Integration of toxicity indices into materials informatics

Environmental legislation has targeted the use of toxic substances for decades, but known human toxicants such as lead and mercury, as well as ecologically toxic pollutants such as copper, are still used in engineered products. Furthermore, numerous new substances, including organic chemicals, are added to industrial manufacturing every year. For many of these chemicals, there is deep uncertainty about their toxicity and environmental fate. Table I reports the magnitudes of toxic releases of selected metals and organic monomers into the environment in the United States, as reported in the EPA’s Toxics Release Inventory (TRI).¹⁴ The adverse consequences of toxic releases are widely recognized as diverse and costly. Coherent procedures that enable design engineers to reduce or eliminate toxic substances in their products will be a major advance toward sustainability.

Table I. Environmental releases (2009) and human health toxicity indices for selected chemicals.Footnote ^a, Footnote ^b

^a Acronyms: IARC, International Agency for Research on Cancer; NTP, National Toxicology Program; PEL, permissible exposure limit; REL, reference exposure level; TLV, threshold limit value; TRI, Toxics Release Inventory.

^b Cells highlighted in rose contain quantitative or qualitative information sufficient to trigger caution in using the corresponding chemicals in manufacturing.

^c High values are preferred for PEL, TLV, and REL.

^d The specific IARC group classifications are as follows: Group 1, carcinogenic to humans; Group 2A, probably carcinogenic to humans; Group 2B, possibly carcinogenic to humans; Group 3, not classifiable as to carcinogenicity to humans; and Group 4, probably not carcinogenic to humans.

In December 2006, MRS Bulletin featured a special issue on the new concept of materials informatics, defined as “the application of computational methodologies to processing and interpreting scientific and engineering data concerning materials.”Reference Rodgers and Cebon¹⁵ As envisioned at that time, materials informatics databases included primarily performance-related attributes and did not account for sustainability- or toxicity-related attributes in a substantial way. In fact, only the article by Cebon and Ashby in that issue mentioned such impacts,Reference Cebon and Ashby¹⁶ acknowledging the need for “environmental impact information,” “recycling information,” and “hazardous materials regulations.” The article quantified the first two attributes simply as “energy content in MJ/kg” and “recycling fraction” and made no attempt to incorporate toxicity metrics.

Since 2006, Ashby and colleagues have further developed their materials selection method and software databases.Reference Ashby^17–20 Although their approach provides an excellent foundation for materials selection, their databases do not yet account for material toxicity, except for the inclusion of a “yes/no” indicator in response to whether the material is approved for skin and food contact by the U.S. Food and Drug Administration.

In conventional materials selection for product development, the primary purpose is to link material and function. This is achieved by focusing on selected material attributes, including mechanical, thermal, electrical, optical, and chemical properties; processing characteristics; cost and availability; and environmental consequences of use. Ideally, the attributes should be quantifiable so that they can be used to screen and rank material alternatives. The design engineer defines material function, which ultimately determines the attributes of interest.

Once the function is defined, a list of viable materials is identified and ordered according to the materials’ combinations of attributes. The conventional method considers both screening and ranking to entail quantitative manipulation of attribute data sets. In practice, though, ranking often requires significant subjective judgment to consider tradeoffs among attributes, as it is not generally possible to identify one material that is best in all aspects.

For sustainable product design, material toxicity data must be incorporated into the screening and ranking processes. Figure 2 illustrates how such data might be integrated into the progression from chemical inventory to materials formulation and product development. In this scheme, assessment and selection of an alternative material can be triggered by information on toxicity, including risks during manufacturing (occupational exposures), product use (general population exposures), and ultimate disposal into the environment at the end of the useful product lifetime (ecotoxicological and human health impacts).

Figure 2. Proposed scheme for integrating toxicity risk assessment data into materials informatics during the manufacture of widely distributed consumer products. Red arrows signify triggers for returning to initial steps to identify less toxic alternatives based on cautionary data gathered through screening processes as product development advances. Such triggers can result from information on toxicity including risks during manufacturing (occupational exposures), product use (general population exposures), and ultimate disposal into the environment (ecotoxicological impacts).

