Learning Objectives/Outcomes: After participating in this CME/CNE activity, the provider should be better able to:

1. Describe the proposed mechanisms of action of esmolol's antinociceptive effect.

2. Explain esmolol's efficacy as a perioperative opioid-sparing adjunct.

3. Interpret esmolol's safety profile and other perioperative properties.

Esmolol is an ultra-short-acting, selective [beta]₁-receptor antagonist with rapid onset (within 2-3 minutes) and easy titratability.1 It is widely used in the perioperative setting in controlling the ventricular rate in supraventricular tachyarrhythmias and in the management of heart rate and hypertension that may occur during laryngoscopy, surgical stimulation, emergence, and extubation.2 Blunting the adrenergic response to stimuli is particularly useful in patients with ischemic heart disease, where the effects of tachycardia and hypertension can be detrimental.3 Esmolol is rapidly metabolized by esterases present in the cytosol of red blood cells and has a short postinfusion elimination half-life of 9 minutes.1,4 Potential adverse effects include bradycardia and hypotension.1 Unlike the risk of bronchoconstriction carried by nonselective [beta]-blockers, esmolol has very minimal respiratory effects and has successfully been used in patients with severe chronic obstructive pulmonary disease.5

Although esmolol has largely been used by anesthesiologists for its cardiovascular properties, its role as an analgesic adjunct has come to light recently after multiple studies have demonstrated its ability to modulate the perioperative pain response and reduce anesthetic requirements.4,6,7

The purpose of this article is to review the current evidence on esmolol's antinociceptive properties and its use as a potential opioid-sparing adjunct.

Analgesia is one of the fundamental principles of perioperative care. Conventionally, this is accomplished with the use of opioids, which are effective broad-spectrum analgesic agents. However, these drugs can lead to an increase in length of hospital stay and have undesirable adverse effects such as respiratory depression, sedation, nausea, vomiting, and constipation.8 In addition, postoperative hyperalgesia can occur with very short-acting opioid use intraoperatively, leading to increased postoperative opioid consumption.9

Some authors suggest that postoperative opioid use and resulting home prescriptions contribute to inappropriate use by some patients and diversion of the prescription pills into the community.9 The trend in anesthesia has therefore been to move toward opioid-sparing techniques to accelerate recovery after surgery and reduce postoperative opioid use.

Perioperative pain control in these cases is accomplished in a multimodal fashion targeting various components of the pain pathway, using drugs such as acetaminophen, nonsteroidal anti-inflammatory drugs (NSAIDS), lidocaine, ketamine, magnesium, and [alpha]₂-receptor agonists. Among the many adjuncts, esmolol may serve a place in these analgesic protocols to further enhance pain control while reducing the use of opioids.

Multiple single-center randomized controlled trials (RCTs) have been conducted over the past 10 to 15 years to examine the efficacy of esmolol as an antinociceptive agent. In recent years, several meta-analyses have emerged collating this broad data set, as well.4,6,7

In this literature review, we will first discuss esmolol's pharmacology and outline proposed mechanisms of its antinociceptive properties. Next, we will review relevant studies and meta-analyses regarding esmolol's efficacy as an analgesic and opioid-sparing adjunct. Finally, we will discuss esmolol's safety profile and other important perioperative anesthetic considerations when using esmolol.

Pharmacology of Esmolol and Proposed Mechanism of Antinociception

The antinociceptive properties of esmolol were first described in rats in a 2001 study that provided direct evidence of esmolol's ability to modulate pain pathways.10 In those experiments, rats were demonstrated to reduce their nociceptive behavior to formalin injection after being subjected to an esmolol infusion. Since then, several studies have demonstrated esmolol's analgesic properties in a perioperative setting across various surgical fields, including gynecology, general surgery, and orthopedic surgery.11-13

The mechanism of action of esmolol as an antinociceptive is still unclear. Several hypotheses have been presented in the literature. Some of the proposed mechanisms involve direct central nervous system (CNS) effects dependent and independent of [beta]-receptors.14-17 However, esmolol's hydrophilic nature may theoretically limit its ability to cross the blood-brain barrier and act directly on CNS structures.18 Other theories involve modulation of the proinflammatory cytokines19 and altered opioid metabolism.20

Yasui et al14 completed a study on rats to test whether esmolol affected the transmission of nociceptive information in the substantia gelatinosa of the spinal trigeminal nucleus. This was done by directly measuring neuronal signals in sectioned samples of the rat brain. The results demonstrated that esmolol selectively increased the frequency of inhibitory neuronal signals in a fashion that was dose-dependent and required the presence of extracellular calcium.

