Cindy Boer

General Discussion | 257 6 Table 1: Osteoarthritis heritability and variance explained in the Genetics of Osteoarthritis Consortium Phenotype Genetic heritability* (H 2 ) GWAS SNV heritability † (h 2 ) Total genetic heritability explained ‡ Hip OA 58% (29%-87%) [3] 12% 21% (14%-43%) THR 73% (66%-78%) [4] 15% 21% (19%-23%) Knee OA 44% (31%-56%) [5] 5% 11% (8%-15%) TKR 53% (31%-75%) [6] 5% 10% (7%-18%) Spine OA 75% (30%-100%) [7] 10% 13% (10%-14%) Hand OA 56% (34%-76%) [7] 4% 6% (5%-11%) Finger OA 42% (26%-58%) [8] 5% 12% (9%-20%) Thumb OA 53% (37%-69%) [8] 4% 8% (6%-11%) *Osteoarthritis genetic heritability as determined by twin studies. † Predicted proportion of the SNV caused genetic heritability explained by the genome-wide signifi- cant SNVs in the GO Consortium ‡Percentage of the total genetic heritability explained by the genome-wide significant SNVs of the GO Consortium and the 95% Confidence Interval(CI)( H 2 /h 2 ). OA: osteoarthritis, SNV: Single Nucleotide Polymorphism, THR: Total Hip Replacement, TKR: total Knee Replacement The genetic architecture of osteoarthritis may explain the current lack of ex- plained genetic heritability. Osteoarthritis genetic architecture can be described by the polygenic model. Under the assumption of polygenic genetic architecture many genetic variants collectively contribute to disease risk[12]. Rare and/or uncommon variations with very large effect sizes cause more severe and/or early-onset diagnoses, in line with the monogenetic(early-onset) forms of osteoarthritis. Meaning that the effect sizes for common and rare variants linked with the non-monogenetic forms of osteoarthritis are likely to be small to very small[12]. This also means that on an individual level osteoar- thritis genetic risk is caused by a (near) unique combination of many rare and common risk variations, each contributing a small effect to the over genetic risk for osteoarthri- tis. Thus, in order to identify these many rare and common variants with small effect sizes, massive and population specific sample sizes are needed to sufficiently increase GWAS power to identify all of them and to fully explain all of the osteoarthritis genetic risk. Phenotyping is essential for GWAS However, increasing sample sizes for GWAS is a double edged sword: it generates more power through larger sample sizes but also more noise due to heterogeneity in the di- verse cohorts[13]. Increasing sample sizes in GWAS also increases the burden and costs of phenotyping. For very large studies (e.g., UK biobank, 23andMe) extensive phenotyp- ing is not feasible, thus more easily and cost effective phenotypes are collected (hospital records, self-reported diagnosis and complaints). Such “minimal phenotyping”[14] has