(P1 can be found here)
Now we can look at the LSS in greater detail.
Remember this is a largely prospective cohort study. The exposed cohort consisted of about 120,000 people and contains a large proportion of survivors who were within 2.5 km of the hypocenters, and a similar-sized sample of survivors who were between 3 and 10 km from the hypocenters and whose radiation doses were negligible. The unexposed cohort consists of non-exposed Japanese folks.
The cohort data is stratified by sex, age at exposure, attained age, city, etc.
And doses are estimated for the exposed INDIVIDUALS in the stratified data set.
The observed effect we're looking for is excess specific cancer (could be stomach, liver, lung, etc.) above that observed for the unexposed cohort.
These take time to manifest and detect epidemiologically. As time goes on, the statistics improve!
We can take a snapshot of this bulk data, but this time stratified by total dose and by effect.
And that is what is done here for "all solid cancers", where solid cancers exclude leukemia and other blood related cancers. Rather than give each individual organ dose for each possible solid cancer, a weighted colon dose is used as a surrogate and all of those organ cancers are called "all solid cancers".
The response is clearly LNT, down below (though barely) 0.1 Gy.
However, this snapshot only goes through 1998. When that table is modernized, expect better overall estimates and more excess cancers in all stratifications.
You can peruse other cancers and effects.
You know you're an LNT-denier if when looking at the "all solid cancers" table, you thought, "I wonder how low below 0.1 Gy, those cancers show up?" It doesn't matter. The curve shows that LNT is the BEST model to fit the data we have. As I've pointed out in previous posts, there will always be a gap between the graph axes and the lowest positive point.
If we were discussing dioxin or second hand smoke, no one would think twice as to whether LNT best described the data. It would be obvious. Some pro-nukes just have an emotional attachment which fogs their objectivity. Epidemiologists don't care. They treat radiation the same way they treat other toxins.
Next: BEIR VII
"The curve shows that LNT is the BEST model to fit the data we have."
ReplyDeleteEven BEIR VII says that the best fit of the data is actually linear-quadratic, not linear.
It is, in their words, computationally convenient to use a linear graph, so they introduce the DDREF. This allows them to use a different linear graph for low dose/rate from the high dose/rate graph. (And neither fits at very high doses.)
I look forward to you discussion of BEIR VII. Perhaps you can give a better explanation than the text does regarding why the linear model is best.
"Even BEIR VII says that the best fit of the data is actually linear-quadratic, not linear."
ReplyDeleteNot true. In a general sense, an LQ model includes an extra term, so it will tend to be more descriptive. In LNT, the quadratic coefficient is set to zero, so some level of descriptiveness is loss.
See Figure ES-1, where both the LNT and LQ models are sketched. There is no statistically significant difference, particularly in the low dose range, which is what the LNT-deniers are so fixated on.
They would have used a DDREF on an LQ curve as well (see Figure 10-2). It's just once it's applied , the LQ and LNT curves are about the same.
The above is not true for leukemia, so BEIR stuck with an LQ fit for it.