Then, they tested the effect of esmolol on inhibitory signals after [beta]-receptor activation with isoproterenol, a nonselective [beta]-agonist, and found no significant change. This suggests that esmolol functions in a [beta]-receptor-independent fashion. Yasui et al14 also found that landiolol, another short-acting [beta]-receptor antagonist, did not produce a similar rise in inhibitory neurotransmission as did esmolol. Inhibitory signals in the substantia gelatinosa physiologically play a role in reducing the transmission of excessive nociceptive signals to the brain.21

Tanahashi et al15 demonstrated that esmolol blocked tetrodotoxin (TTX)-resistant sodium channel in the rat dorsal root ganglion. TTX-resistant sodium channels on small dorsal root ganglion neurons play an important role in nociceptive pain transmission via C fibers.21 These channels are present in multiple parts of the nervous system, both central and peripheral.22

The group recorded whole cell membrane currents in dorsal root ganglion neurons of rats using the patch clamp method. Esmolol and landiolol were applied in increasing fashion and the blockade was tested using depolarizing pulses. Esmolol was demonstrated to inhibit TT-resistant sodium channels similar to lidocaine in a dose-dependent and use-dependent manner, suggesting a potential peripheral-site mechanism of action.15 Landiolol, on the other hand, was demonstrated to block receptor activities only at very high concentrations.

G proteins play a pivotal role in cellular signal transduction and may also be implicated in the central analgesic effects of dexmedetomidine.23 [beta]-blockers have been demonstrated to interact with pertussis toxin-sensitive G protein receptors.17 Activation of inhibitory G proteins can cause membrane hyperpolarization via potassium channels or inhibit neurotransmitter release through the regulation of voltage-gated calcium ion channels.24 Overall, their activation may result in reduced hypersensitivity and pain input.25,26 Hageluken et al17 did not test esmolol specifically but did demonstrate that more lipophilic [beta]-blockers had more of an effect on G protein activation. Given its hydrophilicity, esmolol is less likely to cause G protein activation.

Menigaux et al16 hypothesized that a possible mechanism may be gleaned by studying the effects of esmolol on the arousal/electroencephalogram (EEG)/Bispectral Index (BIS) response. [beta]-receptors are present in various parts of the brain, including the reticular-activating system, and [beta]-agonism, by way of IV epinephrine, has been demonstrated to increase the arousal response.27 [beta]-antagonism centrally has been demonstrated in animals to reduce this response.28 It follows that any noxious stimulus such as laryngoscopy or surgical incision causes increased central catecholamines and arousal, which can be blunted by central action of [beta]-blockade.

Kim et al19 completed a double-blind RCT to analyze the effect of esmolol on inflammatory biomarkers during and after surgery. Patients undergoing laparoscopic gastrectomy for gastric cancer were examined and divided into 3 groups: a clinical dose group, subclinical dose group, and a control saline group.19 The results demonstrated that esmolol infusion attenuated the rise in IL-6 and IL-10 to a significant degree and suppressed IL-4 and C-reactive protein 1 hour after surgery.19 These results are corroborated by prior studies demonstrating increased inflammatory markers caused by [beta]-agonism and animal studies demonstrating propranolol attenuating IL-10 increase in animal models.29,30 These findings together suggest the role of esmolol in modulating the immunologic and inflammatory responses perioperatively and may thus affect pain pathways in that manner.

[beta]-blockers may also alter pharmacokinetics of short-acting opioids; by reducing cardiac output and hepatic blood flow, [beta]-blockers may reduce the elimination clearance of opioids, thus prolonging their effect.20 Avram et al20 studied the effects of a propranolol infusion on the concentration of antipyrine, a pharmacokinetic of lipophilic agents metabolized in the liver. They found that the propranolol infusion resulted in a higher blood concentration of antipyrine in the first few minutes following bolus, suggesting a reduced first-pass metabolism, and early drug distribution among the tissues.20

Postoperative Pain Scores and Perioperative Opioid Use

Many single-center trials,11-13,31-47 followed by several meta-analyses4,6,7 have been published studying esmolol's analgesic properties. There is significant heterogeneity among the single-center trials as esmolol dosing, type of surgery, and outcome measures vary greatly. In addition, control groups were crystalloids, local anesthetic infiltration, or opioids.

