Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
5 LABORATORY METHODS FOR EVALUATION OF TOXIC POTENCY OF SMOKE Assessment of fire hazard requires that toxicity data be available for incorporation into a hazard assessment model. Therefore, methods for evaluation of toxicity are presented in some detail in this chapter. The most common end point used to assess the toxicity of inhaled combus- tion products in animals is death, specifically the LC50, but an equally important consideration in a real fire is the propensity of the smoke to impede or prevent escape before lethal conditions are reached. Toxic events are not measured or detected in LC50 studies; studies that take such events into account could be useful for determining the potential human health consequences of nonfatal exposures and provide a more convincing means of predicting ability to escape from fires. Although a goal of laboratory animal toxicologic studies of inhaled combustion products is to estimate the toxicity of these materials in man, extrapolations to man are usually only qualitative. The first section of this chapter addresses the use of combustion-product toxicity tests for screening purposes. After contrasting chemical and biologic analyses, we discuss animal test methods that use death as an end point and then methods that use nonlethal end points, including factors that can impede escape from fires and nonlethal pulmonary effects that can be extrapolated to human exposures. USE OF COMBUSTION-PRODUCT TOXICITY TESTS: TO SCREEN OR NOT TO SCREEN A screening test should be simple, inexpensive, and valid. One assumes that the discriminations provided by 78
79 test data can ensure some desired degree of safety. E.g., in the case of regulation, the judgment might be a pass/ fail discrimination for approval of product application; in the case of manufacturing, the judgment might be to produce or not to produce an item. The Committee takes issue with the assumption that any regulatory pass/fail judgment can be made on the basis of toxicity screening test data alone. The more complex evaluation of fire properties (fire hazard analysis) remains, in our opinion, a requirement for judgments of suitability of products for specific uses. If products for the same intended use nave Been shown to be very similar in composition and other fire properties, a pass/fail decision that depends on a toxicity test could be justified. Although this appears to be a screening test, it is in fact simply the final point of discrimination in a less formal hazard analysis. For uses with no regulatory component (e.g., a manufacturer's surveillance of products under develop- ment), any chosen test can be used for screening, with specific performance criteria set by the user. CHEMICAL ANALYSIS VS. BIOLOGIC ASSAY Awareness of the possibility of unknown fire hazards came with the Cleveland Clinic fire in 1929, in which the newly developed nitrocellulose film played a major role. For the first recorded time, a major fire produced con- siderable quantities of toxic combustion products in addition to the ubiquitous oxides of carbon. The incident presaged a concern that would become acute with the avalanche of new synthetic materials in the 1950s and 1960s: how to assess the risks associated with combustion products of new materials. Modern analytic chemistry soon made it clear that many pyrolysis products could be generated by materials of relatively simple composition. ~ ~ ~~ ^ · . . _ Boettner and Weiss Identified over bu compounds produced by the pyrolysis of a sample of poly(vinyl chloride), and different species of wood produce different combustion products.~69 Such complexity showed that the toxicity of smoke could not be accurately predicted simply from chemical analysis of the original material. 6 2 - in an actual fire, many different materials are involved, some of which have combustion products not readily predicted by classical chemistry and perhaps not detectable by current techniques. Some of
80 these products could prove highly toxic, and their toxicity could go undetected unless biologic tests were used. One such product was reported in 1975 by Petajan and co-workers at the University of Utah.~77 Eventually identified as a bicyclic organic phosphate, it arose during the combustion of a urethane foam from the reaction of a phosphorus-containing fire retardant with trimethyl- olpropane, one of the components of the urethane formulation. The product was first detected by its acute effects on laboratory animals, which included grand mal seizures; it also produced observable psychomotor effects on several of the human investigators. This incident is often cited as evidence of chemical tests' potential failure to detect unanticipated, and in this case unusually toxic, combus- tion products. The preference for biologic approaches to screening for combustion-product toxicity does not mean that chemical approaches have been ignored. Recent advances in analytic techniques have led to readily available systems that can separate smoke components and identify many of them. 6} A combination of gas chromatography and mass spectrometry, commonly called GC/MS, is a powerful tool for analyzing smoke components. Such gases as carbon monoxide (CO), carbon dioxide (CO2), and oxygen (O2) can also be detected by optical or magnetic methods. The analysis of smoke, however, is much more than a simple problem in gas analysis. Condensation, selective absorption, and stratification make reliable sampling difficult. Smoke is a mixture of gases, solid particles, and liquid droplets, and a careful analysis for even the most pedestrian set of known toxicants requires that all three phases of smoke be checked. This is not a trivial job and generally requires considerable treatment of smoke components before analysis. Spurgeon209 has analyzed the pyrolysis products of 75 aircraft cabin interior materials for nine gases and attempted to correlate incapacitation times, as measured in the Federal Aviation Administration Civil Aeromedical Institute (FAA/CAMI) test method, 6 0 with the chemical profile of these gases. He reported that the observed incapacitation time in rats can be predicted as a linear function of the concentrations of selected gases in the combustion atmosphere. It remains to be demonstrated
81 whether this method has a more generally applicable predictive value. Purser and Woolley,~8 in the United Kingdom, exposed monkeys to sublethal concentrations of individual gases-- hydrogen cyanide (HCN) and CO--and Purser and Grimshaw~86 exposed them to combustion products of wood, poly- propylene, polyurethane, polyacrylonitrile, polystyrene, and nylon. They reported that the test subjects displayed signs of intoxication typical of only one or another of the major gases tested, depending on the conditions of - the burning, and suggested that animal models can be replaced by chemical analysis. However, chemical test methods pose substantial prob- lems. For example, the sensitivity of a test is limited by the detection limits of the analytic instrument. Chemical tests often do not discriminate among the physical forms of a toxicant, such as its adsorption on particles and its occurrence as an aerosol, even though toxicity might depend on form. Correlations between human health response and combustion-product mixture dosage are poorly understood, so the health consequences of exposure to a fire gas at a given concentration over a given duration are poorly predicted. Finally, the human- effects data that have been cataloged are relevant almost exclusively to single chemical exposures, and little guidance is available for prediction of the effects of exposures to mixtures of chemicals. Biologic test methods expose living systems to chew - cals or mixtures of chemicals, to induce changes in the performance of those systems. The results can be useful even if the components of the test material are not identified, but the biologic changes chosen for observa- tion must be measurable and preferably are easy to extrapolate and interpret. The most commonly used end point in bioassays of smoke potency has been death, usually expressed as the concen- tration that causes death of 50% of the exposed animals (LC50) in a specified period. With a bioassay for smoke potency, it is possible to describe a set of conditions under which the predetermined end point is known to occur. The bioassay, then, is the obverse of the chemical analysis: the effect can be defined even if the cause remains unknown.
