Solvents, Ethanol, Car Crashes & Tolerance
How risky is inhalation of organic solvents?
Solvents at Low Concentration
To be useful in the context of risk assessment, relations among data sets must be quantitative, so that the likelihood of adverse effects can be estimated, along with its uncertainties. We therefore applied meta-analytical techniques, which enable quantitative comparisons across studies about relations among variables of interest. The first step in such analyses is to survey the existing peer-reviewed literature to determine the scope of existing data that have sufficient information about the exposure to enable application of a PBPK model, and sufficient detail about the measured effects to permit transforming them to a common scale.
Sufficient information was found for four solvents (toluene, trichloroethylene, perchloroethylene and 1,1,1-trichloroethane). The effects of these solvents were studied on tests including choice reaction time (CRT) in humans and rats, visual evoked potentials (VEP) in rats, behavior motivated by food reward in rats, and behavior motivated by avoidance or escape from electric shock in rats. All exposure information was converted via PBPK models to the common metric of solvent concentration in the brain at the time that the behavior or physiology was measured, and each effect was converted to a “behavioral effect” scale that ranged from 0 (no effect) to 1.0 (maximum possible effect). Each measure was then plotted against the estimated molar concentration of the solvent in the brain, and each resulting dose-effect curve was fit to a logistic function that quantified the dose-effect relation (see Figure 4, left).
Three important results emerged from this analysis. First, statistically significant differences were apparent across test methods, which aligned with the strength of the incentive involved in performing the task. Specifically, no incentives were associated with the VEPs generated in rats (red curves), because the response was driven directly by the visual inputs from the eye to the brain; humans performing CRT tests (blue curves) were instructed to respond as quickly as possible, but no consequences for slow responding were imposed during the test, and no feedback was given regarding the accuracy of the choice or speed of the response; rats working for food (green curves) received food only for correct responses, so food rewards elicited accurate responding; and rats pressing levers to avoid or escape electric shock (gray curves) were highly motivated by the punishment associated with errors. This pattern indicates that the incentives involved in a test greatly affect the sensitivity of the measure: When the cost of an error is high, the effect of the solvent is greatly attenuated compared to tests involving weak or absent incentives.
The second important result is that the four solvents did not differ significantly in either the maximum effect they produced or the amount of the solvent (in molar units) required to cause a given effect size (potency). For this reason, the data for the four solvents were combined within each type of measurement, yielding pooled dose-effect functions in the right panel of Figure 4 and their associated uncertainty bands (colored shading indicates 95 percent confidence limits around the curve).
The third important result was the lack of difference between rats and humans in their sensitivity to the effects of the solvents. The data from the human reaction time studies fell among the data for the rat studies, not on either side of them. This result gave us confidence that the animal and human experiments were directly comparable, and the rats were appropriate models for assessing effects in humans.
Despite the consistency of these results across solvents and species, and the comforting plausibility of the role of different incentives to modulate the effects of exposure, all of these effects were obtained at high concentrations that would far exceed Marcie’s exposure. How are these high-dose studies relevant to real-world exposures?
To address this question, we examined these effects in relation to the well-known effects of ethanol, which is frequently consumed for its pleasant, anxiety-reducing effects. A great deal is known about the acute effects of ethanol, which closely mirror the acute effects of solvents. One of the effects that these compounds have in common is to slow reaction time. A meta-analysis of the effects of ethanol on reaction time in humans yielded the function shown in the top panel of Figure 5, which plots the size of the effect as a function of blood ethanol concentration (BEC) and its associated confidence limits. The analogous relation for the four combined solvents in Figure 4 is plotted as a function of brain solvent concentration in the bottom, left panel of Figure 5.
Because both solvents and ethanol increase reaction time, and because these dose-effect functions are mathematically quantifiable, it is possible to define a function that describes equivalent effects of solvents and ethanol. This dose-equivalence equation is derived algebraically by setting the two dose-effect functions equal to each other and solving for ethanol dose as a function of solvent dose. The resulting function (see Figure 5, right) describes the locus of doses of both ethanol and solvents that produce effects of the same magnitude. For example, equal increases in reaction time will occur at 0.036 gram per deciliter BEC and 10 micromoles of solvent, and at 0.08 grams per deciliter BEC (a typical legal limit of intoxication) and 117 micromoles of solvent.
