HLA-DQB1*05:03 ranks at the top (P = 4·7 × 10<sup>-40</sup> ; odds ratio 12·4) in both subtypes, with significantly different risk allele frequency (RAF<sub>PV</sub> = 34·2% vs. RAF<sub>PF</sub> = 18·8% vs. RAF<sub>control</sub> = 4·4%), whereas HLA-DRB1*14:01 and HLA-DRB1*04:06 are PV specific.
For the HLA variants, we confirmed independent genome-wide level risk associations in HLA-DQB1 and HLA-DRB1, with DQB1*05:03 to be the strongest association with PV (P = 8.59 × 10<sup>-68</sup> , OR = 31.16) and PF (P = 4.84 × 10<sup>-17</sup> , OR = 5.64).
When based on each subtype of HLA-DQB1, DQB1*05:03 and DQB1*03:02 may play susceptibility roles in PV, and DQB1*03:03, DQB1*05:01 and DQB1*06:01 are negatively associated with PV.
Our findings suggest that HLA-DRB1*04:02, DRB1*14:04, HLA-DQB1* 03:02 and DQB1*05:03 alleles and HLA-DRB1*04:02-DQB1*03:02 and HLA-DRB1*14:04-DQB1*05:03 haplotypes are genetic markers for susceptibility for PV, while DRB1*11 allelic group appears protective in Serbian population.