MPs and head tumours: positive data
An overview of the most significant results obtained by the Hardell group in the three pooled analyses of their data through case-control epidemiological studies referring to tumours diagnosed during 1997-2003 is given in Table [1
] With ≥ 10-year MP use or latency, a statistically significant (s.s.) increase (ca 2- to 4-fold) in risk of overall (ipsi- plus contralateral) malign and benign brain tumours and acoustic neuromas is shown after use of analogue and digital cellulars. With cordless phone use, instead, risk is about double, s.s. only for malign brain tumours (Table ).
Results from the case-control studies by Hardell.
As Table shows, the tumour increase is chiefly localized on the habitual-use side of the head (ipsilateral tumours), and is very marked (up to 3-5 times normal incidence) and s.s. for malign brain tumours and acoustic neuromas with cellular phone use, and for astrocytomas and meningiomas with cordless use. The data for overall tumours are lower, though still considerable (up to 2-3 times normal incidence) and s.s., while the risk of contralateral tumours is not s.s., except for astrocytoma following use of cellular phones. According to Hardell, this latter finding results from the fact that the radiation produced by MPs - despite being much lower on the contralateral side - is still significant in the ventricular and subventricular space from which gliomas and (their subtype) astrocytomas originate, such that these can develop also to the contralateral side. Finally, the increase in risk of cerebral astrocytomas and acoustic neuromas, in particular ipsilateral, is higher in the subgroup that started using MPs at an age <20 years, even if the 95%CIs are very broad, owing to the still-limited number of subjects being studied (Table ).
It should be stressed that a greater increase in ipsilateral tumours than in total tumours, but absence of increase in contralateral tumours, is precisely what would be expected in the case of MPs having oncogenic action [4
A detailed analysis of the data from Hardell's seven most recent studies [7
], including the pooled analyses [1
], shows that (see additional file 1
• the percentage participation in the epidemiological study is always very high (84-91%) for both cases and controls;
• the percentages of people exposed are sizeable (mean = 60%, but - in a few studies - up to 70-80%) for both cases and controls;
• MP use is significant: 194 cases used MPs for more than 1000 hours, and 85 for more than 2000 hours, for at least 10 years (i.e. from > 16 to just > 32 min/day);
• the percentages of cases and controls exposed for at least 10 years are 18% and 13% of the total number of exposed cases and controls;
• of the total OR values reported in the above studies, over 90% are > 1, 37% of which are s.s., and the probability of this highly asymmetrical distribution of OR values being due to chance is almost zero (Figure ). This pattern indicates that the results are not due to errors or conditioning in the protocol Hardell used, since in other reports regarding other types of tumour (salivary glands [11
] and testicles [12
]) in MP users - due to the very limited number of those exposed for at least 10 years - no s.s. risk increase is found, nor is there any clear prevalence of OR values > 1. Only for non-Hodgkin lymphomas [13
] a s.s. risk increase is found and the distribution of OR values is shifted towards values > 1 (73%), with low probability of this being due to chance (Figure );
Hardell and Interphone data: percentage of the OR values > 1 or < 1, and percentage of those statistically significant.
• the increased risk in MP users is not limited to gliomas, meningiomas and acoustic neuromas, but involves also other types of head tumour -low grade and high grade astrocytomas, oligodendrogliomas, medulloblastomas, ependimomas, and other/mixed malignant tumours; pituitary adenomas and other/mixed benign brain tumours - which are considered separately [3
In the Hardell group studies [1
] a dose/response relationship and thus the existence of a cause-effect relationship are documented by the fact that:
• the risk of developing tumours is prevalent, if not exclusive, on the head side habitually exposed to MP radiation (ipsilateral);
• the trend for increase in OR as a function of time of MP use is s.s.;
• the risk is higher in rural areas [15
], where the signal required for optimal use of cellular phones is often very limited owing to the low number of base-stations, and the compensatory emission of the cell-phone battery is particularly high (up to 80 V/m or more) compared with urban areas where the signal is almost always optimal, and the battery emission is a minimum (1 V/m or even less);
• the combined use of various types of MP raises the risk of developing head tumours;
• the risk of head tumour is higher in those starting MP use when aged below 20 years [3
] (see also Table ).