82 In summary, the advantages of chemical tests are that many are quick to perform and that they avoid the use of test animals. The main advantage of biologic tests is that they produce data of high validity. The major potential danger of a chemical test is that it could miss unanticipated, and perhaps unusually toxic, combustion products, whereas in principle there is less danger of missing a biologically relevant response in a biologic test. It is important to note, however, that the actual occurrences of toxic combustion products whose formation was not predicted by chemistry CO2, CO, HCN, and 02) are rare. the application of data derived solely from chemical tests to fire hazard models is that most current models are designed to accommodate toxicity data in units of concen- tration. The use of chemical data alone in a fire hazard model would require development and verification of a scheme to summarize and add the various measured concen- (e.g., by measurements of 77 One other problem with t rations in a useful way. Many of these concerns (e.g., over the use of analytic methods as an alternative to animal testing) have been addressed by other groups, such as the European Chemical Industry Ecology and Toxicology Centre7t and the International Standards Organization (B. Levin, personal communication). A toxicity-testing strategy that avoids the uncertain- ties of chemical analysis while exploiting its advantages could follow these steps: · Chemically analyze the test material's smoke for expected major toxicants, such as CO, HCN, and HC1. . Calculate an "expected" LCso for the smoke, on the basis of the response of test animals to the toxicants identified in the chemical analysis. · Perform a bioassay of the material's smoke at, slightly above, and slightly below the expected LC50. If all the important toxicants have been identified in the chemical analysis, this test should be sufficient to confirm the identification and to yield an approximate LC50. If the observed LC50 is very different from the expected LC50, this will also be apparent; in such a case, more extensive bioassays must be carried out.
83 TEST METHODS THAT USE DEATH AS AN END POINT l BIOASSAY OF SMOKE Palt13NCY The chemistry of smoke and therefore presumably its toxic potency change with the conditions of burning. Fires typically evolve through a continuum of general conditions; the temperature, the ventilation, and the makeup of the fuel change over the duration of the process. These changes can modify the potency of smoker 17 ~ 25 and make the development of smoke potency evaluations difficult. The design of a test fire must include an abstract and simplified representation that will not only substitute for actual events, but also provide reproducible test conditions. The fire surrogates that are incorporated into test methods are most plausibly viewed as phases of a fire process, rather than as scaled-down replicas of specific fires. Exposure of a test sample at a fixed temperature has been compared with exposure to a fully developed fire, and exposure at an increasing temperature, to a growing fire. However, a test protocol that incor- porates a single, arbitrarily chosen temperature is subject to question, because of the temperature dependence of smoke chemistry. Fire products generated at one temperature might be relatively innocuous, and products formed from the same material at another temperature might be significantly more potent. 7 Generic toxicity testing is usually designed to reveal the worst-case response to a test agent, and the use of a single, arbitrarily chosen temperature for decomposition of samples is unlikely to achieve this condition. The f i rim surrogate for a test system is improved by choosing ~ _ , ~ a series of fixed temperatures or programed temperature increases for decomposition of test samples. Any exposure system used for evaluation of toxic potency must be considered representational, because actual human exposures vary widely and cannot be fully modeled in a test system. The major categories of expo- sure that have been incorporated in smoke toxicity tests can be described as static, in which smoke collects in a closed compartment, and dynamic, in which smoke streams from its source past the test subjects.4 ~ 07 Whether these differences are important is not known.
84 LIMITATIONS IN USE OF BIOLOGIC DATA . . . The time at which an effect occurs, as measured in these tests, depends on sample size and heat transfer by the furnaces used. None of the tests under discussion provides data that can be directly extrapolated to predict a duration of safety or a time at which human death would occur in an actual fire. If a single test chemical has been thoroughly studied and its mechanism of action in humans defined, a most appropriate animal species can be chosen for further evaluations. In studies of combustion-product potency during which mixtures of unknowns are administered in unknown quantities, the perfect surrogate for humans clearly does not exist. Instead, data must be collected from a well-characterized species that is typical of living systems. The most commonly used organisms are rodents. In the absence of definite, specific biochemical parallels, the use of other animals offers no advantages. The use of an animal (a "typical" living organism) should not be expected to predict the actual dose at which some event would occur in humans, but rather to define relative potency of a series of materials. Because, in the bioassay of smoke, the actual interactions of unknown test materials with biologic pathways are uncertain, it is important to interpret the data with a margin of error for individual, as well as species, differences. A variety of numerical safety factors have been invoked by regulatory agencies for this purpose. Integration of data from any toxicity test into a currently available numerical fire hazard model requires the expression of the test results as concentrations of smoke at which a specific end point is attained. Examples of end points might be change in attention span, lacrima- tion, loss of postural bonus, or, at the extreme, death. For the purposes of the numerical model, the specific biologic end point chosen is not critical. The choice depends on the desired degree of protection, which varies with the user. For some, the desire is simply to ensure escape by fire victims; for others, it might be to ensure continuation of peak performance in the presence of continued smoke exposure.
