Management guidelines assume that results from clinical trials can be generalised, although seldom is data available to test this assumption. We aimed to determine the proportion of patients commencing tumour necrosis factor inhibition (TNFi) who would have been eligible for relevant clinical trials, and whether treatment response differs between these groups and the trials themselves. The British Society for Rheumatology Biologics Register for Ankylosing Spondylitis (BSRBR-AS) recruited a real-world cohort of TNFi-naïve spondyloarthritis patients with data collection from clinical records and patient questionnaires. Participant characteristics were extracted from trials identified from a recent Health Technology Assessment of TNFi for ankylosing spondylitis/non-radiographic axial spondyloarthritis. Descriptive statistics were used to determine the differences, including treatment response, between BSRBR-AS participants who would/would not have been eligible for the clinical trials and with trial participants. Among 2420 BSRBR-AS participants, those commencing TNFi (34%) had shorter symptom duration (15 vs 22 years) but more active disease (Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) 6.4 vs 4.0; Bath Ankylosing Spondylitis Disease Functional Index (BASFI) 6.2 vs 3.8). Of those commencing TNFi, 41% met eligibility criteria for ≥1 of fourteen relevant trials; they reported higher disease activity (BASDAI 6.9 vs 6.1) and poorer function (BASFI 6.6 vs 6.0). 61.7% of trial participants reported a positive treatment response, vs 51.3% of BSRBR-AS patients (difference: 10.4%; 95% CI 4.4% to 16.5%). Potential eligibility for trials did not influence treatment response (difference 2.0%; 9.4% to 13.4%). Fewer patients in the real world respond to TNFi than is reported in the trial literature. This has important implications for the generalisability of trial results, and the cost-effectiveness of TNFi agents.