Genetic variation and response to morphine in cancer patients

Catechol-o-methyltransferase and multidrug resistance-1 gene polymorphisms are associated with central side effects

Authors


Abstract

BACKGROUND

Pain is a common symptom for patients with cancer, and opioids are the treatment of choice for moderate or severe cancer-related pain. Central side effects, such as drowsiness, confusion, and hallucinations, can limit the use of opioids in clinical practice.

METHODS

The authors prospectively recruited 228 cancer patients who received morphine. Clinical data, including pain and side-effect scores, were correlated with genotype data.

RESULTS

Genetic variation in the multidrug resistance-1 gene (MDR-1) was associated with moderate or severe drowsiness and confusion or hallucinations. Patients who carried the common guanosine (G) allele at position 2677 in exon 26 were less likely to experience drowsiness and confusion or hallucinations than patients who carried the variant thymidine or adenosine alleles, which code for alternate amino-acid substitutions (chi-square statistic, 13.3; P = .0003). In addition, genetic variation in the catechol-O-methyltansferase (COMT) enzyme was associated independently with these central side effects. Single nucleotide polymorphisms (SNPs) in intron 1 were associated significantly with central side effects; the most significant was at position −4873G (chi-square statistic, 9.1; P = .003). SNPs in intron 1, defined as haplotype, were present in 10.4% of the population and were associated significantly with central side effects (chi-square statistic, 7.7; P = .005). Genotype data did not correlate with morphine dose or serum morphine or metabolite concentrations in this study.

CONCLUSIONS

COMT and MDR-1 genotypes were correlated with morphine-related central side effects. The authors believe that this work adds significantly to the current understanding of genetic variants that may influence an individual's response to opioids. Cancer 2008. © 2008 American Cancer Society.

Although morphine is the opioid of choice for the treatment of moderate to severe cancer pain,1 from 10% to 30% of patients who receive oral morphine do not have successful outcomes either because of intolerable adverse effects, or inadequate analgesia, or a combination of both.2 This is a significant problem. In clinical practice, dose-limiting side effects prevent further dose escalation and limit the degree of analgesia that can be achieved. Patients who do not receive the desired analgesic effect or who suffer intolerable side effects from morphine often are “switched” to alternative strong opioids. Opioid switching, or changing from morphine to an alternative opioid, is a therapeutic maneuver that is gaining popularity in pain management as a method of improving analgesic response and/or reducing adverse side effects.3, 4 In 1 study, this strategy of opioid switching achieved good pain control with minimal side effects in 96% of patients.5 The commonest side effects that limit dose titration are central side effects, such as drowsiness and confusion.

The Human Genome Project has led to an increase in pharmacogenomic research with the hope that, in the future, drug therapies may be tailored individually. There are well known examples in which monogenic (single gene) traits affect drug metabolism and response.6 In pain, genetic variation in the cytochrome p450 enzyme 2D6 (CYP2D6) affects codeine metabolism and analgesic response.7, 8 However, for most drugs, drug effect and treatment outcomes are determined by the interplay of multiple genes, with no single predominant factor.

Candidate genes that influence both pharmacokinetic and pharmacodynamic factors will be important. Morphine acts through the G protein-coupled μ-opioid receptor. Agonist-induced activation of the receptor inhibits neuronal transmission of painful stimuli through a complex signaling cascade.9 Recently, an important regulator of μ-opioid receptor desensitization has been identified, the intracellular protein β-arrestin2.10–12 In β-arrestin2 knockout mice, morphine analgesia is increased and prolonged.13 Different opioids induce different rates of receptor internalization and desensitization.14 In clinical practice, genetic variation in β-arrestin2 is associated with the need to switch from morphine to an alternative opioid, although genetic variation in the μ-opioid receptor itself did not correlate with morphine response.15

The membrane-bound drug transporter P-glycoprotein is important in regulating drugs crossing the blood-brain barrier. It actively pumps drugs out of the central nervous system (CNS) and, thus, may affect the frequency of opioid-induced CNS side effects.16 Interindividual variability in P-glycoprotein activity is well recognized, and genetic variation in the multidrug resistance (MDR) gene MDR-1, which encodes for P-glycoprotein, alters P-glycoprotein activity.17, 18

Response to a painful stimulus is regulated by interactions between multiple regions within the brain through different neurochemical pathways.19 Catechol-O-methyltransferase (COMT) is an enzyme that metabolizes the catecholamines dopamine, noradrenaline, and adrenaline, which are important neurotransmitters in the brain. There is growing evidence to support interaction between dopaminergic and adrenergic pathways and opioid signaling pathways in the CNS.20–24

The COMT gene is polymorphic, and the most widely studied variant in this gene is the single nucleotide polymorphism (SNP) adenosine (A) to guanosine (G) substitution at nucleotide 1222 (A1222G) (reference sequence 4680), which causes an amino-acid substitution of valine (Val) for methionine (Met) in the transcribed protein. This variant causes a 3- to 4-fold decrease in enzyme activity25 and has been shown to affect μ-opioid neurotransmitter responses to a pain stressor in normal volunteers.24 Individuals who were homozygous for the methionine variant had higher sensory and affective ratings to pain. At baseline, they had increased expression of μ-opioid receptors in multiple regions of the brain; however, in response to pain, they had decreased activation of the μ-opioid because of decreased release of endogenous opioids.

