LettersGerman nuclear plant cancer data disputedTo the editors: I have read with interest the paper by Körblein and Hoffmann (“Childhood Cancer in the Vicinity of German Nuclear Power Plants,” M&GS 1999:6:18-23) based on a study performed by us. The authors are not familiar with the epidemiology literature. They write “to date (August 1999) the new study has not been published in the scientific literature,” although the results were published in 1998 [1]. Both in this publication and in the technical report mentioned in the article, the study design is clearly described but not understood by the authors. The study was performed to validate exploratory results obtained in a former (1980-1990) study on the incidence rates of childhood malignancies in the vicinity of German nuclear power plants and to evaluate the confirmatory results of this previous study. For the evaluation of the former exploratory results only independent data (from 1991-95) can be used. Inclusion of the previous, hypothesis generating data into the analysis--as performed by the authors--corresponds to a vicious circle and produces misleading results. No exploratory results from the previous study could be reproduced by the validation study. This is also true for children with acute leukemia younger than 5 years who were living within a 5 km radius of an installation. An observed relative risk (RR) of 1.39 was not significantly increased (95 percent confidence interval CI: 0.69-2.57). Removing the cases around Krümmel from these data--because the study was triggered by this cluster--results in a relative risk of 1.01. Former confirmatory results could be confirmed again. A pooled analysis of both studies based on 2,390 cases resulted in RR of 0.99 for all malignancies (CI: 0.91-1.07) and of 1.00 for acute leukemias (CI: 0.87-1.16) (children younger than 15 years of age living within a 15 km radius). We therefore conclude that the results did not show significantly increased incidence rates for any subgroup with previously significant exploratory results. Therefore, it appears to be most likely that the previous results were just due to chance. Evaluating the previously confirmatory results with the combined data from the two study periods reassures that incidence rates are not increased in children younger than 15 years who are living within a 15 km radius, either for all malignancies or for acute leukemias. We conclude that at present, in Germany no further investigations of this kind are necessary. Post-hoc subgroup analyses like the ones performed now by the authors, e.g. for pressurized water reactors and boiling water reactors, would need again a validation by independent data since these analyses were performed after inspection of the new data. Before performing our second study we had asked an expert panel for additional a-priori hypotheses which might be analysed with the new data. None of the suggested analyses included the subgroups which were now picked by the authors, although one of the authors had been asked for input to the expert panel. This illustrates the process of data dragging, which is now performed by the authors with the fully published data from our technical report and clearly shows that they cannot draw any valid conclusions from their post hoc analyses. --Prof.
Dr. Med. Jörg Michaelis References 1.
Kaatsch P, Kaletsch U, Meinert R, Michaelis J. An extended study on childhood
malignancies in the vicinity of German nuclear power plants. Cancer Causes and
Control 1998;9: 529-533. [Return to text] The authors respond: In his letter to the editor, Prof. Michaelis writes that we are not familiar with the epidemiological literature and with the design of his study. Further, he accuses us of “data dragging,” which we understand as intentionally selecting a data subset to support our hypothesis. When our analysis was initially submitted to M&GS in summer, 1998, the IMSD study had not been published in the scientific literature. We regret not having added the publication in Cancer Causes and Control to our list of references a year later. Prof. Michaelis’s comments indicate a considerable misunderstanding regarding our approach which was not to verify or challenge the results of the IMSD study, but to look for a possible increase in childhood cancers near commercial nuclear power reactors only. In contrast, the IMSD study included three small research reactors and two reactors which were operated for only a very short time. When combined with the power generating reactors (NPPs), these five sites tend to dilute any possible effect of radiation released from NPPs. Using the IMSD data for cancer rates in children (0-14 yrs.) during 1980-1995 and the statistical method applied in the IMSD study, but with a one-sided significance test, we found a significant 22% increase in childhood cancers in the zone (0-5 km) closest to the nuclear power plants. (We do not understand why the IMSD study used a one-sided test in its 1992 report and then changed to a two-sided test in the later report. This change effectively obscures a significant 49% increase in early childhood leukemia.) The data we subsequently obtained from IMSD on early childhood cancers (0-4 yrs.) revealed an even more pronounced 53% increase in cancer rates (p=0.0034). Our analysis specifically refers to a subset of all childhood cancers, of all nuclear facilities, and of all zones from 0 to 15 km. But we believe this to be a reasonable approach because biologically one expects a greater sensitivity to radiation at a younger age as well as a stronger effect in the direct vicinity of NPPs. Our results indicate that the question of childhood cancers near nuclear power plants deserves further attention. --Alfred
Körblein, |