First, we will briefly review several representative RCTs.11-13 Then, we will discuss the results of the meta-analyses and summarize any conclusions that can be drawn.

Randomized Controlled Trials

Chia et al11 conducted a double-blind, RCT in 97 patients undergoing total abdominal hysterectomy. Patients were divided into either the esmolol group (0.5 mg/kg bolus, followed by 50 [micro]g/kg/min infusion) or the placebo normal saline infusion group.11 A standard set of induction agents (fentanyl, thiopental, succinylcholine) was used on all patients, and intraoperatively, fentanyl or isoflurane was titrated to meet a predefined criteria of inadequate anesthesia depth.11 Results demonstrated reduced intraoperative fentanyl use in the esmolol group (P = 0.006).11 There was also less patient-controlled analgesia (PCA) morphine use in the esmolol group across all studied time periods and cumulatively across 3 days (37.3 +/- 8.4 mg vs 54.7 +/- 11.2 mg, P = 0.005).11 Differences in pain scores and PCA medication adverse effects between the 2 groups were not statistically significant.11

Collard et al12 studied esmolol infusions in 90 patients undergoing elective outpatient laparoscopic cholecystectomies. Patients were randomly divided into 3 groups: fentanyl control group (1 [micro]g/kg at induction and 50 [micro]g every 30 minutes thereafter), esmolol group without supplemental opioids (1 mg/kg bolus at induction and an infusion of 5-15 [micro]g/kg/min thereafter), and remifentanil group (1 [micro]g/kg bolus at induction and an infusion of 0.1-0.5 [micro]g/kg/min).12 Drugs were titrated to maintain heart rate within 20% of baseline and postanesthesia care unit (PACU) pain scores, fentanyl requirements, and length of stay were recorded.12 The results demonstrated that the amount of fentanyl used in the PACU was less in the esmolol group (91.5 +/- 42.7 [micro]g) compared with the remifentanil (237.8 +/- 54.7 [micro]g) and fentanyl groups (168.1 +/- 96.8 [micro]g) (P < 0.0001).12 The esmolol group also left the hospital by 45 to 60 minutes earlier than the remifentanil group (P < 0.004).12 There was no significant difference in PACU pain scores (P > 0.05).12

Haghighi et al13 randomized 82 patients requiring surgical fixation of tibial fractures into an esmolol infusion group (bolus of 0.5 mg/kg, followed by 5 [micro]g/kg/min infusion) and saline control group in a double-blinded fashion. Infusions began 30 minutes before surgery and continued till closure of surgical incision.13 Patients underwent a standardized induction, with isoflurane and nitrous oxide to maintain anesthesia.13 Intraoperatively, BIS levels were maintained between 40 and 60 through titration of isoflurane.13 The visual analog scale (VAS) pain scores were recorded postoperatively and meperidine given for breakthrough pain.13 The esmolol group was found to have lower VAS scores in the immediate postoperative period (P = 0.02), at 3 hours (P = 0.0001), and 6 hours (P = 0.0001).13 In addition, patients were administered less total postoperative meperidine in the esmolol group (P = 0.004).13

Systematic Reviews and Meta-Analyses

Through our literature review, we found 3 systematic reviews and meta-analyses.4,6,7 A total of 31 unique clinical studies were included across all review articles. Gelineau et al4 performed the most recent and comprehensive review. They included 23 RCTs, including the 3 summarized previously.4 In all studies, esmolol was given as a bolus, followed by an intraoperative infusion with variable dosing. Studies were first separated into either esmolol versus crystalloid/local anesthetic infiltration control or esmolol versus opioid control.