85 Although for some purposes it remains desirable to know as much as possible about the chemistry of the test material, that is not required in numerical fire hazard models. Present hazard analysis models call for a unit of smoke concentration. Smoke concentration is not generally measured, but the value can be numerically derived by dividing the mass of sample by the volume of space filled. The resulting value is represented in the tests discussed below as the LC50. (The routine measurement of CO in smoke or of carboxyhemoglobin, COHb, in the blood of exposed animals, however useful such measures are for research purposes, provides no informa- tion of utility to hazard assessment efforts that is not provided with more certainty by the LC50 itself.) Because of the requirement for data expressed in units of concentration, only three test methods are candidates for use with current numerical models of fire hazard: the National Bureau of Standards (NBS) method, the University of Pittsburgh (Pittsburgh) method, and the Deutsche Industrie-Norm. (DIN) 53 436 method. These are the most completely developed and documented of the available methods, have been published in peer-reviewed literature, and have been accepted or are under evaluation as consensus standards. GERMAN DIN 53 436 METHOD The DIN 53 436 method65 was developed in Germany to provide a standardized procedure for the generation of combustion products, as well as for animal exposure. The DIN method uses a dynamic airflow system. The furnace is a quartz tube 110 cm long fitted with a variable-temperature annular oven that moves along the tube at 1 cm/mint The furnace is designed to hold a cylindrical sample of equal volume or weight per unit length. Pyrolysis gases are swept out of the tube and into the exposure chamber by a 100-L/min airstream, with one further dilution possible between furnace and exposure chamber. The test animals are rats, and the duration of exposure (head-only or whole-body) is 30 min. LC50 and the pyrolysis temperatures associated with specific lethal concentrations are determined. Death rates, necropsy results, and blood COHb values are the primary biologic data collected. Continuous chemical monitoring of CO, CO2, and O2 and periodic measurements of HCN, HC1, and other gases are performed.
86 In a series of papers, Klimisch, Kimmerle, and co- workers have described the procedure and results based on exposure of rats to a variety of materials. 122 123 The data indicate that consistent and reproducible results can be obtained with this method within a limited range of temperatures (300-600°C). At temperatures above 600°C, when pyrolysis occurs, the variability of the results increases significantly. This method has been adopted as the German national standard, but has not been tested in the United States. NATIONAL BUREAU OF STANDARDS METHOD The NBS combustion systems 2 9 ~ 3 ° consists of a cup furnace in which heating elements surround a well that contains a 1-L quartz beaker, the sample holder. The exposure chamber is a 200-L airtight box connected directly exposure holders, chamber. to the furnace. All smoke is collected in the chamber. Six rats are restrained in individual with their heads extending into the exposure In this static test, the cup furnace is heated to a temperature 25°C lower than the autoignition temperature of the sample (nonflaming or smoldering condition). A weighed sample is dropped into the cup for decomposition. The animals are exposed to the resulting fumes for 30 min. starting at the introduction of the sample, and observed for survival at 30 min and up to 14 days. The test is repeated with progressively larger samples and additional animals. For estimating the LC50*, the *Although data produced by all the combustion-product toxicity test methods are expressed as LC50s, strictly speaking they are not LC50s, but they can be considered surrogate LC50s. LC50 is defined as the concentration of a toxicant that causes death in 50% of the exposed animals in a specified period; concentration is defined as the relative content of a substance, e.g., milligrams per milliliter or milligrams per kilogram. The "concentration" as NBS uses it is grams of sample charged relative to the volume of the exposure chamber (200 L). The "concentration" as the Pittsburgh method uses it is grams of sample charged. In neither test is the real exposure concentration known. However, both measures are commonly used to represent LCso.
87 exposure concentration is calculated by dividing the sample weight (grams charged) by the chamber volume (200 L). The entire series of tests is repeated with the cup preheated to a temperature 25°C above autoignition (flaming condition). For most uses, no correction is made for the portion of the sample that remains in the beaker on completion of the test, although modification of the system to collect these data would be possible. The resulting data are therefore nominal concentrations, rather than measured concentrations. For summarizing smoke potency, the statistical method of Litchfield and Wilcoxoni 3 ~ iS used to estimate the LC50. Data from the smoldering or flaming condition, whichever is more potent, are used for comparison with data on other materials, if a worst case is being described. Phe nominal exposure concentration, in milligrams per liter, is assumed to remain constant for most of any exposure period, because sample decomposition is expected to be rapid. Routine chemical monitoring of the exposure atmosphere includes evaluation of CO, CO2, and O2. The test atmosphere is sampled continuously during the test and returned to the chamber after filtering and analysis. Approximately 30% of the volume of the chamber is recirculated during the 30-min exposure; for some materials, this could be a source of important error. End points other than or in addition to death have been studied as part of this test system. The available data suggest that the other end points selected were not more sensitive or more useful than lethality alone. \29 (Nonlethal end points are discussed more fully later in this chapter.) UNIVERSITY OF PITTSBURGH METHOD The combustion system of the Pittsburgh methods 6 19 2 0 is a box furnace that is heated at 20°C/min. The exposure chamber is a 2.2-L glass box with ports to allow monitor- ing of the test atmosphere and placement of mice for head-only exposure.
