Learning from our Mistakes in Modeling

After the discussion of the octanol/water partition coefficient in the last blog, I thought this would be a good point to go over what was perhaps my biggest professional surprise and how we learn a lot from our mistakes.   What I am recounting below is my recollection of the facts.   To “blind” the identity of the product some of the numbers are changed but the scientific facts remain unchanged.

I was responsible for a risk assessment for a water treatment chemical that was used to treat large circulating systems within a plant.  The active ingredient in the product was highly lippophillic (had a high octanol water partitioning coefficient) and limited water solubility (about 500 ppm) (see previous blog on this subject).   It was used in the water at about 5 ppm by weight.   The vapor pressure of the active ingredient is relatively low such that the saturation or head space concentration of the pure material (see previous blog on this subject) was only about 1 mg/m³.    The exposure limit was relatively low at 0.4 mg/m³ based on its ability to cause local tissue upper respiratory irritation response seen in rat inhalation studies at relatively low concentrations in air.  

So I ran the models (UNIFAC and modified Raout’s Law):  Even with a large thermodynamic activity coefficient (see previous blog on UNIFAC) the worst case saturation concentration of this active ingredient at 5 ppm in water was a VERY small portion of the exposure limit.   I told my internal clients that I could not see any problem with inhalation exposure and risk from this active in this application. 

Imagine my surprise when we started getting reports of upper respiratory irritation in systems where older actives were being substituted with our new product.   I requested that some air monitoring be done in selected area within a specific plant.   To my complete and utter amazement, some of the values came back around 0.5 mg/m³.   This is 50% of the value for PURE active!   How could this be?   How could a 5 ppm solution in water generate this much airborne active ingredient?   I had to know so I booked a trip ASAP  to visit the site, to see things for myself and repeat the monitoring.

The highest sampled airborne concentration occurred over an open sump that contained about 5000 gals of the treated water.   As soon as I saw the sump, the whole thing started to made sense.   On top of the liquid surface was a 1-2 inch layer of FOAM.   Apparently some part of the plant process resulted in a foaming agent going into the water and the agitation within the process produced the foam.  

Because foam is typically much more "oily" than water, our active was partitioning into the foaming agent and thus into the foam.   Subsequent analysis of the foam showed that it had very high concentrations of our active ingredient.   As the form bubbles “popped” they released aerosol particles to the air that were rich in our active which further concentrated as the water and any other more volatile components evaporated from the aerosol particles because of their high surface area to volume ratio.   The entire mix of aerosol and active vapor produced the high breathing and irritating breathing zone concentrations above the sump.

If there was no foam there was no concentrating of the active ingredient into the foam and no high levels of airborne exposure.   The reality from a product stewardship perspective became clear to me; namely, that when foam was present in some systems that these scenarios where clearly capable of producing unacceptably high airborne concentrations.  

For me, the moral of this story is twofold.  First, one should always have some level of “ground truthing” for the predictions of a model.   That is, do some monitoring.   Those who model should always be open to monitoring.   They are not two separate camps.   The modeling informs the monitoring and the monitoring “ground truths” the models.   Second, we should always be ready to have Mother Nature throw us a curve ball.  We can predict a lot with models but we should always be open to potential game changers from unanticipated factors.