This relation is more than just a curiosity: It allows us to estimate quantitatively effects of solvents in terms of effects of a chemical for which a great deal of public health information is available. For example, a well-documented effect of ethanol ingestion is the propensity of automobile drivers to lose control of their vehicles. Some of these instances cause fatal crashes. The National Highway Traffic Safety Administration (NHTSA) counts these crashes, and most states in the United States measure the BEC of drivers killed in them. The incidence of single-car fatal crashes in relation to the driver’s BEC were compiled for incidents in 1986 and in 1996. These BECs were then compared to the BECs of drivers who had not crashed, but were stopped randomly at checkpoints during the NHTSA’s National Roadside Surveys of 1986 and 1996. This analysis revealed a soberingly steep relation between BEC and the risk of a fatal crash (see Figure 6). This curve shows that the relative risk of a fatal car crash increases by about 25 times at the legal limit of ethanol intoxication and by about 600 times at very high BECs.
We then fit a logistic function to this dose-effect relation (see Figure 7, left), averaging values only for drivers 21 years of age and older, and rescaling the effect as the increase in the annual number of fatalities per thousand exposed drivers. Because we know the internal doses of solvents that are equivalent to these internal doses of ethanol (from the dose-equivalence relation in Figure 5), we can now express the increase in the number of fatal car crashes as a function of the internal solvent concentration (see Figure 7, middle). This plot shows the same car crash data in relation to the concentration of a solvent in the brain, with confidence limits around the predicted values that incorporate the uncertainties in both the dose-effect function for solvents and the dose-equivalence relation between ethanol and solvents.
This relation permits us to explore the consequences of exposure to solvents at low concentrations, by focusing on the low end of the curve. The right panel in Figure 7 replots the relation between brain-solvent concentration and fatalities on a linear concentration scale to show that the function starts from zero, indicating that estimates of effects of these low concentrations do not rely on extrapolation beyond the range of the data, but are in fact constrained by measurements at both ends of the range.
What concentrations of an inhaled solvent will increase fatal car crashes by an amount of concern to public health? Of course, the level of concern is a judgment call, but it can be placed in the context of other exposures that clearly do elevate concern for public health. For example, benzene is known to cause leukemia in humans and in animals. The incidence of death from leukemia has been analyzed extensively in a cohort of about 1,700 workers who were exposed to benzene during manufacture of a synthetic packaging material between 1939 and 1960. This “Pliofilm” material was made by dissolving latex in benzene, which then was removed during production of the film, causing exposures of up to 125 parts per million (ppm) at several steps in the process. The incidence of mortality from leukemia was examined across a 45-year period after exposures stopped, to account for the typical 30-year latency period between exposure and death from the disease. These analyses yielded a cumulative incidence of 11 to 12 deaths per thousand exposed workers.
Given that death, either from leukemia or from a fatal collision, is an unacceptable consequence of chemical exposure, and that the 30-year cumulative incidence of mortality from benzene-induced leukemia is about 12 per thousand, how much acute exposure to a solvent would be necessary to increase the 30-year cumulative incidence of mortality from fatal car crashes to that level? To answer this question, we used a human PBPK model to simulate exposures to toluene that would increase its concentrations in the brain to values associated with increased death from fatal car crashes of about 0.5 per year, or 15 per 30 years.
These simulations demonstrated that it takes surprisingly little toluene to elevate the incidence of fatal car crashes to a level commensurate with the incidence of benzene-induced leukemia. Figure 8 shows the results of simulating a reasonable exposure scenario at three different concentrations of toluene. A healthy person exercising at the rate of a moderate walk (100 watts) and breathing toluene for 2 hours at 0.42 ppm would elevate his or her brain toluene concentration to about 0.45 micromoles. From the function plotted in Figure 8, this concentration corresponds to just over 0.5 annual fatalities per thousand exposed drivers, as shown on the right-hand vertical axis of Figure 8.
This concentration of toluene exceeds estimates of ambient concentrations in the U.S. (about 0.001 ppm), and of estimated personal exposure in the general public, but is lower than the EPA’s Reference Concentration (RfC) for toluene (1.06 ppm), and far lower than typical occupational standards (50 ppm averaged over an 8-hour work day). Note that the RfC (for any chemical) is defined as “an estimate (with uncertainty spanning perhaps an order of magnitude) of a daily inhalation exposure of the human population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious effects during a lifetime”( http://www.epa.gov/iris/help_ques.htm#whatiris). That is, RfCs are derived from information on the hazard of chronic exposure to a chemical and do not account for potential effects of acute exposure, which our analysis identified as another potential risk.