88 Smoke from the furnace is diluted with chilled air and drawn through the exposure chamber. The total airflow and the dilution air are measured. In each test, four male mice are exposed for 30 min. Animals are observed for 10 min after the test exposure. The test is repeated with larger samples and additional test animals, to estimate the LC50 by the method of Thompson and Weil.220 In addition to the LC50, the time required to kill 50% of the animals (LT50) is recorded in this method. Results from this test are most frequently expressed in terms of furnace loading weight. However, because sample weight is recorded continuously-and air and smoke volumes are measured, it is equally convenient to express the results in terms of nominal concentration. 2 0 CO, CO2, and O2 are monitored in air pulled from the exhaust line immediately after animal exposure; this air is not recycled. COMPARISON OF TEST METHODS Acute Toxicity The LC50 protocols from the NBS and Pittsburgh tests are similar in several respects. The primary information is the number of animal deaths within a fixed period. For the NaS test, the time is 14 days after exposure. For the Pittsburgh test, the time is within the 30-min exposure period and 10-min recovery period; the Pittsburgh test has also been used with the longer animal observation period. In each case, the LC50 is calculated with a statistical method that provides confidence limits and the slope of the regression line. For both the NBS and Pittsburgh tests, the limited data available indicate acceptable reproducibility within and between laboratories. Anatomic Changes Necropsy data collected 24 h after exposure in the two tests have not been remarkably informative. With some important exceptions, reported gross necropsy findings have been limited to increased lung weights and corneal opacification.5 ~ 3 ~ ~ 3 9 Anatomic changes are time- dependent and would not be expected to be visible at
89 the times selected for termination of either test. If observation periods were changed for specific research needs, both test protocols could yield more information pertaining to pathologic changes ~ Test Subjects after animal exposure. The choice of test subject constitutes a problem for all toxicity tests. Similarities have been shown between rodent and human in sensitivity to toxic agents, such as CO ~ ~ 4 and HCN. 4 However, in rodents, which breathe through the nose, water-soluble materials might be retained in or absorbed through the nasal mucosa, thereby decreasing lung exposure. Rodents, for this reason, might be less acutely sensitive than other species to the effects of a water-soluble corrosive gas.~9 As a result, products that release a water-soluble corrosive gas as a primary toxicant could be evaluated differently from some other materials. 4 It is necessary to remain alert to this potential problem Physical Test Characteristics The physical characteristics of these two tests need to be carefully compared and contrasted to identify differences in the fire and exposure models used and artifacts to be encountered. The impact of physical characteristics on smoke potency remains mostly unknown. Differences in physical characteristics between the NBS and Pittsburgh tests might make these tests suitable for different uses. The validity and importance of this intuitive conclusion have not been carefully analyzed. Furnace Dimensions The Pittsburgh furnace is large enough (42 L) to accommodate a wide range of sample sizes, configurations, and orientations. The NBS sample holder is much smaller (1 L); as a result, some low-density foams cannot be evaluated in the NBS system. For some products, it might be desirable to use sample holders that allow end products to be tested in an
go orientation similar to that of the final use. It is more convenient to position a sample and a holder in the Pittsburgh apparatus than in the NBS system. Heat Transfer and Oxygen Availability The two test methods approach combustion in very different ways. In the Pittsburgh test, the sample sits on a stage in a relatively large oven; heating is pri- marily convective in the early stages, with radiative heat transfer to the sample becoming increasingly important as the oven's interior surfaces get hotter. In the NBS test, heat to the sample is conductive where the sample is in contact with the hot container; if the preset temperature is high enough, radiation from the surfaces of the container is also important. If the sample is large, heating produces substantial thermal gradients within it, according to its physical form, thermal conductivity, and so on. When low-density materials are used in the NBS test, they can occupy a sizable fraction of the hot container, making local O2 availability a problem. The Committee notes, however, that both these concerns diminish in importance as the size of the sample decreases; testing a product with an LC50 of 5 mg/L with the NBS method requires only 1 g of sample. Even if the material is a low-density foam, 1 g of sample occupies 30-50 cm3, which constitutes only a few percent of the 1-L volume of the cup furnace. The extent, therefore, to which the details of heat and O2 transfer control differences between the smoke toxicities measured by the two methods is expected to be less impor- tant for more toxic materials. To put it another way, the difference is most pronounced when it is least important. Heating Renimen A more important difference between the two tests is probably in the way in which energy is delivered to the sample. In the Pittsburgh test, the material is heated at a constant rate. Heating continues well beyond the ignition temperature; after ignition, therefore, the sample receives energy both from its own flame radiation and from the oven. Thermoplastic materials, once ignited,
