Seasonality in the diagnosis of acute lymphocytic leukaemia.

Literature on seasonality of leukaemia shows conflicting results. We analysed the month of diagnosis of acute leukaemia in East Anglia, UK, for the period 1971-94, which showed a significant 40% summer excess (P < 0.001) for acute lymphocytic leukaemia both in children (P < 0.01) and adults (P = 0.01). Methodology, results and possible aetiological interpretations are presented.

considering the USA as a whole (Walker and Van Noord, 1982).
A later reanalysis (Harris et al, 1987) of these data for acute lymphocytic leukaemia (ALL), taking account of geographical heterogeneity, suggested a more complex pattern with an indication of seasonality in northern areas.
The inconsistency in the results discussed above may itself be informative, reflecting various levels of within population heterogeneity and different patterns of seasonality in possible causative agents. Further studies in more homogeneous populations are therefore of interest. As no recent studies on the seasonality of acute leukaemia are available from the UK, we examined the data from the East Anglian cancer registry for the same.

MATERIALS AND METHODS
The East Anglian Cancer Registry has been in operation in its present form since 1971 and covers a population of two million, the three counties of Cambridgeshire, Suffolk and Norfolk. The registry receives information from hospital departments at diagnosis, together with reports from a variety of sources, including pathology departments and departments of forensic medicine, giving the results of autopsies of cancer patients. Cases first diagnosed at autopsy, i.e. as an incidental finding, are included in the registry's material. A review of all death certificates supplements this information. We included all subjects diagnosed with leukaemia between 1971 and 1994 for analysis in the study. From the registry records the following information was abstracted for each case, namely sex, date of birth, ICD code, date of diagnosis and mode of registration. One hundred and sixty-nine cases (3.8%) were registered based on death certificates only, and we excluded them from the analysis. Of these only eight cases were classified as ALL. The inclusion of all cases notified by death certificates made no difference to the results reported here. We divided the year into summer (May-October) and winter (November-April) as defined by Knox (1964) to facilitate comparison with published studies from the UK. Summer-winter ratios were calculated for children (0-14 years), adults and for the leukaemia subtypes. This apparently crude approach, being based on a specific prior hypothesis, is more powerful than the application of more complex tests for seasonality.  Child  All  21  19  23  23  24  25  25  32  25  27  17  10  Child  Other   4  4  7  5  7  6  3  2  5  5  8  11  Adult  All  18  23  11  15  26  20  21  29  27  19  15  20  Adult  CLLa  113  110  99  105  123  122  136  110  124  122  113  128  Adult  Acute  104  117  124  125  147  116  103  123  134  143  117  136  (other)  Adult  Else  73  65  68  77  79  70  53  61  76  58  68  68 aCLL, chronic lymphocytic leukaemia; child, ages 0-14 years; adult, ages 15 + years. Table 1 gives the number of registered cases of acute lymphocytic leukaemia (ALL) and other leukaemias by month of diagnosis for those under 15 years of age and those aged 15 years and over. Table 2 summarizes the seasonality shown by ALL, which is not evident for other types of leukaemia. There is a 40% excess of ALL in the summer months, as defined by Knox (1964), which is statistically significant in children and adults and highly significant when the results are combined. There is no suggestion of similar seasonality for any other cell types of leukaemia. The corresponding results for month of birth for ALL show no indication of seasonality. There is a suggestion that seasonality is greatest for cases of ALL diagnosed under 6 years of age and that it is greater in the first half of the time period, but neither effect was close to statistical significance (data not shown).

DISCUSSION
The seasonality of time of diagnosis displayed by ALL in East Anglia over the past 25 years is unlikely to be an artifact of the diagnosis process, both because of the rapidity of disease onset and because of the specificity for ALL. No seasonality is seen for other forms of acute leukaemia or for other types of leukaemia. It is also unlikely that the results are due to chance, given the level of statistical significance. The inconsistent pattern seen in previous studies is perhaps not surprising. For many exposures, seasonal variation will depend on climate and a range of population characteristics. In the USA, for example, in the group of ALL patients aged under 20 years of age, a different pattern is displayed in southern and northern states (Harris et al, 1987), although this finding is obscured by the statistical analysis. In the UK, our results are very similar to the early reports of Lee (1962Lee ( , 1963 for England and Wales and for childhood and adult ALL and of Knox (1964) for childhood ALL in Northumberland and Durham. On the other hand, Till et al (1967) found no seasonality for all childhood leukaemia or for childhood ALL in the Greater London area of the UK. The implication of seasonality of date of diagnosis for the behaviour of a putative aetiological agent has been considered in detail previously (Day et al, 1985) for Burkitt's Lymphoma in Africa, for which seasonality has also been described (Williams et al, 1978). In brief, seasonality of diagnosis requires (a) the existence of a causative agent which displays even greater seasonality and (b) a restricted latent period between exposure to this agent and diagnosis of ALL, as the degree of monthly variation in the incidence of ALL is the combination of monthly variation in the causative agent and monthly variation in the latent period. As it appears to be implausible for a long latent period to have a small standard deviation, restricted variation implies that the mean latent period should be short. Modelling the latent period distribution as log normal with unknown mean and variance to fit the observed seasonality of ALL, in a similar way to that done for Burkitt's lymphoma (Day et al, 1985), suggests a mean latent period of less than 2 years. A number of possible causative agents might display strong seasonality, including electromagnetic fields. Given, however, the strong seasonality displayed by a number of viral infections, these results provide considerable support for the viral hypothesis (Kinlen, 1995). Little can be said concerning the exact pattern of seasonality of the putative causes, as the seasonal peak in ALL could result from a peak in exposure at any season of the year, depending on the mean latent period. For future epidemiological studies, these results suggest that attention should be given to exposures, particularly infections, in the 2 years preceding diagnosis.