Studies of the COMT gene have indicated that SNPs in the promoter, intron 1, and the 3′-untranslated region may be linked more strongly to disease,26, 27 emphasizing the importance of haplotype analysis of multiple SNPs across the gene.28 However, correlation between these SNPs and neuronal COMT messenger RNA (mRNA) expression has been inconsistent.29, 30 Diatchenko et al. used 4 SNPs to define a high, average, and low pain-sensitivity haplotype in normal volunteers who had variable pain perception and risk of developing chronic temporomandibular joint pain.31

We wanted to investigate whether or not a functional polymorphism in the MDR-1 gene, which is responsible for the production of differing concentrations of the drug transporter P-glycoprotein, is relevant to the need for cancer patients to switch from morphine to an alternative opioid. Furthermore, it was been demonstrated previously that the COMT gene has influence in relation to pain24, 31 and morphine dose32; thus, it was a double candidate for investigation in our study. From a previous study, we also hypothesized that other factors apart from pain may relate to the need to switch.5 For this reason, a sophisticated database was designed for the current study to include factors that would be markers of central side effects in which the passage of morphine and its metabolites across the blood-brain barrier may be critical.

MATERIALS AND METHODS

The recruitment of cancer patients to this prospective case-controlled study has been described in detail elsewhere.5, 15 Briefly, patients who received morphine for cancer-related pain were identified prospectively by the palliative care team. It is important to stress that all patients received morphine that was titrated to achieve the desired analgesic effect. Those who did not achieve good clinical benefit because of inadequate pain relief and/or unacceptable side effects were switched actively to an alternative opioid, oxycodone. All patients completed a modified Brief Pain Inventory33 and recorded both their degree of pain over the previous 24 hours for 5 separate pain parameters (numerical rating scale from 0 to 10 for average pain, worst pain, least pain, pain now, and percentage pain relief) and side effects they believed were related to morphine (4-point scale: “not at all,” “a little,” “quite a bit,” “very much”). The database included demographic data (age, ethnicity, cancer diagnosis), current opioid use, opioid history, and current medications. Blood samples were taken for hematologic and biochemical analyses and DNA extraction. DNA was extracted by using a modified salting-out method34 and was stored at −20 °C for genotype analysis. The primary analysis compared 2 phenotypes: controls, who had taken morphine for at least 1 month with good pain relief and minimal side effects; and switchers, who, despite adequate dose escalation, had inadequate analgesia or intolerable side-effects on morphine and required switching to an alternative opioid.

Putative SNPs in the COMT and MDR-1 genes were identified from the literature and SNP databases (http://www.ncbi.nlm.nih.gov, http://snpper.chip.org/bio). At the time of commencement of this work, no added information could be gained from the HapMap Project database. Genotypes were determined by using sequence-specific primers in a polymerase chain reaction (SSP-PCR).35 Sequence-specific primers with mismatches at the 3′ end were designed to identify each variant, which, in combination with a consensus primer, produced a product of known size (primer sequences are available from the authors on request). Primer concentrations were titrated to ensure amplification only with exact matching of the primer with genomic DNA. Briefly, 5 μL of primer mix were added to an 8-μL PCR reaction mixture that contained buffer (67 mM Tris base, pH 8.8; 16.6 mM ammonium sulfate; 2 mM magnesium chloride; 0.01% volume/volume Tween-20; Bioline Ltd., London, United Kingdom); 200 mM each of deoxyadenosine triphosphate, deoxythymidine triphosphate, deoxyguanosine triphosphate, and deoxycytidine triphosphate (Bioline Ltd.); 0.32 U of Taq polymerase (Biotaq; Bioline Ltd.); and from 0.01 to 0.1 μg DNA. This 13-μL reaction was dispensed under 10 μL of mineral oil and amplified in a PCR machine (MJ Research, Waltham Mass; PTC-200 machine). Cycling parameters were set as follows: 1 minute at 96 °C; followed by 5 cycles at 96 °C for 25 seconds, 70 °C for 45 seconds, and 72 °C for 45 seconds; followed by 21 cycles at 96 °C for 25 seconds, 65 °C for 50 seconds, and 72 °C for 45 seconds; followed by 4 cycles at 96 °C for 25 seconds, 55 °C for 60 seconds, and 72 °C for 120 seconds. Then, PCR products were electrophoresed on 1.5% agarose gels (Bioline Ltd.) containing 0.14 mg/mL ethidium bromide (Sigma Ltd., Poole, United Kingdom), at 200 V/cm2 in 0.5% Tris borate ethylenediamine tetraacetic acid buffer (Sigma Ltd.). Products were visualized with an ultraviolet illuminator and photographed with a Polaroid camera. The presence of an allele-specific band of the expected size, in conjunction with a control band, was used to identify an allele. A reaction was considered negative if the control band was visualized without a specific band of the expected size.

Genotype and allele frequencies and allele carriage were calculated and checked for Hardy-Weinberg equilibrium. Interactions between SNPs across each gene were considered by constructing haplotypes with the computer programs Phase (http://www.stat.washington.edu/stephens/software.html) and Arlequin (http://lgb.unige.ch/arlequin/). Identification of rare intragene haplotypes can identify genotype errors; therefore, samples that were identified as having rare intragene haplotypes were regenotyped to reduce any errors.36

The database of clinical, laboratory, and genetic data was analyzed using data mining software, Knowledge Studio (http://www.angoss.com). The standard statistical software package Stata (version 8; http://www.stata.com) was used to test prior hypotheses and any hypotheses generated by the data mining. Data mining uses decision tree modeling to examine effects between different variables (including clinical, laboratory, and genetic variables) by using a stepwise linear regression approach. Stata was used to perform the formal statistics for regression analyses.

RESULTS

Clinical data were collected for 228 patients, including 164 controls (responders to morphine) and 64 switchers (required an alternative opioid), as described in detail elsewhere.5 DNA was available for 221 patients. There were no significant differences in age, sex, ethnicity, or tumor diagnosis between switchers and controls (Table 1). The majority of patients were Caucasian (87.3%). For this cohort, switching to an alternative opioid was successful in 57 of 64 patients (89%): Forty-eight patients required a single switch (oxycodone), and 9 patients required additional switch(es) before a good clinical outcome was achieved.

