Whatever happened to combustion toxicity?

Hall, John R Jr

From a field full of questions and widely devergent points of view, combustion toxicity has become a solid component of our scientific approach to fire safety and design.

There was a time not too long ago when toxicity was the hottest topic in the field of fire research and even on the agenda of fire safety generally. A 1983 Fire Journal editorial called it "perhaps the most pow

erful word in fire protection" and cautioned against using it as "emotional shorthand" in what should be technical debates.

Maybe you first noticed the issue of fire toxicity when it formed the basis of 22 proposals in the early 1980s to change the National Electrical Code(R) provisions regarding such products as electrical nonmetallic tubing.

Maybe you noticed it in 1982, when NFPA created a Toxicity Advisory Committee--a new kind of committee for NFPA's codes and standards process--so that all technical committees could have access to a common set of guidelines and working assumptions from a scientific field marked by fast change and intense controversy.

Maybe you attended the 1982 NFPA Annual Meeting, where a full-day session was devoted to toxicity, or one of the many early 1980s fire research conferences where sessions were devoted to the topic.

Maybe you paid particular attention in 1986, when the state of New York required that manufacturers selling construction products in the state file toxicity data based on the University of Pittsburgh test, named for the home base of Professor Yves Alarie, the test's principal developer.

Maybe you haven't thought that much about combustion toxicity in the past five years. In fact, nationwide, the intense focus on that topic and its visibility in fire research and fire safety forums tapered off over that period. Fire research conferences often lacked even a single paper on toxicity, let alone a whole

session. Prominent combustion toxicity specialists lost their departments or their jobs or left the field. Yet, work continued.

Last year, NFPA adopted NFPA 269, Test for Developing To.ic Potency Data, its first standardized test addressing toxicity. Many of the same people also shepherded ASTM Standard E1678 and ISO Standard 13344 to adoption last year, as well. Where once there was no nationally or internationally recognized standard test method for assessing combustion toxicity in the United States, there are now three nearly identical standards.

The adoption of the International Standards Organization (ISO) standard was also the catalyst for a conference in March 1996 in Munich, Germany, organized by Henry Roux, chair of ISO Technical Committee 92, Subcommittee 3, ISO's combustion toxicity committee. That conference was the immediate inspiration for this article, as it served to demonstrate how much we now know and agree on--and how much remains controversial and unresolved.

The purpose of this article is to review the major issues in the field, at least those of great practical significance to fire safety decision-making in the United States. It isn't meant to substitute for the much more detailed and technical toxicity material contained in such references as the Fire Protection Handbook and the SFPE Handbook of Fire Protection Engineering. Rather, this article focuses on those issues that illustrate how the new standards have emerged as solutions to the concerns that originally stimulated work in the field.

Why does toxicity matter?

Fire kills thousands of people every year in the United States and in,iures tens of thousands more. It also destroys billions of

dollars in property every year. Fire is a problem that still matters, even after a century of tremendous progress.

Fire kills and injures in two principal ways, which we tend to refer to as "burns" and "smoke inhalation." Both terms are potentially misleading. Thermal injuries include not only the direct damage to tissues properly described as a burn, but also the heat stress that results when the body's ability to regulate its own temperature through heat exchange with the environment--for example, by sweating--is impaired. The visible, airborne, unburned fuel we think of as smoke is a factor in blocking visibility but not directly in causing injury. Injury results from the intake of the invisible toxic gases in smoke, nearly always by inhalation as opposed to, say, absorption through the skin. Injury also results from reduced access to oxygen due to reduced levels of oxygen in the atmosphere of an affected room.

Smoke inhalation accounts for most fire-related deaths in the United States and for a steadily increasing share of those deaths (see Table 1). The smoke inhalation share was nearly three-fourths in 1992, up from less than three-fifths in 1979, or roughly one percentage point a year. This happened even though smokeinhalation fire deaths have declined in

absolute numbers--and declined even faster relative to population--because burn-related fire deaths have declined much faster than smoke inhalation-related fire deaths.

In simplest terms, therefore, toxicity matters because it's the principal mechanism by which fire kills. Any program that can make it less likely that fire will inflict a fatal toxic dose will save three lives for every one life saved by a program that produces a similar reduction in the chances that fire will inflict fatal burns. Yes, there are delayed deaths due to burns. And yes, some people who die in fires die as a result of the combination of thermal and toxic effects or as a result of burns or smoke inhalation alone. But these and similar points are minor in relation to the big picture, which shows the dominance of toxicity as the cause of death in fire.

