4 children in Fukushima test positive for possible thyroid cancer

But what method was used? sensitivity and specificity? what is the normal rate of "possible signs of cancer" in children from other prefectures? If you tested 385,000 children under 18 from other places and you expect to find 2 to 5 cases this finding would not be that important (after all is the third most common solid tumor cancer type in children), if you expected to find less than one then it is quite worrying.

However, the investigative committee voiced its hesitation in linking the test results to the possible effects of the radiation in the aftermath of March 2011, TBS reported. They noted that the overall level of radiation within the region including the area where the four children were from, was far less than what had been present during the Chernobyl nuclear disaster in 1986.

Flashbacks to Minamata anyone? Downplaying and denying any direct link to the fallout from the Fukushima disaster has to be expected. If the direct link is proven not only TEPCO would be responsible, but also the governement for not evacuating the people.

Sad. May the facts get out to all of Japan. Maybe Shimizu can still be put away.

I wouldn't be surprised if this is just the tip of the iceBerg. I expect a lot more cases from that region. Whether or not it reaches the public is another story!

They noted that the overall level of radiation within the region

Its not the overall level of radiation that gives you early cancer unless levels are so high you can see people getting sick. The cancer comes from the misfortune of having ingested or inhaled radioactive particles, one of the worst lotteries you can win. Sadly, I know that there will be more winners discovered in the future.

Our good friends just lost their little girl to a very aggressive and very rare brain tumor. She was a new born at the time of the radiation leak and her and her family lived in a town fairly close to the plant. It's just a terrible shame that these little lives have to suffer so much. It's a crime that her death will never be acknowledged as being caused by the accident.

However, the investigative committee voiced its hesitation in linking the test results to the possible effects of the radiation in the aftermath of March 2011, TBS reported. They noted that the overall level of radiation within the region including the area where the four children were from, was far less than what had been present during the Chernobyl nuclear disaster in 1986.

So they're hesitant to link it to radiation while simultaneously using the horrendously high radiation levels from Chernobyl as their comparison figures... while the Japanese government continues its "Its safe for kids!!" speeches. Yeah, right.

chikvDec. 26, 2014 - 04:27PM JST But what method was used? sensitivity and specificity? what is the normal rate of "possible signs of cancer" in children from other prefectures? If you tested 385,000 children under 18 from other places and you expect to find 2 to 5 cases this finding would not be that important (after all is the third most common solid tumor cancer type in children), if you expected to find less than one then it is quite worrying.

This wasn't 4 in 385,000 children, it was 4 in 61,000 and all of those 61,000 originally tested negative.

The normal prevalence for pediatric thyroid nodules is irrelevant here because these were individuals who tested negative and then developed nodules between the initial testing (date unspecified) and this second round of testing. This makes comparisons to the general population nearly impossible because you're looking at a sample who initially tested negative and then developed nodules at an unknown rate.

The good news is that if detected early the survival rate for normal pediatric thyroid cancer is very high (upwards of 80%).

The bad news is that this article is nearly completely devoid of any useful information.

It is telling though that the scientists are now "setting the bar" at Chernobyl-level radiation, rather than trying to pretend that there was no significant radiation.

The bad news is that this article is nearly completely devoid of any useful information

http://www3.nhk.or.jp/news/html/20141225/t10014283251000.html

It states that the 4 who tested positive had very little radiation exposure. In addition, those four did not come from the areas of high radiation. Hence, the commission concluded that this recent results had very little if any, to do with the Fukushima disaster.

Anything discovered above 1 is alarming but to get a Japanese to attribute the cause to Fukushima, doubtful and the odds of winning the lottery marginally higher. There will be a lot more in the near future, some of which are not being reported to the public, it was by chance and luck this was even mentioned.

This wasn't 4 in 385,000 children, it was 4 in 61,000 and all of those 61,000 originally tested negative.

That would be data manipulation and only a very corrupt person would think of it as valid. Imagine a company test a new drug, gives it to 100 people but nobody gets cured, then they choose 10 people and give the drug again, 2 gets cured. Should they say that the drug have a 20% efficacy? This is the same case, the rate of appearance have to be reported according to the maximum number of people sampled in order to be useful, ethical and valid.

The normal prevalence for pediatric thyroid nodules is irrelevant here because these were individuals who tested negative and then developed nodules between the initial testing (date unspecified) and this second round of testing. This makes comparisons to the general population nearly impossible because you're looking at a sample who initially tested negative and then developed nodules at an unknown rate.