Gelineau et al4 reviewed 16 RCTs comparing esmolol with crystalloid/local anesthetic infiltration. In 11 of 12 RCTs, esmolol infusions improved pain scores, either by VAS or Verbal Rating Scale, in the initial 0 to 60 minutes upon arrival to recovery.13,31-33,41-48 Most studies demonstrated no significant difference in pain scores after 60 minutes.4 Five of 6 studies demonstrated a significant decrease in required rescue analgesic in the esmolol group.4 Cumulative analgesic dose requirement in the recovery area, whether opioid, tramadol, or NSAID, was reduced in 11 of 12 studies.11,13,31-33,41-47 Three of 4 studies demonstrated that the esmolol infusion arm delayed time to first rescue analgesic dose.13,33,42,46 Meta-analysis of relevant studies demonstrated that the perioperative esmolol group received less morphine equivalents while in the PACU than did patients in the crystalloid control group (standardized mean difference [SMD]: -1.23; P < 0.001).4 Patients randomized to esmolol also had lower VAS pain scores than the crystalloid control group up to 60 minutes after surgery (SMD: -1.25; P < 0.001).4

Gelineau et al4 also reviewed 7 RCTs comparing esmolol with an opioid control. All studies used ultra-short-acting or short-acting opioids such as remifentanil, alfentanil, or fentanyl as the control group.4 When postoperative pain scores were compared, 6 of 7 studies demonstrated no difference or favored the esmolol group when compared with the control; one study favored the opioid group.12,35-40 Three studies reported the proportion of patients requiring rescue analgesics.35,38,39 Of the 3, 1 study favored the esmolol group,38 1 favored the opioid group,39 and 1 reported no difference.35 Five studies reported the cumulative dose of opioids at various time points postoperatively.12,35,36,38,40 Three of 5 reported less opioid consumption in the esmolol group12,36,38 whereas the other 2 reported no difference.35,40 Meta-analysis of relevant studies demonstrated that while in PACU, patients in the esmolol group were administered less morphine equivalents than patients in the opioid control groups (SMD: -1.18; P = 0.048).4 Patients randomized to esmolol had higher VAS pain scores than the opioid control group up to 60 minutes after surgery, but this was not statistically significant (SMD: -1.25, P = 0.207).4

Gelineau et al4 also reviewed intraoperative opioid using studies when opioid dose was titrated to hemodynamic goals; 7 RCTs were included.11,37,42,44,45,47,49 Esmolol infusion was compared with either intraoperative remifentanil37,42,45,47 or other short-acting opioids including fentanyl and alfentanil.11,44,49 Meta-analysis demonstrated that patients in the esmolol group were administered less morphine equivalents than in control groups (SMD: -1.60, P < 0.001).4

In summary, the aforementioned studies largely support the following regarding esmolol given as a bolus, followed by an infusion4:

1. It reduces intraoperative opioid use when compared with an opioid control.

2. It reduces immediate postoperative pain scores when compared with a crystalloid control.

3. It reduces PACU opioid/rescue analgesic use when compared with a crystalloid control and when compared with an opioid control.

These results demonstrate that esmolol may have an important role in a balanced, opioid-sparing anesthetic. Poorly controlled postoperative pain leads to poor quality of life, delays in hospital discharge, and unanticipated hospital admissions after surgery.9,50,51 Enhanced recovery after surgery (ERAS) protocols often call for perioperative opioid reduction to minimize postoperative nausea and vomiting (PONV), ileus, and respiratory depression.52,53 Esmolol may have a useful role in these protocols, specifically in ambulatory surgery or when significant postoperative pain is expected.

Adverse Effects and Perioperative Considerations

Hemodynamics

Some clinicians may be hesitant in using esmolol because of the risk of bradycardia and hypotension. Esmolol is indeed contraindicated in settings of preexisting bradycardia, heart block greater than the first degree, cardiogenic shock, and heart failure.54 In conditions in which the heart is relying on its contractility for circulatory support, [beta]-blockade can depress myocardial function and hemodynamic parameters further. In that regard, the effects of the esmolol on heart rate and blood pressure were observed in most studies.

Some studies report lower mean heart rates and/or blood pressures compared with controls37,41 and others found no changes in one or both of these parameters.11,35 If subjects met criteria for severe bradycardia or hypotension as defined in individual studies, they were given short-acting agents such as atropine or ephedrine and were excluded from study results. Keeping in mind that various dosing regimens were used and patients were largely American Society of Anesthesiologists (ASA) grade 1 or 2, no study reported any clinically significant changes or adverse outcomes because of bradycardia, hypotension, or bronchospastic risk of esmolol.4

It is key to perform a safety and risk assessment on a case-by-case basis before undertaking this anesthetic technique. Esmolol's rapid metabolism and short half-life theoretically implies that any adverse effects should resolve quickly.

