Data were collected from three previously published experimental pain studies in healthy volunteers. The first study (Study A) was a comparative study of oxycodone and morphine in a multi-modal, tissue-differentiated experimental pain model . The second (Study B) was a study of analgesic efficacy of peripheral κ-opioid receptor agonist CR665 compared to oxycodone in a multi-modal, multi-tissue experimental human pain model . The third (Study C) was a study of different effects of morphine and oxycodone in experimentally evoked hyperalgesia . All studies were approved by the local Ethical Committee (A: VN 2002/143, B: VN-20060021, C: N-0070025) and The Danish Medicines Agency (A: 2612–2168, B: 2612–3145, C: 2612–3463).
Each study was a three-arm study, with each subject receiving placebo, oxycodone and either morphine or CR665. Therefore each subject underwent baseline pain sensitivity testing on three occasions, each at least one week apart. Only these baseline pain data, which were assessed before administration of placebo and each of the two study drugs, were included in this study. This study includes data from pain sensitivity tests which were performed in at least two out of the three studies. Thermal skin pain, muscle pressure pain and mechanical visceral pain were analysed. For each study subject, details about age and gender were also recorded.
DNA was available for 19 out of 24 subjects from Study A, 15 out of 18 subjects from Study B and 21 out of 24 subjects from Study C. One subject from Study B was excluded due to non-Caucasian ethnicity, in order to minimise bias due to population stratification in the genetic association analysis. Four subjects were excluded from study C as they had already participated in Study B.
Experimental pain sensitivity testing
Thermal skin pain
Thermal skin testing data were used from studies A and C (total N = 36). A computer-driven heat pain device (TSA-II NeuroSensory Analyzer, Medoc Ltd, Ramat Yishai, Israel) was used for the heat stimulation. A thermode with a surface of 25 × 50 mm was applied to the volar surface of the forearm 10 cm distal from the elbow and the subjects were asked to press a button when the pain tolerance threshold (PTT) was reached i.e. when they could no longer withstand the pain. The temperature was increased from 32°C to a maximum at 52°C at a rate of 1°C/s. When the subjects pressed the button, the thermode was cooled to 32°C and the experiment repeated. Three consecutive measurements were performed and the average was computed.
Muscle pressure pain
Muscle pressure pain data were used from studies B and C (total N = 31). The electronic cuff algometer (Aalborg University, Aalborg, Denmark) consisted of a pneumatic tourniquet cuff, a computer-controlled air compressor, and an electronic 10-cm visual analogue scale (VAS). The compressor (Condor MDR2; JUN-AIR International A/S, Nørresundby, Denmark) was connected to an electric-pneumatic converter (ITV2030; SMC Corp., Tokyo, Japan) and controlled by a computer through a data acquisition card (PCI 6024E; National Instruments, Austin, TX). The pain intensity was recorded continuously on the visual analogue scale (VAS) and sampled at 100-ms intervals. The pneumatic tourniquet cuff was wrapped tightly around the gastrocnemius muscle. The cuff was automatically inflated (compression rate 0.50 kPa/s) until the PTT was reached.
Mechanical visceral pain
Mechanical visceral pain data were used from studies A, B and C (total N = 50). A probe designed for multimodal stimulation of the oesophagus was used. Before the study all subjects were instructed how to use the 0–10 electronic VAS, for the visceral stimulations, where 0 = no perception and 10 = unbearable pain. In order to induce mechanicaloesophageal pain, the oesophageal bag was distended at a constant infusion rate until ‘moderate pain’ intensity ratings (defined in these studies as VAS score of 7) were reached. The volumes (ml) of distension at this point were used for further analysis.
Single nucleotide polymorphism (SNP) genotyping
Genotyping for polymorphisms across OPRM, OPRK and OPRD was carried out using sequence specific primers and polymerase chain reaction. Five polymorphisms in OPRM (rs6912029G/T, rs179997A/G, rs56364C/T, rs9479757G/A, rs533586C/T), eight in OPRK (rs1050415T/C, rs7836120A/G, rs647379T/C, rs1365098G/T, rs701677T/A, rs7824175G/C, rs16918875C/T, rs963549G/A) and five in OPRD (rs1042114G/T, rs533123G/A, rs419335A/G, rs2236857T/C, rs2234918C/T) were included in the study. The details of primers are as previously described . Within each of the genes polymorphisms were selected to try to cover allelic diversity across each gene. Polymorphisms in regions most likely to have an impact on gene function were prioritised (promoter region, exons, intron-exon boundaries, 3-UTR), as were polymorphisms for which there exists published data.
Linkage disequilibrium between pairs of SNPs was tested and haplotypes were constructed using the software programmes PHASE  (http://www.stat.washington.edu/stephens/software.html) and Haploview (Haploview version 4.0, Broad Institute, Cambridge, USA) .
All genotype frequencies were tested for Hardy-Weinberg equilibrium using Chi-square goodness-of-fit test (Haploview version 4.0, Broad Institute, Cambridge, USA). Each genetic variant tested was a bi-allelic single nucleotide polymorphism. The genetic model under which a genetic variant may be assumed to influence the pain sensitivity phenotype is unknown. It is not known whether the variant allele could be a risk-enhancing allele (increases the likelihood of a poor outcome) or a protective allele (increases the likelihood of a good outcome). For this reason allele carriage, frequency of subjects homozygous (AA) + heterozygous (AB) for allele A versus frequency of subjects homozygous (BB) for allele B, was used for the genetic association analysis (Table 1).
Reliability between the three baseline scores for each patient was tested using Cronbach’s α. Values >0.7 were considered to be indicative of reliability. The average of the three baseline pain sensitivity scores were used as the outcome variables in the genetic association study.
A number of statistical methods for examining gene-gene and gene-environment interactions have been proposed including regression, classification and regression trees, neural networking, combinatorial partitioning and multifactor dimensionality reduction . In this study multivariate linear regression was used to investigate the joint effect of the predictor variables (clinical and genetic) on the experimental pain thresholds (dependent variables).
Univariate analyses were carried out to screen for an association between individual clinical (gender and age) and genetic predictor variables and the outcome variables (average pain sensitivity scores). Normally distributed continuous variables (thermal skin pain and mechanical visceral pain) were analyzed using t-tests and non-parametric data (muscle pressure pain) were analysed using Man-Whitney U test. Factors with p<0.1 on univariate analysis (Table 2) were included in the multivariate modelling. Variables with p>0.1 were excluded in order to reduce the number of predictor variables. Variable selection was carried out using a stepwise method. Only factors with p<0.05 were retained in the final model. Non-parametric outcome data (i.e. muscle pressure pain) were logarithmically transformed for regression analysis. A Bonferroni adjustment for multiple testing was applied, dependent on the number of factors included in the final modelling .
Statistical analysis and plots were performed using PASW 18 for Windows (SPSS, Chicago, IL) and GraphPad version 4.02 for Windows (GraphPad Software, San Diego California USA).