Coffee and the risk of osteoarthritis: a two-sample, two-step multivariable Mendelian randomization study

Zhang W, Lei X, Tu Y, Ma T, Wen T, Yang T, Xue L, Ji J, Xue H · Frontiers in Genetics

Coffee intake was associated with higher OA risk across sites, largest for knee OA (OR 2.03, 95% CI 1.57-2.61, p < 0.001) and smallest for self-reported OA (OR 1.03, 1.01-1.05, p = 0.006). By type, only decaffeinated coffee was causally associated with knee OA (OR 4.40, 95% CI 1.71-11.33, p = 0.002) and with self-reported OA (1.13, 1.02-1.26, p = 0.022). The paper's own external validation reproduces the coffee-intake result at every site except the hip (OR 1.18, 95% CI 0.53-2.65, p = 0.685) and reproduces none of the results by coffee type, which the discussion and the conclusion do not mention. Body mass index carries 71.1% of the effect on self-reported OA, 67.9% at the knee, 60.6% at knee-or-hip and 45.1% at the hip; bone mineral density carries 1.5-2.3%, and by the delta method only the knee-or-hip estimate reaches significance (p = 0.047).

Population

GWAS summary statistics for coffee intake and for knee, hip, knee-or-hip, total and self-reported osteoarthritis, with BMI and bone mineral density tested as mediators

Intervention

Genetically predicted coffee intake, including by coffee type

Comparator

Lower genetically predicted intake

Limitations

An odds ratio of 4.40 for decaf and nothing for ground or instant coffee is far more consistent with the instrument capturing who chooses decaf — older, more health-conscious, more likely to have been told to cut down — than with a biological effect of removing caffeine. The paper's own validation set does not carry it, yet the conclusion advises against decaf. That most of the total effect runs through BMI says the same thing in another way. European-ancestry GWAS, with pleiotropy detected for some coffee types.