendran 1995; 395 participants). Subsequently, we decided to further downgrade this outcome to low.
For the outcome of unintended postoperative admission to hospital, there was a concern about imprecision, as indicated by wide confidence intervals indicating appreciable benefit and harm, as well as small sample sizes, in a limited number of analysed studies. There was inconsistency in effect. On account of these concerns, we downgraded this outcome to low.
There were some common pitfalls affecting the risk of bias in this literature. For the majority of included studies, there was insufficient description of measures to ensure random sequence generation and allocation concealment. Similarly, the nature of the intervention and its timing made it possible in many instances that blinding of participants and personnel could be compromised. However, we performed sensitivity analyses of studies at low risk of bias where possible, and it was reassuring that the inclusion of studies at relatively higher risk of bias did not appear to affect our estimate of risk in any outcome.
Potential biases in the review process
In general, we followed the protocols and procedures outlined by Cochrane in order to minimize any procedural bias in this review. In completing the meta‐analysis, we elected in several instances to deviate from the published protocol (Jewer 2016), which may have biased the review. For a complete list of protocol deviations and their justification, see Differences between protocol and review. Deviations that may have significantly biased the review are described below.
As included studies completed outcome assessments at a wide range of times, it was necessary to define time points for pooling of data. This was not identified a priori. We opted to define our time points in accordance with existing definitions of early and late PONV, as described in a previous meta‐analysis (Apfel 2012).
Intervention arms in several included studies prompted post‐hoc decisions on eligibility. Specifically, six papers used dextrose‐containing crystalloids (Cook 1990; Egeli 2004; Keane 1986; McCaul 2003; Ooi 1992; Shin 2007). Previous research has reported that intravenous dextrose may reduce PONV (Dabu‐Bondoc 2013). We decided to include these studies in pooled results and complete a sensitivity analysis where applicable, which suggested a negligible impact on risk reduction across all affected outcomes.
Different studies assessed and reported PON as a dichotomous or a continuous outcome. Since the majority of studies used dichotomous data, these results likely have greater precision, and are accordingly emphasized in this review. To provide an estimate of the clinical reduction in nausea, we decided to report continuous data separately. We did not contact authors for the data required to convert them to dichotomous outcomes.
Clinical heterogeneity was anticipated, therefore we planned to complete several subgroup analyses when there was also statistical heterogeneity (i.e. I2 > 40%). Specifically, we examined the relative volume of supplemental intravenous crystalloid administered between intervention and comparator groups, the timing of administration, and participant age. The subgroups definitions were chosen to best reflect the spectrum of populations and interventions present in analysed data.
The protocol intended to report length of stay in PACU as a secondary outcome. This outcome was not reported in the included studies. This was replaced with an alternative secondary outcome, unintended postoperative admission to hospital after ambulatory surgery, which was reported by three studies (Ali 2003; Cook 1990; Maharaj 2005). We chose this outcome to similarly address the potential system cost of PONV.
Finally, despite our comprehensive search for studies, one unclassified study remains, which may be a source of potential bias.
Agreements and disagreements with other studies or reviews
Prior to our meta‐analysis, the most comprehensive review of supplemental intravenous crystalloid administration for preventing PONV included 15 randomized controlled trials (Apfel 2012). The results of that review demonstrated statistically significant decreases in early, late, and cumulative PON, cumulative POV, late and cumulative PONV, and postoperative antiemetic administration. Pooled effect sizes for early and late POV and early PONV suggested a risk reduction, but 95% confidence intervals could not rule out a type I error.
Our meta‐analysis furthers the work completed in that review. By identifying new publications and completing a thorough grey literature search up to August 2018, we have included 26 additional studies, more than doubling the number of participants. This allowed for a more highly powered analysis, which explains the increased precision of our results when compared to the previous meta‐analysis. Improved power also likely explains why some outcomes, specifically early and late POV, were found to have significant risk reductions, when this was not the case in the prior analysis (Apfel 2012).
Forest plot of comparison: 1 Supplemental IV crystalloid administration for preventing PONV versus control, outcome: 1.5 Risk of overall PON (when cumulative nausea events were explicitly reported for the entire study period), as measured by the presence of subjective nausea, reported dichotomously or based on a study‐defined dichotomous threshold on a continuous scale such as a VAS.
Forest plot of comparison: 1 Supplemental IV crystalloid administration for preventing PONV versus control, outcome: 1.9 Risk of pharmacologic treatment for PONV.
Forest plot of comparison: 1 Supplemental IV crystalloid administration for preventing PONV versus control, outcome: 1.6 Risk of cumulative POV.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 1 Risk of cumulative PON.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 2 Risk of early PON.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 3 Risk of late PON.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 4 Risk of early PON, reported as a continuous outcome.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 5 Risk of late PON reported as a continuous outcome.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 6 Risk of cumulative POV.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 7 Risk of early POV.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 8 Risk of late POV.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 9 Risk of pharmacologic treatment for PONV.