Postoperative Nausea and Vomiting

Considering the risk of PONV is always important before administering perioperative medications. PONV is distressing for patients and may cause delays in discharge and increased health care costs.55,56 Thiruvenkatarajan et al57 performed a meta-analysis of 439 patients across 8 RCTs and concluded that perioperative esmolol may reduce incidence of PONV when compared with short-acting and ultra-short-acting intraoperative opioids. Even though the incidence of PONV was lower in the esmolol group, time spent in recovery was not reduced. RCTs by Hwang, Moon, and Gokce also studied PONV as a secondary outcome when esmolol was compared with placebo32,45 or remifentanil infusion.37 All 3 studies demonstrated decreased perioperative opioid use in the esmolol group. However, in contrast to Thiruvenkatarajan et al,57 there was no difference in PONV incidence.

The mechanism of action of esmolol as a potential antiemetic is unclear. In studies comparing esmolol with perioperative opioids, the decreased use of intraoperative opioids in the esmolol group, a well-studied risk factor for PONV, is a logic conclusion. In addition, Gupta and Chaudhary58 demonstrated that [beta]-blockade may have direct antiemetic properties. The role of esmolol as an adjunct for PONV is still unclear, and more high-quality RCTs and reviews are needed.

Effect on Electroencephalogram

The electroencephalogram (EEG) is a neurophysiologic monitor that detects electrical activity in the brain and may be used to monitor depth of anesthesia while a patient is anesthetized.59 Bispectral Index or BIS monitoring and Entropy monitoring are processed EEGs that are approved for perioperative use as devices to prevent recall in patients and titrate anesthetic agents accordingly. Menigaux et al16 studied the effect of esmolol infusion on BIS monitoring during endotracheal intubation after induction of anesthesia. After induction, vitals and BIS levels were trended before laryngoscopy.16 Interestingly, BIS values remained unchanged between the 2 groups before laryngoscopy but during intubation, BIS levels rose by 40% in the control group versus 8% in the esmolol group (P < 0.01).16 This suggests that esmolol may not have direct anesthetic properties but instead blunts the sympathetic response.

Johansen et al60 also found that when given as a bolus followed by infusion, esmolol decreased EEG parameters by 40% (BIS 37 +/- 6 to 22 +/- 6 [mean +/- SD]) compared with placebo during elective surgery. As demonstrated by Berkenstadt et al,61 a single bolus dose does not seem to have the same effect.

Based on the aforementioned studies, esmolol may potentiate sedation specifically during a sympathetic response and only in the setting of a continuous infusion. More high-quality studies are needed to understand the intricacies of esmolol's effect on EEG.

Requirements for Anesthetic Agents

Studies comparing the total requirement of volatile anesthetic during an esmolol versus placebo infusion demonstrate varying results. Bhawna et al41 used isoflurane and found that when titrated to response entropy and state entropy, there was no difference in volatile requirement when an esmolol infusion was used versus placebo during lower abdominal surgery (P = 0.385). Moon et al32 used sevoflurane and found that when titrated to BIS and systolic blood pressure, the esmolol group used significantly less sevoflurane by percent volume than placebo (P < 0.01). Johansen et al60 used isoflurane and compared how esmolol alone, alfentanil alone, and a combination of alfentanil with esmolol effected minimum alveolar concentration (MAC) requirements. Results demonstrated that esmolol alone had no effect on MAC requirement, alfentanil decreased MAC requirement by 25%, and a combination of high-dose esmolol and alfentanil decreased MAC by almost 42%.60 More homogenous studies are needed to understand esmolol's effect on common volatile agents.

Whether or not esmolol decreases propofol requirements is unclear. Watts et al6 reviewed 3 studies analyzing the effect of an esmolol bolus and infusion on the induction dose of propofol. Their results demonstrated that the induction dose of propofol was decreased by a mean of 0.53 mg/kg (P = 0.0001).6 Johansen et al62 also found that high-dose esmolol (1 mg/kg bolus, then 250 [micro]g/kg/min infusion), but not low-dose esmolol (0.5 mg/kg bolus, then 50 [micro]g/kg/min infusion), reduced propofol concentration at which 50% of patients would not move to surgical stimulus (Cp50) (P < 0.04). However, Orme et al63 found that esmolol did not reduce Cp50 in their study with propofol-only anesthesia, suggesting that esmolol may act synergistically with opioids to reduce propofol requirements.