Toxicity is also an important, though less dominant, cause of nonfatal injuries. For example, during the five years from 1989 to 1993, 42 percent of nonfatal home fire injuries were attributed to smoke inhalation alone, while 29 percent were attributed to burns alone, 17 percent to burns and smoke inhalation, and 13 percent to other effects.1 What's more, the airborne gases and other products of incomplete combustion in fire are increasingly recognized as significant factors in property damage.

How much does toxicity matter?

Just because toxicity is the principal mechanism by which fire kills, it needn't mean that a more detailed understanding of how it kills is necessary, or even useful, as a means to greater fire safety. For a pattern in fire experience to be useful programmatically, there has to be something you can do about it.

For example, men are much more likely to die in fires than women, but no one (I hope) would propose widespread sex changes as a fire safety program. Less facetiously, nearly every fire safety program you can define--educational, technological, whatever--would be expected to work roughly as well on women as on men. And while one could define a program to target men--news releases aimed at Esquire and not at Redbook, say--such priority-setting exercises would be counterproductive. Why ignore half your audience when you don't have to and it's cheaper not to?

Toxic effects are present in all fires. There's no point in changing the materials we use if it only changes how we die in fires but not whether we die in fires. In fact, this may be a major reason the field of toxicity has lost some of its prominence. Much of what we learned about the mechanisms of fatal toxic injury in fire pointed to strategies outside combustion toxicology as the best avenues of reducing fire risk.

Before we can elaborate on this point, however, we must step back and examine how fires create toxic effects and how the recently adopted standard test methods measure toxicity.

You are a carbon-based life form

Fans of Star Trek or other popular science fiction may recall references to the fact that life on Earth is carbon-based. In nearly all fires, the principal fuels are organic material, from renewable resources like wood and cotton to fossil fuels like natural gas and petroleum and its derivatives, including plastics. Organic molecules are rich in carbon atoms, and fire is a chemical process that releases energy by throwing oxygen atoms into the mix. Therefore, nearly any hostile fire will result in reduced oxygen and enhanced levels of carbon monoxide (CO) and carbon dioxide (C0^sub 2^), all relative to ambient conditions before the fire started.

Other toxicants require the presence of other kinds of atoms, which aren't present in all fuels. Hydrogen cyanide (HCN) needs nitrogen (N) to go with the hydrogen (H) and carbon (C) found in all organic material. Hydrogen chloride (HCI) requires chlorine (Cl) atoms. And so on.

Let's tie this back to strategy implications. If the toxic effects of fire are mostly due to oxygen deprivation or inhalation of carbon monoxide or carbon dioxide, then there's little point in trying to improve fire safety by strategies specific to toxicity, because you can't have a fire without a resulting oxygen deprivation in the surrounding atmosphere or production of carbon monoxide and carbon dioxide. You can change materials to make them harder to ignite in the first place. You can change materials so they will join the fire more slowly--say, by reducing flame spread over a surface. You can change materials to reduce the

peak heat release rate or slow the rate at which fire reaches the peak, which will slow the developing thermal threat from fire and hold down the quantity of toxicants released into the air. But you can't get rid of carbon monoxide or carbon dioxide.

You can get rid of hydrogen cyanide, hydrogen chloride, and many of the other toxicants. Therefore, a critical question for toxicity-oriented fire safety strategies is whether you gain anything by narrowing the common toxicants in fire to those you can't avoid.

That question, in turn, depends upon a better understanding of the complicated relationships between the type of fire and the type of fuel that eventually lead to toxic effects on people in fires, based somewhat on the type of people. It may help to consider this as a sequence in space and time.

Before a fire begins, the fuels that will feed it exist in the form of often-complex products in a building, its furnishings, and its finishes. Each product may have many components, which, in turn, may be made of many different materials. If a fire begins on an upholstered

couch, the exposed fabric covering is likely to be the first thing ignited. Only later does the filling material, a very different composition than the fabric covering, ignite. And in between, the fire may have to burn through an interliner of yet a third composition. Once the filling material ignites, however, it will be, by mass, the largest fuel source of the three, by far. As the three parts burn, joining the fire at different stages, they each release different toxic gases. As the fire intensifies, the rate of release of toxicants also increases.

When the fire has significantly reduced the room's oxygen concentration, particularly if limited ventilation keeps the fire from drawing in enough fresh air for complete combustion, the same materials will release different proportions of toxicants. In a fire in which plenty of air is coming in through open doors and windows, for example, there is plenty of oxygen. Oxygen from the air combines with carbon from the vapor rising from the pyrolized--that is, heated--materials serving as fuel, and the combination is a carbon monoxide molecule. Further reactions add an additional oxygen atom to some of these CO molecules and produce carbon dioxide molecules. When oxygen is scarce, many of the chemical reactions stop short of that last step from carbon monoxide to carbon dioxide, and the ratio of carbon monoxide to carbon dioxide can increase dramatically.

