Opioid genetics: the key to personalized pain control?

Authors

  • R Branford,

    1. Department of Medicine, Royal Marsden NHS Foundation Trust, London, UK
    2. Department of Clinical Genomics, National Heart and Lung Institute, Imperial College, London, UK
    Search for more papers by this author
  • J Droney,

    1. Department of Medicine, Royal Marsden NHS Foundation Trust, London, UK
    2. Department of Clinical Genomics, National Heart and Lung Institute, Imperial College, London, UK
    Search for more papers by this author
  • JR Ross

    Corresponding author
    1. Department of Clinical Genomics, National Heart and Lung Institute, Imperial College, London, UK
    • Department of Medicine, Royal Marsden NHS Foundation Trust, London, UK
    Search for more papers by this author

  • Nothing to declare.

Corresponding author: Dr Joy Ross,

Department of Medicine,

Royal Marsden NHS Foundation Trust,

Fulham Road, London SW3 6JJ, UK.

Tel.: +02078082761;

fax: +02078118132;

e-mail: joy.ross@rmh.nhs.uk

Abstract

There are now several strong opioids available to choose from for the relief of moderate to severe pain. On a population level, there is no difference in terms of analgesic efficacy or adverse reactions between these drugs; however, on an individual level there is marked variation in response to a given opioid. The genetic influences to this variation are complex, and although current research has shown some promising results, these have not been replicated across larger studies and as such the ultimate aim of personalized prescribing remains elusive. If personalized prescribing could be achieved this would have a major impact at an individual level to facilitate safe, effective and rapid symptom control. This review presents some of the recent positive advances in opioid pharmacogenetic studies, focusing on associations between candidate genes and the three main elements of opioid response: analgesic, upper gastrointestinal and central adverse reactions.

The fast, effective and safe relief of pain is a priority in clinical medicine. Opioids are routinely used in the treatment of moderate or severe acute and chronic pain. Morphine is the archetypal opioid and has been used successfully for thousands of years. There are now many alternative opioids available, for example, fentanyl, oxycodone and hydromorphone, each with comparable efficacy at a population level [1, 2]. At an individual level, there is wide variation in clinical response to these drugs in terms of both analgesic efficacy and adverse reactions. For example, up to 30% of patients do not respond well to morphine, the majority of whom then achieve a better clinical outcome by switching to an alternative opioid [3, 4]. Common adverse reactions to opioids include central side effects such as drowsiness, confusion and hallucinations and upper gastrointestinal side effects such as nausea and vomiting.

Pharmacogenomics is the study of how genetic variation may influence the pharmacokinetic and pharmacodynamic profile of a drug and thereby impact on an individual's response and is central to the idea of personalized medicine. Pharmacodynamics is the study of a drug's mechanisms of action, encompassing the role of different receptors, second messengers and downstream effects of a drug once a receptor is activated. If pharmacodynamics describes what the drug does to the body, then pharmacokinetics explains what the body does to the drug, i.e. mechanisms of absorption, distribution, metabolism and elimination. Interindividual variation in response to opioids will also be influenced by a patient's susceptibility to or perception of pain; therefore, many more candidate genes from pain signaling and modulatory pathways will also influence overall response.

Animal studies, in particular gene knockout studies [5], give valuable information to identify candidate genes; however, translational research into the clinical setting can be challenging. For pain research, normal volunteer studies, controlled studies in post-operative pain and cancer-related pain represent a spectrum of increasing variability and ‘noise’ with increased scope for confounding factors that further dilute the power of a study to detect a true association. In addition, small patient numbers and different response phenotypes make comparisons between studies difficult. This thematic review summarizes the evidence, primarily from clinical candidate gene association studies since this is the approach taken to date, on opioid response. First, we consider analgesic response and second the two most common classes of adverse reactions: upper gastrointestinal and central adverse reactions.

Opioid response phenotypes

Several opioid response phenotypes have been used to investigate potential genetic elements behind clinical response to opioids. Primarily, studies have focused on the beneficial analgesic effects; however, more recently, the adverse effects such as nausea, vomiting and confusion and drowsiness have also been studied. It is likely that different mechanisms contribute to these elements of opioid response. Table 1 summarizes current positive genetic associations with the three key opioid response phenotypes in the clinical setting.