91 usually burn relatively quickly, so the amount of such materials exposed to the high postignition heat load is comparatively small. In contrast, materials that contain fillers, which char or otherwise decompose more slowly after ignition, might lose most of their mass later. Douglas fir, for example, loses about 80% of its weight in this relatively high-energy environment. 6 The NBS test provides no such high-energy period. The room-temperature sample is dropped into the cup furnace, which has been preheated to just below or just above the ignition point. Early heating is very fast in comparison with that in the Pittsburgh test; the sample surface goes from 25°C to 300-500°C rapidly, instead of at 20°C/min. However, it is never given energy at a rate appreciably greater than that required to ignite it. If ignition occurs, the sample is irradiated by its own flame, but there is no increasing energy flux from an exterior source. Measurements of large-scale test fires indicate that heating rates immediately preceding ignition are likely to be very steep, perhaps tens of degrees per second, unless energy flux from an exterior source has been sufficient to preheat the sample surface well in advance of the flame's arrival. The external component of flux that a combustible sample receives is greatest near and after flashover. At earlier stages, the fire spreads as unignited combustible material is heated to its ignition point by an adjacent flame. From the preceding analysis, it appears that the thermal conditions of the NBS test more closely resemble those of a young fire while it is still growing. In contrast, the Pittsburgh test's thermal conditions before sample ignition are more similar to those near and after flashover. As noted below, neither test provides enough O2 to guarantee a well-ventilated fire. Combustion Chamber Atmosphere The atmosphere surrounding the sample in the furnace is important for the availability of O2 to the combus- tion process. Because of the relative sizes of the cup furnace and the sample, the atmosphere in the NBS cup can be low in O2 in some tests and higher in others. In
92 the Pittsburgh system, the atmosphere has a greater likelihood of being rather uniform, despite changes in sample size. However, according to the rule of thumb that a well-ventilated fire needs 5 or more times the air required for stoichiometric burning and the stoichio- metric O2 requirements for burning LC50 samples of Douglas fir in the two systems, both systems are "under- ventilatedn; consequently, fluctuations in chamber atmospheres might be relatively unimportant. Exposure System The static exposure chamber of the NBS test is representative of conditions in which smoke accumulates and mixes as the fire progresses, and it allows inter- actions among products that might be given off sequen- tially. The dynamic exposure system of the Pittsburgh test can be considered representative of human exposure to a moving stream of fire gas. The gases tend to be delivered to the animal as a series of concentration peaks, with low potential for interactions between fire products. This could have a toxic impact different from that of the mixed atmosphere of a static system, but the differences have not been studied. Physical Artifacts Artifacts inherent in both methods have been iden- tified, but not quantified. In the Pittsburgh test, the surface-to-volume ratio of the delivery tube from furnace to exposure chamber is high; that allows deposition of potentially toxic products on the tube surface. Moreover, the introduction of cold air into the smoke stream, which cools and dilutes the furnace products, might alter the chemical composition of the smoke. Some products might condense or precipitate out of the stream, because of the sudden change in temperature. In the NBS test, the gas monitoring procedure requires the removal and recycling of gas at 2 L/min for 30 min. for a total of one-third of the chamber volume. The air is recycled back into the chamber only after it passes through a series of filters and analytic equipment. The resulting change in test atmosphere has not been quan- tified. An additional possible artifact associated with
93 this system is the interaction of thermal decomposition products with the surfaces of the plastic exposure chamber. Because of the long residence time in the static chamber, it is possible for substantial quantities of reactive products to adhere to the inner surfaces of this chamber.45 Comparison of Data from NBS and Pittsburgh Tests Two kinds of comparison are of interest: broad classification and the actual measured values of toxic potency. The former is of interest if the tests are to have utility as screening devices; the latter is related the "c-;m=P" Of overall hazard. ~- ~1= =~= a_ _~ In comparing test data, it is important to point out that lack of agreement of test values implies not that either test is in error, but only that they are different. Considering the sub- stantial differences in the test characteristics, lack of agreement should not be surprising. A published range of LC50 values obtained with the Pittsburgh test for a wide variety of end products is 126 g (starting weight) for gypsum board, which is decomposed minimally by heat, to 2.7 g for a PVC pipe.~9 The data were not corrected for residue weight. The same series of products tested with the NBS method resulted in a range of 45 to 1.8 g (225 to 9 mg/L). More and less potent materials have been studied in each test System.4-6 130 i39 Anderson and co-workerst9 have compared LC50 values obtained by the two methods on various materials and concluded that they agree fairly well for thermoplastic materials, but not for materials that can char or other- wise leave behind substantial residue on decomposition. This conclusion is consistent with the observations made above on the differing thermal environments of the two methods. Actual values of toxic potency can be compared only if the data can somehow be normalized. Traditionally, this has been done by comparing the measured potency of a material with that of some reference material measured in the same test. ~ ~ ~ ~ ~ The results of Such a comparison, when Douglas fir as the reference, show relatively poor agreement; most materials are more toxic when evaluated
94 by the Pittsburgh method.5~ However, the Committee believes that Douglas fir is a poor choice for a reference material, because its decomposition appears to be particu- larly susceptible to the differences noted between the two tests, and agreement of test data would be assessed differently if a less variable reference material were chosen. Because time of exposure is so different between static and flow-through systems, comparisons might be more informative if time-weighting were used. Alexeeff and Packham~ have calculated L(Ct)sos from published data on the Pittsburgh test for cases in which information on sample weight loss was also available. In many cases, LC50s obtained on the same materials by the NBS test are also available, but the needed corresponding data on weight loss generally are not. In their absence, a limiting assumption is that all the material loaded in the NBS test is rapidly converted to smoke, so the nominal LC50 would be the actual concentration encountered, and it would be experienced for the entire 30-= n exposure period. The