Meanwhile, the released toxicants have to move to the occupants to exert any effect. Those produced early in a fire will drift slowly from their point of ori

gin. They won't travel very far or very fast, and the resulting concentrations are very rarely lethal (more on that later). Toxicants produced later in the fire, particularly after flashover, will be pushed rapidly outward by the intense heat, making way for fresh air and more oxygen, which are pulled in as if by a pump. As the gases and soot and other products of fire travel from fire to exposed person, some of them will be deposited on intervening surfaces, where they can't injure anyone, unless or until those surfaces become involved in fire. Other toxicants may interact chemically, producing different proportions in the toxicant stew.

An occupant may first interact with a fire by turning away from visible smoke. Such an action may lengthen the individual's escape time, as a more difficult or at least less familiar route is substituted for the normal route. At this point, the fire will still have no physical effect on the person.

The first quantities of gas actually inhaled may not stop a person from escaping, but they may make his or her evacuation activities less efficient and less effective. Asphyxiant gases, like carbon monoxide, first cause confusion, dizziness, and fatigue. Irritant gases, like hydrogen chloride, sting the eyes, leading to tears. And gases like carbon dioxide may cause more rapid breathing, speed

ing the effect of every other toxicant. The victim is still moving, still trying to escape, but he or she can't see clearly, finds breathing painful, and may even become so disoriented as to move toward more severe conditions. Next comes incapacitation, possibly unconsciousness. And finally, if rescue doesn't arrive in time, comes death.

At each stage, the individual's health may accelerate or delay the onset of a toxic effect. A heavy smoker, for example, is already somewhat loaded with carbon monoxide, and a heavy alcohol or drug user will already be mentally impaired.

But how would you describe those who were blocked by visible smoke until hydrogen chloride impaired their vision, further delaying their escape as carbon dioxide accelerated their breathing until hydrogen cyanide incapacitated them and carbon monoxide provided most of the dosage that ultimately proved lethal? In such a case, what actually killed the victims? And what kind of test or analysis could capture all the facets of this complicated picture, when and if they apply?

How burning products lead to toxic hazards

The first stage in the creation of toxic effects from a burning product is the ignition of the product. Being hard to ignite--both initially, as the first fuel in a

fire, and secondarily, as a major fuel source later in the fire--may just be the most important fire safety property a product can have, but it tends to be excluded from combustion toxicity tests. Toxicity tests start measuring after the product has been ignited.

The second stage in the creation of potentially lethal toxic effects from a burning product is to involve the product fully in the fire. If the product is a large item, such as a piece of upholstered furniture or a wall or floor covering, the speed at which it becomes fully involved may depend, in part, on the rate of flame spread over the surface of the product. Again, toxicity tests usually expose only a small specimen of the product and so aren't designed to measure flame spread rates.

Suppose the product consists of multiple layers of very different materials, such as a piece of upholstered furniture, or is part of a system consisting of different layers, such as a carpet and its pad. In that case, the toxic gases released may be quite different in both quantity (per minute of burning) and quality (how dangerous a given amount of gas is) for each layer. Many combustion toxicity tests address this aspect by exposing a specimen of the product-- like a core sample from the ground--to

fire as it would be in actual use. By contrast, some tests, such as the University of Pittsburgh test, grind a sample into pellets and don't capture the effects of layering.

In fact, layer effects aren't easy to capture properly. Suppose you had a cyanide brick wrapped in an outer covering that burned very slowly. If that covering were thick enough and sufficiently resistant to flame, you might be able to say with confidence that the cyanide brick would never become part of the fire until the whole building had burned down around it--and you might call that safe. But how durable would the covering have to be to give you such confidence, knowing what was inside? In fact, some products, like seats on city buses, are so subject to vandalism that they're tested using specimens in which the layers have been exposed.

The issue of how to treat the complex structure of products and the early stages of fire provide the first, but not the only, reason toxic hazard assessment may require mathematical analysis. If the process of involving the product fully in the fire can't be captured in a test procedure, that process must be addressed in some other manner, such as by modeling or calculation. Otherwise, fire safety decisions on product toxicity would have to be made strictly on the quantity and potency of gases released by the burning product when fully in

volved, a potentially misleading focus on characteristics that might not be important in real fires.

Toxic potency and supertoxicants

Note that I said both the quantity and the potency mattered. Typical toxic gases generated in fires are produced in widely differing quantities, and the amount of toxic gases needed to kill you differ, as well. Toxic "potency" refers to the quantity of gas required to produce a toxic effect, in this case, death.