Table 1. Demographic, Biochemical, and Hematologic Data for the 228 Patients Recruited for This Study
VariableControls, n = 164Switchers, n = 64Total, n = 228
  • SD indicates standard deviation; GI, gastrointestinal; HGB, hemoglobin; WBC, white blood cells; γ-GT, γ-glutamyl transpeptidase.

  • *

    Diagnosis 2 patients multiple tumors.

  • P < .05 (unpaired t test).

Mean age ± SD, y56.5 ± 1459.1 ± 11.557.2 ± 13.4
Sex, no. of patients (%)
 Men72 (43.9)34 (53.1)106 (46.5)
 Women92 (56.1)30 (46.9)122 (53.5)
Ethnicity, no. of patients (%)
 Caucasian143 (87.2)56 (87.5)199 (87.3)
 Non-Caucasian21 (12.8)8 (12.5)29 (12.7)
Diagnosis, no. of patients (%)*
 Breast36 (22)11 (17.2)47 (20.6)
 Sarcoma26 (15.9)6 (9.4)32 (14)
 Lung20 (12.2)11 (17.2)31 (13.6)
 Urogenital15 (9.1)11 (17.2)26 (11.4)
 Upper GI tract18 (11)8 (12.5)26 (11.4)
 Head and neck14 (8.5)6 (9.4)20 (8.8)
 Gynecologic17 (10.4)3 (4.7)20 (8.8)
 Lower GI tract8 (4.9)6 (9.4)14 (6.1)
 Hematologic11 (6.7)1 (1.6)12 (5.3)
 Melanoma4 (2.4)3 (4.7)7 (3.1)
 Pancreas4 (2.4)1 (1.6)5 (2.2)
 Prostate3 (1.8)2 (3.1)5 (2.2)
 Other6 (3.7)3 (4.7)9 (3.9)
Blood levels, mean ± SD
 HGB, g/dL11.4 ± 1.711.5 ± 1.711.4 ± 1.7
 WBC, ×10−9 g/dL9.8 ± 5.911.3 ± 10.910.2 ± 7.7
 Platelets, ×109/L319 ± 167357 ± 173330 ± 169
 Sodium, mmol/L136 ± 3.6135 ± 3.7135.7 ± 3.7
 Potassium, mmol/L4 ± 0.54.2 ± 0.44.1 ± 0.5
 Urea, mmol/L5.3 ± 3.34.7 ± 2.45.2 ± 3.1
 Creatinine, μmol/L75.4 ± 23.478 ± 24.376.1 ± 23.6
 Alanine transaminase, mmol/L37 ± 42.626.8 ± 24.434.2 ± 38.5
 Alkaline phosphatase, mmol/L164.8 ± 259.4164.2 ± 226.6164.7 ± 250.1
 γ-GT, mmol/L110.1 ± 201.9115 ± 143.3111.5 ± 187
 Albumin, g/L28.6 ± 6.229.9 ± 6.629 ± 6.3
 Calcium, mmol/L2.2 ± 0.22.2 ± 0.12.2 ± 0.2

Switchers scored higher for individual morphine-related side effects compared with controls (Table 2), as predicted. Stepwise logistic regression with switch/control as the dependent variable demonstrated that, when pain and side-effect scores were examined, “% pain relief” was the strongest predictor of the need to switch (odds ratio [OR], 0.95; 95% confidence interval [CI], 0.93–0.97; P < .0001) followed by the central side effects drowsiness (OR, 2.45; 95% CI, 1.45–4.13; P = .001) and confusion and hallucinations (OR, 2.45; 95% CI, 1.19–5.05; P = .015). Dry mouth (OR, 4.56; 95% CI, 1.74–11.9; P = .01) and nightmares (OR, 3.58; 95% CI, 1.04–12.25; P = .04) also were retained in the model. Those patients with moderate or severe side effects (“quite a bit” or “very much” on a 4-point scale) were most likely to switch. The numbers of patients with moderate or severe nightmares were small (16 of 228 patients).

Table 2. Morphine-related Side Effects
Side effect/score*No. of patients (%)Chi-squarePOR [95%CI]
Controls, n = 164Switchers, n = 64
  • OR indicates odds ratio; CI, confidence interval.

  • *

    Side effects were scored as 0 (none), 1 (a little), 2 (quite a bit), or 3 (very much). Data shown represent the number of patients (%) for each group.

  • Significant (P < .05).

Drowsy  47.3<.00013.20 [2.13–4.79]
 0–193 (56.7)18 (28.1)   
 256 (34.2)14 (21.9)   
 315 (9.2)32 (50)   
Confusion  38.8<.00014.80 [2.71–8.50]
 0–1145 (88.4)34 (53.1)   
 217 (10.2)19 (29.7)   
 32 (1.2)11 (17.2)   
Drowsy and confused  18.2<.00017.80 [3.52–17.31]
 0–1157 (95.7)51 (79.7)   
 26 (3.7)6 (9.4)   
 31 (0.6)7 (10.9)   
Nightmares  16.7<.00015.48 [2.01–15]
 0–1159 (97)53 (82.8)   
 25 (3)7 (10.9)   
 30 (0)4 (6.3)   
Nausea  10.2.0061.96 [1.26–3.05]
 0–1136 (82.9)43 (67.2)   
 220 (12.2)10 (15.6)   
 38 (4.9)11 (17.2)   
Constipation  3.2.3551.25 [0.96–1.62]
 046 (28.1)14 (21.9)   
 141 (25)13 (20.3)   
 243 (26.2)17 (26.3)   
 334 (20.7)20 (23.7)   
Dry mouth  18.2<.00013.92 [1.85–8.30]
 0–1157 (95.7)51 (79.7)   
 26 (3.7)6 (9.4)   
 31 (0.6)7 (10.9)   
Itch  2.6.451.50 [0.85–2.65]
 0157 (95.7)59 (92.2)   
 13 (1.8)1 (1.6)   
 23 (1.8)2 (3.1)   
 31 (0.6)2 (3.1)   
Myoclonus  5.3.0712.77 [0.88–8.72]
 0160 (97.6)60 (93.8)   
 14 (2.4)2 (3.1)   
 20 (0)2 (3.1)   

