Pain project - MR analyses

To: JH
From: AKV
Subject: MR pain
Date: 2018-08-01

Follow-up from previous work:

Task: explore relationship between statin use and mi/stroke (since most with a stroke or mi will be on a statin)

Table 1. Proportion of MI and stroke sufferers taking a statin at baseline

No statin Statin
Had MI 1,928 (16.6%) 9,666 (83.4%)
Had stroke 2,354 (30.7%) 5,303 (69.3%)

Collinearity diagnostics are not well-defined for categeorical predictors, but I fitted models without MI and stroke and the effect of statins is very stable.

Task: Explore association of statin use and pain constructs in full UKB sample (not QC'd subsample)

Table 2. Regression of musculoskeletal pain on statin use and other individual-level characteristics in full UKB sample (blockwise addition of predictors)

Effect of statins
n OR
Just statins 500268 1.361
+ age and sex 500268 1.406
+ BMI 498073 1.289
+ SBP and DBP 497621 1.278
+ alcohol and smoking 495392 1.246
+ stroke, MI, and angina 494490 1.148
+ hypertension & diabetes 493231 1.058
+ aspirin & bp meds 493231 1.071

Table 3. Regression of general pain on statin use and other individual-level characteristics in full UKB sample (blockwise addition of predictors)

Effect of statins
n OR
Just statins 501778 1.263
+ age and sex 501778 1.358
+ BMI 499514 1.203
+ SBP and DBP 499055 1.199
+ alcohol and smoking 496108 1.173
+ stroke, MI, and angina 495062 1.077
+ hypertension & diabetes 493784 1.005
+ aspirin & bp meds 493784 1.003

Table 4. Regression of arthritic pain on statin use and other individual-level characteristics in full UKB sample (blockwise addition of predictors)

Effect of statins
n OR
Just statins 501778 1.387
+ age and sex 501778 1.319
+ BMI 499514 1.149
+ SBP and DBP 499055 1.140
+ alcohol and smoking 496108 1.120
+ stroke, MI, and angina 495062 1.064
+ hypertension & diabetes 493784 1.020
+ aspirin & bp meds 493784 1.030

Question for JH: Other known/possible risk factors for pain: SES, employment status, physical activity/exercise, mental health, sleep, nutrition, race - should I be considering these as well?

Task: examine prevalence of pain by two functional SNPs: rs12916 and rs17238484

**Notes:**

  • Dosage was recoded to represent number of LDL lowering alleles
  • rs17238484 has non-integer dosage values, so frequency table is not included

Table 5. Baseline prevalence of musculoskeletal pain, by dosage in rs12916

Allele count No MSK pain MSK pain Overall P-value (Chi-Square)
0 40,044 (16.1%) 14,085 (16.0%) 54,129 (16.1%) 0.550
1 119,365 (47.9%) 42,436 (48.1%) 161,801 (48.0%)
2 89,573 (36.0%) 31,617 (35.9%) 121,190 (35.9%)

Table 6. Association of functional SNPs and musculoskeletal pain

SNP Participants EAF Cases Controls OR (95% CI) P
rs12916 337120 0.599 88138 248982 1.000 (0.989,1.011) 0.997
rs17238484 337120 0.775 88138 248982 1.005 (0.992,1.018) 0.430

Task : Create polygenic risk score using SNPs from Ference et al.

Fig1. Distribution of HMGCR polygenic risk score using SNPs from Ference et al. (2016)

Fig1. Distribution of HMGCR polygenic risk score using SNPs from Ference et al. (2016)

UKB participants below the median on the HMGCR GRS were slightly more likely (1% absolute difference) to use a statin compared to those above the median (Table 7).

Table 7. Proportion of UKB participants who reported using a statin, stratified by genetic LDL-lowering risk (below or above the median)

GRS score No statin Statin Overall P
Below median 142,421 (83.1%) 29,044 (16.9%) 171,465 <.001
Above median 139,621 (84.1%) 26,416 (15.9%) 166,037

Table 8. Association of genetically determined statin use (using median-split GRS) for each of the pain constructs

Phenotype Participants Controls Cases OR (95% CI) P
Musculoskeletal pain 337120 248982 (73.9) 88138 (26.1) 1.004 (0.989,1.020) 0.609
General pain 337533 133781 (39.6) 203752 (60.4) 0.998 (0.984,1.012) 0.748
Arthritic pain 337533 198232 (58.7) 139301 (41.3) 0.999 (0.985,1.013) 0.878

Table 9. Associationof genetically determined statin use (using continuous GRS) for each of the pain constructs

Phenotype Participants Controls Cases OR (95% CI) P
Musculoskeletal pain 337120 248982 (73.9) 88138 (26.1) 1.018 (0.964,1.075) 0.525
General pain 337533 133781 (39.6) 203752 (60.4) 0.998 (0.950,1.047) 0.919
Arthritic pain 337533 198232 (58.7) 139301 (41.3) 1.000 (0.953,1.050) 0.986

Table 10. Baseline Characteristics of Participants, According to HMGCR Genetic Score (N = 337,534)

Characteristic Below median score (N=171,480) Above median (N=166,054) P-value
Male - no. (%) 79,426 (46.3%) 76,841 (46.3%) 0.801
Age - mean (yrs) 56.9 ± 8.0 56.8 ± 8.0 0.003
BMI - mean??? 27.3 ± 4.7 27.5 ± 4.8 <.001
Biomarkers
Systolic blood pressure - mean (mm Hg) 138 ± 19.0 138 ± 19.0 0.627
Diastolic blood pressure - mean (mm Hg) 82.3 ± 10.1 82.3 ± 10.1 0.390
Ever smoked - no. (%) 77,238 (45.2%) 75,264 (45.5%) 0.084
Ever drank - no. (%) 166,077 (96.9%) 160,769 (96.9%) 0.819
Previous disease - no. (%)
MI 3,967 (2.32%) 3,832 (2.31%) 0.909
Angina 5,449 (3.18%) 5,181 (3.12%) 0.336
Stroke 2,576 (1.50%) 2,628 (1.58%) 0.058
Hypertension 46,415 (27.1%) 44,723 (27.0%) 0.371
Diabetes 8,046 (4.70%) 8,177 (4.93%) 0.002
Prior medication - no. (%)
Blood pressure 35,607 (20.9%) 34,509 (20.9%) 0.964
Aspirin 24,027 (14.1%) 22,983 (14.0%) 0.176

Table 11. SNPs used in HMGCR risk scores (Ference vs Katerina's diabetes work)

Ference et al. Katerina
rs12916 rs12916
rs5909 rs3857388
rs2303152 rs10515198
rs2006760 rs12173076
rs17238484 rs7711235
rs10066707