Sunday, September 25, 2011

Laundered Shop Towels: 3 - Is it mean to ask for a median?

There are five rules that I would like to put forth regarding any scientific work used to describe something as being bad or good.
  • Rule Number 1: Always read the report the findings/recommendations were based on.
  • Rule Number 2: Always look at the data used.
  • Rule Number 3: Always check the assumptions used to derive the model that drives the conclusion.
  • Rule Number 4: Compare apples with apples and oranges with oranges.
  • Rule Number 5: Always make sure the model and equations reflect reality.
This includes spending time to scrutinize what I am using to defend my position as well.  Go to the links, read the references I am using and quoting.  Do this not just to keep me honest, but to point out any mistakes I may have made.

I'm only human, and so are the three authors of the 2003 and 2011 Gradient Laundered Shop Towel Study.

Because we are human we make mistakes.  We blunder, we miscalculate, we make poorly reasoned assumptions, we put forth facts that are not properly vetted, we think we know what we're talking about when we really don't, and sometimes we purposely mislead for reasons unrelated to the pursuit of knowledge.

In any case, what we generate and hold out as "Scientific" or "Peer Reviewed" may be used to further a particular way of thinking or an agenda.  It may even be used to help sell disposable shop towels.

Kimberly-Clark's dire warning to workers is based on an "exceedance ratios" which was calculated by Gradient:
"Concentrations of metals in laundered shop wipes can result in exposures (as evaluated using the methodology presented in this report) which exceed toxicity criteria for certain metals."
 "Why take the risk" Kimberly-Clark warns workers who are are currently using laundered shop towels.

Is there a risk?  It all comes down to how exposure risk via a laundered shop towel was calculated - the methodology used to derive the black and red bars shown in Kimberly-Clark's graph:

Source
This graphically illustrates to a reader that the "exceedance ratio" for certain metals is huge, therefore the risk to the worker for adverse health effects, specifically cancer, is equally huge as well.  And in case the graph does not convey that message...:

Source

All of this concern (Google "Laundered Shop Towels") is based on a study performed by a "renowned" company with "expertise" in such studies.  But despite what the Supreme Court may have said, Gradient is not a person who is human.  Gradient did not write this paper and make these assumptions, three humans did.

The contention that a worker should be concerned, that a worker should be presented with a question of why take the risk?, is resting solely on the work performed by three educated and experienced researchers.

If their study's conclusions are valid, it will stand up to scrutiny.  Its assumptions will be judged sound, and its conclusion deemed reasonable.  That's how it works when you put a paper out for publication, even if it will only appear in a trade journal.  And it particularly must stand up to scrutiny when you present it to an association of Environmental Health & Safety Professionals who are in the business of protecting employee health. (1)  You had better be right on this if I am going to change from laundered shop towels to disposable ones.

The reason I spend my free time writing this blog on subjects such as this, is to help me understand what is going on regarding a particular topic.  I am also an EHS educator, a career I am passionate about.  I understand stuff because someone took the time to explain it to me along the way.  So this is payback.

If you Google "Laundered Shop Towels" you will find the first page of search results trumpeting the same message Gradient's study concluded.  Laundered shop towels are deemed bad because these three researchers wrote a paper where they show via a model and calculation where a worker could consume more than 3,600 times the Proposition 65 safe intake level for lead.  One after another, these web sites parrot the same message about "toxicity" and "risk."

And they do so because they have not read the study.  They don't look at the data used, they don't question the assumptions, they can't understand what values can be compared with one another, and they don't stop to ask the fundamental question: does the model and calculation reflect reality?  They assume it is true because Gradient put it out there and these three highly educated and experienced researchers wrote it.  End of story.

In my last post, I showed how Gradient's "Towel to Hand" transfer rate was derived, and how there is no data showing if heavy metals remaining on a laundered shop towel can - or do - transfer to the hand.  That assumption that they do transfer is the foundation behind all the "exceedance values" they calculated.  That assumption is based on other studies that looked at dislodgeable dust and pesticides.  Will a laundered shop towel transfer heavy metals to the hand like pesticide/dust will be transferred to the hand from carpet?