The biological plausibility of the oncogenic action of the EM radiation emitted by the MPs is supported by a considerable amount of experimental data [17
]. This radiation, in fact, can produce a variety of effects able to cause or contribute to the neoplastic cell transformation:
• genetic alterations (DNA damage, chromosomal aberrations, micronuclei, sister-chromatid exchanges and gene mutations) in cells irradiated in vitro
(including germinal and cerebral cells) [20
]; in animals exposed in the laboratory [22
] or natural environment [24
], and in MP users [25
• induction of DNA reparative synthesis and alterations in transcription of DNA, activation of oncogenes and other epigenetic effects [26
• alteration of the blood-brain barrier permeability and brain neuron damage [28
• induction of heat-shock proteins and apoptosis that stress living cells [30
• reduction of melatonin synthesis and activation of Fenton's reaction which increase the concentration of free radicals and peroxides able to damage the DNA [20
• alteration of functionality, count and form of sperms in MP users whose phone stays on and in their trouser pockets during the call [33
It should be noticed that many studies on biological effects of MP use are negative, but for the major part were conditioned having been funded by the cell-phone industry [35
] (see Discussion).
MPs and head tumours: negative data
Between 2000 and 2002 three case-control studies were published - two were funded by MP companies [36
], while in one no information was given about how the study was funded [38
]. The findings indicated no increase in risk of brain or acoustic nerve tumour associated with MP use. However, there was complete absence of subjects exposed for at least 10 years, and the maximum latency period was only 4-5 years (see additional file 2
On this basis, it is small wonder that there is a complete absence of increase in brain or acoustic tumour risk; quite the contrary: most OR values (67-85% of 122 total ORs ≠ 1) were < 1, and the probability of this being chance is very low or almost zero (Figure ).
Since 2004, 17 case-control epidemiological studies have been published under the Interphone project launched by the International Agency for Research on Cancer (IARC) in 2000 [39
], and overall are considered to lack any evidence for increase in head tumours in MP users. However, examination of all the above "negative" studies shows that there are bias, confounding factors, and errors in the methodological approach and the data processing and presentation. These factors include (see additional file 3
• the low participation of cases or controls: ≤ 50% [43
], ≤ 60% [45
], ≤ 70% [39
], not even given [37
• the low percentages of exposed cases or controls: ≤ 30% [36
], ≤ 40% [47
], ≤ 50% [39
], ≤ 60% [40
], not even given [44
• the low percentages of cases or controls exposed for ≥ 10 years: 0% [36
], ≤ 5% [39
], ≤ 10% [40
], not even given [44
• the inadequate definition for "normal use of cell phones" as "at least 1 phone call per week, for at least 6 months". Therefore, if a risk exists, it is "diluted" because of the dominance, in the examined sample, of subjects exposed too little: the average daily use of MPs in subjects considered "exposed" by Interphone is just 2-5 minutes per day, very scarcely representative of the intensive use made of cellphones today;
• the failure to include cordless users who, although exposed, are included among the non-exposed. The Interphone authors justify the exclusion of cordless users through the postulation that the intensity of the EM emission of this type of MP should be irrelevant and in any case much lower than emissions from cell phones, but in fact quite the reverse is true [1
], to the extent that significant increases in the incidence of malign and benign brain tumours are found by Hardell also in those using only cordless phones (Table ).
