April 30, 2012

by: Lorne Trottier


The following is an analysis of the paper entitled: "Mortality by neoplasia and cellular telephone base stations in the Belo Horizonte municipality, Minas Gerais state, Brazil" (*1) by Dode et al, Science of the Total Environment 409 (2011) 3469 - 3665. This paper examined mortality from "neoplasia" (which means cancer in the sense given by the authors') in the vicinity of cellular telephone base stations. The study concludes that the risk of death from cancer increases in close proximity to these base stations.

The authors of this paper are clearly biased on the issue of electromagnetic fields (EMF) and health. The first sentence of their abstract states: "Pollution caused by the electromagnetic fields (EMFs) of radio frequencies (RF) generated by the telecommunication system is one of the greatest environmental problems of the twentieth century". The authors also state that they signed the notorious alarmist "Porto Alegre Resolution". Like many other alarmist papers, the authors selectively describe only papers that support their conclusions, most of which are not accepted by mainstream scientists. They fail to list a single reference from the preponderance of scientific literature that find no harmful effects from EMF and that contradict their thesis.

This geographic epidemiological study was carried out in the municipality of Belo Horizonte in Brazil for the period from 1996 - 2006. The study was based on the following data: census and demographic statistics, base station location and date of installation, and cancer mortality data. The study used this information to tabulate mortality vs. distance from the nearest base station. The main conclusion of the study was that the relative risk of death from cancer was 1.35 or approximately 35% for people living within 100 meters of a base station compared with the overall average for the municipality.

Epidemiology and Confounders

This finding has been seized upon by a number of alarmist groups who are using it to question the safety of wireless technology. It is well known that epidemiologic studies have a number of major limitations and can be unreliable. Even in a well designed study, many other factors besides distance from a base station could be associated with higher cancer incidence. These other factors are what are known as "confounders".

The authors of the study concede that the following confounders were not taken into account: "it is not possible to consider individual characteristics, such as food and life habits, activity level, smoking, self-medication, individual pathologies, or genetic factors". The authors also state that: "Age and sex did not appear to be a confounder in this study. In Belo Horizonte municipality, like all of Brazil, the population suffers from a demographic transition characterized by the aging of the population, and this is a possible confounder for all chronic degenerative diseases". Other factors the authors failed to take into account include socio-economic status and local environmental pollutants. In fact the authors failed to take into account any of the potential confounders.

Exposure Assessment

In epidemiology, exposure assessment is critical for evaluating a risk factor. In this study, the authors failed to conduct a proper assessment of exposure to EMF. They simply assumed that exposure varies with distance to the nearest base station without any detailed verification. There are many sources of EMF in an urban environment such as cell phones, Wi-Fi networks, and broadcast TV and radio transmitters. Studies by Röösli et al. in Europe, showed that exposure from other cell phone users in the environment was comparable to that from base stations. And, according to the IARC exposure data (*2), base station EMF exposure is typically less than most other sources in the environment, and several orders of magnitude less than what cell phone users experience.

Furthermore, in most urban areas the cellular base stations are so densely packed, that the distance to any one station is meaningless. As the number of base stations in an area is increased, the average power of each station is reduced to minimize inter-zone interference. Furthermore, most base stations use directional antennas in which most of the EMF energy passes overhead. The energy from a given antenna is usually highest at a distance which is typically well beyond 100 meters. For all these reasons, the exposure does not necessarily vary in any predictable way as a function of distance from the nearest station.

Age Distribution

We will consider several alternate explanations for the results of this study. One of these is the age factor. While the authors state that "age...did not appear to be a confounder", they concede that "this is a possible confounder for all chronic degenerative diseases". Table 3 in the paper illustrates the striking differences in mortality as a function of age. For example the percentage of mortality in the 20 - 29 year ago group was 0.95% vs. 28% in the 70 -79 year age group. This is an increased risk of 30. This confirms that cancer is primarily a disease of the aged. It is possible that at least a portion of the mortality difference between the districts might be due to differences in the age distribution of the population.
Table 3 from (*1). This table shows the percentage of deaths from cancer by age group

The difference in mortality between districts is shown in Table 4. The Central-South region had an accumulated incidence of 5.83 deaths per 1000 vs. 2.05 per 1000 in the Barreiro region. This is a relative risk of 2.84. As a highly simplified example, assume that the Barreiro district consists only of people in the 20 - 29 year age group. Then the 2.84 times higher mortality in Central-South can be explained if 6.4% (
1) of the population is in the 70 -79 year age group, while the rest are 20 - 29. In epidemiology, comparisons must be age adjusted. This study did not adjust for age distribution.

Table 4 from (*1) This table shows accumulated deaths 1996 - 2006 from cancer by district

According to the authors, the Central-South district is an affluent residential neighborhood. Older people are wealthier than young people. Wealthier people live longer, so an affluent community will tend to have an older population. There could well be a high proportion of retired people living there. All these factors would tend to make the population of the Central-South district older than other districts. It is plausible that age demographics might explain at least part of the higher cancer mortality in this district.

Selection Bias

Selection bias is another possible confounder for the study results. Fig. 5 from the study is a "fluxogram" of how the cancer deaths used in the study were selected. The top of the diagram shows that the total number of cancer deaths in Belo Horizonte were 22,493 from 1996 - 2006 (11 years). The total population of the municipality was 2,238,332 as given in Table 5. This gives an overall annual cancer death rate of 91.3 (2) per 100,000. This seems low, since it is only 54% of the death rate from cancer in Canada which is167 per 100,000 (*3 P17). The authors do not discuss this apparently low overall cancer death rate.

Fig. 5 From *1 "Fluxogram" of cancer deaths from 1996 - 2006

The authors then "select" 12,094 cancer deaths out of 22,493 or 53.7% of the total for their study using the following poorly defined criteria: "Total amount of deaths select according to literature revision".
Table 1 in their paper lists the types of cancer that were retained for the study - which include all the major types. In Canada the cancer types listed account for 77.6% of cancer deaths (*3 P17). Again the authors do not discuss this apparently low percentage of deaths from all of the major forms of cancer. Some comparison with national Brazilian or international cancer statistics would have been useful as validation of the quality of these statistics.

Finally the cancer deaths "included" for further analysis in the study is further reduced to 7,191 based on the criterion of "Deaths from positive exposure from first license". In other words, cancer deaths are counted only after a base station begins to operate and in zones in close proximity to the station. A total of 4,093 (3) cancer deaths were excluded from further analysis due to these criteria. This is in addition to the 10,390 (4) deaths already excluded in the first step.

This exclusion will skew the results towards an association. For example, most of cancer deaths in the Central-South region were counted in the study because of the large number of base stations located there (Fig 8). But many cancer deaths in districts such as
Barreiro were excluded because of the relatively low number of stations (Fig 8). Table 4 lists deaths in Central South as 1459 and deaths in Barreiro as 451. The actual deaths in these districts are not given, but it is likely that they are similar.

A related factor is that since the Central South district has so many base stations, many of these will have been in operation for a longer time than in other districts. Because of this longer time period, more deaths would be counted in the Central South district.

The selection bias of counting more deaths in district the Central South district because of the greater number of stations and years of operation and not counting many deaths in Barreiro because of distance from a base station, probably accounts for most of the reported difference in the number deaths between districts
    Except where noted all images on this web site are taken from the Wikipedia commons
    All trademarks are the property of their respective owners                                                                    Copyright 2009 EMF & Health