To investigate the change in adult T-cell leukemia incidence between 1983 and 1992 and to evaluate the sensitivity of the nationwide adult T-cell leukemia survey, we estimated adult T-cell leukemia incidence in the Kyushu district, southern Japan, where adult T-cell leukemia is endemic. The incidence of adult T-cell leukemia was calculated from the difference between Kyushu and the rest of Japan in mortality from malignant lymphoid neoplasms, i.e., Kyushu's excess rate was assumed to be due to adult T-cell leukemia. In Kyushu, average annual adult T-cell leukemia cases aged > or = 20 years were estimated for men as 252 during the period 1983-87 and 341 during 1988-92, and for women as 201 and 246 respectively. The age-adjusted mortality rate tended to be higher in the latter period [6.29 per 100000 (95% confidence interval 5.59-7.00) vs. 5.25 (4.60-5.90) in men, and 3.33 (2.85-3.80) vs. 3.18 (2.71-3.66) in women]. By contrast, the registered number of adult T-cell leukemia cases nationwide during 1988-93 was only 35% (203/587) of the estimated number, and the number of registered versus estimated cases decreased with age, especially when cases were > 60 years old. In conclusion, the estimated adult T-cell leukemia incidence for 1983-92 increased in the latter half of the period. The estimation suggests that 65% of adult T-cell leukemia cases might be missed by a nationwide survey, and older cases were more likely than younger ones to be missed.