To give you a sense of how much typical fire gases differ in potency, it takes about 18 times as much carbon dioxide to kill you as carbon monoxide. It takes about 25 times as much carbon monoxide to kill you as hydrogen cyanide. Now suppose there were a toxicant many times more potent than hydrogen cyanide.

For years, a major theme in combustion toxicity was the search for a "super

toxicant." Early on, one research group reported testing a material whose smoke was so toxic that just one breath might be fatal. Another group tested an item that proved frightening not so much because it was unusually potent, but because its manner of killing was so grisly. In this case, chemical reactions among the gases produced by burning an experimental product yielded an extremely potent neurotoxin that wasn't part of the makeup of any of the ignited materials. Exposed laboratory rats suffered seizures before their deaths that left everyone who saw the tapes of the experiment shaken.

Every toxic hazard assessment procedure since those early days has been judged first and foremost on its ability to identify supertoxicants, if and when any should appear. However, real supertoxicants have proven to be extremely rare, even in the laboratory. Most of the handful of materials and products found to produce supertoxicant smoke were

never available commercially, and the rest were withdrawn before they could pose a significant threat.

Some materials initially suspected of posing a supertoxicant threat proved not to upon further analysis. Perhaps the best known example was polytetrafluoroethylene (PTFE), perhaps best known through Dupont's Teflon(TM). Early tests identified a substance of unusually high toxic potency in the smoke from burning PTFE, but later work showed that chemical interactions in the smoke kept the quantity of supertoxicant released extremely low, no matter how much PTFE was burning.

This episode illustrates a larger lesson that repeatedly emerged from combustion toxicity work: Toxic hazard is usually driven far more by other factors--particularly mass loss and energy release rates but also resistance ignition and flame spread--than by toxic potency.2 In fact, as early as the 1984 preliminary report of NFPA's Toxicity Advisory Committee, experts have sought to underscore the importance of these two rates in smoke toxicity hazard. That fact runs counter to what many people intuitively expect to be true.

Even in the fire safety field, most of us are lay people when it comes to com

bustion toxicity, and, as such, we react as lay people do when considering the relative importance of dose versus substance. Think back to when you first encountered the words "cyanide" and "neurotoxin" in this article. Didn't you get a bit of a shiver down your spine just thinking about those substances? Did you have the same reaction to "carbon monoxide"? Probably not. If supertoxicants had proven to be the main problem in fire toxicity, toxic potency would have been the key to fire safety, but that isn't the picture that has emerged from two decades of research.

So, if the supertoxicants aren't what's killing people in fires, what is?

What kinds of fires kill people by toxicity?

Returning to the match between how fire kills through toxicity and how the new standardized toxicity tests operate, recall that the toxic gases released in a fire depend not only on what parts of the product are burning, but also on what fire conditions are driving the burning. A more intense fire, such as a post-flashover fire, will greatly increase the rate at which toxicants of all types are formed and released, but its effect on carbon monoxide will be especially pronounced because the depleted oxygen will shift the balance of gas production between carbon monoxide and carbon dioxide. The test and analysis should reflect the fire conditions pro

ducing the deadly toxicants in fatal fires.

What kinds of fires are those?

The combustion toxicologists who first agreed on the different types of fires that were important to the subject identified six. Three were nonflaming types, of which one was smoldering. One was a free-burning fire. And two were fully developed fires, which involved temperatures associated with flashover and postflashover conditions. One of the two involved very low oxygen levels--under 5 percent, compared to the 20.9 percent characteristic of normal ambient airand the other involved oxygen levels in the 5 to 10 percent range.

The defining characteristics of these six categories involve some information that isn't provided in fire department reports on fire conditions. However, enough data are available to allow us to say that the three nonflaming types of fire collectively account for only a tiny fraction of fatal U.S. fires. Most fatalities in the United States that are due to toxicity alone occur in fires that spread flames beyond the room of origin (see Table 2), and that is usually taken to indicate that flashover was reached in the room of origin.

This pattern varies from country to country. In the United Kingdom, for instance, fatal fires are much more likely to

be confined to the room of origin, possibly reflecting the much greater use of closed doors to separate rooms in British homes to help save on heating costs. But in the United States, the principal fatal fire is a fire that reaches flashover.

Along the way to flashover, a fire may spend considerable time in any or all of the other phases. But fires that peak at those smaller phases are statistically much less likely to result in fatalities. Thus, it appears that the toxicants produced at or after flashover are the critical elements that are killing most people. How can this fact be reflected in a standardized toxicity test or analysis? The answer is one of the major points of difference among the three new standards.