Table 1. Summary of positive opioid response genetic association studies in the clinical setting
GenePolymorphismOpioidStudy populationnClinical effectReference
Analgesic response
OPRM1rs1788871 (A118G)MorphineCancer pain patients207Increased dose requirementKlepstad et al. [51]
  MorphineCancer pain patients145Decreased pain reliefCampa et al. [34]
  MorphinePost-operative patients80Increased dose requirementChou et al. [49]
  MorphinePost-operative patients147Increased dose requirementChou et al. [48]
  MorphinePost-operative patients588Increased dose requirementSia et al. [47]
  FentanylLaboring women223Decreased dose requirementsLandau et al. [95]
  MixedChronic pain patients352Increased painLotsch et al. [35]
COMTrs4680 (G472A)MorphineCancer pain patients207Decreased dose requirementsRakvag et al. [66]
ABCB1rs1045642 (C3435T)MixedChronic pain patients352Decreased dose requirementsLotsch et al. [35]
 rs1045642 (C3435T)MorphineCancer pain patients145Increased pain reliefCampa et al. [34]
CYP2D6Poor metabolizersCodeinePost-hysterectomy patients11Decreased analgesiaPersson et al. [14]
 Poor metabolizersTramadolPost-operative patients300Decreased analgesiaStamer et al. [18]
CYP3ACYP3A4*1GFentanylPost-operative patients143Decreased dose requirementsZhang et al. [26]
 CYP3A5*3 and CYP3A4*1GFentanylPost-operative patients203Decreased dose requirementsZhang et al. [25]
TNFrs1800629 (G308A)MixedCancer pain patients140Decreased pain reliefReyes-Gibby et al. [77]
IL6rs1800795 (G174C)MixedCancer pain patients140Increased dose requirementsReyes-Gibby et al. [77]
Upper gastrointestinal adverse reactions
OPRM1rs1799971 (A118G)MorphinePost-operative patients102Decreased nauseaKolesnikov et al. [54]
 rs1799971 (A118G)TramadolOsteoarthritis160Decreased nausea and vomitingKim et al. [24]
 rs 1799971 (A118G)MorphinePost-operative patients588Decreased nausea and vomitingSia et al. [47]
COMTrs4680 (G1947A)MorphinePost-operative patients102Decreased nauseaKolesnikov et al. [54]
HTR3Brs1672717MixedCancer pain patients1579Decreased nausea and vomitingLaugsand et al. [67]
CYP2D6Extensive metabolizersTramadolOsteoarthritis154Increased nausea and vomitingKim et al. [24]
UGT2B7rs7439266 *2 (C802T)MorphineCancer pain patients32Decreased nausea and vomitingFujita et al. [94]
ABCB1G2677T/A and C3435T diplotypeMorphinePost-operative patients74Decreased nausea and vomitingCoulbault et al. [37]
Central adverse reactions
OPRM1rs1799971 (A118G)MorphinePost-operative102Decreased sedationKolesnikov et al. [54]
MDR1rs2032582 (G2677T/A)MorphineCancer pain patients228Decreased drowsiness/confusionRoss et al. [38]
COMTrs740603 (A-4873G)MorphineCancer pain patients228Decreased drowsiness/confusionRoss et al. [38]

Analgesic response

Several different phenotypes have been used to compare analgesic response to opioids. These include: dose requirements to achieve pain control and patient-reported measures of pain relief. Outcomes have been measured on continuous visual analog scales, discrete numerical rating scales and binary groupings of variables, for example, severe/non-severe pain. Most studies investigate clinical responses to a single opioid; however, some group several opioids together and calculate ‘dose equivalents’, which adds an extra layer of complexity and ‘noise’ given that dose conversions are imprecise and highly variable between individuals [6].

Upper gastrointestinal adverse reactions: nausea and vomiting

Nausea is one of the most common reasons to discontinue morphine treatment and trigger switching to an alternative opioid [4]. The ‘vomiting center’ in the medulla oblongata receives input from four major areas: the chemoreceptor trigger zone (CTZ), the gastrointestinal tract, the vestibular apparatus in the temporal lobe, and the cerebral cortex (Fig. 1). The emetic effects of opioids are thought to be due to multiple mechanisms involving the classical opioid receptors, principally through stimulation of the CTZ, inhibition of gut motility, and stimulation of the vestibular apparatus. At the CTZ, μ-and δ-opioid receptors are activated, and signaling to the vomiting center occurs via dopamine D2 receptors and serotonin (5-HT3) receptors [7]. Genetic variation has been described in genes coding for these receptors.