Human health impacts and ecotoxicity considerations

In the United States, hazard data are reported in material safety data sheets (MSDSs), which focus on occupational exposure limits and physical traits such as flammability and explosiveness. The potential human health concerns identified in MSDSs provide a reasonable template, but a compilation of MSDS data across multiple material options does not currently exist. We provide in Table I such an aggregation, derived from sources independent of the material suppliers, of hazard-trait values for a select group of metals and monomers.

Based on the data in Table I, it is apparent that the three attributes considered, permissible exposure limit (PEL), threshold limit value (TLV), and reference exposure level (REL), do not rank the materials equivalently. For instance, the four metals of greatest concern when selected on the basis of PEL are beryllium, cadmium, arsenic, and silver. Although beryllium, cadmium, and arsenic are also prioritized by TLV, cobalt is now fourth on the list. The REL values prioritize silver and nickel, followed by cobalt, lead, and mercury.

These differences in priority are to be expected because these three quantitative measures of occupational hazard derive from differing objectives. Specifically, PEL defines the legal limit in the United States for exposure of an employee to a chemical substance or physical agent, typically as a time-weighted average, whereas TLV is the level to which it is believed a worker can be exposed to a chemical substance day after day for a working lifetime without adverse health effects. In comparison, REL, the concentration at or below which adverse health effects are not likely to occur, is intended to establish a reference point to gauge potential effects over a lifetime of exposure. RELs account for particularly susceptible subpopulations and are based on the most sensitive adverse health effects reported in the toxicological literature. Published REL values typically include margins of safety, such that exposure of human populations to chemical levels that exceed the REL does not necessarily mean that adverse health impacts will occur. Because PELs, TLVs, and RELs capture different aspects of the intersection of chemical toxicity and human exposure, it is deemed valuable to develop integrative approaches through which all three measures are used to generate a more robust index of risk when screening and ranking material alternatives relative to their potential for adverse human health effects.

Measures of potential human health impact have also been developed that are specific to cancer causation, or carcinogenicity, reflecting the high societal investment in curbing such diseases. In the laboratory, carcinogenicity is assessed most often through long-term rodent studies.²¹ Although toxicological tests can quantitatively measure acute toxicity, clarifying potential chronic effects requires long-term monitoring of health and, ultimately, a full pathological analysis of the animal tissues and organs.

Substantial time and resources are needed to fully characterize the carcinogenic potential of a substance, and data availability can be problematic. Furthermore, interpretation of results requires extensive training in toxicology and related subdisciplines of public health. Moreover, despite the tedious and expensive process of generating chemical toxicity profiles based on animal studies, uncertainties remain in how to establish safety thresholds for human and ecological health.Reference Krewski, Anderson, Mantus and Zeise²² It is common to summarize the results using simplified carcinogenic classification schemes, two of which are briefly described here.

The IARC, part of the World Health Organization (WHO), publishes the series Monographs on the Evaluation of Carcinogenic Risks to Humans, which has been influential in the classification of carcinogens.²³ During the classification process, interdisciplinary working groups review published toxicity studies and evaluate the weight of evidence that exposure to a given substance can potentially cause cancer in humans. On the basis of these reviews, substances are assigned to one of five groups based on carcinogenic risk.

The National Toxicology Program (NTP) of the U.S. Department of Health and Human Services produces a biennial report on carcinogens that includes newly reviewed substances as well as all those listed previously.¹¹ This scientific and public-health document identifies and discusses agents, substances, mixtures, or exposure circumstances that could pose a carcinogenic hazard to populations. The NTP classifies carcinogens into only two groups: those that are known to be human carcinogens and those that are reasonably anticipated to be human carcinogens.

These two classification schemes can be used to screen and rank material alternatives, as shown in Table I. They are consistent for known carcinogens but exhibit some discrepancies for anticipated carcinogens. Thus, the recommendation is to use multiple data sets for a more robust analysis.