ISO 13344 doesn't specify the furnace, or "fire model" as they call it, using the word "model" not to refer to a computer model but to mean that the test furnace "models" a certain type of real fire. Instead, ISO 13344 tells the user to identify the types of fires relevant to his or her application, then select and operate a furnace that will properly represent those fires. This is a tall order if the fires of interest are fla:shover or postflashover fires.

Most small-scale test furnaces can't produce the heat flux levels characteristic of flashover. Small-scale test specimens of many products would burn up so quickly under flashover conditions that reliable time-dependent measurements might be difficult or impossible to

obtain. In most small-scale toxicity test apparatus that use animals, the animals are so close to the furnace that toxicants produced under flashover conditions will tend to kill them through heat effects, rather than the toxic effects the apparatus is intended to measure. And finally, large-scale toxicity test apparatus, which could produce flashover-level heat flux values more easily and move test animals farther from the fire, tend to be prohibitively expensive for routine use and subject to more variability among supposedly identical tests than can easily be accommodated.

NFPA 269 and ASTM E1678 both specify the furnace, which should mean greater consistency among users. The apparatus they specify uses a cylindrical test chamber. If you were to look down the cylinder, you would see the ends of the two rectangular radiant heaters at about 10 o'clock and 2 o'clock. This apparatus has many advantages for test purposes, but it isn't designed to apply flashover-level heat fluxes. Consequently, NFPA 269 mandates an analytical correction to the test results to adjust carbon monoxide levels to reflect flashover. ASTM E1678 doesn't mandate such a correction but is considering it and warns against any direct use of raw toxic potency data to assess toxic hazard, as does ISO 13344.

The NFPA 269 analytical solution is not without its critics. Some combustion toxicologists believe the available data don't

clearly support the specific correction in the standard or at least involve more questions and problems than are involved in the ISO-recommended approach of operating the furnace at flashover conditions. There's also an administrative or philosophical objection that says test method standards should limit themselves to the test procedure, while other standards should be written for any analytical adjustments required to apply the test data for a particular purpose. At this point, it appears likely that all three organizations will continue to discuss and refine their recommendations on how to adapt and apply the data to toxic hazard.

Note that we now have several points at which combustion toxicity assessment is likely to require mathematical analysis. More are coming. But this may be a good place to mention the advantages, given the apparent unavoidability of calculation within toxicity, of connecting toxicity to the larger exercise of overall fire hazard assessment.

Concerns about animal testing

From the beginning, combustion toxicologists have had to answer questions about the use of animals to represent human response to toxicants. Earlier discussions focused on similarities and differences between the responses of human beings and responses of particular animals or on other issues regarding the suitability of particular animals in laboratory work. Reactions of rats and mice to inhaled gases of the narcotic type, such as carbon monoxide and hydrogen cyanide, have been shown to model human reactions quite well. Mice are smaller, more vulnerable animals than rats, which can pose some special problems. For example, mice are more likely to die due to the stress of being in the apparatus, even before exposure to toxicants. Responses of rats to irritant gases, such as hydrogen chloride, aren't as good a model of human responses, though most experts consider them adequate, and this helps to explain why some experimenters have used monkeys or baboons.

Another reason is the difficulty of devising human-like tests of nonfatal physical or mental impairment in rats, for those tests that focus on nonfatal toxic effects. Suppose you want to detect significant impairment well short of fatal injury. You would teach your animal subjects a behavior, then see how severe a dose of toxicants is required before they stop performing the behavior. If the behavior is too difficult, some ani

mals won't be able to learn it to begin with. If the behavior is too easy, even very impaired animals might be able to continue performing it. And what do these animal-specific behaviors mean for people, anyway?

In the past decade, these concerns have been overtaken by the greatly increased visibility and political influence of groups opposed, on ethical grounds, to any exercises that deliberately inflict pain or death on animals, for any purpose. These groups inspired a search for methods that would at least minimize, and possibly even eliminate, the use of animals.

Fractional effective doses and N-gas models

More than a decade ago, several concerns converged to set the stage for two similar concepts of toxic hazard assessment, which have since come to dominate work in the field. One concern was practical and financial. Literally hundreds of potentially harmful gases and other components are now known to be created in fire. Just to assess each component, alone and in interaction with other components, would mean tens of thousands of experimental series--and no assurance that even that unaffordable volume of work would be enough to answer all possible questions.

A second concern extended beyond the issue of toxicity. If modeling were needed to properly address the complex issue of fire toxicity, then toxicity was also needed to construct comprehensive

packages to model all aspects of fire hazard.