Figure 1.

Simplified diagram of nausea and vomiting pathways including published genetic associations [24, 47, 54, 67, 94]. CHRM3 (cholinergic receptor, muscarinic 3 gene), HRT3B [5-HT (serotonin) receptor 3B gene], OPRM1 (mu opioid receptor gene), CYP2D6 (cytochrome P450 2D6), UGT2B7 (UDP-glucuronosyltransferase 2B7), and ABCB1 (multidrug resistance gene encoding P-glycoprotein).

Central adverse reactions: drowsiness, confusion and hallucinations

Central adverse reactions to opioids include drowsiness, confusion, hallucinations and nightmares, which often occur together in clinical practice. Prospective studies indicate that sedation or drowsiness is observed in between 20% and 60% of patients receiving oral morphine for chronic cancer pain [8]. Drowsiness and confusion are also one of the most common reasons for patients to stop morphine treatment and trigger opioid switching [4]. The mechanisms behind central adverse reactions to opioids have been debated. Pharmacological studies suggest a role for κ-opioid receptors in mediating the dysphoric and sedative effects of opioids [9]. Other theories suggest that different opioid metabolites contribute to central adverse reactions [10].

Candidate genes

Pharmacokinetics

Opioid metabolism

There is no universal metabolic pathway for the metabolism of opioids, and different enzymes from the cytochrome P450 and UDP-glucuronosyltransferase (UGT) systems are important for different drugs. In cytochrome P450, the two common pathways are 2D6 and 3A.

Cytochrome P450 2D6

The cytochrome P450 enzyme 2D6 (CYP2D6) is involved in the metabolism of several opioids including codeine, tramadol and oxycodone (Fig. 2). Over 70 CYP2D6 alleles have been described, which directly affect the final protein including both single nucleotide polymorphisms (SNPs), deletions, insertions, and copy number variation [11]. The sum effect of variation on phenotype has been classified into four major groups: poor metabolizers, intermediate metabolizers (IMs), extensive metabolizers (EMs) and ultra-rapid metabolizers.

Figure 2.

Major metabolic pathways for (a) codeine and morphine, (b) oxycodone and (c) tramadol [93].

Codeine is partially (10%) metabolized to morphine by CYP2D6 [12]. Ten percent of Caucasians are poor metabolizers and experience little analgesia from codeine [13, 14]. At the other end of the spectrum, 3% of Caucasians are ultra-rapid metabolizers and have a higher incidence of codeine-related adverse reactions [15]. There are case reports of fatal neonatal opioid toxicity in children who are breastfed by CYP2D6 ultra-rapid metabolizing mothers who have ingested codeine [16]. Recent data suggest that CYP2D6 phenotype also plays a role in clinical response to tramadol and oxycodone by significantly altering ratios of parent opioid to its more active metabolite [17, 18].

The clinical relevance of the activity of the main metabolites of oxycodone has been debated. Noroxycodone is inactive in contrast to oxymorphone, which has a greater analgesic potency compared with oxycodone [19, 20]. It is currently unclear whether variation in CYP2D6 activity, either through drug inhibition and/or through genetic variation, significantly alters the clinical efficacy of oxycodone [17, 21, 22]. In one study of 450 cancer patients receiving oxycodone, although CYP2D6 metabolizer status influenced the ratios of oxycodone metabolites in the expected direction, this did not translate into differences between analgesic response [23].

CYP2D6 is important in the metabolism of tramadol to an active metabolite O-desmethyltramadol. In Korean patients taking tramadol for osteoarthritis of the knee, CYP2D6 metabolizer status was assessed and classified as EMs or IMs. IM had 3.4-fold lower odds of nausea/vomiting than EM, suggestive of a relationship with O-desmethyltramadol levels (p = 0.0051) [24].

There have been no positive studies linking 2D6 genetics to opioid-induced central side effects. One study of cancer patients receiving oxycodone observed no significant difference in cognitive function, as measured by the Mini Mental State, between those with different CYP2D6 metabolizer status [23].

