Arsenic in Rice: Part 11 - Dose-response is the cornerstone of toxicology

From 2010 Draft IRIS we learn:

In response to comments from NRC and SAB, a slightly different approach to estimate cancer risks for U.S. populations is being used. In the following analysis, arsenic concentrations corresponding to an additional 1% lifetime cancer incidence (effective dose; ED01 values) above “background” are derived for each endpoint [bladder/lung cancer].

A little primer on EDs...

NRC

 Using the Morales et al. data, the EPA IRIS reports:

Depending on the model used and the comparison population used in the analysis, the effective dose at the 1% level (ED01) estimates ranged from 21 to 633 ppb for male bladder cancer, and from 17 to 365 ppb for female bladder cancer. The lung cancer risk for males was found to be slightly higher than the bladder cancer risk, with ED01 estimates ranging from 10 to 364 ppb.  The risk for female cancer tended to be higher than that of males for each cancer type. For lung cancer, female ED01 estimates ranged from 8 to 396 ppb.

What this means is that the concentration required to obtain an additional 1% cancer incidence for male bladder cancer is either 21 ppb or 633 ppb, depending on what model you use.

Why this is important, this variation of 21 or 633 needed to get the same result - an additional 1% cancer incidence for male bladder cancer, shows up here (draft IRIS):

Also derived are lowest effective dose (LED01) values, which represent the lower confidence limits on the dose corresponding to a one percent lifetime incidence risk in the U.S. population.  [R]isk estimates are derived based on a linear extrapolation from the points of departure (LED01s for lung, bladder, and combined cancers) because the [Mode of Action] MOA for inorganic arsenic is unknown.

What this means, if I am understanding it correctly, is that depending on what model is used, an ED01 is produced.  This ED01 is a statistically derived number and will have a 95% Confidence Interval produced, and the lowest number of that range will produce the LED01.  That number is what is used to produce the Slope Factor which is used to produce the risk of excess cancers which is then used to determine the "safe" dose - the line in the sand - the threshold.

You can see these values, the ED01 and LED01, in Table 5-3:

Draft IRIS Page 131

I can't really do the math here any justice, but what you can see from this table is that the LED01, which is the lower confidence interval, and is used to produce the line (linear extrapolation) from which the Slope Factor is determined, which represents the theoretical potency, which determines the risk (1 in 1,000,000 or 1, 10,000) was based on the lower values spat out by the different models used to calculate the ED01.

...in the house that Jack built!

Confused?  Me too, kind of.  What this shows is that when calculating the Slope Factor the lowest concentrations of arsenic that showed came out of the models were used.  This being the case, with a range of ED01 anywhere from 21 ppb or 633 ppb, for example, means that the risk calculated from these Slope Factors are extremely high.  With that in mind, a serving of rice that has inorganic arsenic 5 ppb over the 5 ppb threshold established by New Jersey is probably no more hazardous for lung/bladder cancer than if it was 5 ppb below that value.

Because the Slope Factor is dependent on a linear extrapolation of the LED01 the numbers used to calculate the ED01 must accurately reflect an additional 1% lifetime cancer incidence above “background.”  

EPA reports in the draft IRIS that these numbers were from 21 ppb or 633 ppb for male bladder cancer depending on what model was chosen.  How far of, then, the Slope Factors are from reality (actual incidence of excess bladder/lung cancer) is anyone's guess.

What there is no guessing about is this.  Eating a serving of rice containing 9.6 μg every day - for 70 years - would not be "troubling," "worrisome," "cause for concern," or "potentially harmful."

But don't take my word for it.  Let's look at what the National Rural Water Association wrote about this draft IRIS report:

The most likely  [Mode of Action] MOA for arsenic correspond to sublinear dose-response models, which when fitted to the available data, suggest much lower cancer risks (by a factor of up to 200) at exposure levels at or below the MCL when contrasted to the EPA’s estimates from the linear model.

..and:

Since the MOA could not be established, the EPA is continuing in its application of the linearized, non-threshold model of dose-response, despite a lack of evidence of increased cancer incidence at typical environmental levels of exposure in the U.S.

...and:

The data of Morales et al. for bladder cancer in a Taiwanese population has been graphed. The arsenic concentration is in μg/L, and the y-axis is lifetime probability of bladder cancer. The solid line is the best-fitting linear model::

Which brings us back to Dr. Honeycutt with the TCEQ:

USEPA used lung and bladder mortality data from Morales et al. (2000) for the dose-response assessment for the final draft SFo.  Morales et al. (2000) uses these mortality data to calculate standardized mortality ratios (SMRs) and notes,

• “Although the computed SMRs display a large amount of noise, there appear to be higher SMRs at high exposure levels compared to exposures in the lower range, especially for bladder and lung cancer.”

Dr. Honeycutt responds:

To say that there is “noise” in the SMRs over the eight exposure categories is an understatement. 

Dose-response is the cornerstone of toxicology, but the lung and bladder mortality data (SMRs) from Morales et al. provide a poor basis for dose-response assessment as a dose-response is not apparent and not monotonic. 

Breaking the data down into the form of age-specific person-years at risk and cancer deaths does not improve the basis for dose-response assessment; it only obscures the lack of a good dose-response which is readily apparent from examination of the SMRs.

I have whipped this dead horse on the Slope Factor as much as I can.  Let's look at it another way.  Do we see the number of bladder cancers estimated by the Slope Factor derived from the Morales et al. data?

I addressed this once before in a previous post.  I'll do it again with a bit of a different spin.

Next Post: Arsenic in Rice: Part 12 - Look No Closer Than Your Own Backyard