Oral morphine doses ranged from 10 mg/24 hours to 1260 mg/24 hours. There was no difference in the 24-hour dose of morphine between switchers and controls: Switchers had lower serum morphine and M3G concentrations with a trend toward lower M6G concentrations (P = .009, P = .03, and P = .07, respectively). In agreement with other studies, no correlation was observed between average pain scores and analgesic dose, serum morphine or metabolite concentrations per se, or metabolite ratios. However, “confusion or hallucinations” correlated with the morphine/M6G ratio (P = .05); and, despite the small numbers of patients experiencing myoclonus (n = 8), this was associated with 24-hour morphine dose and serum morphine and metabolite concentrations (P = .005, P = .009, P = .007, and P = .009, respectively).

Assays were developed for 15 putative SNPs in the COMT gene: Thirteen SNPs spanning a 29-Kb region were polymorphic in the Caucasian population (Fig. 1; for primer sequences, see Table 3). The genotype and allele frequencies and the allele carriage for each SNP are tabulated in Table 4. Allele frequencies for SNPs in intron 1 were significantly different in switchers compared with controls (−7370 cytosine [C]/G: chi-square statistic, 3.7 [P = .05]; −7053G/ adenosine [A]: chi-square statistic, 7.2 [P = .007]; −4873A/G: chi-square statistic, 9.1 [P = .003). Haplotypes were constructed from the SNP data for both the Caucasian population and the total population. In the Caucasian population, the frequencies of the 2 most common haplotypes were 35% and 23.8%, respectively. Thirteen haplotypes were present at a frequency of >1% and described 90% of the population (Table 5).

Figure 1.

Schematic diagram of the catechol-O-methyltransferase (COMT) gene. Fifteen single-nucleotide polymorphisms were examined; 13 were polymorphic in this cohort and are indicated in red. A indicates adenosine; G, guanosine; C, cytosine; T, thymidine; cis, cysteine; ser, serine; val, valine; met, methionine; COMT-MB, membrane-bound COMT; COMT-SOL, soluble COMT.

Table 3. Primer Sequences for Single Nucleotide Polymorphisms in the Catechol-O-Methyltransferase Gene
RegionSNP referenceSNP (ATG+)*NucleotidePrimer sequence
  • SNP indicates single nucleotide polymorphism; rs, reference sequence; A, adenosine; T, thymidine; G, guanosine; C, cytidine; Cons, consensus; UTR, untranslated region.

  • *

    Position in relation to the ATG start codon (sequence available at: http://snpper.chip.org/bio).

Promoterrs2097603−21958ATGGGGACAACAGACAGAAAAGT
  GGGGGACAACAGACAGAAAAGC
  ConsACTCCTCTGGCGGAAAGGAA
Intron 1rs737866−19941ACACAAAAATCCCTGGCTGGAA
  GACACAAAAATCCCTGGCTGGAG
  ConsTGTGGAGCCCCAGACTCAG
  (C) AACAAAAACCCCTGGCTGGAA
  (C) GACAAAAACCCCTGGCTGGAG
Intron 1rs7287550−18074CGGCGGGATAATCGCTTGAACT
  TGCGGGATAATCGCTTGAACC
  ConsGGCTTGGCTTCCTAACCTCG
Intron 1rs174680−15051CCCTTACCTGTCACTGGAGATA
  TCCTTACCTGTCACTGGAGATG
  ConsGTCATTGGCAGAAGGGTATTC
Intron 1rs7290221−7370CACCCATTGAGCCAGGCATG
  GACCCATTGAGCCAGGCATC
  ConsTCACCTCTTCTCAGGTGTCAC
Intron 1rs5746849−7053ATGCAAGAAGCCCTGCCCTTT
  GTGCAAGAAGCCCTGCCCTTC
  ConsTCACCTCTTCTCAGGTGTCAC
Intron 1rs740603−4873ACACTGAGGATGCCCTCACA
  GACTGAGGATGCCCTCACG
  ConsCTTCCATCCGTTAGCACCTGA
Intron 2rs6269−98ATGAACCTTGCCCCTCTGCA
  GTGAACCTTGCCCCTCTGCG
  ConsACCGTCAGACCCTGTGAGT
Exon 3rs6270A101CTCGTTCCAGCCGATAAGGC
  GTCGTTCCAGCCGATAAGGG
  ConsTGGTGAGTGGGTCTGTCCTG
Exon 3rs4633186CCAGCGCATCCTGAACCAC
  TGCAGCGCATCCTGAACCAT
  ConsAGCCCTATCTGGGCATATCC
Exon 3rs2239393379AGGAGAAGCTGTTATCACCCCA
  GGAGAAGCTGTTATCACCCCG
  ConsCACGATCTTGCCTGTGCAG
Exon 4rs46801222GATGGTGGATTTCGCTGGCG
  AGATGGTGGATTTCGCTGGCA
  ConsAGGGTTCTGGGATGACAAGG
Exon 5rs1656311767CTTGGGGCTCACCTCCAAG
  TGTTGGGGCTCACCTCCAAA
  ConsACCCTGCACAGGCAAGATC
Intron 5rs1746994409CTGACACTGTCGCTTCTCCAC
  TTGACACTGTCGCTTCTCCAT
  ConsCAATCTGGTCATCAGGGATAC
3′UTRrs1657286974CGATGCAGTGCTGGTTTCTGC
  TGGATGCAGTGCTGGTTTCTGT
  ConsACTTGGGCTCACTGGTCTGT
Table 4. Catechol-O-Methyltransferase Genotype and Allele Frequencies in 221 Patients, Including 158 Controls Who Responded to Morphine and 63 Switchers Who Did Not Respond to Morphine
SNP: GenotypeGenotype frequencyAlleleAllele frequencyAllele carriage
Controls (n = 158)Switchers (n = 63)Controls (n = 158)Switchers (n = 63)Controls (n = 158)Switchers (n = 63)
  • SNP indicates single nucleotide polymorphism; A, adenosine; G, guanosine; C, cytosine; T, thymine; UTR, untranslated region.

