Shunichi Yamashita's Graph Alteration

At the Annual Meeting of the National Council on Radiation Protection & Measurements (NCRP) held in Bethesda, Maryland, on March 11, 2013, Shunichi Yamashita, the Radiation Risk Management Advisor for Fukushima Prefecture and then vice president of Fukushima Medical University, gave a keynote address.

NCRP http://www.ncrponline.org/.
Yamashita's keynote address video http://echo.colostate.edu:8080/ess/echo/presentation/d6ddb666-85bd-48a3-8d83-a691910906be
Yamashita's keynote address PowerPoint presentation PDF http://www.ncrponline.org/Annual_Mtgs/2013_Ann_Mtg/Yamashita.pdf

His PowerPoint presentation included a version of a graph from the 2005 study by Cardis et al., Risk of Thyroid Cancer After Exposure to 131I in Childhood, linked here http://jnci.oxfordjournals.org/content/97/10/724.long. Yamashita was one of the co-authors of the study, and the graph he referred to is "Fig. 2 Comparison of odds ratios (ORs) predicted by the best-fitting risk models with categorical odds ratios estimated in 11 dose categories," on page 729.

On page 727 of the Cardis study, it is stated:
"Figure 2 shows the variation in odds ratios as a function of dose level. A strong dose-response relationship was observed (P<.001); the odds ratio appeared to increase linearly with dose up to 1.5-2 Gy and then to plateau at higher doses. Statistically significant increases in risk were associated with all radiation dose categories greater than 0.2 Gy.
The statistical models that best describe these data are the linear excess relative risk model up to 1 Gy, the linear excess relative risk model up to 2 Gy, and the linear-quadratic excess relative risk model over the entire dose range. As shown in Fig. 2, however, the latter model tended to underestimate risks up to 2 Gy. "

Here's Yamashita's version, which is slide 12 of his PowerPoint presentation titled "Risk of Childhood Thyroid Cancer around Chernobyl." 

　 　　　　 

Here's the original Cardis version:

　　　　　
 
It appears that he omitted the curve 1, the ERR model - linear-quadratic dose-response model over the entire dose range, which was considered one of the best-fit models. Is it because it "tended to underestimate risks up to 2 Gy"?

The videotape of Yamashita's keynote address revealed the following statement (sic) associated with this slide:
"According to other joint project on the case control studies, it's also clearly shown the increase of thyroid cancer by in dose-responsibly by radioactive iodine.   These data also reconfirmed by United States and Belarus and United States-Ukranian cohort project recently. This is really important to understand the dose-responsiveness of how much they received thyroid dosing."

Apparently he used the slide to show the dose-response of thyroid cancer risk.

Alfred Korblein, a physicist from Germany who has analyzed post-Chernobyl and post-Fukushima birth and mortality data, plotted the data points from Yamashita's graph as well as the original Cardis version, as shown below.

   


A question remains unanswered as to the reason behind Yamashita's decision to leave out the curve 1. In addition, a question arises as to if it's appropriate for him to cite the original study, as he did in the slide, when the graph has been "altered" by him.

Is this ethical as a researcher?

A consensus by a group of international researchers was that it probably wasn't. Cardis was contacted with the above information multiple times, but she never responded.

When asked what he thought of Yamashita's action of making his own graph yet citing the Cardis study as a source, Korblein said, "I would call it deception or fraud."