Cytochrome P450 3A

The CYP450 3A superfamily of enzymes is involved in the metabolism of 50% of all known drugs. 3A4 and 3A5 are clinically important in adults, and a number of functional polymorphisms have been identified [25, 26]. Some substrates including opioids, e.g. oxycodone and fentanyl can be metabolized equally by 3A4 or 3A5, and therefore a defect in one enzyme may be compensated for by the other [25].

The frequencies of CYP3A polymorphisms are highly variable between different ethnic groups. The CYP3A4*1G polymorphism has been implicated in decreased enzyme activity and fentanyl consumption in post-operative pain control [26]. A further study looked at the interaction between 3A4 and 3A5 genetic polymorphisms, and although 3A5 variation was not independently important, interactions between 3A4 and 3A5 polymorphisms were additive and significantly influenced post-operative analgesia.

UDP-glucuronosyltransferase 2B7

The hepatic isoenzyme UGT 2B7 is primarily responsible for morphine metabolism. Levels of UGT expression vary widely within the population [27]. In vitro work has suggested that functional genetic variants in UGT2B7 are linked to altered levels of mRNA expression [28, 29] and enzyme activity with differential metabolite production [28]. The main metabolites of morphine, morphine-3-glucuronide (M3G) and morphine-6-glucuronide (M6G), account for approximately 50% and 10% of metabolites, respectively [30]. M3G binds poorly to opioid receptors, and although it lacks analgesic properties, it has several neuroexcitatory effects and may be responsible for hyperalgesia, allodynia and myoclonic jerks [31].

Unlike M3G, M6G has been shown to have analgesic qualities and has been used as an analgesic agent in its own right. In vitro and animal studies suggest that M6G is more potent than its parent morphine because of higher efficiency of receptor activation. Clinical studies have linked genetic variation to differences in morphine/metabolite ratios [32], but not to overall clinical response to morphine [33].

Multidrug resistance gene

The multidrug resistance gene or ATP-binding cassette subfamily B, member 1 (MDR1 or ABCB1) encodes P-glycoprotein. P-glycoprotein is a membrane transporter important in regulating drugs across the blood–brain barrier and actively pumps drugs out of the central nervous system (CNS). Carriage of the ABCB1 3435T allele has been associated with increased pain relief from morphine in cancer-related pain [34] and decreased morphine equivalent daily dose (MEDD) in mixed chronic pain population [35].

Zwisler et al. studied ABCB1 polymorphisms and oxycodone analgesia and adverse reactions in an experimental pain study; the variant alleles 2677A and 3453T were protective against nausea and vomiting [36]. However, in a post-operative pain study, use of anti-emetics for morphine-related nausea and vomiting was decreased in patients who were homozygous for the 2677GG/3435CC diplotype [37].

Carriage of the A allele at position 2677 of ABCB1 has been reported to be protective of central side effects, i.e. drowsiness and confusion in patients treated with morphine for cancer-related pain [38]. P-glycoprotein is a transmembrane transporter that actively pumps drugs including morphine across the blood–brain barrier. Functional variants affecting activity of the transporter may influence drug concentrations and parent drug/metabolite ratios in the CNS and consequently central adverse reactions; G2677T/A has been shown to be functional and linked to altered expression of P-glycoprotein in vivo [39, 40].

Pharmacodynamics

Opioid receptors

Opioid receptors are the target receptors of all strong opioids and are found throughout the nervous system. There are three different types of classical opioid receptor: mu (μ), kappa (κ) and delta (δ). They are all G protein-coupled receptors with an extracellular N-terminus, seven transmembrane domains and an intracellular C-terminus. The three receptors share a high degree of homology with most variation found in the extracellular loops and N-terminal domains [41, 42]. The extracellular loops are particularly important as they determine ligand binding, and there is differential binding of the common opioids. Opioid receptor subtypes generated by alternative splicing have also been classified through pharmacological studies [43]. The classical receptors are also thought to interact and form heterodimers with each other and other G protein-coupled receptors [43-46].

Mu opioid receptor

Knockout studies in mice show that the μ-opioid receptor is essential for morphine-induced analgesia [5]. Genetic variation in the μ-opioid receptor gene (OPRM1) has been associated with variation in opioid response in different settings including acute post-operative pain [47-49], chronic non-cancer pain [35, 50] and cancer-related pain [34, 51].