That is an important bit of missing information.  So we are asked to assume the rags behave similar to carpet and the rate of transfer is 13%.  Gradient has nothing to back that up, it's just an assumption - one of many - based on other studies they looked at.

So for the sake of moving forward, and in the absence of any other data, lets assume the Tt/h is 13%.

Gradient claims that the worker's hands will contact 75% of the towel's surface area and the "Load (mg/cm2)" on that 75% of the surface will be transferred onto the hand at an "efficiency" rate of 13% (Tt/h ratio).  If you look at the calculation Gradient uses:

Gradient 2011 Paper

...the amount of heavy metals on 75% of the towel's surface that contacts the hand is reduced by 13%.  Since we are going to assume that the lead on the towel can dislodge onto the hand, we will also assume that up to 13% of the lead on the towel's surface the worker's hand comes in contact with, will now be transferred onto the hand.

The black and red bar graph at the beginning of this post shows the "exceedance ratios" Gradient determined using the calculation shown above.  That calculation spits out an intake based on variables that are derived from assumptions and hard data they use.

The "Intake" Gradient calculated for a worker is based on a contaminant "Load" per towel.  In other words, how much of these heavy metals are on the towel before the employee uses it?  Their model is based on the assumption that the towel is contaminated with heavy metals and that these metals will be transferred to the employee's hand - and then into the mouth (intake).

So the first critical step in this model's calculation is to find out how much heavy metals are available for transfer.  Let's look at the data and the decisions that were made regarding that data.

In the black and red bar graph, it graphically shows the "maximum" risk and "mean" risk that was calculated when compared to "various health-based criteria."

The black bar is based on the "mean" and the red bar is based on the "maximum" concentration of heavy metals detected in/on the rag.  This is important because those bar graphs show how far above an acceptable intake the worker could be exposed to.  That number drives the warning: "why take the risk."

That intake value is dependent on the mean and maximum concentrations Gradient found to be in/on the laundered towel.  Based on these lab reports, the mean and maximum heavy metals in/on a laundered towel are assumed to be available to be transferred (Load) to the employee's hand (Tt/h) and then transferred into the employees mouth (intake).  You need to have intake to have a risk, you need to have transfer to get it into the mouth, and you need to have exposure to a certain amount of a contaminant to determine the degree of risk.  So the mean and maximum heavy metal values Gradient calculated are critical.

Let's look at the mean contaminant concentrations Gradient reported. (Rule Number 2: Always look at the data used).  This table is copied directly from their study.

Gradient 2011 Paper
You see those red arrows I put there?  Those are there to show you which contaminant results had Standard Deviations higher than their mean.  This shows that there is high variability in the data, which impacts the true average - mean - concentration that actually is present.

The concentrations Gradient used to derive the Load are skewed, which means they are not symmetric, Therefore the "mean" concentration was not appropriate to calculate the Load because the data does not behave in a Gaussian (bell curve - normal distribution) fashion.  The mean they calculated has been influenced by a few very high concentrations and is much higher than what would be normally found if you were to sample thousands of towels.

Source

Here is what the EPA has to say about skewed data:

EPA
At this point I am skeptical that the mean values Gradient calculated and reported accurately represent the average concentration of heavy metals that would be found.  I am pretty sure they are too high since the lower end was bottomed out at 1/2 the detection limit.  Look at the range they report for lead.  The 2nd to last column is the US detected range and the last column is Canada.

2011 Study

There are statistical accepted methods to work around this problem of having "hot spot" data influence the average concentration.   This is where a Statistician comes in handy.  To obtain a proper and statistically sound average concentration the median should have been used or the data transformed (e.g., lognormal).  So right there, the first real number plugged in to Gradient's equation - "Load" - is biased higher than what would be found under normal conditions.  How much higher?  That would require the individual data points for each of the 26 different towel's sampled.  I've requested this from Gradient and have not received it as of the date of this post.

The decision to use values that are biased taints the actual results one is looking to find.. If the Tt/h ratio is higher than it most likely is, and the Load is higher than it most likely is, well you can see the problem.  We keep compounding the errors which drives the intake upwards.

But those black bars pale in comparison to how the red bar maximum "exceedance ratios" were calculated.


Next Post: Laundered Shop Towels: The red bar and the "three sigma" rule

No comments:

Post a Comment