Identification of seasonal variation in the diagnosis of acute myeloid leukaemia: a population-based study

Abstract

Until now, the role that seasonal factors play in the aetiology of acute myeloid leukaemia (AML) has been unclear. Demonstration of seasonality in AML diagnosis would provide supportive evidence of an underlying seasonal aetiology. To investigate the potential seasonal and long-term trends in AML diagnosis in an overall population and in subgroups according to sex and age, we used population-based data from a Spanish hospital discharge registry. We conducted a larger study than any to date of 26 472 cases of AML diagnosed in Spain between 2004 and 2015. Using multivariable Poisson generalized linear autoregressive moving average modelling, we found an upward long-term trend, with monthly incidence rates of AML annually increasing by 0.4% [95% confidence interval (CI), 0.2%-0.6%; p = 0.0011]. January displayed the highest incidence rate of AML, with a minimum average difference of 7% when compared to February (95% CI, 2%-12%; p = 0.0143) and a maximum average difference of 16% compared to November (95% CI, 11%-21%; p < 0.0001) and August (95% CI, 10%-21%; p < 0.0001). Such seasonal effect was consistent among subgroups according to sex and age. Our finding that AML diagnosis is seasonal strongly implies that seasonal factors, such as infectious agents or environmental triggers, influence the development and/or proliferation of disease, pointing to prevention opportunities.

Keywords: acute myeloid leukaemia; diagnosis; infection; leukaemias; seasonality.

FIGURE 1

Decomposition of the standardized monthly incidence rates of AML diagnoses in Spain from 2004 to 2015. Nine separate time‐series decompositions are depicted using data for all cases (panel titled 'Overall Population') and for cases stratified by sex and age. Each panel includes the observed series (named 'data') and its three additive components (i.e. trend, seasonal and remainder) obtained from a robust STL (Seasonal and Trend decomposition using Loess) decomposition with flexible trend and fixed seasonality. The grey bars to the right of each panel show the relative scales of the components. Each grey bar represents the same length but because the plots are on different scales, the bars vary in size.

FIGURE 2

Fitted values from each final Poisson generalized linear autoregressive moving average (GLARMA) model. Time‐series plots depict observed counts (black dashed line) of monthly cases of acute myeloid leukaemia (AML) and predicted counts (red smooth line) of monthly AML cases using GLARMA for the overall population in Spain and stratified by sex and age from 2004 to 2015.