L(Ct)50 for the NBS test would then be the product of 30 min and the measured LC50. Values obtained by Alexeeff and Packham are compared with those obtained with the NBS test in Table 5-1. With two exceptions (cotton fabric and Douglas fir), the values obtained with the NBS test are systematically higher than those obtained with the Pittsburgh test. If the NBS test results were corrected for nonvolatile residues and for the time it takes for the samples (particularly the large ones) to decompose, the agreement would probably be much improved. COMPARISON OF TEST METHODS WITH GUIDELINES FROM 1977 NATIONAL RESEARCH COUNCIL REPORT The previous National Research Council Comb ttee on Fire Toxicologyt 6 4 presented guidelines for combustion- product toxicity testing, which provide another point of reference for comparing the NBS and Pittsburgh methods. The reader should note, however, that those guidelines were to address the development and improvement of methods for "first-level screening of materials." The present Committee is addressing methods that provide data to be incorporated in hazard analysis, not screening, of
95 TABLE 5-1 Comparison of L(Ct)50s Materiala PTFE wire* C-PVC pipe* Thin wire* PVC conduit Intumescent paint Paper walkover ABS pipe Nylon-carpet foam backing Vinyl walkover Mineral-base ceiling tile Latex paint Nylon-carpet jute backing Asphalt felt Wood-base tile Cotton fabric Reprocessed paperboard Douglas fir Gypsum wallboard L(Ct)508 L(Ct)so (NBS)~9 85 127 5b 212 230 260 280 298 313 384 403 465 486 486 561 883 1~440 1~521 76b 480 24ob 885 615 1/125 855 3 r 240 l'179 3 r 390 5~280 lr710 648 11371 387 1 r 5 ° ° 873 6r750 aAll materials except those marked with an asterisk reported to leave substantial residue after burning. bBased only on weight of insulation. materials. The Committee believes that toxicity data alone--e.g., data from screening tests--are not sufficient for the complete and accurate assessment of fire hazard. The first guideline recommended the use of both pyrolysis and flaming decomposition conditions. Both are used in the Pittsburgh and NBS procedures. The discussion preceding this recommendation, however, assumed that changes in airflow would accompany different burning conditions. As currently used, both tests have kept ventilation constant at the fire site. The second guideline recommended the use of specific test animals and exposure conditions: exposure of rodents for 15-30 min at temperatures not exceeding 35°C and with
96 O2 at no less than 16%. Both tests comply with this guideline to some extent. Because of the dynamic conditions of the Pittsburgh test, which result in the exposure of animals to different combustion products at different times, it has been run for a 30-min interval, but the time of exposure to specific smoke components cannot be fixed. Because of the addition of air for dilution, O2 is seldom low. In the NBS static system, a 30-min test has been selected. Concentrations initially rise and then are relatively stable for the remainder of the 30 min. O2 in the animal exposure chamber is monitored, and O2 is added if the concentration falls below the desired point. Each test provides for cooling smoke before animal exposure. The third guideline recommended inclusion of a suitable measure of incapacitation, followed by 2 weeks of observa- tion for behavioral and physical changes. The recommenda- tion regarding incapacitation has not been successfully met by either test. Incapacitation tests, designed to detect and measure a dose-related effect on some behavior, were found to be incapable of detecting such an effect at a time after exposure much different from the time when death occurs; and the behavioral and physical monitoring selected did not prove satisfactory for screening purposes. 2 9 Although tests that have been developed were not successful in providing the required evaluations, the Committee considers that this remains a desired test end point. The 2-week observation period is incorporated in the NBS test; the Pittsburgh test calls for a 10-min postexposure recovery time. The fourth guideline pertained to evaluation of the test atmosphere. In both tests, temperature, CO2, CO, and O2 are monitored routinely. Humidity and smoke density are not monitored. Other gases are measured when specific questions warrant it. The previous committee recommended that data be compared with equivalent test results from reference materials, rather than being used as absolute values. Douglas fir has been used as a reference material in both procedures, but is not specified by either. NBS, which has done considerable study with Douglas fir, recommends that it not be used as a reference material.) 29 The same conclusion has been drawn by others.~9 No other product has been recommended as a reference material for these tests.
97 TEST METHODS THAT USE NONLETHAL END POINT S FACTORS THAT IMPEDE ESCAPE Circumstantial evidence suggests that many fire deaths are related to victims' failure to escape before lethal conditions are encountered (perhaps evidence of incapaci- tation) and that characteristics of the smoke might have been responsible. As a result, the earlier National Research Council Committee on Fire Toxico~ogy~64 recommended that small-scale animal test protocols include a measure of the loss of ability to escape, termed n incapacitation." "Incapacitation," however, was not defined, and the methods later developed for measuring incapacitation, some of which are described in this chapter, have reflected the various investigators' interpretations. The characteristics of smoke that might impede or prevent escape cover a wide range of effects, from relatively minor to severe, including: · Blocking of visibility, which makes escape routes more difficult to find and use. . "Burning" and tearing of eyes, burning of nasal passages, and respiratory irritation (coughing and choking), which, even if not serious, could slow escape by causing distraction, discomfort, and panic. . Pharmacologic and toxicologic properties that impair sensorimotor function, alertness, and judgment. . Psychologic responses that cause panic, which could result in "freezing" and inappropriate choices of actions. . Physiologic effects on other organ systems, especially the respiratory system, that would render a person incapable of life-saving actions. Of these characteristics, the two that have received the most experimental attention are impairment of sensori- motor performance and sensory irritation. Various methods for measuring these effects are described briefly below.
