Risk Assessment: Summary and Review of Chapter 32 of Environmental Health: From Global to Local. Scott Bartell (Editor: Howard Frumkin, 2005), Pages 940-960

     I have written about risk assessment previously. With this post, I want to reinforce some of that writing with a similar assessment of risk assessment (wink) from an Environmental Health textbook. Quantitative calculation of risk is a predictive approach that relies on science as much as possible but also may rely heavily on statistical techniques and probabilities.

     Here, risk assessment is defined as “the process of identifying and evaluating adverse events that could occur in defined scenarios.” Kaplan, in 1997 described it as an attempt to answer three questions: What can happen? How likely is it to happen? And what are the consequences if it happens? People who live in, work in, or visit an environment that is contaminated may be at risk for health impacts. Water, wildlife, food, soil, and air may become contaminated environments. Environmental health risk assessment can be seen as a quantitative framework for evaluating and combining evidence from toxicology, epidemiology, and other disciplines that can aid our understanding of risks. The author notes that risk assessment itself is not a science but a synthesis of scientific data that is utilized to inform policy. Thus, it is a mixture of science and judgment.

     An example used in this chapter is the ingestion of chloroform as a byproduct of water disinfection by chlorine compounds. U.S. drinking water levels of chloroform average 1 to 90 ug/L. Although chlorination of drinking water is “one of the most effective public health interventions ever conceived,” chloroform in sufficient quantities may cause or contribute to cancer. Risk assessment is primarily used to help determine acceptable limits for specific pollutants in air, water, and living beings, and also to determine what emissions levels are acceptable from industry and machines like automobiles.

     The 1983 book by the National Research Council (NRC), Risk Assessment in the Federal Government, which I utilized in my previous analysis, also known as the “red book,” explains risk assessment as a composite process of four elements: 1) hazard identification, 2) dose-response assessment, 3) exposure assessment, and 4) risk characterization. The first three inform the synthesis, which is risk characterization.

 

 

     Hazard identification often relies on evidence from toxicology and epidemiology. Hazard identification can be simple or complex. It is simpler if only one contaminant is present. It can be complex when two or more contaminants are present. He notes that early risk assessment efforts focused heavily on cancer risks. EPA keeps a registry of toxic substances and their known hazards.

     Dose-response assessment seeks to describe a quantitative relationship between exposure and diseases. This is usually a quantitative model of the toxic response from exposure to a known quantity of pollutants. Dose-response curves, or models may be generated for different species and extrapolated for other species based on a better-understood curve from another species. The author describes more about dose-response modeling later in the chapter.  

     Exposure assessment, according to the NRC “includes the estimation or measurement of the magnitude, duration, and timing of human exposures to the agent of concern.” Exposure assessments can be difficult to conduct, and routes of exposure and levels of exposure must be plausible and preferably accurately quantifiable. That is not always the case. Summary values, such as time-averaged exposure rates are often used. Default assumptions are sometimes made that may not fit the exposure at hand.

     Risk characterization involves combining the information from the previous assessments to estimate the response and the probability of response to the exposure from the hazard. Mathematically, the exposure often holds more weight than the hazard in the response. Thus, the response probability is compared to the “background response” from an unexposed person or animal. These are summarized as relative risk, additional risk, attributable risk, and excess risk.

Since risk usually involves significant unknowns, its study involves calculating statistical probabilities with potentially high margins of error. Risk characterization should involve identifying and discussing qualitative uncertainties.

     A recurring uncertainty is whether health impacts at high exposures can be used to predict the health impacts of lower exposures. This is known as low-dose extrapolation. He notes: “Risk estimation should actually be called low-dose interpolation when data from both lower and higher doses are used to fit the dose-response model.” He also notes that due to the inherent uncertainties, some of which are hard to see, quantitative uncertainty analysis has come to be utilized more than qualitative uncertainty analysis.

     Risk management involves developing strategies, communicating them to relevant parties, and making decisions on how to address the issue. NRC points out that risk assessment and risk management should be kept separate but also be mutually informative.

     De minimus risk refers to levels of risk that are acceptable or not statistically significant from a social perspective. For example, one in a million is often used as a threshold for cancer risk.