And then there was the growing concern about animal testing.

These concerns were, and are, addressed by the fractional effective dose (FED) and N-gas models. N-gas is better designed to limit the testing burden, while FED is better designed to fit easily with the other parts of a fire hazard analysis modeling package.

The N-gas approach starts with the premise that most toxic effects from a burning product are due to the same small number of gases. (It was dubbed the N-gas model because the originators weren't sure how many gases would turn out to be important, only that the number (N) would be small.) You can burn the product, measure the rate of release of each of the N-gases, and combine their effects based on experimental resuits.

FED is similar to the N-gas approach but has some important differences. FED can include far more toxic gases but uses a simpler model to combine their effects. FED first measures the quantities of certain gases released by a burning product, then converts each measure into a fraction of the total "dose" required to kill someone, based on the substantial body of data already assembled on lethal levels of the major toxicant gases. This calcula

tion assumes that it doesn't matter how long it takes for the dose to be applied-- or, to put it mathematically, the effect of a certain value of concentration times the time of exposure is the same, no matter what the actual concentration and time are. This assumption is called Haber's Rule, and while it has limitations, it's valid across a wide range of toxicants and exposure times. FED also assumes that the major toxicants are all roughly additive. Thus, half a lethal dose of carbon monoxide plus half a lethal dose of hydrogen cyanide would be just enough to kill you, for example.

These two assumptions are critical to the compatibility between the FED approach and other elements of a fire hazard analysis model. Fire growth models can be assembled to generate a description of the toxic gas concentrations for every location in a building at every moment of a fire's development. Human behavior models can be used to generate a location for every occupant at every moment. FED's assumptions allow the toxic impact to be calculated as if every occupant were wearing a little dosimeter. At time t, occupant J picks up one second's worth of exposure to whatever toxic gas concentrations exist at his location L. One second later, at time t+1, occupant J has moved a step to location L', where he or she picks up one second's worth of exposure to toxic gas concentrations that

may be different because his or her location has changed, the fire is one second older, or both.

FED, then, calculates the total toxic effect on a person as the sum over the different toxicants of the fraction of a lethal dose received for each toxicant. Each toxic gas has an identified exposure, in parts per million concentration, that's just enough to kill if that gas is the only toxicant to which a person is exposed. If you get into the details, you'll see references to LC50. LC^sub 50^ is a toxic potency measure that refers to the lethal (L) dose of concentration (C) that in a specified time (usually 30 minutes) will kill half (50 percent) the laboratory animals. LC^sub 50^ values for a specific toxicant gas are calculated statistically from the results of experiments that killed animals through exposure to that one gas alone. A low LC^sub 50^ value is bad, because it means a small amount of gas is enough to kill.

Now that we've already done the experiments that killed animals and gave us LC^sub 50^ values for the major toxicant gases, N-gas and FED are two ways to develop an estimated LC^sub 50^ value for a whole product, whether new or existing, without killing any more animals.

Because the FED model is linear, therefore simpler, it's easier to use as an example. In an FED analysis, you burn the product for 30 minutes, measure how much gas is released for each major toxicant gas, and divide the quantity of each gas released by 30 minutes to give the average concentration (C) of

gas. Then you divide each of those average concentration values by the appropriate LC^sub 50^ to create a fractional effective dose value for each toxicant gas. And finally, you sum the FED values for the gases to produce a total FED value for the whole product.

An FED calculation capturing only carbon monoxide, hydrogen cyanide, and hydrogen chloride would look like this in equation form:

In this expression, C(CO) is the average concentration of carbon monoxide over 30 minutes.

N-gas and FED are ingenious constructions that spare animals, permit quicker and cheaper toxicity calculation, and facilitate the linkage of toxicity data with other fire data in performance-based fire hazard or risk calculations. But--and you knew there'd be a but--they rely on a number of simplifications and assumptions, all of which remain the subject of discussion, even dispute. What follows are a few notable examples.

First, there are literally dozens of known toxicant gases, and if you want

to capture, say, 95 or 99 percent of the toxic effect, you can't limit your model to a small number of gases.

Even if you use the FED approach, some gases known to be toxic don't have established LC^sub 50^ values. And even for those that do, there's some variability in the values experimenters have found and some variability across people in what it takes to kill. (Remember, LC^sub 50^ is a lethal concentration for 50 percent of the population for a particular toxic substance. It says nothing about how different lethal concentrations may be for the most and least vulnerable members of the population.) Thus, toxicity calculation is far from a cookbook operation. You have to know a great deal to set up the test and interpret the results properly.