98 OBSERVATIONAL METHODS l Observational methods have enjoyed considerable success in various screening programs. However, when applied to the evaluation of smoke toxicity, they have a potentially serious limitation: the presence of smoke imposes a restriction on the exposure-observation environment that could compromise other aspects of the test protocol, and the exposure-observation unit must be small enough so that smoke obscuration does not constitute a major problem. Hilado and co-workers99~~0 4 have been the major proponents of observational methods that use time to incapacitation as the observed end point. They define time to incapacitation (Ti) as the time to the first observation of loss of equilibrium (staggering), collapse, or convulsions. They also record time to death (Td), the time to cessation of movement and respiration. In the exposure protocol that they have used most extensively, a 1-g sample of material is heated in the absence of forced external air, and the temperature is increased from 200 to 800°C at 40°C/min. Data abstracted from several reportsi00~~02 104 showed that the Td:Ti ratios for the 15 materials tested were all less than 2.0:1, except that for poly(vinyl chloride), which was 2.8:1. The correlation between Ti and Td for these 15 materials was 0.89. Thus, the two end points apparently would provide similar information for ranking materials. Motorized Activity Wheels Several investigators have used motorized activity wheels to measure the capacity of rats or mice to perform a motor act during exposure to products of thermal degradation 6 0 ~ 5 9 ~ ~ 4 6 ~ 9 3 2 JO Crane et al 6 0 tested rats in motor-driven exercise wheels housed in an exposure chamber. The rats were exposed to smoke from a 0.75-g sample heated at 600°C for up to 10 min in a Lindberg tube furnace with recirculating flow. Ti was taken as the time when they could no longer walk and began to slide or tumble. Td was recorded when respiratory and body movements could no longer be seen. The average difference between Ti and Td for 71
99 materials was 7.3 min. and the average Td:Ti ratio was 2.0:1. For 15 of the materials, there were no deaths during the 30-min exposure period. An incapacitating combustion or pyrolysis product that is not lethal will generate what appears to be a very large difference between Ti and Td; however, if there is no Td (no lethality, then differences and ratios are meaningless or indeterminate. Indeed' the time to effect for any end point cannot be considered adequate as a measure of smoke toxicity if only one concentration is evaluated. 4 Hind-Leg Flexion ; In this procedure, developed by Packham and co-workers, 7 ~ 7 5 ~ 7 0 a rat's hind leg is wired so that a shock is delivered if the foot touches a metal plate below it, and the task is to keep the foot raised. The rat is considered incapacitated when it can no longer avoid the shock. This test was adopted by NBS as its measure of incapacitation, thoroughly evaluated, and incorporated into an interlaboratory study to examine the reliability of the NBS test protocol. 2 9 ~ 3 0 Com- parisons of effective concentrations (EC50s) for hind-leg flexion and 14-day LC50s led to the conclusion that the hind-leg flexion response did not provide any great increase in sensitivity over the 14-day LC50. Indeed, because of delayed deaths, the 14-day LC50 was more sensitive than the hind-leg flexion response for three of the 11 materials in the nonflaming mode and for one of the 12 materials in the flaming mode. Moreover, the materials were ordered in essentially the same way by both end points. The correlations were 0.88 and 0.95 for the nonflaming and flaming modes, respectively. As a result of these studies, NBS eliminated the hind-leg flexion response from its protocol. Sensory Irritation and Physiologic Stress The use of plethysmography to measure sensory irrita- tion in laboratory animals was developed by Alarie and co-workers 2 3 5 2 ~ ~ ~ 2 A mouse is exposed to the products of thermal decomposition of a material in a chamber into which its head protrudes (described earlier as the Pittsburgh method). The rest of its body is sealed in a plethysmograph that continuously records
100 respiratory and other body movements before, during, and after the test period. Two quantitative measures are obtained from the plethysmographic records--sensory irritation and physiologic stress. For sensory irrita- tion, the maximal decrease in respiratory rate in each test is recorded and plotted against the amount of material decomposed in that test. The amount of material decomposed that is associated with a 50% decrease in respiratory rate (i.e., the RD50) is estimated from the concentration-response curve obtained from a series of such tests. Physiologic stress (stress index, SI) is obtained from the plethysmographic records minute by minute throughout the exposure and recovery periods. The SI reflects the severity of a series of physiologic adjustments that are made to compensate for the reduction in breathing rate and apneic periods. 2 ~ For most materials, respiratory depression in mice is seen at much lower concentrations than is death.5 The correlation between the RD50 and the LC50 was only 0.14, indicating relative independence of the two end points. Kane et al. 2 compared the RD50s for 11 sensory irritants with effects reported in the literature for humans and animals. They proposed that the RD50 is equivalent to an intolerable degree of sensory irritation for humans and would probably cause incapacitation within 3-5 min.3 28 ll2 They also proposed that use of 10 times the RD50 would be lethal or cause severe injury to the respiratory tract. This prediction has been verified. 4 ~ Prediction of safe exposure of humans to 40 industrial chemicals from RD50 values found with mice was excellent (correlation, 0.92).7 Furthermore, Potts and Lederer~t used this model to examine the tolerability of an RD50 of red oak smoke in humans for 3 min. They used a strain of mice (HAICR) with different sensitivity from that of the Swiss Webster mice used by Alarie.4 At the RD50 concentration for HAICR mice, humans reported irritation, but this concen- tration was not intolerable and certainly not incapac- itating. However, at RD75, Potts and Lederer~ reported that the irritancy was high and that the person inhaling the smoke believed that it could not be tolerated for 3 min. but whether incapacitation would be caused by a 3-min exposure was conjectural.
