Objective. To study the HRQOL in veterans with prevalent total knee arthroplasty (TKA) or total hip arthroplasty (THA) and compare them with age- and gender-matched US population and control veteran population without these procedures. Methods. A cohort study and cross-sectional survey on veterans obtained demographics and HRQOL with Short-Form 36 for veterans (SF-36V). Veterans were categorized into: primary TKA; primary THA; combination group (≥ 1 primary and/or any revision TKA/THA); and control population (no THA/TKA). Multivariable regression compared the physical and mental component summary scores (PCS and MCS scores, respectively) in each group. Results. Response rate was 58% (40 508/70 334): 531 with TKA, 254 with THA, 461 constituted the combination and 39 262, the control group. Mean PCS scores in veterans with THA, TKA, and combination group were 2 s.d. lower than the US mean (29.5 ± 0.8; 30.1 ± 1.1 and 27.1 ± 0.8). MCS scores were similar to the US mean (47.3 ± 0.9; 49.1 ± 1.2 and 45.6 ± 0.9). Compared with controls, significantly more veterans in TKA, THA or combination groups had multivariable-adjusted PCS ≤ 30 (55, 64, 71 and 76%; P ≤ 0.0001); similar proportion had MCS ≤ 30 (15, 12, 8 and 16%; P = 0.29); and mean scores on SF-36 physical domains (P ≤ 0.0011), but not mental/ emotional domains (P = 0.01) were statistically and clinically lower. Conclusions. Profound physical HRQOL deficits exist in veterans with TKA/THA and in combination group compared with age- and gender-matched general US population and with veteran controls. In these groups, these deficits are not attributable to differences in sociodemographics, comorbidity and healthcare access/utilization. Arthroplasty status may be a surrogate for poorer HRQOL and worse outcomes. Future studies are indicated to determine HRQOL deficit causes and interventions to improve HRQOL in patients with arthroplasty. © The Author 2008. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved.