Time Trend and Age-Period-Cohort Effects on the Incidence of Patients with Adult T-Cell Leukemia in a Population-Based Study in Japan, 1991-2010

Background: Adult T-cell leukemia-lymphoma (ATL) is an aggressive T-cell malignancy, which is exclusively associated with human T-cell leukemia virus type I (HTLV-1) infection. Japan is a well-known endemic area of ATL, with approximately 400 new cases and 1,000 deaths annually. A recent Japanese study reported that the trend in the age-standardized incidence rate of ATL has remained relatively stable. However, age-standardized rate may mask important epidemiological characteristics related with age and birth-cohort. To date, several epidemiological studies reported the interesting birth-cohort patterns on the incidence of lymphoid malignancies. However, little has been known about the effects of age and birth-cohort on the time trend in the incidence of ATL.

Methods: A dataset of patients with ATL (ICD-O-3 code 9827) was obtained from the Nagasaki Prefecture Cancer Registry that contains all cancer data from a population of approximately 1.5 million since 1985. To calculate the average annual percent change (AAPC) for crude incidence rate and age-standardized (world) incidence rate (ASRw) (per 100,000), the Joinpoint Regression Program with Hudson’s method was used. The effects of age, calendar-period, and birth-cohort on the time trends of the ATL incidence were evaluated by performing the age-period-cohort (APC) analysis using the R-package Epi. For the APC analysis, ten 5-year age groups and five 5-year calendar periods were set, resulting to create 14 birth cohorts. Parameter estimates of the APC models were obtained by fitting five models (age, age-drift, age-period, age-cohort, and age-period-cohort). In the full model of the age-period-cohort model, period and birth cohort effects were evaluated as relative risk (RR). This work was supported in part by the Health Labour Sciences Research Grant (H26-sinkoujitsuyouka-ippan-013).

Results: A total of 1,971 patients (1,085 men and 886 women) who were diagnosed during 1991-2010 were analyzed. The crude incidence rate showed a slightly increasing tendency from 8.2 (year 1991) to 9.7 (year 2010) for men (AAPC, +0.6, 95%CIs, -1.6 to 1.8) and 5.7 (year 1991) to 8.4 (year 2010) for women (AAPC, +1.3, 95%CIs, -0.2 to 2.8), without statistically significant. Meanwhile, the ASRw showed a significantly decreasing tendency from 5.2 (year 1991) to 3.7 (year 2010) for men (AAPC, -2.1, 95%CIs, -3.2 to -1.1) and from 3.5 (year 1991) to 2.5 (year 2010) for women (AAPC, -1.8, 95%CIs, -3.3 to -0.3). The age-specific incidence rate was highest in patients aged 70 years or older, particularly the rate in patients aged 80 years or older drastically increased after year 2003. In the APC analyses, both the age-cohort model and age-period model were statistically significant for men (P<0.0001 for both) and women (P=0.002 for both), indicating that both a cohort effect and a period effect were the major effects on the incidence of ATL. The RR among various cohorts showed a drastic difference of the cohort effect on the ATL incidence from a higher risk in cohorts before 1931 but a lower risk in cohorts after 1931 in men, and from a higher risk in cohorts before 1940 but a lower risk in cohorts after 1940 in women. The RR among various periods was almost stable in 1, but becomes high after the period 2000 in both sexes.

Discussion and Conclusion: The earlier birth-cohort and the recent calendar period were significantly associated with the incidence of ATL beyond the age effect. The birth cohort effect reflects the exposure to environmental factors, thus the effect can be explained by the higher HTLV-1 infection in early generation. Sex difference in the cohort effect may be reflected the way of life, biological, and behavioral factors. Continued ATL incidence and careful analysis of period effects are of importance to elucidate this unique epidemiology of ATL.