Cost

It remains unclear whether routine use of esmolol as an opioid-sparing adjunct would be cost effective. Compared with labetalol, a nonselective [beta]-adrenergic and [alpha]₁-adrenergic blocker, esmolol is typically about 5 to 10 times more expensive.64,65 For reference, at our local county hospital, one 100 mg/10 mL vial of esmolol costs about $12. The cost effectiveness of implanting a [beta]-blocker initiation algorithm in the setting of abdominal aortic aneurysm (AAA) surgery was studied by Fleisher et al.66 They found a cost savings when calculating the cost of uncomplicated AAA versus complicated AAA despite using relatively expensive [beta]-blockers such as esmolol, atenolol, and bisoprolol. The cost effectiveness of esmolol as a potential opioid-sparing agent is not well studied and represents a significant gap in the literature.

Given that the economic burden of opioid overdose, abuse, and dependence in the United States is approximately $78.5 billion, it is not unreasonable to hypothesize that use of esmolol may be cost effective in the appropriate setting.67

Conclusion

The effects of esmolol as an anesthetic adjunct are well studied. The literature demonstrates that when used as a combined bolus and continuous infusion, esmolol reduces intraoperative opioid use, reduces immediate postoperative pain scores, and reduces PACU opioid use.

Some individual RCTs also support esmolol's role in decreasing anesthetic requirements, reducing PONV, improving PACU discharge time, and reducing opioid use and pain scores beyond the immediate postoperative period. It is important to emphasize that these outcomes have not been demonstrated for a bolus dose of esmolol alone but rather a bolus dose combined with an infusion.

Most studies conducted are single-center RCTs with small sample sizes, which limits the strength of some of the conclusions that can be drawn. Several comprehensive systematic reviews and meta-analyses have also been completed, despite the significant heterogeneity in manner of study design, data collection, outcome measures, and esmolol dosing regimens between individual RCTs.

Overall, these data demonstrate an improvement in postoperative pain control and a reduction in opioid use in a combined bolus and infusion dose of 0.5 to 1 mg/kg and 50 to 250 [micro]g/kg/min.

Esmolol's safety profile is evident in the studies, most of which did not report adverse outcomes related to administration of the drug. Many investigators report stable hemodynamics throughout the duration of the infusion, and there are no reports of intraoperative recall.

Several possible mechanisms have been proposed to explain esmolol's anesthetic and antinociceptive properties involving the inflammatory pathways and the peripheral and central nervous systems. However, the exact mechanism remains the subject of ongoing investigations.

Based on our review of the literature, we believe that esmolol should be considered as a part of any anesthetic plan if the patient is undergoing noncardiac surgery, if there is concern for significant postoperative pain, and if there are no contraindications to its use.

Given the patient demographics in the RCTs on this review, we recommend avoiding this technique in patients at the extremes of age (younger than 18 years or older than 85 years); ASA grade 3 or above; or with hepatic failure, renal failure, cardiac failure, diabetes mellitus, morbid obesity (body mass index > 40) and severe reactive airway disease; or in surgeries that involve large hemodynamic and fluid shifts.

Patients undergoing all manner of surgeries could potentially benefit from this technique, with benefits demonstrated in laparoscopic, gynecologic, and orthopedic surgeries. Elective open abdominal procedures that are typically enrolled in an ERAS protocol are ideal candidates. If bradycardia (pulse <40) or hypotension (mean arterial pressure < 60) is encountered as an effect of esmolol, then our recommendation would be to treat using short-acting agents such as atropine or ephedrine, and esmolol infusion dose adjustment. If bradycardia or hypotension persists, then discontinuing the infusion can be considered.

Finally, given the concern for esmolol masking intraoperative awareness by way of blunting the hemodynamic response, we recommend that this technique be used in conjunction with EEG monitoring such as the BIS monitor.

Going forward, more studies are needed to understand how esmolol affects the depth of anesthesia, volatile and intravenous anesthetic requirements, and the cost effectiveness of its routine use as an opioid-sparing adjunct. There also needs to be a consensus on drug dosing, and efficacy should be studied when used in combination with other multimodal analgesic agents.

References