  • *

    Significant (P < .05).

−21958 A/G: Promoter
 AA0.290.39A0.540.630.790.87
 AG0.500.48G0.460.370.710.61
 GG0.210.13     
−19941 A/G: Intron 1
 AA0.480.53A0.710.730.940.92
 AG0.460.39G0.290.270.520.47
 GG0.060.08     
−18074 C/T: Intron 1
 CC0.560.42C0.750.670.940.92
 CT0.380.50T0.250.330.440.58
 TT0.060.08     
−15051 C/T: Intron 1
 CC0.580.47C0.750.690.930.92
 CT0.350.45T0.250.310.420.53
 TT0.070.08     
−7370 C/G: Intron 1
 CC0.30*0.16*C0.54*0.40*0.79*0.65*
 CG0.50*0.48*G0.46*0.60*0.70*0.84*
 GG0.21*0.35*     
−7053 A/G: Intron 1
 AA0.21*0.35*A0.54*0.60*0.79*0.84*
 AG0.50*0.48*G0.46*0.40*0.71*0.65*
 GG0.29*0.16*     
−4873 A/G: Intron 1
 AA0.220.34A0.48*0.59*0.740.84
 AG0.520.50G0.52*0.41*0.780.66
 GG0.260.16     
−98 A/G: Intron 2
 AA0.360.29A0.610.580.860.87
 AG0.490.58G0.390.420.640.71
 GG0.140.13     
186 C/T: Exon 3
 CC0.200.26C0.470.510.770.76
 CT0.530.50T0.530.490.800.74
 TT0.260.24     
379 A/G: Exon 3
 GG0.140.13G0.390.420.640.71
 AG0.490.58A0.610.580.860.87
 AA0.360.29     
1222 A/G: Exon 4
 AA0.260.24A0.530.490.810.74
 AG0.550.50G0.470.510.740.76
 GG0.190.26     
4409 T/C: Intron 5
 TT0.890.94T0.950.961.000.98
 TC0.110.05C0.050.040.110.06
 CC0.000.02     
6974 T/C: 3′UTR
 TT0.870.94T0.940.971.001.00
 TC0.130.06C0.060.030.130.06
 CC0.000.00     
Table 5. Catechol-O-Methyltransferase Haplotype Frequencies for 221 Patients
HaplotypeAllele at individual SNP position (5′ to 3′)Haplotype frequency %: no. of chromosomes (%)
SNP 1SNP 2SNP 3SNP 4SNP 5SNP 6SNP 7SNP 8SNP 9SNP 10SNP 11SNP 12SNP 13CaucasiansTotal population
−21958: Promoter−19941: Intron 1−18074: Intron 1−15051: Intron 1−7370: Intron 1−7053: Intron 1−4873: Intron 1−98: Intron 2186: Exon 3379: Exon 31222: Exon 44409: Intron 56974: 3′UTRControls, n = 276Switchers, n = 110Controls, n = 316Switchers, n = 126
  1. SNP indicates single nucleotide polymorphism; UTR, untranslated region; G, guanosine; A, adenosine; C, cytosine; T, thymine; X, residual haplotypes with frequency <1%.

1GACCCGGATAATT101 (36.6)34 (30.9)111 (35.1)35 (27.8)
2AGCCGAAGCGGTT64 (23.2)28 (25.5)74 (23.4)33 (26.2)
3AATTGAAATAATT29 (10.5)16 (14.5)30 (9.5)16 (12.7)
4GACCCGGGCGGTT16 (5.8)3 (2.7)17 (5.4)4 (3.2)
5AATTGAAGCGGTT11 (4)6 (5.5)12 (3.8)7 (5.6)
6AATTGAAACAGTT3 (1.1)2 (1.8)5 (1.6)5 (4.0)
7AATTCGGACAGCC4 (1.4)3 (2.7)4 (1.3)4 (3.2)
8AATTCGGATAATT5 (1.8)1 (0.9)6 (1.9)2 (1.6)
9AGCCCGGATAATT6 (2.2)06 (1.9)0
10GACCGAAGCGGTT2 (0.7)3 (2.7)2 (0.6)4 (3.2)
11AGCCGAAATAATT3 (1.1)1 (0.9)3 (0.9)1 (0.7)
12AATCGAAATAATT03 (2.7)04 (3.2)
13AATTCGAGCGGTT3 (1.1)1 (0.9)3 (0.9)1 (0.7)
X29 (10.5)9 (8.2)43 (13.6)10 (7.9)

Five SNPs spanning a 96.8-Kb region were genotyped for the MDR-1 gene (for primer sequences, see Table 6). Four SNPs have been identified as functional variants in other studies, and the other in intron 7 has a high allele frequency in Caucasians.37, 38 The genotype and allele frequencies and the allele carriage for each SNP are tabulated in Table 7. Haplotypes were constructed from the SNP data, and haplotype frequencies agreed closely with those observed in the Caucasian population studied by Kroetz et al.38, 39 (Table 8). There were no significant differences in allele or haplotype frequencies when comparing switchers and controls. MDR-1 genotype did not correlate with morphine dose or serum morphine or metabolite levels (data not shown).