Comparison 1 Supplemental IV crystalloid administration for preventing PONV versus control, Outcome 10 Risk of unintended postoperative admission to hospital.
| Supplemental IV crystalloid compared to comparator IV crystalloid volume for preventing postoperative nausea and vomiting. | |||||
| Patient or population: participants aged 6 months or older undergoing surgical procedures under general anaesthesia | |||||
| Outcome | Illustrative comparative risks* (95% CI) | Relative effect | № of participants | Certainty of the evidence | |
| Assumed risk | Corresponding risk | ||||
| Risk with comparator IV crystalloid | Risk with supplemental IV crystalloid | ||||
| Risk of PON, defined as the presence of subjective nausea, reported dichotomously or based on a study‐defined dichotomous threshold on a continuous scale such as a VAS | Cumulative events, explicitly reported for the entire study period | RR 0.62 | 1766 | ⊕⊕⊕⊝ | |
| 482 per 1000 | 183 fewer per 1000 (120 to 236 fewer) | ||||
| Early events, occurring in the first 6 hours postoperatively | RR 0.67 (0.58 to 0.78) | 2310 (20 RCTs) | ⊕⊕⊕⊝ | ||
| 307 per 1000 | 101 fewer per 1000 (68 to 129 fewer) | ||||
| Late events, occurring at the time point closest to or including 24 hours postoperatively | RR 0.47 (0.32 to 0.69) | 1682 (17 RCTs) | ⊕⊕⊕⊝ | ||
| 187 per 1000 | 99 fewer per 1000 (58 to 127 fewer) | ||||
| Risk of POV, reported dichotomously by any discrete episodes of vomiting | Cumulative events, explicitly reported for the entire study period | RR 0.50 | 1970 | ⊕⊕⊕⊝ | |
| 295 per 1000 | 147 fewer per 1000 (109 to 177 fewer) | ||||
| Early events, occurring in the first 6 hours postoperatively | RR 0.56 (0.41 to 0.76) | 1998 (19 RCTs) | ⊕⊕⊕⊝ | ||
| 106 per 1000 | 47 fewer per 1000 (25 to 63 fewer) | ||||
| Late events, occurring at the time point closest to or including 24 hours postoperatively | RR 0.48 (0.29 to 0.79) | 1403 (15 RCTs) | ⊕⊕⊕⊝ | ||
| 68 per 1000 | 35 fewer per 1000 (14 to 48 fewer) | ||||
| Risk of requiring pharmacologic treatment for PONV, reported dichotomously as the use of any medication intended to treat nausea or vomiting during the postoperative period | Cumulative events, explicitly reported for the entire study period | RR 0.62 | 2416 | ⊕⊕⊕⊝ | |
| 284 per 1000 | 108 fewer per 1000 (68 to 139 fewer) | ||||
| Risk of unintended postoperative admission to hospital, reported dichotomously as admission to an inpatient unit of a participant after an intended ambulatory surgical procedure | Cumulative events, explicitly reported for the entire study period | RR 1.05 | 235 | ⊕⊕⊝⊝ | |
| 288 per 1000 | 14 more per 1000 (66 fewer to 124 more) | ||||
| Risk of suffering a serious adverse event, reported dichotomously as the occurrence of any of: admission to high‐dependency unit, postoperative cardiac or respiratory complication, or death | Cumulative events, explicitly reported for the entire study period | ‐ | This outcome was not reported for included trials | ‐ | |
| ‐ | ‐ | ||||
| * For all outcomes, the assumed and corresponding risks (and their 95% CI) are based on the proportion of events in the comparator and intervention groups, respectively. CI: confidence interval; PON: postoperative nausea;POV: postoperative vomiting; RR: risk ratio; VAS: visual analogue scale | |||||
| GRADE Working Group grades of evidence | |||||
| 1Downgraded one level due to risk of publication bias, following inspection of funnel plot. 2Downgraded two levels due to imprecision and inconsistency. | |||||
| Study | Timing | Comparator group | Intervention group(s) |
| Preoperative | RL 2 mL/kg | RL 15 mL/kg | |
| Preoperative | No crystalloid bolus | RL 10 mL/kg | |
| Intraoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Intraoperative | RL 4 mL/kg | RL 10 mL/kg | |
| Intraoperative | RL 20 mL/kg | ||
