Exploring Interaction between Genetically Predicted Body Mass Index and Serum 25-Hydroxyvitamin D Levels on the Odds for Psoriasis in UK Biobank and the HUNT Study: A Factorial Mendelian Randomization Study

Abstract

Mendelian randomization (MR) studies show that higher body mass index (BMI) and lower 25-hydroxyvitamin D (25[OH]D) increase psoriasis risk. The combined effect of these factors has not been explored using factorial MR. Using cross-sectional data from UK Biobank (n = 398,404) and The Trøndelag Health Study (n = 86,648), we calculated polygenic risk scores for BMI and 25(OH)D to estimate ORs for psoriasis using 2 × 2 and continuous factorial MR. We quantified additive interaction by estimating relative excess risk due to interaction. We also performed traditional observational analyses in UK Biobank. There were 12,207 (3.1%) participants with psoriasis in UK Biobank and 7794 (9.0%) in The Trøndelag Health Study. In 2 × 2 factorial MR, we found no evidence of relative excess risk for psoriasis due to interaction between genetically predicted higher BMI and lower 25(OH)D, neither in UK Biobank (relative excess risk due to interaction = -0.01, 95% confidence interval = -0.08 to 0.07) nor in The Trøndelag Health Study (relative excess risk due to interaction = -0.04, 95% confidence interval = -0.14 to 0.06). The same was observed in the continuous factorial MR and observational analyses. In conclusion, this study did not find evidence of interaction between BMI and 25(OH)D on the risk of psoriasis. Given minor differences in measured BMI and 25(OH)D between the factorial groups, small effects may have been undetected.

Keywords: BMI; Interaction; Mendelian randomization; Psoriasis; Vitamin D.

Figure 1

Factorial Mendelian randomization using a 2 × 2 design. This image was adapted from Rees et al (2020). BMI-PRS denotes the PRS for BMI, and vitD-PRS denotes the PRS for 25(OH)D. 25(OH)D, 25-hydroxyvitamin D; BMI, body mass index; PRS, polygenic risk score.

Figure 2

OR estimates for psoriasis from factorial MR analyses in UKB and HUNT. (a) Unscaled associations. (b) Associations in UKB scaled to represent a 5 kg/m² increase in BMI and 10 nmol/l decrease in 25(OH)D. Upper panel (for a and b): Results from 2 × 2 factorial MR. Participants are separated in 4 groups on the basis of median PRS for BMI and 25(OH)D. RERI (95% CI) was estimated to −0.01 (−0.06 to 0.09) in UKB and −0.04 (−0.14 to 0.06) in HUNT. Meta-analysis of the UKB and HUNT result showed RERI (95% CI) of −0.015 (−0.046 to 0.048). Interaction exists if RERI ≠ 0. Lower panel (a and b): Results from continuous factorial MR, including the 2 PRSs on a continuous scale as well as the cross-product between PRSs. RERI (95% CI) was estimated to 0.00(-0.02, 0.02) in UKB and 0.00(-0.03, 0.02) in HUNT. Meta-analysis of the UKB and HUNT result showed RERI (95% CI) of 0.00 (−0.017 to 0.017). The symbols # denotes PRSs for BMI and 25(OH)D. In 2 × 2 analysis, PRSs were dichotomized at the median value and combined to 4 groups. OR estimates derived from logistic regression models adjusted for age, sex, genetic batch, and principal components 1–20. The symbol † denotes that continuous factorial MR included the PRS for BMI, PRS for 25(OH)D (both standardized to have a mean of 0 and SD of 1), and their cross-product. The symbol ‡ denotes the ORs for the additive term of the continuous variables calculated by exponentiating the sum of estimated beta values for PRS for BMI (b1), PRS for 25(OH)D (b2), and their cross-product (b3) (ie, OR = exp[b1+b2+b3]). N = total number of participants. 25(OH)D, 25-hydroxyvitamin D; BMI, body mass index; CI, confidence interval; HUNT, The Trøndelag Health Study; MR, Mendelian randomization; PRS, polygenic risk score; RERI, relative excess risk due to interaction; UKB, UK Biobank.

Figure 3

Flow charts showing the study sample selection. Left panel: UKB. Right panel: HUNT. The symbol ∗ denotes that participants with ICD9 codes 696.2 parapsoriasis, 696.3 pityriasis rosea, 696.4 pityriasis rubra pilaris, 696.5 other and unspecified pityriasis, or 696.8 other psoriasis and similar disorders were excluded. N = total number of participants. HUNT, The Trøndelag Health Study; ICD9, International Classification of Diseases, Ninth Revision; UKB, UK Biobank.

Figure 4

Venn diagram showing the source of psoriasis diagnosis in HUNT. Numbers are shown as count (%). HUNT, The Trøndelag Health Study; ICD10, International Classification of Diseases, Tenth Revision; ICD9, International Classification of Diseases, Ninth Revision; ICPC-2, International Classification of Primary Care, 2nd Edition.