Beyond human health concerns, potential impacts to the environment must also be addressed in materials selection. Table II provides information on two ecological toxicity indices for the materials considered previously. The ecotoxicological indices are the EC₅₀ value, which is the effective concentration of a substance in water that will cause irreversible damage to one-half of a population, and the fish bioconcentration factor (BCF), which is the ratio by which the concentration of a substance in fish is enhanced relative to the surrounding water and thus a measure of persistence and bioaccumulation potential.

Table II. Ecological toxicity indices for selected chemicals.Footnote ^a, Footnote ^b

^a Definitions: EC₅₀, effective concentration of a substance in water that will cause irreversible damage in one-half of a population; BCF, bioconcentration factor by which the concentration of a substance in fish exceeds its concentration in the surrounding water. High values are preferred for EC₅₀, and low values are preferred for BCF.

^b Cells highlighted in rose contain values sufficient to trigger caution in using the corresponding chemicals in manufacturing.

On the basis of EC₅₀ values, silver, mercury, and copper are of particular concern, whereas the fish BCFs for thallium (100,000), selenium (50,000), mercury (5000), and zinc (2000) are disturbingly high. In general, BCFs greater than 1000 are considered high, and values less than 250 represent low to negligible risk, with intermediate values being considered moderately risky. The bioaccumulation of metals in fish tissue can produce cascading impacts up the food chain, ultimately affecting wildlife and human health, as demonstrated for mercury in edible fish.²⁸ Again, the recommended approach is to use multiple indices to obtain a more complete representation of material hazards.

A new classification scheme, developed by the United Nations and called the Globally Harmonized System of Classification and Labelling of Chemicals (GHS), is intended to eliminate discrepancies resulting from nation-to-nation variability and thereby replace classification schemes such as IARC and NTP. With GHS, it will ideally be possible to systematically classify chemicals on the basis of multiple attributes such as carcinogenicity, as well as acute and chronic ecotoxicity.²⁹ Because of its novelty, however, few substances have yet been fully classified within the GHS system.³⁰

Integrative approaches

In addition to hazard-trait values and classification schemes, such as those described in the preceding section, a variety of tools have been developed that integrate several metrics to generate numerical indicators of human health and ecological toxicity. These methods include

• • TRACI, Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts (EPA);Reference Bare, Norris, Pennington and McKone^31, Reference Bare³²

• • USEtox^TM, the UNEP–SETAC consensus toxicity mode1;Reference Rosenbaum, Bachmann, Gold, Huijbregts, Jolliet, Juraske, Koehler, Larsen, MacLeod, Margni, McKone, Payet, Schuhmacher, van de Meent and Hauschild^33, Reference Huijbregts, Hauschild, Jolliet, Margni, McKone and van de Meent³⁴

• • CHEMS-1, Chemical Hazard Evaluation for Management Strategies (EPA);Reference Swanson, Davis, Kincaid, Schultz, Bartmess, Jones and George³⁵ and

• • TPI, Toxic Potential Indicator (developed by the Fraunhofer Institut für Zuverlässigkeit und Mikrointegration, Hannover, Germany).Reference Nissen, Griese, Middendorf, Müller, Pötter and Reichl³⁶

A comparison of the main features of these methods is presented in Table III. Unlike the binned cancer classification schemes described in the preceding section, these four methods provide quantitative values unique to each substance (Table IV). To highlight the value of such methods, we describe and compare TRACI and TPI. TRACI represents a set of life-cycle impact-assessment modules that generate three characterization factors [cancer potential, noncancer potential (other human diseases), and ecotoxicity] in each of four environmental media (air, water, ground-surface soil, and root-zone soil). Six of these 12 factors are listed in Table IV for selected TRI metals.³⁷

Table III. Comparison of measures for selected alternative assessment tools.Footnote ^a

^a Models: TRACI, Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts; USEtox, UNEP–SETAC consensus toxicity mode1; CHEMS-1, Chemical Hazard Evaluation for Management Strategies; and TPI, Toxic Potential Indicator.

Table IV. Selected TRACI characterization factors and TPI scores for TRI metals.Footnote ^a, Footnote ^b

^a Acronyms: TPI, Toxic Potential Indicator; TRACI, Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts; TRI, Toxics Release Inventory.

