Hairy cell leukaemia: a heterogeneous disease?

Abstract

The US National Cancer Institute's Surveillance, Epidemiology and End Results program was used to develop aetiological clues for hairy cell leukaemia (HCL). Descriptive techniques (age-adjusted incidence trends, age-specific incidence rates (IR), and age distributions-at-diagnosis) were supplemented with mathematical models (two-component mixture, generalized linear regression, and age-period-cohort). There were 2856 cases of HCL diagnosed during 1978-2004 (IR 0.32/100,000 person-years). IRs were nearly 4-fold greater among men than women and more than 3-fold higher for Whites than Blacks. Temporal trends were stable over time. Age-specific IRs increased rapidly until approximately 40 years then rose at a slower pace. The age-specific IR curves reflected bimodal early- and late-onset age distributions-at-diagnosis (or density plots), with some variation by gender. Among both men and women, a two-component mixture model fitted the data better than a single density or cancer population. Age-period-cohort models confirmed statistically significant age-related effects after full adjustment for temporal trends (calendar-period and birth-cohort effects). In summary, age incidence patterns (rates and bimodal densities) suggested that HCL is a heterogeneous disease, consisting of at least two underlying subgroups and/or cancer populations by age-at-onset. Distinct early- and late-onset HCL populations may reflect different age-related causal pathways, risk factor profiles, and/or stem cells of origin.

Fig 1

Age-adjusted temporal trends of HCL diagnosed in SEER-17 (1978–2004) according to gender, with P value for trend interaction by gender.

Fig 2

Age-specific incidence and age-distribution of HCL diagnosed in SEER-17 according to gender (1978–2004). (A) Age-specific IRs among males and females, with P value for trend interaction by gender. Age density plots among (B) males and (C) females (95% confidence limits were applied with bootstrap re-sampling techniques).

Fig 3

Age-period-cohort model of HCL diagnosed in SEER-17 (1980–2004). (A) Age effects as rates for the reference period. (B) Period effects as relative risks for the reference period. (C) Cohort effects constrained to be 0 on average with 0 slope. In a sensitivity analysis, age effects were robust for all constraints or parameterizations. Dashed vertical line – reference line. CI, confidence interval.