Drug treatment for nociceptive musculoskeletal pain (NMP) follows a three-step analgesic ladder, starting from non-steroidal anti-inflammatory drugs (NSAIDs), followed by weak or strong opioids until the pain is under control. Here, we conducted a genome-wide association study (GWAS) of a binary phenotype comparing NSAID users and opioid users as a proxy of treatment response to NSAID using data from the UK Biobank. We aim to find the common genetic variants associated with pain treatment response in the general population.Type of data uploaded in this repositoryUK Biobank is a large-scale biomedical database and research resource containing in-depth genetic and health information from half a million UK participants (https://www.ukbiobank.ac.uk/). The database is globally accessible to approved researchers undertaking vital research into the most common and life-threatening diseases. As the raw data is quite large and only available upon application to UKB, we only provide the results from our analysis, which is also described here: medrxiv and currently in revision in a scientific journal. In the dataset, you will find the association of 9,435,994 SNPs genetic variants with the pain treatment response (PTR) phenotype. This dataset is not applicable to be opened with Excel and can best be opened on a cluster computer or using specific software.SubjectsThe UK Biobank is a general population cohort with over 0.5 million participants aged 40–69 recruited across the United Kingdom (UK). We derived a phenotype as a proxy for the pain treatment response to NSAIDs by using recently released primary care (general practitioners', GPs') data, which contains longitudinal structured diagnosis and prescription data. To define the PTR phenotype, we first extracted all nociceptive musculoskeletal pain (NMP) treatments and diagnoses from the GP data. NMP diagnosis was primarily selected from the chapters on musculoskeletal and connective tissue diseases and relevant symptoms or signs from other chapters in the Read codes (versions 2 and 3). See Supplementary data 1 on medrxiv for the diagnosis codes included in this study. Secondly, pain prescriptions (NSAID and opioid) were extracted from the GP data using the British national formulary (BNF), dictionary of medicines and devices (dmd), and Read code (version 2) for data extraction. An overview of the extracted medication codes is provided in Supplementary data 2 on medrxiv. Only participants with an NMP diagnosis record and a pain prescription record occurring on the same date were included for analysis to ensure that we would only include pain treatment for NMP.PhenotypeBased on the information of NMP and pain prescriptions from the UK biobank, a dichotomous score was used for the binary (case/control) PTR phenotype: NSAID users were defined as controls and opioid users as cases. Two additional quality control (QC) steps were applied. First, participants with only one treatment event were removed to safeguard the inclusion of only participants with relatively long-term treatment. Second, a chronological check was applied for the first prescription of each ladder to ensure that the treatment ladder was correctly followed, i.e., initial NSAID use was followed by weak or strong opioids. Participants that were not treated according to this order were removed.SNP genotyping and quality controlGenotyping procedures have been described in detail elsewhere [PMID: 30305743].The third-release genotyping data were used for analysis (see https://biobank.ctsu.ox.ac.uk/crystal/label.cgi?id=100319).Participants passing quality control were included for analysis. QC steps for the samples included removal of participants with (1) inconsistent self-reported and genetically determined sex, (2) missing individual genetic data with a frequency of more than 0.1, (3) putative sex-chromosome aneuploidy. Participants were also excluded from the analysis if they were considered outliers due to missing heterozygosity, not white British ancestry based on the genotype, and had missing covariate data. Note that when we fit the linear mixed model in GCTA, it reminded us that the number of closely related participants was low. Therefore, we didn't further remove the related individuals in the sample.Routine QC steps for genetic markers on autosomes included removal of single nucleotide polymorphisms (SNPs) with (1) an imputation quality score less than 0.8, (2) a minor allele frequency (MAF) less than 0.005, (3) a Hardy-Weinberg equilibrium (HWE) test P-value less than 1 × 10−6, and (4) a genotyping call rate less than 0.95.Genome-wide association analysisA GWAS for binary PTR phenotype was conducted using a linear function in GCTA [38] for markers on the autosomal chromosomes, adjusting for age, sex, BMI, depression history, smoking status, drinking frequency, assessment center, genotyping array, and the first ten principal components (PCs). The following variables from the UK Biobank data set were used for the covariate definition: (1) depression history, which was defined as "YES" if depression records were found in self-reported, inpatient hospital or GP data, and (2) drinking frequency, which was derived from data field 1558: "Daily or almost daily" or "Three or four times a week" was defined as high drinking frequency, other values except for "Prefer not to answer" were defined as low drinking frequency.