It does not matter because we are not interested for now about the time that took them to develop them (for that yes, you would need a cohort followed for the same amount of time repeating the same number and type of tests). In this case we are interested on incidence, and for that it is not important if you get tested 100 times in a year or just one time today. We just need the number of cases expected and the number of cases found (if the normal rate for children under 18 is 1 out of 100,000 then this finding is actually normal)

So why is thyroid cancer in Fukushima the highest in Japan? It seems to me that these researchers are pretty useless if they are not able to attribute this statistically high level of cancer to nothing!

chikvDec. 27, 2014 - 07:00AM JST It does not matter because we are not interested for now about the time that took them to develop them (for that yes, you would need a cohort followed for the same amount of time repeating the same number and type of tests). In this case we are interested on incidence, and for that it is not important if you get tested 100 times in a year or just one time today.

Who is this "we" you are referring to? Are you implying you are one of the researchers?

And if you are then you need a refresher course in research methodology and statistics. What the researchers have done in this case is known as "cherry picking", deliberately narrowing the sample to individuals who were initially healthy.

And the time scale is important because incidence of a condition can be measured many ways, either life-time, over a specific period of life (e.g. childhood incidence), or annual incidence. They're all dramatically different figures.

If you are one of the researchers then I weep for the state of academia in Japan.

Who is this "we" you are referring to? Are you implying you are one of the researchers?

Obviously "we" as the public that is interested in the results of the investigation, reading the article in the newspaper.

What the researchers have done in this case is known as "cherry picking", deliberately narrowing the sample to individuals who were initially healthy.

Please check your sources on methodology because cherry picking IS a form of data manipulation, that is why your clarification of being "4 in 61,000" is unacceptable. Also, the importance of the narrowing is not that the individuals in the later group were initially healthy (all individuals tested were healthy) the manipulation that you suggested was wrong because it ignores the 385,000 initially tested as if they were never included in the testing and only included those that for any reason were considered more susceptible of getting the disease. Saying that the real rate is "4 in 61,000" is cherry picking, a form of improper data manipulation

And the time scale is important because incidence of a condition can be measured many ways, either life-time, over a specific period of life (e.g. childhood incidence), or annual incidence. They're all dramatically different figures.

You don't need the time scale because the most important factor to analyze is a comparison of raw incidence in the age group, for now nobody really cares about annual incidence or how fast the tumors grew, the most important data to get is how many cases you would expect to find usually in this age group (the Ministry of Health, Labour and Welfare have this information) and how many cases you actually got (not only "possible" but confirmed cases). If the rate is 10 times more than normal then that is enough to sound a big alarm instead of waiting until you get a cohort tested on the same schedule as the children in Fukushima. And if the rates are similar or even lower then you can begin to design studies to see if there is any other difference, it will take years to get proper controls so there is no meaning on waiting that long to get a significative comparison.

"DIAGNOSED with POSSIBLE thyroid cancer"? Care to be any more vague? Either they have cancer or they don't. Which is it?

Humans are much more tollerant of radiation than initially belived. We survive all sorts of things.

chikvDec. 27, 2014 - 10:44PM JST

What the researchers have done in this case is known as "cherry picking", deliberately narrowing the sample to individuals who were initially healthy.

Please check your sources on methodology because cherry picking IS a form of data manipulation, that is why your clarification of being "4 in 61,000" is unacceptable.

Now this is a classic straw man attack. What I actually wrote was, "This wasn't 4 in 385,000 children, it was 4 in 61,000 and all of those 61,000 originally tested negative.".

I stated very clearly the total sample, the criteria for the sub-sample and the size of the sub-sample. Your attempt to misrepresent my statement is incorrect, as was your original statement that it was 4 in 385,000.

The reason for this is very simple. Some of those 385,000 DID test positive, if memory serves me correctly it was a figure in the low teens, but I'd need to go back and find the original paper, but for argument's sake let's say it was 16. In the subsequent tests of 61,000 who initially tested negative a further 4 developed signs of thyroid cancer. That brings the total to 20 suspected cases of thyroid cancer out of 385,000. But if we extrapolate from the 61,000 the real number of cases will probably be closer to 40 in 385,000.

Therefore your statement that it is 4 in 385,000 is incorrect, by about 10 times!

You don't need the time scale because the most important factor to analyze is a comparison of raw incidence in the age group, for now nobody really cares about annual incidence or how fast the tumors grew, the most important data to get is how many cases you would expect to find usually in this age group

Again you're mistaken. The normal incidence would represent children who have had 18 years to develop cancer. The children from Fukushima have only had about 3 years at the higher radiation levels. We can see that even after 3 years the children from Fukushima are showing higher cancer rates than average, but we can't compare that against the historical data because that would be an inaccurate comparison. It would be a low-ball figure.

