From Medscape Medical News
February 5, 2010 — Cell phones have become an integral part of everyday life, but concerns about their safety persist. A meta-analysis, published in the November 20 issue of the Journal of Clinical Oncology, which found evidence linking cell phone use to an increased risk for tumors has since attracted criticism.
The meta-analysis found that, overall, the use of cell phones was not significantly associated with the risk for tumors in a random-effects model meta-analysis. Compared with people who never or rarely used a cell phone, the odds ratio (OR) for the overall use of cell phones was 0.98 for malignant and benign tumors (95% confidence interval [CI], 0.89 - 1.07).
However, it also found that cell-phone use of 10 years or longer was associated with a risk for tumors in 13 studies that reported this association (OR, 1.18; 95% CI, 1.04 - 1.34).
The authors, led by Seung-Kwon Myung, MD, MS, from the National Cancer Center in Goyang, South Korea, write that the results of the studies included in their analysis varied widely, and appeared to depend on who conducted and funded the research and the controls used for bias and errors.
"We found a large discrepancy in the association between [cell] phone use and tumor risk by research group, which is confounded with the methodologic quality of the research," they noted.
However, in subsequent letters to the editor, published online January 25 in the journal, the meta-analysis was criticized for having methodologic flaws and issues and a number of limitations.
No Conclusive Evidence
A number of epidemiologic studies have reported associations between the use of cell phones and malignant or benign tumors of the brain and of the head and neck, non-Hodgkin's lymphoma, and testicular cancer. Results have been inconclusive or even contradictory. But as previously reported by Medscape Oncology, a report released in August 2009 by the International Electromagnetic Field Collaborative suggested that the regular use of cell phones can result in a "significant" risk for brain tumors.
Soon after the release of that report, a US Senate hearing on the health effects of cell-phone use was held; it concluded that more and better research is needed to determine if there is a risk to human health. Several nations have decided not to wait for additional data, have issued warnings about the use of cell phones, and advise taking precautionary measures. In the United States, the state of Maine is considering legislation that would require placing warnings on all cell phones.
Long-Term Use Associated With Benign Tumors
The meta-analysis was conducted to investigate the association between cell-phone use and the risk for malignant and benign conditions. It included 23 case–control studies, and involved 37,916 participants (12,344 patient cases and 25,572 control subjects).
They observed that the studies included in their meta-analysis varied in quality, and some that suggested there was little to no risk did not correct for bias. In 8 high-quality studies that used blinding, a significant positive association (harmful effect) was observed; in a fixed-effects analysis of 15 studies that did not use blinding, a significant negative association (protective effect) was observed.
A subgroup meta-analysis by methodologic quality showed a significant positive association in the high-quality studies (OR, 1.09; 95% CI, 1.01 - 1.18), whereas a negative association was observed in the low-quality studies. In that subgroup meta-analysis, cell-phone use of 10 years or longer was significantly positively associated with the risk for benign tumors but not for malignant tumors.
"Distinct Pattern" Seen
Another factor affecting the validity of the meta-analysis is the funding sources of each research group, "because it is possible that these may have influenced the respective study designs and results," the authors write. In subgroup meta-analyses by research group/funding source, they observed a "distinct pattern" in their findings. A positive association was seen in 7 studies, conducted by Swedish oncologist Lennart Hardell, MD, and colleagues, that were independent of industry funding. Conversely, a negative association was seen in the industry-funded INTERPHONE studies. No association was seen in studies conducted by other groups.
In an analysis of 15 studies involving brain tumors, a significant association was not observed. A preventive effect was observed for meningiomas, and the authors note that this effect was largely the result of a decreased OR in the INTERPHONE studies. A significant negative association was found when studies involving benign brain tumors were analyzed and, again, this was largely the result of a decreased OR in the INTERPHONE studies, report the authors.
They note that all of the studies conducted by Hardell et al used blinding to the status of patient cases or control subjects and were categorized as having a high methodologic quality, whereas most of the INTERPHONE studies and studies by other groups did not use blinding and were "thus categorized as having low methodologic quality."
The authors conclude that they "found a large discrepancy in the association between [cell] phone use and tumor risk by research group, which is confounded with the methodologic quality of the research," and that "our findings should be confirmed in prospective cohort studies to provide a higher level of evidence."