The most studied SNP in OPRM1 is the non-synonymous exonic SNP A118G, rs1799971. The variant G allele has been associated with increased doses of morphine to achieve pain control in cancer patients [34, 51] and patients following surgery [47-49]. Conversely, the common A allele has been associated with increased analgesia from morphine in cancer-related pain [34]. However, a recent meta-analysis of opioid pain studies showed no overall association with increased pain and only weak associations with increased morphine dose requirements in homozygous carriers of the variant G allele [52]. The functional significance of A118G remains unclear [50].

The largest genetic association study of opioid response to date, European Pharmacogenetic Opioid Study (EPOS), included 112 SNPs in 25 genes, including OPRM1, OPRK1 and OPRD1 [53]. This study explored genetic associations with oral equivalent morphine dose requirements in 2294 patients taking a variety of strong opioids for cancer-related pain, but did not identify associations with any of the 112 SNPs in 25 candidate genes tested in both development and validation analyses [53].

Two studies of morphine in the post-operative period and one of tramadol for osteoarthritis have reported an association with the variant G allele of rs1799971 and less nausea/vomiting [24, 47, 54]. The A118G genotype was not associated with fentanyl-induced post-operative nausea and vomiting in another similar sized study of 165 Chinese women who had undergone gynecological surgery [55]. Meta-analysis of genetic association studies in opioid response shows only a weak association with nausea in homozygous carriers of the variant G allele of rs1799971 [52].

In patients receiving morphine post-operatively, carriers of the variant G allele rs1799971 had less sedation [54]. Although the functional significance of this polymorphism remains unclear, the pattern of less analgesia, but also less adverse reactions (upper gastrointestinal and central) suggests reduced receptor sensitivity to morphine.

STAT6

STAT6 is an important transcription factor activated following stimulation by TH2 cytokines such as IL-4 [56]. IL-4 has been shown to induce upregulation of μ-opioid receptor expression. A functional SNP in the binding site for STAT6 in the promoter region of OPRM1 has been shown to reduce its trans-activating potential by 50% [57]. Polymorphisms in STAT6 have also been implicated in overall response to morphine and opioid switching [58].

β-Arrestin 2

β-Arrestin 2 is an intercellular protein that is involved in μ-opioid receptor inactivation and internalization [59]. Opioid receptor agonists differentially trigger receptor phosphorylation and recruitment of β-arrestin 2, with consequent receptor internalization [60, 61]. It has been shown that β-arrestin 2 gene (ARRB2) knockout mice achieve prolonged analgesia from morphine treatment, even at normally subanalgesic doses [62]. In addition, desensitization and tolerance to morphine are not observed in these animals [63]. Conversely, rats in which β-arrestin is overexpressed experience little or no analgesia from morphine [64]. Polymorphisms in ARRB2 have been associated with overall response to morphine and opioid switching [58].

Modifying systems

Catechol-O-methyltransferase

The enzyme catechol-O-methyltransferase (COMT) metabolizes catecholamines; therefore, changes in activity may modify adrenergic/noradrenergic and dopaminergic neurotransmission. The most commonly studied SNP in the COMT gene is Val158Met, rs4680, which results in a substitution of valine to methionine. This change has a significant effect as the enzyme activity is reduced by three- to fourfold. Zubieta et al. first suggested that the Val158Met polymorphism was associated with increased pain sensitivity and higher μ-opioid system activation in prolonged experimental pain, reflecting endogenous opioid activity [65]. Rakvag et al. subsequently reported an association between the Val158Met polymorphism and increased morphine dose requirements in cancer-related pain [66].

In subanalyses of EPOS, 1579 patients reporting nausea and vomiting scores, symptoms were correlated with genotype [67]. Three COMT SNPs were found to be weakly associated with less nausea/vomiting: rs165722C, rs4633T, and rs4680G, although the significance was lost after correcting for multiple testing [67]. Genotype data were incomplete for some of the SNPs tested, which would have reduced the power. COMT metabolizes catecholamines such as dopamine, an important neurotransmitter in the area postrema and vomiting center. Dopamine D2 receptor antagonists such as domperidone are used as anti-emetics. COMT inhibitors increase dopaminergic activity, and therefore nausea and vomiting are prominent side effects [68].

Another study of morphine following abdominal surgery reported that carriers with the variant A allele of rs4680 had significantly lower nausea scores and when in combination with the variant G allele of rs1799971 had reduced requirement for anti-emetic treatment [54].