On the positive side, the basic FED approach is so robust that you can use it for other toxic effects besides death, and you can even use it for heat effects and oxygen deprivation. In the latter case, the "concentration" is the gap between the normal oxygen concentration in ambient air, which is 20.9 percent, and the actual concentration, which will be lower--possibly much lower--due to fire effects. The N-gas approach can handle oxygen deprivation, too.

Second, the effects of some toxicants aren't just additive. The leading example is carbon dioxide, which can accelerate the breathing rate. In other words, the

simplified model used in the FED approach is known to be wrong for at least one major toxic gas. This is a powerful argument for the N-gas approach. Efforts have been made to modify the linear FED model just for carbon dioxide, and the standards provide modified FED formulas for consideration. But there is as yet no consensus on how best to make those modifications. The ISO standard uses a different formula from that used in the NFPA and ASTM standards. More important, there's no consensus on whether the FED model is accurate enough with just that one modification.

In the end, both approaches tend to need some animal testing for confirmation, if for no other reason than there's still the possibility of a hitherto unrecognized toxicant. The ISO standard allows, and the NFPA and ASTM standards require, experiments with animals to confirm the results. Experiments may be designed to burn just enough product to deliver a dose that is lower or higher than the dose calculated as the just-lethal dose delivered by the burning product. A lowerdose test should reveal any significant toxicants not captured in the calculation, and it may not kill any animals. A higher-dose test will kill animals unless the calculation has somehow shown the product to be much more toxic than it really is. NFPA and ASTM standards require a confirmato

ry bioassay, as these experiments are called, but it's optional in ISO because concerns over avoidable animal deaths have more political clout in Europe, which dominates ISO proceedings, than in the United States, home to NFPA and ASTM.

None of these concerns disqualify the use of models, and clearly, these recognized organizations don't regard them as reasons to delay the use of models. But they are further evidence that toxicity assessments must be made by people who are knowledgeable in combustion toxicology.

Where are the victims?

As I've said, most of those who die in fires in the United States as a result of toxicity aren't located in the room of fire origin. This is less true of burn victims (see Tables 3 and 4), and it isn't true in some other countries, such as the United Kingdom.

The principal reason for mentioning this fact again is that it underscores the need for additional large-scale test work and other research to give us a better understanding of how the mix of toxic gases changes as they move over multiroom distances. Hydrogen chloride is the most prominent, but not the only, toxicant known to deposit from the air to available surfaces during transport over these kinds of distances. This means that the quantity of hydrogen chloride released from the burning product, which can be readily measured in a small-scale test, may overstate the quantity of hydrogen chloride actually delivered to a typical victim.

What's more, transport over several meters provides plenty of time for chemical reactions among the airborne toxicants. This may be the least understood issue in combustion toxicity.

In the end, is it all just CO?

Nearly two decades ago, Walter Berl and Byron Halpin of Johns Hopkins University analyzed 463 fire deaths in Maryland and provided the first statistical evidence in the United States that toxicity might be almost entirely a matter of carbon monoxide.3 They found that:

* 48 percent of the fatalities they studied could be explained by carbon monoxide alone;

* 26 percent could be attributed to significant carbon monoxide in combination with one or more other toxicants or other factors, such as preexisting heart disease, that made victims more susceptible to the effects of carbon monoxide;

* 18 percent could be attributed to burns or other thermal injuries alone; and

* 8 percent involved either miscellaneous causes not related to thermal or toxic injuries, or unknown causes.

Berl and Halpin found no deaths that could be attributed solely to toxicants other than carbon monoxide, and only one-fourth of the deaths, at most, involved other toxicants, even as contributing factors. Their approach started with the hypothesis that every death should be evaluated first in terms of carbon monoxide poisoning. This may understate the contribution of other toxicants or other factors to death. In addition, they used highly simplified criteria to determine whether carbon monoxide was sufficient to kill, was a contributor to death but not sufficient by itself, or was not a factor. This requires some explanation.

One of the principal functions of blood is to bring fresh oxygen to the different parts of the body, especially the brain. It does this through a chemical process by which inhaled oxygen com

bines with the hemoglobin of the blood at the lungs and is then taken where it's needed and released by the hemoglobin to the other tissues. Unfortunately, given a choice, blood is hundreds of times more likely to combine with carbon monoxide than oxygen. That combination is called carboxyhemoglobin, and its relative presence in the blood means that the blood is carrying carbon monoxide, which is poisonous even if there's plenty of breathable oxygen available in the atmosphere, and isn't carrying needed oxygen.