Critique and Response
In a letter to the editor, Andreas Stang, MD, Andrea Schmidt-Pokrzywniak, PhD, and Oliver Kuss, PhD, all from the Martin-Luther-Universität Halle-Wittenberg, Halle (Saale), Germany, expressed their concerns about some of the methodology in the meta-analysis. In particular, they point out that the study authors used the Newcastle-Ottawa Scale (NOS) to assess the quality of published case–control studies, but that "the validity of this checklist approach is at best unknown or doubtful."
Dr. Myung and colleagues, in their published response, "acknowledge that several methodologic issues raised by Stang et al are important for our main conclusions," and agree that the NOS is not an ideal quality-assessment tool for case–control studies because it has not yet been fully validated. However, they note that other checklists used to assess the quality of observational studies in systematic reviews have not been fully validated either. Of these tools, they considered the NOS to be comprehensive for case–control studies, and selected it for their quality assessment.
A second letter from Jack T. Rowley, PhD, from GSM Association in London, United Kingdom, and Michael J. Milligan, secretary general of Mobile Manufacturers Forum in Brussels, Belgium, rejected the "suggestion that there was any influence by our organizations on the scientific conduct of the INTERPHONE studies." They acknowledge that, in some cases, partial funding was provided by the GSM Association and Mobile Manufacturers Forum, but explain that funding was also received from noncommercial sources.
They also point out that a higher quality score was given to the Hardell group of studies, even though other analyses have raised methodologic questions about them.
Funding sources can influence research in subtle ways.
In response, the meta-analysis authors write that although they would prefer to believe that the design and conduct of the INTERPHONE studies were not influenced by the cell-phone industry, "nonetheless, we recommend that research on the topic of [cell]-phone use and health should not be funded by the industry because funding sources can influence research in subtle ways, and to preserve the credibility of the research it is important to avoid even the appearance of a conflict of interest."
However, they agree that the Hardell studies might have some methodologic limitations and acknowledge that they did not consider information and recall bias or validation of the self-reported indices of cell-phone use because the NOS does not include these items.
In a third letter, Florence Samkange-Zeeb, MPH, from the Universitätsmedizin der Johannes Gutenberg-Universität Mainz in Germany, and colleagues note that the authors "give no rationale" for pooling studies that cover a broad spectrum of heterogeneous and biologically diverse cancer sites, for which the localized exposure from the use of cell phones is completely different.
They also did not "understand why blinding is used as the major quality criteria," and found it surprising that there was no mention of the concerns that have been raised about the Hardell studies. As an example, they write, there have been questions about selection of the target population and the way the response rates were defined.
"In our opinion, this meta-analysis is an example of what happens when authors may have the technical skills [to perform] a meta-analysis, but are unfamiliar with the topic," they write.
The study authors responded by noting that there is "substantial precedent for pooling studies to examine the association of a potential risk factor with tumor risk in biologically diverse sites." They point out that researchers have pooled study results that reported the association between smoking and various heterogeneous diseases, including diabetes, cardiovascular disease, and cancers, to determine morbidity.
They reiterate the importance of blinding in this analysis, because "blinding and comparable response rates among case and control groups constituted the most important items in quality assessment using the NOS." These 2 items differentiated the 2 largest groups of studies in the meta-analysis, which were the Hardell and INTERPHONE studies.
"One of the problems in case–control studies is bias due to measurement error caused by the retrospective approach to measuring the predictor variable, and a recommended strategy for avoiding measurement error is blinding," they add.
Regarding the Hardell series, they agree with the critics that the differences in the methodologies, analyses, and presentation between the Hardell and other studies should have been explored. These issues and other potential confounding factors in the Hardell studies need to be explored in further studies, they write, and this issue was already discussed in their paper.
"We had no vested interest in the outcome of our meta-analysis" they point out.
The study was supported in part by the Centers for Disease Control and Prevention. The authors have disclosed no relevant financial relationships.
J Clin Oncol. 2009; 27:5565-5572. Abstract
J Clin Oncol. Published online January 25, 2010. Abstract, Abstract, Abstract, Abstract