Interpretation of epidemiological studies on the relationship between mobile phone use and cancer.

The latency period for malignant brain tumors averages over 20 years. Since widespread mobile phone use did not begin until the mid-1990s, virtually all patients included in case-control studies had already developed an undetected tumor before they started using mobile phones. Therefore, a tumor-initiating effect is unlikely. The authors hypothesize that mobile phone use accelerates the growth of pre-existing tumors, resulting in earlier diagnoses. There are three possible mechanisms of action on a pre-existing tumor: 1) the tumor becomes more aggressive, 2) a regressing tumor recurs, or 3) the tumor grows faster. A central element of the study is the age-incidence function. This function shows how many new cases occur per 100,000 people in each age group. If an agent accelerates tumor growth, the entire curve shifts to the left along the age axis toward younger ages. In adults, this curve typically rises with age. A condition that would not be diagnosed until age 55 in non-users may occur as early as age 52 in users. However, in childhood and adolescence, the curve declines, meaning the incidence of cancer decreases with increasing age. A condition that would not be detected until age 16 in non-users is detected at age 15 in users. Consequently, fewer cases are recorded at a fixed comparison age among users than among non-users. (In the above examples, arbitrary numbers were used; editor's note.)

The authors modeled the age-incidence function using age-dependent degeneration probability, Weibull-distributed latency, and survival probability (SEER data from 1992 to 2000). They examined three datasets: a meta-analysis of four case-control studies in adults, the MOBI-Kids study in 10- to 24-year-olds, and the Million Women Study in perimenopausal women. The extent of the shift was expressed as a proportion of the duration of use (32% corresponds to a 3.2-year shift for 10 years of use).

For adults, a 32% shift corresponds to a pooled odds ratio (OR) of 1.22. For MOBI-Kids, a 20% shift yields predominantly values below one, including a significantly reduced value in the middle age group (15 to 19 years). For the Million Women Study, whose curve initially rises and then declines, a 32% shift results in a hazard ratio of 0.98. The published value for gliomas was 0.89.

The authors conclude that the standard interpretation is flawed as soon as an agent begins to act after tumor development begins. According to the standard interpretation, a relative risk (RR) > 1 indicates risk, whereas an RR < 1 indicates no risk. All three datasets are consistent with increased tumor progression. The common objection that a sharp increase in users should have led to a visible rise in incidence does not hold up either. According to the authors' model, such an increase would fall within the normal range of annual incidence fluctuations. The authors draw specific conclusions regarding non-ionizing radiation protection. Since the tumor-promoting effect only persists while vulnerable tissue is exposed to mobile phone radiation, usage recommendations should aim to minimize the exposure duration and intensity to these tissues. They specifically identify bone marrow, meninges, and brain tissue. Consequently, the authors criticize the limitations of RF dosimetry. SAR is only relevant for vulnerable tissues, not for adjacent tissue types. Currently, SAR is averaged across all tissue types. The authors argue that research should focus more on cellular growth and intracellular signaling pathways, and less on genotoxicity.

Editor’s note:

The key finding of this publication has significant implications for interpreting epidemiological studies. If mobile phone radiation affects a pre-existing tumor, the direction of the risk measure depends solely on the slope of the age-incidence function. In other words, an increased odds ratio indicates tumor promotion when the age-incidence function increases, as it does for adults between 20 and 70 years of age. Conversely, if the function decreases, as it does for adolescents aged 15 to 19, then a decreased odds ratio suggests tumor promotion. This undermines the argument that an absence of incidence trends proves safety. At the same time, it supports the requirement that exposure parameters, particularly the SAR value, be tissue-specific rather than averaged across adjacent tissues. The widespread assumption that carcinogenic agents pose a greater threat to public health than tumor-promoting agents is a fallacy. Undetected precancerous lesions occur far more frequently than diagnosed tumors, so even a slight shift toward malignant development can strongly impact disease rates. (RH)