     Safety assessment involves the determination of safe exposure levels, focusing on what is safe rather than what is harmful. It used to be conducted by determining no-observed-adverse-effect-levels (NOAELs) and divided by 'uncertainty factors.' He notes that NOAELs have largely been replaced by model-based estimates where the extra risk is 1%, 5%, or 10%. Dose-response modeling and risk management decisions regarding thresholds of acceptable risk are used to arrive at these “benchmark doses.”

     The risks of any activity must be weighed against the benefits. He again gives the example of water system chlorination which has been extremely beneficial for human health even though there are some risks of cancer from chlorination byproducts like chloroform. The risks of not chlorinating are deemed to be much higher than the risks of chlorinating, which are still there. Risk-benefit analysis also seems (to me) to be a natural inclination for humans who want to protect themselves from unnecessary harm. It must be done logically, reasonably, and comprehensively in order to get the most accurate assessment.

     Cost-benefit analysis is a common feature of risk analysis. It is required that any new federal environmental law be put through some form of cost-benefit analysis. In cost-benefit analysis abatement costs are compared to a metric known as willingness to pay, referring to how much one is willing to pay for a certain level of risk. Deaths avoided may be compared to the statistical value of a human life metric, which itself varies depending on the wealth of the society. It can be messy and controversial, but it does provide a comparative metric.

     Decision Analysis, also called Alternatives Analysis, is a consideration and comparison of multiple options and scenarios. Best management practices may be arrived at by comparing methods. Decision analysis is comparing options and overlaps with cost-benefit analysis.

      The Precautionary Principle is a method that is popular in some places and for some issues. The author quips that while the Precautionary Principle is often seen as an alternative to risk assessment, it can also be seen as a risk management strategy. He mentions it being applied in cases of global warming, bioterrorism, and genetic engineering. There are many arguments that point out its downsides including that it does not give value to benefits, that it is often applied randomly pre-emptively by exploiting uncertainty, that it leads to costs and harms (my examples: banning golden rice preventing needed Vitamin A from being ditributed, banning glyphosate in France leading to lower crop yoelds, going full organic farming in Sri Lanka which resulted in severely reduced yields, Japan shutting down nuclear plants after Fukushima which made heat costs and pollution rise, and banning DDT, preventing its use in preventing malaria when applied to mosquito screens).

     Next, Bartell returns to dose-response modeling and its use in risk assessment. These models may be mechanistic, biologically based, or biologically motivated. These models often rely on toxicology data. Monotonic dose-response models are expressed via a 'tolerance distribution' as shown below. The second graph is an example of three dose-response model types utilizing the same data.

 

     Uncertainty Analysis was deemed an important method in the NRC red book. It can be qualitative or quantitative. One way it is quantitative is by giving a range or distribution of reasonable risk estimates. Interval analysis is a type of uncertainty analysis that compares best-case and worst-case scenarios and estimates the risk twice, once based on each scenario. Probabilistic risk analysis is a type of uncertainty analysis that uses probability distributions. These are statistical methods such as Monte Carlo simulations that are often used in probability analysis for environmental risk assessment. There are limitations to probability analysis and often it does not reduce uncertainty enough. Often, it can only suggest plausibility.

     Criticisms of risk assessment include the observation that it is based on science and subjective judgment, with judgment being what is criticized. It is a similar argument to one that deems that scientific experts can inform policy but should not make policy. Epidemiology studies are often criticized due to vagueness. Conservative default assumptions as the EPA has advocated for in the past in risk assessment have also been criticized. Dose-response models have been criticized when low-dose effects on humans are extrapolated from high-dose effects on rodents.

     He notes that for ethical reasons, environmental epidemiology is often observational rather than experimental. Additionally, epidemiological data may be lacking due to no adverse events having happened to have been studied, poor exposure knowledge, and other factors. I have seen epidemiological studies on possible health effects of fracking that were inconclusive, be given media headlines that suggested otherwise. Thus, there can be a media aspect to risk assessment and risk communication. It is sometimes argued that risk assessment has morphed into risk management. Again, it is not the objective aspect but the subjective aspect of risk assessment that is criticized. He notes that there will likely always be disagreement about risk assessment and it will continue to be debated as it always has been.