• the relative prevalence in controls exposed over the non-exposed subjects which is due to the fact that, there being no blind protocol, the subjects interviewed knew what was the purpose of the study. Therefore, MP users willingly elect to participate in the study, aware of its goals, while non-users tend to decline. This "selection bias" is recognized by the Interphone authors themselves, but in their view it does not cause reduction in OR of more than 10% [44
], which is true for the overall Interphone data, but in some studies this bias alone can result in a more significant reduction in OR assessment: ≥ 15% [45
], ≥ 25% [39
], ≥ 30% [37
], ≥ 50% [36
] (see additional file 4
In the negative cohort studies [58
], where exposure is based simply on the fact that subjects work for an industry that produces MPs or are mobile telephony company subscribers (i.e. without the need to question participants), and where the illness/mortality incidence data estimate is based on linkage of data from people exposed with data from national tumour registers, there is an overabundance of data showing a reduced risk of those exposed - in all subjects [58
] or only in males [59
] - often s.s. and with very little probability of being chance (Figure , see additional file 5
). In these studies the s.s. reduction in risk even concerns organs that most certainly cannot be irradiated during the calls, in particular lung, stomach, liver and pancreas, and also the mortality from all causes, cardiac problems, liver cirrhosis, and car accidents. Clearly, the above "healthy worker effect" is due to systematic methodological errors and bias, e.g. the low prevalence of long exposures or latencies or the inadequate definition of cellphone use.
In the negative case-control studies [36
] the combination of all the above factors leads to strong underestimation of the risk, and together act such that the majority of OR values are < 1, often s.s. (Figure ):
• in the 17 Interphone studies, out of 1084 OR values different from 1, 76% are < 1 and only 24% are > 1:
• the prevalence of OR values < 1 is extremely unusual: = 100% [44
], ≥ 90% [41
], ≥ 80% [45
], ≥ 70% [39
], ≥ 60% [43
• the probability of this asymmetric distribution of OR being chance in 6 of these studies is low [39
], while in another 5 [41
], as in the overall data, it is practically zero;
• Lloyd-Morgan [62
] applied a probability test to a distribution identical to that above, obtained by examining a lower number of OR values from 11 of the Interphone studies 76% OR < 1 and 24% OR > 1), and found the probability of this being chance to be 6.2 × 10-20
• even more extraordinary, the OR values in 4 studies fall off with increased duration of exposure to MPs and/or latency time [36
Discarding the idea of this being due to a protective effect from head tumour risk effected by MP use (not supported by experimental data - indeed, not even the Interphone authors support it), the only explanation can be found from a strong reduction in the assessment of risk resulting from the methodological errors present in the Interphone protocol.
The Interphone researchers themselves have published various studies on the methodological bias and flaws present in their work [44
]. Most of the errors are attributed to the fact that the exposure is assessed on the basis of the data self-reported by participants in the case-control study ("recall errors"): in particular, it has been claimed that the increased risks reported in some studies (Table ) could be due to cases blaming MPs as the cause of the disease. However, recently Hardell published the results of a case-control study on mortality (not incidence) due to malignant brain tumours in subjects who had used MPs and died before the interview could be performed, and found that use of analogue or digital cell-phones gave a s.s. increased risk, highest in the > 10 year latency group (OR = 2.4; 95%CI = 1.4-4.1), increasing with cumulative number of lifetime hours of cellular use and being highest in the > 2000 h group (OR = 3.4; 95% CI = 1.6-7.1) [64
Increased OR values in the Interphone studies on relationships between MP use and head tumours.
Hardell versus Interphone
The low number of cases with ≥ 10 years latency in the above negative studies is confirmed by data given in the last Interphone Study Results update [65
• only 54% of overall cases with "regular since ever use" (≥ 1 call/week for ≥ 6 months);
• only 5% of overall cases actually exposed for ≥ 10 years;
• only 2% of overall ipsilateral actually exposed for ≥ 10 years;
• while OR < 1 predominate in data referring to "regular use" of cell-phones (85%, of which 22% s.s.), the OR distribution clearly shifts towards values > 1 for only ipsilateral tumours with ≥ 10-years of cellular use or latency (86%, of which 25% s.s.), with the percentage of s.s. OR > 1 decreasing to 12% for total tumours and falling to 0% for contralateral tumours (Figure , see additional file 6
Moreover, in some of the Interphone studies s.s. increases in risk for ipsilateral tumours are quite common in people having used MPs since or for ≥ 10 years, and - more generally - even when there is no significant evidence of risk, a clear increase in OR values is often seen considering the figure for ipsilateral rather than total tumours, while there is a net fall for just contralateral tumours (Table ). Taking into consideration the systematic underestimation of OR values in the Interphone studies, this is a clear indicator of probable carcinogenic risk.
