Interesting read.
Though I'm not a statistician or analytic methodologist, I've worked in market research for 15 years, so I think I have a good understanding of surveys and studies.
Most of the studies noted on
that page are problematic: when comparing cancer incidence in two groups (cell phone users vs non-users), other factors will probably play a major and possibly more important role. On average, a cell phone user will have a different lifestyle than a non-user. Globally, a non-user is probably more likely to live in a rural area, have a lower income, lower education, be more religious, have a less varied diet, less access to healthcare etc. To illustrate non-scientifically: consider the cell phone penetration in New York City versus rural China, and the vastly different lifestyles. Then consider the lifestyles, as a group, of average users versus average non-users.
Likewise, the reverse may also be true: using a cell phone may have effects on income, education, healthcare, diet etc. These factors will probably also be present (though maybe to a lesser extent) in national or regional studies.
The correlation between these factors and health must be accounted for. To counteract such effects, statisticians often weight ("balance") their data, making certain "scarce" participants more important. But this introduces problems of its own, and you can only go so far. It is impossible to predict, measure and account for all these factors.
This argument can be used both for and against cell phones. I don't mean to take sides, I only mean to evaluate and clarify.