The histological prevalence of beta-2 microglobulin amyloidosis (A beta 2m) was evaluated in a prospective study of joint samples obtained at autopsy in 54 patients on hemodialysis (HD) for 2 to 163 (median 47) months, aged 20 to 80 (median 63) years at HD onset. Carpal tunnel syndrome surgery or radiological signs of A beta 2m were present in 2 and 4% of them, respectively. A control group of 34 patients without end-stage renal disease, autopsied during the same period was used as a reference. The 153 sampled joints (1 to 8, median 2 per patient) were sternoclavicular joints (N = 77), shoulders (N = 35), knees (N = 28), others (N = 13). A beta 2m was diagnosed (positive Congo red with typical birefringence and positive immunostaining of deposits for beta 2m) in 26 of 54 (48%) patients. Prevalence reached respectively 21%, 33%, 50%, 90% and 100% within two years, after 2 to 4 years, 4 to 7 years, 7 to 13 years and more than 13 years HD. The calculated sensitivity of the various joints for A beta 2m detection is significantly higher (P < 0.03) for sternoclavicular joints (97%) and knees (91%) than for shoulders (57%). Multivariate stepwise logistic regression with discriminant analysis identified both HD duration (P = 0.0008) and age at HD onset (P = 0.0093) but not diabetic nephropathy (P = 0.23) or gender (P = 0.25) as independent risk factors for A beta 2m. The probability of joint A beta 2m was quantitated as a function of age and HD duration. In conclusion, A beta 2m may be observed in the large joints early after HD onset. Overall prevalence reaches 48% of the patients on HD for a median of 47 months. It is much higher than that reported on the basis of clinical or radiological evidence. The sternoclavicular and knee joints are more frequently (P < 0.03) involved than the shoulder. The easily accessible sternoclavicular joint therefore appears to be the best site for the early detection of A beta 2m. Both HD duration and age at HD onset, but not diabetic nephropathy, are independent risk factors for A beta 2m.