French geocap study confirms increased leukemia risks in young children near nuclear power plants

Abstract

Dear editor, The new French study1 (Geocap) of childhood leukemia near nuclear power plants (NPPs) is the fourth European study of its kind. It follows studies with similar findings in Germany,2, 3 Great Britain4 and Switzerland.5 The noteworthy result of the French study is a statistically significant increase in leukemia in children below age 15 in 2002–2007 within 5 km of 19 French NPPs. This finding was determined in two separate ways. First, by a comprehensive nationwide case–control study (odds ratio = 1.9); and second, by a geographical study [standardized incidence ratio (SIR) = 1.9]. However, the authors also state that they only found increases in 2002–2007 but not between 1990 and 2001 nor over the whole study period, 1990–2007. But, their Table 2 in fact shows an SIR of 1.1 for children below 15 and an SIR = 1.4 for children below age 5 for 1990–2007 which, due to small numbers, are not statistically significant. The authors close their article stating: “Overall, the findings call for […] collaborative analysis of all the evidence available from multisite studies conducted in various countries.” This is relatively straightforward to perform, so we pooled the data from the above four countries and studied the distance dependency of leukemia incidence by regression analyses. The French geographical study provides numbers of observed (O) and expected (E) cases in 1990–2007. The German KiKK study2 itself is a case–control study which cannot be used for a pooled analysis, because it does not provide information on expected cases. But, after the KiKK study, an additional geographical study3 was performed in the same study region and for the same time span. The British COMARE 14 report4 focuses on acute leukemia plus non-Hodgkin lymphoma, but it also contains data for acute leukemia only. The Swiss study is a cohort study without numbers of expected cases, but these can be calculated from information in the study. In addition to data on residence at birth, the Swiss study contains data on residence at diagnosis. We used the latter in our pooled analysis as the other three studies also used residence at diagnosis for determining distances from NPP sites. Our pooled analysis of acute leukemia in children under 5 years in Great Britain (1969–2004, 0–25 km, SIR = 1.30),4 Germany (1980–2003, 0–30 km, SIR = 1.41),3 Switzerland (1985–2009, 0–15 km, SIR = 1.40)5 and France (1990–2007, 0–20 km, SIR = 1.37)1 yields an overall SIR of 1.37 in the 5 km zone (95% CI: 1.09–1.71, p = 0.008). We also found an SIR of 0.95 (0.89–1.01) at distances r = 5 km. From the ratio of the two SIRs, a relative risk (RR) of 1.37/0.95 = 1.44 (p = 0.0034) is obtained. An excess risk is only found in the 5 km zone in the four countries. Most SIRs are smaller in the 5–10 km distance zones than farther away (see our Figure 1). The likely reason is that the spontaneous leukemia incidence in rural areas where most NPPs are situated is lower than the national average. A comparison of the combined SIR in the 5 km zone (1.37) with the combined SIR in the 5–10 km zone (0.91) yields an RR = 1.52 (p = 0.004). Thus, only using SIR and not RR as a measure of risk may systematically underestimate the real levels of risk in the vicinity of NPPs. Leukemia incidence near nuclear plants in Germany (D), Great Britain (GB), Switzerland (CH) and France (F), and results of joint Poisson regressions with a linear (dotted line) and a linear-quadratic (solid line) dependency on reciprocal distance. To study the shape of the distance dependency of leukemia risk, we performed a joint Poisson regression of the four datasets, using a linear as well as a linear-quadratic dependency on reciprocal distance. The linear-quadratic model yields a better fit to the data than the linear model (Fig. 1). The best fit, judged by the Akaike information criterion, is obtained with a model estimating the excess rate in the 5-km zone relative to the rate in the remainder of the study area. The result is an RR = 1.44 (p = 0.0018). To conclude, the new French data provide valuable additional evidence for an increased leukemia risk near NPPs. Unfortunately, the study is marred by important information only being found in a table at the bottom of the article and by the authors' incorrect statement that over the whole study period no increased leukemia risk was found. The new data fit squarely within the preexisting epidemiological evidence of increased childhood leukemias near NPPs in three other European countries. Over these four multisite studies, a consistent pattern of increased incidences of childhood leukemias near NPPs is clearly emerging. Yours sincerely, Alfred Koerblein Ian Fairlie Alfred KOERBLEIN, Ian FAIRLIE.