Table 6. Primer Sequences for Single Nucleotide Polymorphisms in the Multidrug Resistance-1 Gene
RegionSNP referenceSNP exon/nucleotide*NucleotidePrimer sequence
  • SNP indicates single nucleotide polymorphism; rs, reference SNP; C, cytosine; A, adenosine; G, guanosine; T, thymine; Cons, conserved; Ala, alanine; Ser, serine; Thr, threonine.

  • *

    The SNP position is given relative to the start of the nearest exon (Exon/Nucleotide).

Exon1rs32136191b/12CAAGCCTGAGCTCATTCGAGC
  TAAGCCTGAGCTCATTCGAGT
  ConsTGATTCCAAAGGCTAGCTTGC
Exon 6rs12021686/+139CCAAAATTCCTTCTAAGCAGCAAC
  TACAAAATTCCTTCTAAGCAGCAAT
  ConsTACTCAGTGAGGCTGGATATG
Exon 12rs112850312/1236CCCTGGTAGATCTTGAAGGGC
  TTCCTGGTAGATCTTGAAGGGT
  ConsCAGGAGGATCTTGGGGTTG
Exon 21rs203258221/2677GGTTTGACTCACCTTCCCAGC
  TTAGTTTGACTCACCTTCCCAGA
Ala893Ser/Thr ATAGTTTGACTCACCTTCCCAGT
  ConsCAGACTCCCTCTGGAATTCAA
Exon 26rs104564226/3435CGTGGTGTCACAGGAAGAGATC
  TGTGGTGTCACAGGAAGAGATT
  ConsTTCCAGTTCTCCTATCCCAG
Table 7. Multidrug Resistance 1 Genotype and Allele Frequencies in 221 Patients, Including 158 Controls who Responded to Morphine and 63 Switchers who Did Not Respond to Morphine
SNP: GenotypeGenotype frequencyAlleleAllele frequencyAllele carriage
Controls (n = 158)Switchers (n = 63)Controls (n = 158)Switchers (n = 63)Controls (n = 158)Switchers (n = 63)
  1. SNP indicates single nucleotide polymorphism; T, thymine; C, cytosine; G, guanosine; A, adenosine.

  2. * Significant (P < .05).

1b/−129: Exon 1
 TT0.960.89T0.980.941.001.00
 TC0.040.11C0.020.060.040.11
 CC0.000.00     
6/139: Exon 6
 TT0.200.21T0.450.410.710.62
 TC0.510.41C0.550.590.800.79
 CC0.290.38     
12/1236: Exon 12
 CC0.280.37C0.550.560.810.76
 CT0.530.40T0.450.440.720.63
 TT0.190.24     
26/3435: Exon 26
 TT0.310.29T0.570.560.830.83
 TC0.520.54C0.430.440.690.71
 CC0.170.17     
21/2677: Exon 21
 GG0.270.30G0.520.480.780.67
 GT0.480.33T0.450.470.700.67
 TT0.200.27A0.030.050.060.10
 GA0.030.03     
 AT0.020.06     
 AA0.010.00     
Table 8. Multidrug Resistance-1 Haplotypes Based on 5 Single Nucleotide Polymorphisms*
HaplotypeAllele at individual SNP position (5′ to 3′)Haplotype frequency, % chromosomes
SNP 1SNP 2SNP 3SNP 4SNP 5All 5 SNPs4 SNPs (1, 3, 4, and 5)
−129: Exon 1b+139: Exon 6+1236: Exon 12+2677: Exon 21+3435: Exon 26Controls (n = 276)Switchers (n = 110)Total (n = 386)Kroetz, 200339 (n = 400)
  • SNP indicates single nucleotide polymorphism; T, thymine; C, cytosine; G, guanosine; A, adenosine.

  • *

    Four published SNPs have been associated with multidrug resistance 1 gene function and a fifth SNP is common in Caucasians and defines the rarer haplotypes. Haplotype frequencies are compared with those published by Kroetz, 2003.39

1TTTTT39.137.340.541
2TCCGC32.230.932.332.5
3TCCGT15.911.815.312
4TCCAC2.25.53.12
5CCCTC1.85.52.84.5
6TTTTC2.50.9
7TCTTT1.11.8
8TCCTT0.71.8
9TTTGC1.400
10TCTTC0.70.9
11TCCTC0.41.8
12TTTGC0.40
13TCTAT0.40

The decision to switch opioids is based both on analgesic response and side-effect profile. Data mining revealed significant associations between central side effects (drowsiness, confusion and hallucinations, nightmares) and genetic variation in MDR-1 and COMT. No such link was observed with other side effects, for example, nausea, constipation, and dry mouth (data not shown). The strongest associations with central side effects were observed in patients who had symptoms that were either moderate or severe and who experienced both drowsiness AND confusion or hallucinations.

For MDR-1, alleles 21/2677G and 12/1236C were associated significantly with a decreased level of “drowsiness and confusion or hallucinations” (21/2677G: chi-square statistic, 13.3 [P < .0003]; 12/1236C: chi-square statistic, 8.3 [P < .004]). For COMT, the strongest association with a protective effect against “drowsiness and confusion or hallucinations” was the SNP in intron 1 −4873G (chi-square statistic, 9.1; P < .003), and a strong association also was observed with haplotype 3 (chi-square statistic, 7.7; P < .005). The SNPs in the promoter region and those in intron 1 define this haplotype; the presence of a T allele at position −15051, thus, shows a significant but weaker association (chi-square statistic, 4.4; P < .04). In addition, haplotype 1 has a weak correlation with “drowsiness AND confusion or hallucinations” (chi-square statistic, 4.9; P < .03). There was no correlation between these genotypes and pain (Table 9).