In cancer patients receiving morphine, the common A allele at position −4873 of the COMT has been reported as protective of central side effects [38, 54]. The effect of this allele was independent of and additive to the ABCB1 2677A allele and shows the importance of considering interactions between multiple genes.

HTR3B

In EPOS, three SNPs in the 5-HT (serotonin) receptor 3B gene (HTR3B) were found to be associated with nausea/vomiting on preliminary analysis. Carriers of the rs1176744G, rs3782025T and rs1672717T were associated with less nausea/vomiting [67]. Notably, the association with the G allele of rs1672717 remained significant when corrected for multiple testing. Activation of 5-HT3 receptors in the gastrointestinal tract or CTZ is pro-emetic.

Cytokines

Cytokines are a heterogeneous group of small soluble glycoproteins or polypeptides secreted by immune and non-immune cells and are vital to the coordination of the immune system and the inflammatory response. Cytokines may be broadly classified as pro-inflammatory [tumor necrosis factor-α (TNFα), interleukin 6 (IL-6), and interleukin 8 (IL-8)] or anti-inflammatory (IL-10, IL-4, and TGFβ).

In animal models, administration of pro-inflammatory cytokines induces pain behavior and administration of anti-inflammatory cytokines produces analgesia [69, 70]. In humans, an imbalance between serum pro-inflammatory and anti-inflammatory cytokines has been observed in several painful conditions such as painful peripheral neuropathy, fibromyalgia and chronic regional pain syndrome [71-73]. Endogenous and exogenous opioids clearly interact with the cytokine system. In animal models, spinal administration of morphine stimulates the release of pro-inflammatory cytokines by glial cells in the CNS. These pro-inflammatory cytokines inhibit acute opioid analgesia and contribute to the induction of opioid tolerance after repeated administration [74, 75].

In clinical pain studies, SNPs in cytokine gene promoters (IL8, IL6 and TNF) have been associated with both cancer pain severity and morphine dose [76, 77]. Polymorphisms in the promoter regions of genes may either interrupt or generate additional transcription factor sites thereby altering expression profiles [78]. An exaggerated level of pro-inflammatory cytokine expression can augment or maintain pain states.

Tumor necrosis factor-α

TNFα is the archetypal pro-inflammatory cytokine. TNFα has been implicated in many cancer-related symptoms such as fatigue, cachexia and pain [79]. Several functional SNPs in the TNF promoter have been described, although their actual functional significance remains controversial [78, 80]. The variant A allele at position −308 (rs1800629) has been associated with increased expression [81] and a variety of inflammatory and infective disorders [82, 83]. The rs1800629A allele has also recently been associated with increased pain severity and reduced opioid response on follow-up in one study of 140 lung cancer patients receiving supportive care [77].

Interleukin 6

IL-6 is a pro-inflammatory cytokine with an important role in the pathophysiology of inflammatory pain [84]. IL6 knockout mice show reduced opioid responses, with diminished analgesia to restraint stress and morphine administration. The development of tolerance to morphine-induced analgesia is also more rapid in these mice [85]. A functional SNP in the IL6 promoter rs1800795 (−174G/C) has been associated with MEDD in lung cancer pain using a variety of different opioids; the variant C allele linked to higher morphine dose equivalents [77].

Interleukin 8

IL-8 is a pro-inflammatory cytokine secreted by macrophages, endothelium and epithelial cells and acts as a chemoattractant to recruit neutrophils to sites of injury and inflammation. A functional SNP in the IL8 promoter, rs4073 has been associated with more severe pain at diagnosis in one study of 156 Caucasian pancreatic cancer patients, but did not correlate with opioid response to a variety of opioids when compared using MEDD [86].

Discussion

Do opioid genetics hold the key to achieving the goal of personalized pain control? There is clearly a considerable way to go to translate current research into a tool for clinical practice. Future studies need to learn lessons from those already conducted, and promising initial results need to be properly validated in new sample sets. The challenge of differing phenotype definitions and reduction of ‘noise’ from other clinical variables needs to be addressed. Initial opioid genetic studies were directed toward predicting opioid dose requirements, and this is particularly difficult when grouping opioids as dose conversion ratios are imprecise. Recently, different elements of opioid response are being more extensively explored including analgesia and important opioid-induced adverse reactions such as nausea/vomiting and drowsiness/confusion.