The meta-analysis of Ahlbom [4
] includes some of the US studies [36
] and some of Hardell's earlier data (1999, 2001, 2002, not quoted in the present review) on risks of glioma, clearly lacking cases with ≥ 10-year latency time. Moreover it reports from Hardell [8
] only data selectively chosen for subjects with "ever/never use" (> 5 year latency) but not, among those with 10 years since first use, the much more significant increases of risks, although these are clearly indicated in Hardell's paper analogue, digital, and cordless phone use. This meta-analysis shows data overall lacking any indication of carcinogenic risk, but underlines the absolute incompatibility between the two data sources: 83% of Hardell's risk data (OR) are > 1, 43% of which s.s., while the Interphone data are largely < 1 (73%), 11% of which s.s. (Figure , see additional file 6
In contrast, the meta-analyses of Hardell [14
], Kundi [5
], and Khurana [67
] including the literature data on ipsilateral head tumours in people having used MPs since or for ≥10 years - and so also part of the Interphone data [40
] - show large and s.s. increases (100%) for the risk of ipsilateral astrocytomas with high level of malignancy, and sizable and s.s. increases (50-140%) for the risk of acoustic neuromas (Table ). These increases are smaller than those found by Hardell in the pooled analyses of his data alone (Table ), being "diluted" with the Interphone data corresponding to the requirements indicated above. Indeed, by separating the overall OR data of these meta-analyses according to their source [5
], only Hardell's OR data are systematically > 1 (90-100%), 50-90% of which are s.s., whereas Interphone data include 50-70% of OR < 1, a proportion of which (up to > 20%) are s.s. Moreover, when only ipsilateral data are considered [67
], even 100% of the Interphone OR are > 1, 29% of which are s.s. (Figure , see additional file 7
Results of the meta-analyses by Hardell, Kundi, and Kurana including Interphone data (≥ 10 year latency)
Data from Hardell and Interphone meta-analyses: percentage of the OR values > 1 or < 1, and percentage of those statistically significant.
Also our meta-analyses of the literature data (Figures , , ), limited to subjects with ipsilateral tumours and MP latency ≥ 10 years (see additional file 8
), show sizable and s.s. increases in risk of only ipsilateral acoustic neuromas (over 70%) and astrocytomas (almost 60%) compared to subjects not exposed to MP radiation, but it should be noted that the overall figure for these meta-analyses is strongly conditioned by the inclusion of the Interphone data. The results of our meta-analyses confirm the need to identify the head tumour localisation relative to the habitual head side of MP use, which is exposed to 97-99% of the radiation; therefore, the failure to identify the ipsilaterality of tumours adds an additional "dilution factor" to the risk evaluation.
Meta-analyses on data on gliomas after ≥ 10-year latency.
Meta-analyses on data on meningiomas after ≥ 10-year latency.
Meta-analyses on data on acoustic neuromas after ≥ 10-year latency.
Instead, the meta-analyses by Lloyd-Morgan [62
] and Kan [69
], limited to a subset of the Interphone data and without analysing tumour laterality or restricting reference to cases with MP use or latency ≥10 years, show a prevalence (75%) of OR values < 1, partly (33%) s.s. for the Interphone data, and an equal split of values < 1 and > 1 for Hardell's data, 100% of those > 1 being s.s (see additional file 7
). The same pattern is shown by Lahkola meta-analysis [70
], based on a "targeted" choice of data from the first US studies [36
] and even from a cohort study [59
], as well as from certain Interphone [41
] and Hardell studies [8
] (these latter data for the main part selectively chosen). In fact, Lahkola [70
], besides including Hardell's earlier data (1999, 2002, not quoted in the present review) clearly lacking cases with ≥10-year latency time, calculated "through the pooling of different exposures or tumor categories" moderate risks for >5-year latency of neuromas plus meningiomas and of malign intracranial tumours from Hardell [8
], whereas the original Hardell's much higher risks of meningiomas, neuromas, and malign brain tumours for >10 year latency were not included in Lahkola's meta-analysis.
