Population mixing and the incidence of childhood leukaemias: retrospective comparison in rural areas of New Zealand

BMJ 1996; 312 doi: (Published 11 May 1996) Cite this as: BMJ 1996;312:1203
  1. John D Dockerty, research fellowa,
  2. Brian Cox, senior research fellowa,
  3. Barry Borman, epidemiologistb,
  4. Katrina Sharples, lecturera
  1. a Department of Preventive and Social Medicine, University of Otago Medical School, P O Box 913, Dunedin, New Zealand
  2. b Central Regional Health Authority, PO Box 10097, Wellington, New Zealand
  1. Correspondence to: Dr Dockerty.
  • Accepted 9 January 1996

In 1988 Kinlen hypothesised that childhood leukaemia occurs as a rare response to a common infection and that population influxes into rural areas with higher proportions of susceptible individuals favour epidemics of the infection and increases in the incidence of leukaemia.1 Most ecological studies of childhood leukaemia in relation to this issue have supported the hypothesis.2 Our study was conducted to test whether population influxes into rural areas of New Zealand were associated with increases in the incidence of childhood leukaemias.

Subjects, methods, and results

Forestry developments, including the building of a pulp and paper mill, occurred in the adjacent Rotorua, Matamata, and Whakatane areas of New Zealand's North Island in the 1950s. These developments led to population growth and the establishment of the new towns of Kawerau and Murupara. In this study boundaries were defined for the Rotorua, Matamata, and Whakatane areas (study areas) based on local government divisions. Care was taken to ensure that the areas remained as equivalent as possible over the years. The population of the combined study areas increased greatly at the time of the forestry developments compared with a smaller increase in the rest of New Zealand (fig 1). Similar increases occurred in each individual area (Rotorua, Matamata, and Whakatane).

Fig 1
Fig 1

Rate ratios for childhood leukaemia (ages 0-14, study areas versus the rest of New Zealand). Middle years of the time periods were census years, and population increases are average annual increases for all ages since previous census

Childhood leukaemia registrations for New Zealand residents (1949-83, ages 0-14) were obtained from the National Cancer Registry. On the basis of the street address at diagnosis (or occasionally the domicile code) each child was classified as living in one of the study areas or in the rest of New Zealand. Quinquennial age specific incidence rates were calculated for childhood leukaemias in the study areas (combined) and in the rest of New Zealand. Age adjusted incidence rate ratios (Mantel-Haenszel method3) with 95% confidence intervals were also calculated. To assess differences among the trends in the age specific rates we used the likelihood ratio test for heterogeneity.3

During 1949-83 there were 1134 registrations for childhood leukaemias. The age adjusted rate ratios (study areas versus the rest of New Zealand) were not significantly raised during or after the greatest influx (1951-6; fig 1). For ages 0-4, 5-9, and 10-14 the rate ratios were under 1.8, and none was significantly raised. There was no significant heterogeneity in the rate ratios between time periods for any age group, but for ages 5-9 and 10-14 there were few registrations in the study areas. Subsequently we ran analyses for the 15 years 1949-63. The age adjusted rate ratio was 0.84 (95% confidence interval 0.50 to 1.44). The age specific rate ratios were: 0.92 (0.45 to 1.87) for ages 0-4 years, 0.97 (0.40 to 2.37) for ages 5-9, and 0.37 (0.06 to 2.43) for ages 10-14.

The study areas were very rural in the early 1950s, but prior population mixing would have occurred in Rotorua through tourism. Nevertheless, the age adjusted rate ratio (1949-63) was not raised in any of the three areas (including Matamata and Whakatane, where tourism was negligible).


This study does not provide support for Kinlen's hypothesis, and there are several possible explanations. There were few affected children in the study areas, so a small effect might have been missed. Bias may have been introduced by unmeasured confounders. Community size and population density are important in determining whether an infectious disease will maintain itself,4 and differences between the communities in this study and those studied in the United Kingdom may have produced differences in transmission of infection and different results. The relevant infectious agent may even have been absent from New Zealand in the 1950s. Alternatively, the reported relation between population mixing and childhood leukaemia could have noninfective explanations.5 Leaving speculation aside, the only firm conclusion of this study is that it does not support the population mixing hypothesis.

We thank DCG Skegg, JM Elwood, LDB Heenan, and LJ Kinlen for their helpful comments.


  • Funding Cancer Society of New Zealand. JDD received salary support from the Health Research Council of New Zealand.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract