Ankylosing spondylitis (AS) has been shown to produce exercise limitation and breathlessness. The purpose of this study was to investigate factors which may be responsible for limiting aerobic capacity in patients with AS. Twenty patients with no other cardio-respiratory disease performed integrative cardiopulmonary exercise testing (CPET). The results were compared to 20 age and gender matched healthy controls. Variables that might influence exercise tolerance, including pulmonary function tests (body plethysmography), respiratory muscle strength (MIP, MEP) and endurance (Tlim), AS severity assessment including chest expansion (CE), thoracolumber movement (TL), wall tragus distance and peripheral muscle strength assessed by maximum voluntary contraction of the knee extensors (Qds), hand grip strength and lean body mass (LBM), were measured in the patients with AS and used as explanatory variables against the peak VO2 achieved during CPET. As subjects achieved a lower peak VO2 than controls (25.2 +/- 1.4 vs. 33.1 +/- 1.6 ml kg-1min-1, mean +/- SEM, P = 0.001). When compared with controls, ventilatory response (VE/VCO2) in AS was elevated (P = 0.01); however gas exchange indices, transcutaneous blood gases and breathing reserve were similar to controls. AS subjects developed a higher HR/VO2 response (P < 0.01) on exertion but without associated abnormalities in ECG, blood pressure response or anaerobic threshold. The AS group experienced a greater degree of leg fatigue (P < 0.01) than controls at peak exercise. Although the breathlessness scores (BS) were comparable to controls at peak exercise, the slopes of the relationship between BS and work rate (WR) [AS 0.054 (0.1), Controls 0.043 (0.06); P < 0.05] and BS and % predicted oxygen uptake [AS 0.084 (0.18), Controls 0.045 (0.06); P < 0.01] were steeper in the AS subjects. There was weak association between peak VO2 and vital capacity (r2% 12.0), MIP (11.8) but no association between Tlim, CE, Wall tragus distance or TL movement. The strongest association with aerobic capacity was between measurements of peripheral muscle strength (Qds; r = 0.75; hand grip; r = 0.47) accounting for 53% (P < 0.001) and 23.5% (P < 0.01) of the total variance in peak VO2, respectively. The addition of LBM to Qds in the regression model significantly improved the explained variance to 78.3% (P < 0.001). This study shows that peripheral muscle function is the most important determinant of exercise intolerance in AS patients suggesting that deconditioning is the main factor in the production of the reduced aerobic capacity.