The reasons underlying the discrepancy between Hardell's positive data and the negative Interphone findings are seen clearly by close examination of the latest articles from the two groups. Hardell [71
] carried out a new meta-analysis, which took into account the Interphone data as well as his own: while the data overall do not show any increase in head tumour risk in MP users, limiting the meta-analysis to just ipsilateral tumours in individuals with ≥ 10-year latency, a s.s. increase in risk is found for gliomas (OR = 1.9; 95%CI = 1.4-2.4) and for acoustic neuromas (OR = 1.6; 95%CI = 1.1-2.4). Furthermore, while in Hardell's data > 90% of OR values are > 1, for the main part (>50%) s.s., this is the case for the Interphone data only when the analysis is limited to ipsilateral tumours in individuals with ≥ 10-year latency: 90% of OR values > 1, 22% of which s.s. (Figure , see additional file 9
In the first "official" Interphone Study Group [72
], considering gliomas and meningiomas, the prevalence of OR values < 1 is notable (almost 80%), over 30% of these being s.s. (Figure ). Obviously, also this study is characterized by the usual bias and flaws:
• the low participation of cases (78% for meningiomas: range 56-92%; 64% for gliomas: range 36-92%), and especially of controls (53%: range 42-74%);
• the low median lifetime cumulative call time: 75 h for meningiomas (median: 2 h/month, i.e. 4 min/day), and 100 h for gliomas (median: 2.5 h/month, i.e. 5 min/day);
• the low percentage of cases with ≥ 10 y since the start of ipsilateral MP use: 3% of meningiomas, and 6.5% of gliomas.
However, analysis limited to subjects with "highest cumulative call times" shows a marked prevalence of OR values >1 [90% for meningiomas, and 100% for gliomas (20% s.s.)] (Figure , see additional file 9
). Moreover, given the selection bias due to the under-representation of never users among controls, an analysis was carried out with short-term users as controls. In this analysis, the OR values for glioma are almost all > 1, 30% of which s.s. with a dose-response relationship, showing that those who used MPs for ≥ 10 years are twice as likely to develop a brain tumour, especially in the ipsilateral side (OR = 1.96; 95%CI = 1.22-3.16) compared to total tumours (OR = 1.82; 95%CI = 1.15-2.89) and contralateral tumours (OR = 1.25; 95%CI = 0.64-2.42) (Table five of the Interphone text, and Table in its Appendix 2 online), just what is expected in the case of MPs having oncogenic action. This should rule out the possibility of the increase in risk of head tumours in high MP users, and indeed the increase in gliomas, acoustic neuromas and parotid gland tumours reported in some Interphone studies (Table ), being due to methodological bias and confounding factors.
The conclusive report from Interphone [72
] was accompanied by a commentary [73
] whose title is very telling - "Call me on my mobile phone ... or better not? - a look at the Interphone study results", which pointed out some of the chief bias highlighted in the present report. Our analyses strongly reduces the uncertainty of the response to the quoted question: "better not" !
And indeed, even some Interphone authors have expressed disagreement with the reassuring interpretation of the Interphone results, which essentially indicates a lack of cause for alarm [72
]. In September 2009, before the US Senate [74
], Sadetzki defended the validity of her results showing an increase in risk of parotid tumour in strong cell-phone users, particularly in rural areas [53
]. Finally, an editorial by Cardis, former Interphone coordinator, and by Sadetzki - also under a highly significant title "Indications of possible brain-tumour risk in mobilie-phone studies: should we be concerned?" [75
] - gives a careful discussion of a selection of Hardell's main papers [1
], noting that these show an increase in cerebral tumour risk in people using MPs for relatively long periods, and recognizes that the Interphone research contains a number of bias that lead to large underestimation of the risk values, among which some of those highlighted in the present report.
On the other hand, the editorial points out a number of observations supporting the risk:
• a 40% increase risk for glioma in the highest decile of cumulative call time;
• the increase of risk with time since start of use, suggesting a true effect of mobile-phone use;
• the increased risk of tumours in the temporal lobe in the highest decile of cumulative time.
The authors conclude that "the overall balance of the aboved-mentioned arguments suggests the existence of a possible association" between MP exposure and increased head tumour risk.