[helene_t]

The Dutch newsite nu.nl (which is not particularly trustworthy but also not particular biased, maybe comparable to the British "Telegraph") recently sited Eurostat for the claim that "Netherlands has the second highest number of cancer deaths in the EU". http://www.nu.nl/lif...-in-europa.html

I used to have a lot of respect for Eurostat so I thought that this was just yet another journalist not being able to read numbers. But to my dismay, Eurostat actually did write it almost the same way as nu.nl: http://ec.europa.eu/...ncer_statistics

The problem with this is that they divide the number of cancer death with the total number of deaths, coming to the technically correct but very misleading statement (nu.nl) that "more than three in ten Dutch people die from cancer". This is not a good way of measuring the risk of getting cancer in different countries since it is influenced by several irrelevant differences between countries. For example, high fertility will decrease this number because if the child and/or the mother dies during childbirth it is usually not from cancer. More generally, a fastly growing population or high death risks for non-cancer reasons will keep the number down.

Eurostat at least emphasis what numbers they are talking about ("the share of deaths from cancer in the total number of deaths exceeded 30.0 % in the Netherlands, Slovenia and Denmark"), while nu.nl just spins it to "only in Slovenia, relatively more people die from cancer".

I find this pretty shocking. The relevant numbers are available form the World Health Organization: http://apps.who.int/...in.A864?lang=en

How on Earth can Eurostat let someone, who doesn't have the slightest cue about epidemiology, write a webpage about this topic? If it was some obscure statistics about historical trends in bubble gum colours then I could understand that they would let some random intern cook up something, but this is probably one of the statistics that draw the most public interest. So they really should get it right.

[kenberg]

Since 100% of the people who die must die of something, I suppose that if the numbers are larger for cancer then they are smaller for something else. If we take these statistics at face value then lowering the rate for cancer deaths (good) could be seen as increasing the rate for heart attacks (bad). So yes, one should approach such statistics with caution. I wouldn't dismiss the results as irrelevant, but caution seems right.

But then ratios are always tricky, including WHO data where the number of cancer deaths is divided by the population. Alzheimer's mostly strikes the old, and we are finding many ways to keep people alive longer, so I imagine the percentage of people contracting Alzheimer's is going up. The percentage of 80 year olds with Alzheimer's need not be going up, rather the percentage of the population that is 80 is going up. Cancer is not quite as clearly age related, but it would not surprise me to find that the percentage of deaths at age 60 that are due to cancer is larger than the percent of deaths at age 30 that are due to cancer. These things can get a bit grizzly. Men my age, if they are diagnosed with prostrate cancer, are often advised against chemo-therapy on the grounds that usually such cancers are slow growing so the patient will probably die of something else first anyway. A friend, diagnosed some five years back when he was already 80+, is now undergoing aggressive chemo since the rest of him is quite healthy so he failed to get with the program and die, as predicted, from alternative causes.

At any rate, I trust that nobody thinks living in The Netherlands causes cancer.

Having data is nice. Data analysis should always be approached with a huge bag of salt.

[barmar]

This seems to be a common way to report statistics about death, disease, and injury. How often do we hear something being called "the leading cause of death" within some particular context?

While it's true that the reason one cause has a high percentage is that we've reduced the incidence of some other cause, does that mean we should discount the severity of the new leader? Whatever condition is currently in the lead is probably the one most deserving of effort. If we reduce it, something else will take the lead, and then we should increase our efforts on that condition, and so on.

If cancer causes a higher proportion of deaths in the Netherlands than other countries, we should try to understand why this is so. If it turns out that it's just because they have a much lower incidence of other diseases, that's good news. But if it were because the Dutch are heavy smokers, that would suggest that a campaign against smoking is called for.

Of course, there are other ways to report these statistics and compare countries. Instead of the percentage of deaths, we could look at the number of cancer deaths per capita.