Berl and Halpin used two simple thresholds: 30 percent carboxyhemoglobin for a significant carbon monoxide contribution to death and 50 percent carboxyhemoglobin as sufficient to cause death due to carbon monoxide alone. In fact, very fit people can survive higher levels, and people with pre-existing health prob

lems might die from lower levels. Furthermore, a brief intake of a very high concentration of carbon monoxide can cause cardiac arrest, leading to death with a very low level of carboxyhemoglobin.

A study by Marcelo Hirschler showed that people who die in situations known to involve only carbon monoxide--such as suicides or accidental victims of motor vehicle exhaust--display the same statistical distribution of levels of carboxyhemoglobin as do those who die in fires, once you adjust for age differences.4 This strongly suggests that, for nearly all fire victims killed by toxic gases, the carbon monoxide is enough to cause death.

But that's not the final word. Remember what we said earlier: A victim may be blocked by one fire effect, slowed by a second, incapacitated by a third, and killed by a fourth. That would be one death with one fire effect showing a

lethal level in the body, but four fire effects contributing to the death.

You can imagine such a sequence, but does that really happen often? One reason some combustion toxicologists continue to pursue this possibility is that there have been a number of full-scale room fire tests, using typical room furnishings, that showed hydrogen cyanide building to incapacitating levels well before carbon monoxide could have any effect. The sequence-of-fire-effects hypothesis is at least one way to reconcile the clear pattern of fire victim autopsy data, which point to carbon monoxide, with the findings of these fire tests, which paint a more varied picture.

At the same time, many of the fire tests that show early incapacitating lev

els of hydrogen cyanide also show early lethal levels of hydrogen cyanide, which isn't consistent with autopsy data. It isn't easy to create conditions, even in the laboratory, that directly demonstrate the plausibility of the sequence-of-effects hypothesis. So it may be, in the end, that these fire tests, however welldesigned and -run, simply do not represent real fatal fires in some important ways not yet identified.

After listening to a series of presentations on these myriad bodies of data and competing hypotheses, I found one thing, at least, was clear: There must, and will, be considerably more research on combustion toxicity before we have a fully satisfactory picture of all the key mechanisms and factors.

What's next

ASTM and ISO are both starting to develop standards to choose the fire scenarios, set the test conditions most relevant to real hazards, and interpret the data. NFPA, which developed its standard in a technical committee devoted to fire tests rather than combustion toxicity, as ASTM and ISO did, already addressed some of those issues in the test standard by requiring the analytical adjustment of carbon monoxide levels. More elaborate provisions might be the province of NFPA's performance-based code initiative.

If we now know how to measure toxic effects and analyze and evaluate their transport and impact, the time may be ripe for more detailed work on modeling the production of those effects in the first place. Some exciting work at the National Institute of Standards and Technology has identified and quantified the four mechanisms by which carbon monoxide is principally generated, for example.

From a strictly scientific point of view, the past decade's progress in this highly specialized field of fire safety sciences has been most remarkable even more so

as it's taken place without the intense public attention that existed a decade ago. And if the answers have implied less leverage on effective fire safety strategies than was originally anticipated, that doesn't minimize the value of having this part of the problem so much better understood--and of having tools that can identify unusual toxic hazards at the design stage, should they appear.

Whatever happened to combustion toxicity? It changed from a field full of fundamental questions and hotly contested, widely divergent points of view to a solid component of our new scientific approach to fire safety and design. Decades of work by dedicated, creative pioneers have produced consensus on what science requires. Congratulations to all who brought

us to this major milestone.

1. Donna M. Slayton and Alison L. Miller, Patterns of Fire Casualties in Home Fires by Age and Sex, Quincy, Mass.: NFPA Fire Analysis and Research Division, September 199a. 2. Richard G. Gann, Vytenis Babrauskas, Richard D. Peacock, and John R. Hall, Jr., "Fire Conditions for Smoke Toxicity Measurement," Fire and Materials, Vol. 18 (1994), pp. 193-199. 3. Walter G. Berl and Byron M. Halpin, Human Fatalities from Unwanted Fires, Baltimore, Md: Johns Hopkins University, Applied Physics Laboratory, Fire Problems Program, APL/JHW FPP TR 37, December 1978. 4. Marcelo M. Hirschler, "Carbon Monoxide and the Toxicity of Fire Smoke," Carbon Monoxide and Human Lethality, ed. Marcelo M. Hirschler, London: Elsevier Science Publishers, 1993.

John R. Hall, Jr., is assistant vice president of Fire Research and Analysis at NFPA.

Copyright National Fire Protection Association Nov/Dec 1996
Provided by ProQuest Information and Learning Company. All rights Reserved