Selection Bias Due to Differential Participation in a Case–Control Study of Mobile Phone Use and Brain Tumors

Mobile Phones and Head Tumours: A Critical Analysis of Case-Control Epidemiological Studies

“…The percentage of cases or controls exposed for at least 10 years in the above studies is 0% [16,[18][19][20], less than 5% [21,25,27,28], less than 10% [23,24,26,29], not given [30].…”

“…• the Interphone protocol considers cordless phone users as not exposed, while it is documented that the rf radiation emitted by cordless can even exceed the intensity of a cell phone [34], so much so that Hardell documents s.s. increases in the risk of meningiomas and acoustic nerve neuromas also in people using only cordless [1][2][3][4] [30], more than 90% [22,33], more than 80% [18,19,24,25,32], more than 70% [20,21,23], and the probability of this asymmetric distribution of risk values -which seems to indicate a protective effect -being chance is very low in some of these studies [21,23,26,27,30,31], while in other studies [19,22,24,25,32,33], as in the overall data, is practically zero. In Hardell's studies, over 90% of the risk values are above 1 (with 37% s.s.), and the probability of this distribution -supporting a carcinogenic effect due to MP use -being due to chance is almost zero.…”

“…The editorial of Cardis and Sadetzki takes further the remarks by Saracci and Samet, and leaves little doubt about the relevance of our criticisms, which we document on as follows ( Table 1): • of the 17 Interphone studies [16,[18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33], fewer than 10% of the exposed cases and controls had completed at least 10 years of cellphone latency or continued use, which means that over 90% had an inadequate exposure time. Since in most of the tumours 95%CI superior limit < 1 for OR<1, and 95%CI inferior limit > 1 for OR>1; 2 ORs decrease together with the increase of exposition and/or latency; n.r.…”

“…These data obviously are barely significant relative to today's intense use of cellphones, especially by those adopting them for work purposes. In Hardell's studies, MP use is reported to be more than 1000 hours for 194 cases and more than 2000 hours for 85 cases over at least 10 years, so that the average use of MPs ranges from over 16 to just more than 32 minutes per day for at least 10 years [1-4]; • in Interphone, the participation of the identified cases or controls in the study is low: 50% [24,30,31], 60% [18,20,28,32,33], 70% [16,19,21,23,25]. In Hardell's studies, participation is always much higher (88-90%) for both cases and controls [1,2]; • owing to the non-blind Interphone protocol, there is often a reduced participation in the study by the nonmobile users initially selected -in particular controls who are not affected by tumours, naturally less interested in the aims of the research than regular users, especially cases affected by tumours -and this represents a further factor of "dilution" of risk estimates.…”

“…In Hardell's studies, participation is always much higher (88-90%) for both cases and controls [1,2]; • owing to the non-blind Interphone protocol, there is often a reduced participation in the study by the nonmobile users initially selected -in particular controls who are not affected by tumours, naturally less interested in the aims of the research than regular users, especially cases affected by tumours -and this represents a further factor of "dilution" of risk estimates. This "control selection bias" is recognised by the Interphone authors themselves [30], but in their view it does not cause reduction in estimated risk by more than 10%, which is true for the overall Interphone data, but in some studies this bias alone can result in a more significant reduction in risk assessment: more than 15% [18,28], more than 25% [21][22][23], and even more than 30% [20,33]. In Hardell studies, this bias does not occur, as the percentage participation is basically equivalent for the exposed and non-exposed cases and controls [1,2].…”