| Preoperative | NS 1 to 2 mL/kg | NS 15 mL/kg | |
| Intraoperative | NS 4 mL/kg | NS 10 mL/kg | |
| Preoperative | RL 2 mL/kg | RL 12 mL/kg | |
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Preoperative | NS 2 mL/kg | NS 20 mL/kg | |
| Preoperative | No crystalloid bolus | RL 20 mL/kg | |
| Preoperative | RL 20 mL/kg with 1 g/kg dextrose | ||
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Postoperative | No crystalloid bolus | D5RL 60 to 120 mL/hour | |
| Intraoperative | RL 10 mL/kg/hour | RL 30 mL/kg/hour | |
| Intraoperative | No crystalloid bolus | RL 1000 mL | |
| Intraoperative | RL 10 mL/kg/hour | RL 30 mL/kg/hour | |
| Intraoperative | RL 6 mL/kg/hour | RL 18 mL/kg/hour | |
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Preoperative | No crystalloid bolus | RL 10 mL/kg | |
| Intraoperative | No crystalloid bolus | RL 4 mL/kg/hour | |
| Preoperative | RL 15 mL/kg | RL 40 mL/kg | |
| Intraoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Mixed | No crystalloid bolus | Intraoperative RL 1000 mL then postoperative 1000 mL D5W | |
| Preoperative | No crystalloid bolus | RL 900 to 1000 mL | |
| Preoperative | RL 5 mL/kg/hour | RL 30 mL/kg/hour | |
| Preoperative | RL 10 mL/kg | RL 30 mL/kg | |
| Preoperative | RL 2 mL/kg per hour fasted | RL 3 mL/kg per house fasted | |
| Intraoperative | No crystalloid bolus | RL 1.5 mL/kg per hour fasted | |
| Intraoperative | D5RL 1.5 mL/kg per hour fasted | ||
| Preoperative | No crystalloid bolus | RL 1000 mL | |
| Preoperative | RL 1.5 mL/kg per hour fasted | RL 15 mL/kg | |
| Preoperative | RL 2 mL/kg per hour fasted | RL 2 mL/kg per hour fasted then RL 10 mL/kg | |
| Preoperative | No crystalloid bolus | RL "maintenance" rate per hour fasted (maximum 1000 mL) | |
| Preoperative | No crystalloid bolus | 20 mL/kg 4% dextrose/0.18% saline solution | |
| Intraoperative | NS 10 mL/kg/hour | NS 1000 mL bolus then 10 mL/kg/hour | |
| Preoperative | RL 10 mL/kg | RL 20 mL/kg | |
| Preoperative | RL 30 mL/kg | ||
| Preoperative | RL 2 mL/kg | RL 20 mL/kg | |
| Preoperative | No crystalloid bolus | RL 30 mL/kg | |
| Preoperative | NS 1.5 mL/kg/hour | NS 1.5 mL/kg/hour then RL 5 mL/kg | |
| Intraoperative | NS 1.5 mL/kg/hour then RL 5 mL/kg | ||
| Intraoperative | No crystalloid bolus | RL 1000 mL | |
| Intraoperative | NS 10 mL/kg/hour | NS 20 mL/kg/hour | |
| Preoperative | Plasmalyte 2 mL/kg | Plasmalyte 20 mL/kg | |
| Intraoperative | RL 2 mL/kg | RL 18 mL/kg | |
| D5RL: dextrose 5% in Ringer's Lactate; D5W: dextrose 5% in water, NS: normal saline, RL: Ringer's Lactate | |||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Risk of cumulative PON Show forest plot | 18 | 1766 | Risk Ratio (M‐H, Random, 95% CI) | 0.62 [0.51, 0.75] |
| 2 Risk of early PON Show forest plot | 20 | 2310 | Risk Ratio (M‐H, Random, 95% CI) | 0.67 [0.58, 0.78] |
| 3 Risk of late PON Show forest plot | 17 | 1682 | Risk Ratio (M‐H, Random, 95% CI) | 0.47 [0.32, 0.69] |
| 4 Risk of early PON, reported as a continuous outcome Show forest plot | 5 | 415 | Mean Difference (IV, Random, 95% CI) | ‐16.38 [‐21.81, ‐10.96] |
| 5 Risk of late PON reported as a continuous outcome Show forest plot | 5 | 415 | Mean Difference (IV, Random, 95% CI) | ‐9.62 [‐14.91, ‐4.32] |
| 6 Risk of cumulative POV Show forest plot | 20 | 1970 | Risk Ratio (M‐H, Random, 95% CI) | 0.50 [0.40, 0.63] |
| 7 Risk of early POV Show forest plot | 19 | 1998 | Risk Ratio (M‐H, Random, 95% CI) | 0.56 [0.41, 0.76] |
| 8 Risk of late POV Show forest plot | 15 | 1403 | Risk Ratio (M‐H, Random, 95% CI) | 0.48 [0.29, 0.79] |
| 9 Risk of pharmacologic treatment for PONV Show forest plot | 23 | 2416 | Risk Ratio (M‐H, Random, 95% CI) | 0.62 [0.51, 0.76] |
| 10 Risk of unintended postoperative admission to hospital Show forest plot | 3 | 235 | Risk Ratio (M‐H, Random, 95% CI) | 1.05 [0.77, 1.43] |
