In some cases, the simple answer does not suffice. Chemists, regulatory officials, etc. want to know which additional gases may contribute a non-negligible amount to fire toxicity. No unique answer is possible, since the circumstances of each fire are different. Certainly, different materials produce different combustion products; for example, a polymer with no chlorine atoms would not produce hydrogen chloride among its products. Having said this, valid reasons could be seen for wanting a "shopping list" of gaseous agents. Thus, I created such a shopping list. This was published in the first edition of NIST's fire model HAZARD I. While such a list will always be subject to debate, the advantage of the NIST one was that it had to undergo the very rigorous review to which NIST publications are subjected. Thus, instead of being just one scientist's opinion, it had the consensus of the fire toxicity research and the fire modeling groups and the informed consent of higher management.

In some cases, human data are available from industrial accidents and such. In most cases, it is necessary to use animals as surrogates. These must be selected so that the biological system which is affected by a particular toxin is one which functions relatively similarly in man. For example, hydrogen chloride inhalation data from mice have generally been considered to be a poorer prediction of human fatality than data from rats, due to respiratory system differences. In all cases, of course, the numbers are only estimates and can never be actually confirmed on humans.

The time of exposure is also important for determining the effects from toxic gases. In general, a higher concentration allows only a shorter time, for the same biologic effect to be reached. If the concentration is exactly hyperbolic, that is, (concentration)x(time)=constant, then this gas is said to follow Haber's Law. Almost any gas shows deviations from this simple relationship, but for rough estimating purposes it can be used if better data are unavailable. For fire toxicity data, the exposure period normally used is 30 minutes.

Users often ask "Why do you compile data for lethality instead of data for incapacitation?" This is a legitimate question, since usually we would like to know for how long a person will be all right in fire, not just when the morgue has to be called. The reason why such a compilation was not developed by NIST is a practical one. It is usually possible to obtain reliable measurements of animal lethalities. By contrast, the concept of incapacitation does not even have a workable definition. The conditions of incapacitation depend greatly on the task which is to be done. Incapacitation can also be affected by alcohol intoxication, and by various medical conditions; these have not been adequately studied in combination to derive usable guidance. The NIST recommendation to users was derive a hypothetical incapacitation level (EC₅₀) by reducing the LC₅₀ values by a suitable margin, say 2- to 4-fold.

As a small sidelight, while the values of EC₅₀ will be expected normally to be several-fold lower than the corresponding LC₅₀ values, occasionally the opposite is seen. During the early 1980s when the so-called USF toxicity test was being evaluated, many strongly lethal gas mixtures were documented to have unexpected large EC₅₀'s (the test used mice and measured EC₅₀ levels only). These turned out not to be measurement errors, but systematic problem formulation errors. The animals which showed the surprisingly high EC₅₀ values were, in fact, the "walking dead." They were able to perform the tasks required of them in the EC₅₀ testing procedure while they were already mortally injured.

By inspecting the list, one can readily see that CO is not greatly toxic and that there are many more gases found in fires which are notably more toxic. The reason why CO is considered the primary agent is simply due to its copious generation by all fires. The importance of any toxic gas species to a particular fire must reflect both its toxicity and its actual concentration in that particular fire. This is the fractional effective dose concept and algorithms are available for using it. Such data cannot be tabulated in a handbook fashion, however, since the concentrations are completely specific to the fire in question.

It may be noted from the references that most of the data are from the 1980s and earlier. This can largely be explained by the fact that major fire toxicity research groups at NIST, the Southwest Research Institute, Hungtingdon, and other institutes were closed or dispersed around the start of the 1990s. This was largely due to 'a job well-done,' but budgetary considerations were also an issue.

Burn and thermal exposure injuries have been extensively studied, but there is little data on combinations of thermal and toxic gas injuries. Some arithmetic scaling laws have been proposed, but there is no satisfactory validation of them. At the moment, the situation is viewed non-interactive. In fire modeling, it is common to compute thermal and toxic exposures and to declare as governing the first one where limit values are reached.