101 OTHER METHODS Several other methods have been used to investigate sublethal effects of products of thermal degradation. Available data are not sufficient to compare them adequately with tests that use death as an end point, nor are the data that most tests produce expressed in terms of concentration. Therefore, they are described only briefly here. Although these tests do describe behavioral or biochemical changes that result from exposure to combustion-products, it is not clear how they correlate with human health effects described in Chapter 4. Unsignaled-Shock Avoidance/Escape Several investigators have used this procedure to monitor the behaviorally disruptive effects of thermal- decomposition products. The general procedure, known as Sidman avoidance, 2 01 involves training rats to press a lever to postpone a scheduled, but unsignaled, painful foot shock ti.e., avoidance) or to terminate the shock if it is received (escape). Acquire and Annau~43 used the procedure to evaluate the effects of exposures to the smoke from a flexible polyurethane foam. They found significant increases in shocks received in 4- and 8-g exposures, both during exposure and during a 30-min recovery period. Trends in the same direction were seen in 2-g exposures, but the differences were not significant. The effects seen could be attributed to the amounts of CO generated. Sette and Annau200 used the method to investigate the interactive effects of pure CO (1,666 ppm) and heat. Heat alone and rm =1~^ ~a.l==A on increase in shocks received, and the ~ v Ha ~ ~ ~ ~ ~ ~ _ _ ~ ~ ~ ., ~ combination appeared to be additive. Russo et al.l9 6 compared the effects of the smoke from polyimide and flexible polyurethane foams. The thermally less stable polyurethane foam caused greater performance decrements at the lower heating temperature at which greater amounts of CO were evolved. The reverse was true of the thermally more stable polyimide foam. Water-Reinforced Task McGuire and Annaul43 compared the effects of the smoke from a flexible polyurethane foam on licking
102 behavior. Water-deprived rats were trained and stabilized on a water-spout licking task. The authors found disrup- tion of performance at 1.0-g exposures. Although exposure to the smoke from 0.1 g did not cause a significant decrease in the total number of responses during the 30-min exposure, a clear and progressive decrease began after 15 min and progressed for about 10 min. The authors suggested that the temporary decrease in licking behavior was caused by irritant components of the smoke to which the rats had adapted by the end of the exposure period. Rotorod with Electrified Grill Floor . This procedure has elements in common with methods that use motorized exercise wheels and shock avoidance/ escape. Hartung et al.9t trained rats to walk on a rod 3 in. (7.6 cm) in diameter rotating at 6 rpm. The incentive for remaining on the rod was the presence of an electrified grill floor under it. A rat was considered incapacitated during exposure when it fell off the rotorod and remained on the electrified grill floor for at least 2 min. after which the current was turned off in that compartment. Variability in Ti was relatively low, and that permitted statistically significant differences to be found among materials with the use of only eight or 12 rats in each group. This method was also used by Mitchell et al.~46 in full-scale tests of the effects of combustion products of natural fiber and synthetic polymeric furnishings. Their results showed that the smoke from synthetic polymeric materials impaired performance faster than smoke from natural-fiber materials. Multisensory Conditioned Pole-Climb Avoidance In this procedure, rats were trained to escape foot shock by climbing or pulling a response pole suspended from the ceiling of the test chamber. They were then trained to avoid the aversive foot shock, which was preceded by one of three warning stimuli--a light, a tone, or a nonaversive electric current on the floor. All of a series of 10 flexible, flame-retarded poly- urethane foams 7 2 ~ ~ 4 caused impairment of the avoidance response, but none was incapacitating (as defined by loss of the escape response). Nor were these materials, in
103 the exposure ranges tested, associated with acute mortality during the exposures and 30-min recovery phases. However, delayed deaths attributed to pulmonary or cardiovascular complications) 7 ~ followed exposure to eight of the 10 materials. In tests with five poten- tial aircraft interior materials, 6 6 almost all deaths occurred within the exposure-recovery test period. The correlation between the concentration of smoke that caused a 50% decrease in avoidance and LC50 for this small series of materials was 0.91. Thus, each end point would predict the other, and they would provide about the same answers with respect to the relative toxicity of smoke. Analysis of Use of Bronchoalveolar Lavage Fluid to Detect Acute Nonlethal Lunn Toxicity Many short-term tests have been developed to assess the potential toxicity of various materials that might be inhaled by humans. These include the use of isolated pulmonary alveolar macrophages, 2 2 s bacterial cultures,~5 and cultured mammalian cells. 4 7 These tests have proved useful in rapid screening of a large number of materials, but the general approach of each test does not allow monitoring of the integrated response of the whole animal to the inhalable agent in question. One in viva method, which is also relatively rapid, uses biochemical and cytologic evaluation of bronchoalveolar ravage (BAL) fluid from the lungs of exposed animals to detect lung damage. Many investigators have used this technique to evaluate the potential chronic lung toxicity of inhaled materials, on the basis of the hypothesis that animals exposed by inhalation to a pulmonary toxin will suffer subtle acute lung damage that can be measured by various biochemical and cytologic changes. A correlative hypothesis is that inhaled toxicants will cause specific types of damage to respiratory tract tissues and cells, altering various biologic and biochemical processes that can be measured with assays of BAL fluid. Sampling the bronchoalveolar region of the respiratory tract by saline washing (ravage) of lungs is relatively simple. The ravage is usually done on excised lungs in toxicity screening tests. However, it can be done in viva if the experimental objectives of the study require repeated sampling of BAL fluid.
104 Changes in BAL fluid characteristics are postulated to be specific and indicative of later-developing pathologic alterations in respiratory tract tissues.9 4 A number of studies have shown good correlation between these early toxicity indicators and later-developing lesions. The application of these methods to combustion-product toxicity testing programs might also prove useful. Not only do analyses of BAL fluid potentially provide infor- mation on the development of chronic lung disease, but these techniques might provide a highly sensitive method for ranking the acute toxicity of inhaled combustion products. SUMMARY Efforts have been made to use analytic chemistry to predict the toxic potency of smoke. However, chemical analysis is not now an acceptable substitute for bioassay of smoke toxicity. Biologic tests of smoke toxicity are most useful in defining relative toxicity of combustion products; direct extrapolation to humans is seldom appropriate. For an accurate assessment of fire hazard, toxicity data alone are not sufficient, but should be incorporated into a fire hazard assessment. The two biologic methods for evaluating the lethal potency of smoke that have been compared here appear to provide data that can be incor- porated into a numerical model for fire hazard evaluation. Neither test successfully addresses the possible sublethal effects of smoke exposure. Each test yields reproducible results. These bioassays represent different fire and exposure conditions, but the relationship between small-scale test conditions and real fire conditions is not well understood. Analysis of toxicity data on nonlethal end points in animals exposed to products given off by different burning materials might provide data for the prediction of potential health effects of these products in humans. These analytic methods could also prove useful for application to overall hazard assessment and selection of materials of greater safety with respect to fire. However, at present none has been developed and validated to the extent necessary for incorporation into such an assessment.