Let me give you a simple example. Yumiko is a 17 year-old from Fukushima. In here life-time she has had 14 years at normal radiation and 3 years at the higher radiation levels. Comparing her against the normal incidence for 17 year olds would show a small increase in her chances of cancer, but it would be inaccurate to say, "Ah, see, the situation isn't so bad". Then we have Akari, who is a 3 year-old from Fukushima, and has had 0 years of normal radiation and 3 years at the higher radiation levels. You compare the 3-year-old's data against the national average and conclude that the risk is much higher for 3 year-olds... completely ignoring the fact that the critical factor here is NOT the national average, but rather the period of exposure and year-on-year increase.

What you're trying to do would be similar to telling someone that smoking one cigarette produces the same increased chance of cancer as smoking 3 000 cigarettes over 10 years.

The reason for this is very simple. Some of those 385,000 DID test positive

If that is the case then yes, the second figure (4 in 61,000) would be correct because that would be a different statistic, and the total would be (around) 20 in 385,000. My mistake would be because in this article there is no mention of any other positive cases, the title is not "4 more children test positive".

The normal incidence would represent children who have had 18 years to develop cancer. The children from Fukushima have only had about 3 years at the higher radiation levels.

But had 15 years to develop it at the normal rate, that would mean any increase in those 3 years would be enough to show differences. Thyroid cancer is not such a common thing that a couple of cases would mean a fraction of a percentage point, most likely a couple of cases is enough to be significant. And of course at the time of reporting the real meaning of the difference should also be explained in simple terms for those not interested in analyzing it for themselves.

Then we have Akari, who is a 3 year-old from Fukushima, and has had 0 years of normal radiation and 3 years at the higher radiation levels. You compare the 3-year-old's data against the national average and conclude that the risk is much higher for 3 year-olds... completely ignoring the fact that the critical factor here is NOT the national average, but rather the period of exposure and year-on-year increase.

But calculating the accumulated number of cases by each age group is trivial, you only need to add the accumulated incidence of cases for as many years as the age you want to analyze and you get the figure you can use to compare. For example for 5 year old children you add the national rate of incidence of 5 and younger of this year, the incidence of 4 and younger of last year, 3 and younger of the year before and so on. That way you can get an accurate number of cases expected and observed without needing any extra observation.

Eventually we will need to know if the speed of development is higher or other details that will need extra studies in children of normal background but for now it is not necessary.

What you're trying to do would be similar to telling someone that smoking one cigarette produces the same increased chance of cancer as smoking 3 000 cigarettes over 10 years.

Not at all, a more appropriate comparison would be saying that smoking even one cigarette would increase the chance of disease and that can be identified when compared with the non-smoking population (that would be the case if the smoking effects are stochastic like radiation instead of deterministic, which I am not sure)

chikvDec. 28, 2014 - 12:12PM JST If that is the case then yes, the second figure (4 in 61,000) would be correct because that would be a different statistic, and the total would be (around) 20 in 385,000. My mistake would be because in this article there is no mention of any other positive cases, the title is not "4 more children test positive".

... so your argument rests on the fact that you seriously assumed that in 385,000 children originally tested there wasn't a single case of thyroid cancer? ... words cannot express how incorrect that assumption would be.

The normal incidence would represent children who have had 18 years to develop cancer. The children from Fukushima have only had about 3 years at the higher radiation levels.

But had 15 years to develop it at the normal rate, that would mean any increase in those 3 years would be enough to show differences. Thyroid cancer is not such a common thing that a couple of cases would mean a fraction of a percentage point, most likely a couple of cases is enough to be significant. And of course at the time of reporting the real meaning of the difference should also be explained in simple terms for those not interested in analyzing it for themselves.

... no. Just no. You wanted to compare the incidence level in other prefectures (relatively normal radiation levels - although that assumption would need to be verified) against the Fukushima kids, and from that obtain ... nothing, because as I pointed out the statistics are asymmetrical over time. Thus time is a major factor... a factor that you denied being important.

Then we have Akari, who is a 3 year-old from Fukushima, and has had 0 years of normal radiation and 3 years at the higher radiation levels. You compare the 3-year-old's data against the national average and conclude that the risk is much higher for 3 year-olds... completely ignoring the fact that the critical factor here is NOT the national average, but rather the period of exposure and year-on-year increase.

But calculating the accumulated number of cases by each age group is trivial, you only need to add the accumulated incidence of cases for as many years as the age you want to analyze and you get the figure you can use to compare. For example for 5 year old children you add the national rate of incidence of 5 and younger of this year, the incidence of 4 and younger of last year, 3 and younger of the year before and so on. That way you can get an accurate number of cases expected and observed without needing any extra observation.

No, because as I pointed out the time factor is asymmetrical and your statistics will be NONSENSE if you do not account for that asymmetry.

Eventually we will need to know if the speed of development is higher or other details that will need extra studies in children of normal background but for now it is not necessary.

