SM, perhaps you could explain to those of us in the class who are apparently only skimming the lecture notes, how one 'extrapolates' multiple data points that deviate from the regression, such as that seen in Fig. 4 of the paper currently under discussion?
http://www.rrjournal.org/doi/pdf/10.1667/RR2629.1

The LNT is not 'intuitively correct', in fact your analogy to car crashes shows us that a Threshold model is more 'intuitive'.

Having worked in a nuclear medicine department for 9 years, I have some insights into the real problems of low level radiation exposure.

Calcium, zinc and magnesium valence two human structural/skeletal elements can be activated by low level radiation causing structural imperfections unrelated to any hormesis effects.

This primarily affects bone and teeth with predictable weaknesses or dislocations developing over time. Teeth in particular are problematic here. Bacterial overload from accelerated decay patterns can have disaterous knock on effects in the digestive and circulatory systems with tertiary involvement of the repiratory system due to bacterial plaque build up and immune response inflammation.

No cellular genetic compensatory effect (hormesis) is able to undo such broad-scale anatomical structural changes.

This is where the rubber hits the road in terms of lowering the Low radiation dose health standards.

World radiation panels will ignore skeletal and dental aspects of low level radiation exposure at their peril.
Perhaps some of these proponents already have plaque build-up in their brain.

QED

KAEP. I gave a link to a source showing about 80,000 extra bowel cancers annually from dietary change in Japan. I missed the bit in your post where anybody actually got sick or died from these "disastrous" low level knock on perils. Do you actually have any data that is relevant?

The fact that you can detect a change due to radiation, or eating chilli, or pumping iron, or jogging, or breathing cooking fumes says a great deal about detection methods but not much about whether the change is detrimental or beneficial. For that you need epidemiology.

Bugsy,

Refer to figure 5 and table 9

Given that an annual background dose is 0.0024G and the excess cases for a dosage of less than 0.005 is 2 out of 40 000 people and nearly 5000 cancer deaths, one can conclude that for lowest dosages it makes no difference what so ever.

For dosages between 0.005G and 0.1G (40 years back ground dosage) there are 49 "excess" cancer deaths among 30 000 people from nearly 4000 cancer deaths.

on fig 4 the bottom 14 plots are based on 51 deaths from 70 000 people over nearly 50 years.

No wonder the results are considered tenuous, especially considering the cancer deaths caused by external factors such as red meat consumption can have a greater effect.

SM: Perhaps I'll repeat my question, are they 'extrapolated'? And if so, how?

Where did you find your information on the bottom 14 doses (did you mean 4)?

Table 9 says that for all the doses up to 0.2Gy, there were 97 excess cases of solid cancer and 73 excess cases of noncancer diseases that were attributable to the dose.

For these doses, the total number of deaths was actually over 9000 for solid cancer and over 37,700 for non-cancer out of over 74000 subjects. Your figures don't quite add up.

You can argue that the relative risks are minimal, but the issue I took was that you stated the conclusions were based on 'extrapolations' of higher doses, you still haven't explained how you 'extrapolate' the data points seen in Fig 4 and Fig 5, given that they deviate from the regression (the actual extrapolation).

Bugsy,

Sorry, typo, I started saying something below 1.0 and forgot to modify.

A dose of 0.1G is 40yrs background dosage and calling it a low dosage is ridiculous, let alone 0.2G. The data in fig 4 below 0.5G is just noise could be just as easily represented by a horizontal line at an ERR of 0.03. If the expected rate was out by just a few percent (i.e. with additional external factors), the threshold model would be more applicable.

To reiterate, I have never said the LNT model is not applicable, just that the data collected in this study does not statistically support it.