Table 9. Association Between Genetic Variation in Catechol-O- Methyltransferase and Multidrug Resistance 1 With Pain and Central Side Effects
GenePositionPain score*Central side effects
Chi-squarePChi-squareP
  • MRD-1 indicates multidrug resistance-1 gene; G, guanosine; C, cytosine; COMT, catechol-O-methyltransferase gene; A, adenosine; Val, valine; Met, methionine.

  • *

    For pain scores, patients scored their pain on a numerical rating scale from 0 to 10 using the validated Brief Pain Inventory. Data are shown only for average pain in the previous 24 hours. For the cluster analysis, patients were divided into 3 groups with pain scores from 0 to 3, from 4 to 7, and from 7 to 10.

  • Patients with moderate or severe drowsiness and confusion or hallucinations.

  • Significant (P < .05).

MDR-121/2677G0.26.87713.3.0003
MDR-112/1236C0.03.9848.3.004
COMT−18074C0.91.6344.4.04
COMT−15051C1.67.4343.7.05
COMT−7370C5.13.0772.9.09
COMT−7053A0.37.837.2.007
COMT−4873A0.54.779.1.003
COMT1222A (Val158Met)0.22.8970.003.956
COMTHaplotype 10.58.754.9.03
COMTHaplotype 34.07.1317.7.005

Further analysis confirmed that the genotypes for MDR-1 and COMT were associated independently with “drowsiness AND confusion or hallucinations.” In each instance, analyses were performed on the total population and on the Caucasian population alone to ensure that such associations held true independent of ethnicity. On univariate and multivariate analyses for the Caucasian cohort, both MDR and COMT were associated significantly with reduction in central side effects (univariate: MDR-1 21/2677G, P = .0003; COMT −4873G, P = .003; multivariate: MDR1 21/2677G, P = .002; COMT −4871G, P = .003). The percentages of patients with moderate or severe central side effects who had both, 1, or neither genetic susceptibility marker(s) are shown in Figure 2. Patients who had COMT −4873G were less likely to be drowsy and confused, and having the common G allele at 21/2677 in MDR-1 also was protective. Thus, the greatest proportion of patients with drowsiness and confusion was observed among patients those with the variant A allele at COMT −4873 and a variant T or A allele at 21/2677 in MDR-1.

Figure 2.

The percentage of patients with moderate or severe central side effects according to combinations of protective alleles (for the multidrug resistance 1 gene [MDR1] and the catechol-O-methyltransferase gene [COMT]). Patients who carried neither protective allele were significantly more likely to have central side effects (chi-square statistic = 21.8; P < .0001).

DISCUSSION

Patients with cancer often have limited prognoses, and it is important to achieve adequate pain control as quickly as possible with minimal side effects. Therefore, it would be beneficial to be able to individualize opioid therapy to achieve this objective. The current report provides data on 2 candidate genes, MDR-1 and COMT, which influence clinical response to morphine. It is noteworthy that these genotypes independently predict the frequency of morphine-related central side effects, which are the most common reason for failure to obtain adequate analgesia from morphine, precluding dose escalation. In total, 18 SNPs were investigated in this study, suggesting a maximum Bonferroni correction of 18. Because of the tight linkage disequilibrium between many of these SNPs, the correct Bonferroni according to the method of Svejgaard et al.40 is 13. The results remain significant at a level of P < .05 even if the maximum correction is applied.

It is clear that the interaction between genetic factors that influence response to morphine will be complex. Haplotype analysis has enabled the interactions of multiple SNPs within a gene to be considered. In addition, combining robust clinical data with individual SNP and haplotype data is vital to examine the relative contributions of individual genetic components to the overall picture and to begin to use the model to identify which individuals are likely to respond to a drug. The use of data-mining techniques to analyze robust clinical and genetic databases allows an examination of confounding variables in this diverse population (eg, tumor types, time to death).

Our hypothesis based on other studies involving the COMT gene was that it would be related to the need to switch, because the need to switch depended mainly on pain control. In support of this hypothesis, we observed an association, albeit weak, between polymorphisms in COMT and the need to switch in our patients. However, we also observed that, although pain was the primary reason for switching, other central side effects, such as confusion, drowsiness, and nightmares, also were important. Therefore, we investigated whether the COMT poly-morphism went with switching in those patients who had pain that was critical or central side effects that were critical. The results clearly showed that the COMT polymorphism was related to the central side effects and not to the pain. It is important to stress at this point, however, that this is not the same functional polymorphism that has been described by others (val158 met). Furthermore, all of our patients had analgesic dose titrated against analgesic effect.

The functional polymorphisms in MDR-1 were not related to the need to switch. One of these, however, was associated significantly with central side effects (P = .0003). Therefore, COMT and MDR-1 are associated with central side effects and not with pain per se. Although the numbers were small, MDR-1 also was linked to a further central side effect: nightmares.

