INTRODUCTION. Lumbar intervertebral disc degeneration (LDD) is an age-related trait and its association with low back pain (LBP) is modest in epidemiological studies. Other factors may modify the LDD-LBP association. LBP is a complex trait under highly polygenic control, yet no studies have examined genome-wide gene-environment interactions relevant to the LDD-LBP relationship. This study examined the association between spine magnetic resonance imaging-determined LDD and LBP in subgroups defined by genetic predisposition to pain. We hypothesized that a polygenic risk score (PRS) for genetic pain propensity influences the LDD-LBP association.
METHODS. This study used data from the UK Biobank (UKB) and TwinsUK. The analytic protocol was posted at https://osf.io/9xyde/before analyses were conducted. First, we conducted a genome-wide association study (GWAS) of multisite chronic pain irrespective of back pain, defined as the number of non-back pain locations (head, face, neck/shoulder, stomach/abdomen, hip, and knee) with duration ≥3 months in UKB participants of European ancestry (EA). We developed a PRS in a separate subsample of 30,000 EA UKB participants using PRSice2. This PRS was considered a general proxy for “genetic propensity to pain”. Next, we stratified TwinsUK participants into 4 equal strata of genetic propensity to pain. A “basic” model examined the association between a summary score of LDD (LSUM) and the dependent variable of ever having had LBP-related disability lasting more than 1 month (LBP1), with adjustment for age and sex only. A “fully adjusted” model examined LSUM and LBP1 with adjustment for age, sex, PRS strata, interaction terms for LSUM x PRS strata, and twin status. To avoid collider bias, other factors such as BMI were not included. We estimated odds ratios (ORs) and 95% confidence interval (CIs). We examined interactions of LSUM x PRS strata, using a p-value <0.05 as the threshold for statistical significance.
RESULTS. The GWAS was conducted in 347,538 UKB EA participants, the PRS was developed in 30,000 UKB EA participants, and the PRS was used in 645 TwinsUK EA participants. Among TwinsUK participants, when using the basic model including age, sex, and LSUM, the risk of LBP1 was OR 1.8 per unit of LSUM (95% CI 1.4 -2.3). In the fully-adjusted model, only age (OR 0.96 per year [95% CI 0.93-0.98]; p=0.001) and the interaction term for highest quartile of genetically-predicted propensity to pain and LSUM (1.8 [95% CI 1.1-3.2]; p=0.03) were statistically significantly associated with LBP1. This indicates that the LSUM-LBP1 association is significantly larger among participants in the highest quartile of genetic propensity to pain, compared to those in the lowest quartile, corresponding to an OR=2.6 for the LSUM-LBP1 association in this subgroup. Female sex (OR 4.1 [95% CI 0.9-20]; p=0.08) and the LSUM main effect (OR 1.4 [95% CI 0.9-2.2]; p=0.15) were not significantly associated with LBP1.
CONCLUSIONS. Genetic propensity to pain is an effect modifier of the LDD-LBP relationship among TwinsUK participants. The overall association of LDD and LBP may be driven by this relationship in those people who are at the highest genetic propensity to pain.