Accurate statistics are ALWAYS necessary. There is no excuse for bad statistics when the data is available. It can be updated as more data comes in, but one cannot publish bad data.

What you're trying to do would be similar to telling someone that smoking one cigarette produces the same increased chance of cancer as smoking 3 000 cigarettes over 10 years.

Not at all, a more appropriate comparison would be saying that smoking even one cigarette would increase the chance of disease and that can be identified when compared with the non-smoking population (that would be the case if the smoking effects are stochastic like radiation instead of deterministic, which I am not sure)

... but in your sample on your hypothetical paper entitled, "Increased health risks from smoking just one cigarette" you're including chain smokers of 10 years.... yeah, that's not going to get published even if the reviewer is your mom. Your statistics would be complete nonsense.

Just accept the fact that you are mistaken and move on. I just hope you're not involved in research in any way...

So what is normal rate for thyroid cancer? Have they tested 385,000 children in other areas of Japan? Without reference numbers, it is pure speculation to link these four cases to the nuclear accident.

so your argument rests on the fact that you seriously assumed that in 385,000 children originally tested there wasn't a single case of thyroid cancer? ... words cannot express how incorrect that assumption would be.

And why is that? the information on the article (and your first response) did not include any mention about other cases, it is believable that would be the case as long as it is the only information you got. Of course if you like to assume one result independently of evidence it is untinkable that other people wait to have the data first, quite enlightening.

no. Just no. You wanted to compare the incidence level in other prefectures (relatively normal radiation levels - although that assumption would need to be verified) against the Fukushima kids, and from that obtain ... nothing, because as I pointed out the statistics are asymmetrical over time. Thus time is a major factor... a factor that you denied being important.

The statistics are not asymmetrical, one variable is, (the exposure to radiation), that is the normal situation on every study where you compare two groups. In some the time of exposure will be important, in others (such as in the case I am making) as long as you get a difference there is value in reporting. You are really confused about what a variable is.

No, because as I pointed out the time factor is asymmetrical and your statistics will be NONSENSE if you do not account for that asymmetry.

Absolutely not, you can get a difference in incidence of a disease in two populations grouped by age, one will have exposure to radiation and the other not, honestly you have not seen ever any paper like that? I seriously worry about how you judge information without enough experience. In this case you have the extra factor of knowing exactly the time of exposure until diagnosed. Every group would be compared easily "background 18 yo incidence vs 3 years of exposure 18yo incidence"

There is no need to account for that asymmetry because it is precisely the effect of that difference what you want to measure. Eventually you will want to measure exactly how strong is that effect, but the first step is to corroborate if you have it at all.

Accurate statistics are ALWAYS necessary. There is no excuse for bad statistics when the data is available. It can be updated as more data comes in, but one cannot publish bad data.

Comparing with the background incidence would not be inaccurate, it is totally valid and a very common thing to do on biological studies, even when you don't exactly how long the affected population has been exposed or to what degree. It is good data. I worry deeply about any kind of input you may have on some unfortunate researcher on biological sciences, your lack of imagination on how to make a worthwhile comparison between groups with one single difference because of imaginary problems with the methodology would result in terrible advice.

IF you want a different comparison (between how long does it take to develop a tumor from undetectable to a certain size for example) Then yes, you need a cohort measured the same as the Fukushima children because you don't know how fast those tumors grew until you detected them, but for simple, plain incidence there is no need to wait a few years to calculate any excess on one of those groups and only then inform that yes, it appears that exposure to radiation appears to be more dangerous than not having it. Informing "we have 20 cases" is meaningless, informing "we have 20 cases when the normal situation is having 2" is much more informative, this seem like a terribly complicated problem to you so I suggest to check a text on medical statistics to see more examples on how it is valid.

but in your sample on your hypothetical paper entitled, "Increased health risks from smoking just one cigarette" you're including chain smokers of 10 years.... yeah, that's not going to get published even if the reviewer is your mom. Your statistics would be complete nonsense.

If I can demonstrate that the effect begins at one cigarette of course it will be published, If like the case on Fukushima we know exactly the degree of exposition I can prepare a beautiful graph showing a dose-dependant relationship on the effect beginning at the single cigarette I would most likely get a first tier journal . What you seem to not understand is that showing a difference between exposed and not exposed do not mean you will not be able to show levels of differences according to amount of exposure. Also that you don't need new observations to compare different levels of exposure against zero exposure because zero exposure is by definition the baseline from where the differences will appear (or fail to appear)

Just accept the fact that you are mistaken and move on. I just hope you're not involved in research in any way

My only mistake was rely on the information of the article to assume only 4 detections, on the other hand, your lack of understanding on how how to determine if a biological effect is present makes me hope that you never reviewed in such a flawed way any manuscript on biological sciences.