EPOS, the largest genetic association study of opioid response to date, included 2294 cancer patients from 17 countries and tested 112 SNPs in 25 genes [53]. This study clearly showed that the logistics of conducting a large international multicenter study in a palliative care setting could be achieved. The patients recruited were on a variety of strong opioids, and the primary outcome was oral ‘equivalent morphine dose requirement’. The study did not identify any associations with the SNPs tested in both development and validation analyses [53], and this may be in part because all opioids were included together and dose conversion equivalents between opioids is not an exact science [87]. Overall opioid response has a number of dimensions including analgesia and adverse reactions, some of which may limit titration and tolerated opioid dose.

When further subanalyses were conducted on another more directed phenotype, nausea and vomiting scores, positive genetic associations were observed in COMT, HTR3B and CHRM3, with one association in HTR3B, rs1672717, standing to the rigors of multiple testing [67]. This highlights the importance of the choice of phenotype to investigate. In addition, patients with cancer-related pain are a heterogeneous group, and thus failure to replicate some of the findings of studies in a more tightly controlled post-operative setting might be expected.

Pain experience and opioid response are complex traits and are likely to be the product of a myriad of gene–gene and gene–environment interactions, which may each be positive or negative. Some studies have begun to explore the interactions between genetic variants from more than one gene; however, this has so far been limited to two candidate SNPs at a time [34, 88]. The concept of gene–gene/environment interactions or epistasis provides a huge challenge for the future of opioid genetics, both practical and analytical. Such work would increase the required sample sizes exponentially and must therefore be balanced with tighter phenotype definitions.

Population-based association studies aiming to correlate genotypes and phenotypes of complex traits, including pain, have had mixed success, and reproducibility of results has remained low. In experimental pain, twin studies have suggested that up to 60% of the variability in responses to painful stimuli is genetically determined; however, genetic and environmental factors are only moderately correlated across pain modalities, suggesting different genes influence different types of pain. In cancer pain, pain is often of mixed etiology and type and consequently genetic influences may not be clearly identified in clinical studies [89].

There are several factors that contribute to the required sample size in genetic association studies including: the prevalence of disease/trait in the general population, the frequency of the susceptibility allele and its effect size, and the number of SNPs to be tested. The lower the frequency of the susceptibility allele and the lower the effect size, the larger the sample size required. Complex traits are likely to be influenced by multiple genetic variables all with small or modest effect sizes. Any variant strongly associated with a disease or trait is likely to be rare [90]. It is also important to consider the number of SNPs to be tested, because significant findings should stand correction for multiple testing, although on balance the effect size of the susceptibility allele is more important [91]. In general, therefore large sample sizes, possibly of many thousands, are preferable in the study of complex traits. The sample sizes in the studies described in this review are small, and therefore many associations, particularly with small effect sizes, may not have been identified.

For complex diseases, two-, three- or even higher-level interactions might be expected, which would require large datasets and highly sophisticated analytic tools. A number of techniques have been proposed to examine these interactions including regression, recursive partitioning approaches, multifactor dimension reduction and Bayesian model selection techniques [92]. These statistical tools have not yet been applied in a meaningful way to opioid genetics but will need to be employed once suitable datasets are available.

Although the candidate gene approach has yielded some promising results, these are limited by what is already known in terms of potential mechanisms and pathways influencing opioid response. In other fields, genome-wide association (GWA) studies are increasingly being utilized. GWA arrays can type as many as 1 million SNPs across the genome to provide the highest possible coverage of common genetic variation. Associations generated from GWA studies may not have any direct causal relevance and are more likely to be in linkage disequilibrium with underlying causative variants, which must then be identified. Clearly, future studies should employ a variety of approaches to further the knowledge base.

Conclusion

The field of opioid genetics has grown rapidly over recent years however continues to yield more questions than answers. The majority of studies to date have focused on single genes or SNPs, with the potential for gene–gene interactions only just starting to be investigated. It has become clear that there are different elements to overall opioid response in terms of analgesia and different adverse reactions, which may have both overlapping and divergent mechanisms.

The jury is out to whether the holy grail of true personalized prescribing of opioids is achievable. What is certain is that to move forward considerable investment and collaborative working is required to increase sample sizes and to ensure datasets can be meaningfully combined.

Ancillary