Genetic variation in the COMT gene was associated with central side effects from morphine in this cohort. However, the known, well studied functional SNP (Val158Met) was not implicated, and intron 1 was identified as the region of interest. This region is proximal to the promoter for the soluble form of COMT and, thus, may be important in regulating gene expression through alteration of transcription factor binding sites. When this sequence is used to interrogate the transcription factor binding program (http://mbs.cbrc.jp/research/db/TFSEARCH.html), the A allele is conducive with binding of the transcription factors upstream—stimulatory factor, N-myc, and C-myc. This binding is lost with mutation of the A allele to G and, thus, may explain a functional role. Another group also reported a significant association between this SNP, but not with other SNPs in COMT, with acute postsurgical pain after dental extraction.41

In addition, in the center of this region is a 4 base-pair tandem repeat. Therefore, we hypothesize that variation in this region may cause the production of a splice variant of the soluble-COMT protein and that this variation is in linkage with the neighboring SNPs that were highlighted by the current study. This hypothesis is supported by the results from a recent study by Diatchenko et al., who investigated 6 SNPs across the COMT gene. Those authors correlated combinations of SNPs (haplotypes) with high, average, and low pain sensitivity in humans to both experimental pain and the risk of developing a common musculoskeletal joint pain, temporomandibular joint disorder.31 However, the SNPs that defined the haplotypes covered intron 2 to exon 4 and did not include intron 1 or the promoter regions. That study demonstrated a difference in enzyme activity but no difference in mRNA production by comparing low and high pain sensitivity haplotypes. Therefore, the authors concluded that the likely functional difference would be a problem with translation, ie, abnormal protein production from the mRNA. This would support the findings of our current study indicating that intron 1 is the functional region of interest in the COMT gene.

One other pain study has analyzed the influence of the common functional variant in the COMT gene, Val158Met, on the efficacy of morphine in a cohort of patients suffering from cancer-related pain.32 In that study, morphine doses and serum concentrations of morphine and morphine metabolites were compared between the genotype groups, and a weak association was observed between morphine dose and genotype (P = .025). That finding was not confirmed in our similar sized cohort.

Genetic variation in the MDR-1 gene was associated with central side effects to morphine in this cohort and the SNP in exon 21 (G2677T/A) showed the strongest association. The drug transporter P-glycoprotein, encoded by the MDR-1 gene, has a broad range of substrates, including anticancer drugs and different opioids. The first study reporting MDR-1-related pharmacokinetic data was by Kim et al.,42 who correlated the serum concentrations of the MDR-1 substrate fexofenadine with genotype in normal volunteers and observed a significant association with both G21/2677TA and C26/3435T SNPs. Siegmund et al. reported that, for single and multiple doses of the MDR-1 substrate talinolol, the G21/2677TA SNP, but not the C26/3435T SNP, was important. Carriers of the TT/TA variants had elevated areas under the serum concentration-time curve compared with those for carriers of at least 1 wild-type G allele. The T and A variants at this position in exon 21 result in substitution of an alanine for a serine or threonine, respectively, at amino acid position 893. Other groups have confirmed the functional importance of this SNP.43, 44

Alteration in function of the P-glycoprotein transporter, thus, could alter the relative amounts of morphine and its glucuronides in the CNS, thereby influencing the prevalence and severity of morphine-related central side effects. No differences were observed in serum morphine levels between genotypes, but this simply may reflect the known poor correlation between serum and CNS levels of morphine.

This study also demonstrated that genetic variation in the MDR-1 gene is independent of and additive to that of the COMT gene in predicting morphine-related central side effects. Interaction between the opioid pathway and the noradrenergic and dopaminergic neurotransmitters, metabolized by COMT, is well known. Zubieta et al. demonstrated alteration in neuronal opioid receptor expression in normal volunteers, depending on COMT genotype.24 Those volunteers also demonstrated variation in their response to experimental pain with altered neuronal endogenous opioid production. When exogenous opioids are given, P-glycoprotein function (ie, MDR-1 genotype) will alter the relative amounts of morphine and its glucuronides in the CNS. However, this will have a greater effect if, in addition, the basal expression of opioid receptors is up-regulated (because of the COMT genotype).

Hence, patients who do not have the protective allele for MDR-1 may have a poorly functioning P-glycoprotein42–44; therefore, we hypothesis that this would result in increased CNS levels of morphine for a given dose. These patients also will have a lower M6G/morphine ratio, resulting in larger dose requirements for a given analgesic effect; and this, in turn, may precipitate central side effects. For those patients with abnormalities of COMT, basal opioid receptor levels will be altered24; thus, the combination may further precipitate an increase in central side effects. One case report that involved deletion of the COMT gene in a patient with velocardiofacial syndrome reported severe central side effects with low doses of opioids.45

Further work is needed now to study the 4 base-pair tandem repeat in intron 1 of the COMT gene. This cannot be typed by using SSP-PCR. Although another group has analyzed this region with microsatellite analysis,46 preliminary attempts by our group to conduct similar analyses have been unsuccessful to date, because we have been unable to fully correlate microsatellite and direct sequencing data when patients are heterozygous for different alleles. Thus, chromosome separation will be required before genotyping or sequencing can be performed in this instance.

Understanding the complexity of opioid response will be an ongoing challenge. Previous work by our group and others has examined other candidate genes that may influence response to different opioids. Although genetic variation in the μ-opioid receptor itself does not correlate with response to morphine, β-arrestin2, which modulates opioid receptor function, also has been correlated with opioid response.15 Genetic variation in the enzyme-metabolizing morphine has not been correlated with either serum morphine levels or analgesic response to date15, 47; and other enzymes, such as the cytochrome P450 system, are being evaluated with respect to other opioids. The current results demonstrate the importance of MDR-1 and COMT genotype in influencing morphine-related side effects. Individual factors may influence either sensitivity to pain, or analgesic response to an opioid, or the frequency of opioid-induced side effects. Therefore, ongoing advances in both data mining and modeling techniques will be essential in helping us to combine the strands of information generated to achieve the objective of individualized opioid therapy.

Acknowledgements

We acknowledge Dr. D. Rutter and the research nurses at Royal Marsden Hospital who contributed to the recruitment of patients for this study

Ancillary