^b Table cells highlighted in rose contain values sufficiently high, within the relevant risk category, to trigger caution in using chemicals for manufacturing consumer products. In general, low values are preferred for all characteristics reported in the table.

To produce quantitative measures of cancer and noncancer risks associated with specific chemicals, TRACI relies on a closed-system, steady-state chemical fate and exposure assessment model called CalTOX, developed by the Exposure and Risk Analysis Group at Lawrence Berkeley National Laboratory. CalTOX combines the toxicity of a chemical with the amount released into the environment over a given period of time to calculate its human toxicity potential, which is expressed relative to baseline values for benzene (carcinogens) or toluene (noncarcinogens), two chemicals for which extensive toxicological data linking exposure to disease in a linear dose–response model exist. Thus, the output data are toxicity equivalencies that support comparisons of the impacts of different chemical emissions into a given environment.Reference Hertwich, Mateles, Pease and McKone³⁸ Similarly, the ecotoxicity potential outputs from TRACI represent toxicity-weighted chemical emission concentrations that allow for the comparison of the ecological impacts of a new chemical to those of a relatively well-characterized chemical, namely, the herbicide 2-4-dichlorophenoxyacetic acid (2,4-D).

In comparison, TPI requires the user to provide selected hazard data (i.e., for human health impact, environmental impact, and physical safety concerns) from which it derives a single indicator by integrating metrics for human health, ecological health, and physical/safety traits.³⁹ To generate TPI scores such as those reported in Table IV, one must input data from original, authoritative, and valid sources.

All available alternative assessment techniques encounter methodological limitations, including complexity, lack of transparency, and limited data sets even for materials used frequently in engineered products (metals, polymers). Despite these limitations, these methods can be used to incorporate material toxicity into the screening and ranking of materials and thereby enable sustainable product design.

In one example, Lam and colleagues used TRACI to evaluate toxic releases from printed circuit board manufacturing facilities as a strategy to avoid pollution.Reference Lam, Lim and Schoenung⁴⁰ Although the environmental releases of toxic chemicals such as methanol, glycol ethers, and dimethylformamide dominated reported air and water emissions, toxicity potentials and risk-screening scores showed that lead and copper actually generate the highest environmental impacts. Therefore, effective strategies for pollution prevention include recovering copper from spent etchant and employing lead-free substitutes for solder and other components.

Materials substitutions to reduce toxic impacts have also been studied for specific electronic products. For example, Lim and SchoenungReference Lim and Schoenung⁴¹ evaluated the end-of-life toxicity potential from the heavy-metal content in flat-panel display devices that are replacing conventional cathode-ray-tube (CRT) devices. That study focused on plasma and liquid-crystal-display (LCD) televisions, LCD computer monitors, and laptop computers. The human health and ecotoxicity potentials were evaluated by combining data on the respective heavy-metal contents and the characteristic factors in TRACI. The results showed that the leading contributors to the toxicity potentials are lead, arsenic, copper, and mercury. Although the heavy-metal content in newer flat-panel display devices produces lower human health toxicity potentials than that in CRTs, the newer devices are worse in terms of ecological toxicity impacts because of the mercury in the fluorescent backlights in LCD televisions and the copper in plasma televisions.

Similar integrative approaches have been used to evaluate cellular phones and light-emitting diodes to pinpoint specific material constituents that exceed hazard thresholds specified in risk-based policy regulations. These materials could be targeted by product designers and manufacturers for substitutions to make safer, environmentally sustainable products.Reference Lincoln, Ogunseitan, Saphores and Shapiro^42–Reference Lincoln, Ogunseitan, Saphores and Shapiro⁴⁴ In concert, these studies demonstrate the usefulness of integrative methods such as TRACI and TPI that generate simple quantitative values combining both human health and ecotoxicological impact rankings. However, much research remains to be done to make such models consistent in terms of the quality and format of raw data inputs and the translation of the output into guidance for the selection of sustainable alternative materials for designers and manufacturers of consumer products.