Exposure Guidelines for Residential Indoor Air Quality

3.0 Derivation of Guidelines and Recommendations

3.1 Data Base for Derivation of Exposure Guidelines

Since research into the health effects of residential indoor air quality is at an early stage, there is a dearth of reliable information on the health effects that result from exposure to the low levels and mixtures of contaminants likely to be found. In most cases, therefore, the Working Group relied upon the results of laboratory experiments using animals, clinical studies with human volunteers, and epidemiological investigations of urban air pollution and the occupational environment. The results of epidemiological and clinical studies are the most relevant for establishing acceptable levels of exposure of humans to air pollutants. Nevertheless, the application of each of these types of study involves a number of assumptions and hence uncertainty in the derived dose-response relationships.

3.1.1 Epidemiological Studies

Most of the relevant epidemiological studies of populations are observational (non-experimental) in nature; that is, the allocation of individuals into study groups on the basis of exposure is not under the control of the investigator. Such observational studies can be further classified as descriptive (cross-sectional) or analytical (cohort or case-control) studies. In cohort studies, exposure and outcome are monitored over time; as a result, the quality of evidence obtained from such longitudinal investigations is generally considered to be superior to that from cross-sectional studies, in which populations are monitored at one time. However, the results of all observational studies must be evaluated against the following features of study design:

(a) Estimation of Exposure. In most observational studies conducted to date, pollution data are usually obtained from one or several outdoor monitoring stations; however, the exposure burden can vary greatly among individuals living in the same neighbourhood because of local climatic conditions and special features of the indoor environment. For example, for pollutants generated mainly in the outdoor environment (oxidants, sulphur dioxide), indoor concentrations will normally be less than those outdoors; however, indoor levels (from indoor sources) may have greater temporal peaks with concomitant effects on health. Exposures to some pollutants (nitrogen dioxide, carbon monoxide) are not well represented by ambient air measurements if there are significant indoor sources of these pollutants. Only in very recent studies have investigators attempted to take such factors into account in estimating exposure.

(b) Role of Confounding Variables. In observational studies of populations exposed to air pollutants, a host of confounding variables (e.g., socioeconomic status, smoking, occupational exposure, meteorological factors), many of which have greater effects than air pollution, must be considered.

(c) Measurement of Outcome. There is substantial variation in the method of measurement of many health-state indicators in the studies conducted to date; such indicators include lung function, hospital admissions and frequency of symptoms. In many of the studies, outcome is ascertained by questionnaire, and responses may be biased by the way and conditions under which the questions are asked.

Even in those studies where the design is acceptable, interpretation of the results is complicated. For example, it is often difficult to attribute effects observed in populations exposed to air pollution to a single contaminant. It is possible that the contaminant investigated may serve as an indicator, or surrogate, for the effects of another contaminant or a combination of contaminants. Also, since exposures are not subject to the manipulation of the investigator, it is difficult to determine whether mean or peak concentrations, variability or some other aspect of air pollution is the most important determinant of health effects.

In summary, those epidemiological studies considered most relevant for developing the guidelines have the following features:

• longitudinal study design;
• adequate control of appropriate confounding factors;
• some attempt to take individual variations in exposure into account (e.g., area of residence, gas stove use).

Epidemiological investigations of the effects in the general population are considered to be the most relevant. Studies of persons exposed to airborne pollutants in the workplace may not reflect potential problems of the general population, since the young, the elderly and other high-risk groups are not accounted for. Moreover, exposure periods and the mixture of pollutants will be different from those in the home.

3.1.2 Clinical Studies

Clinical studies are generally, though not always, conducted in controlled laboratory environments. These studies probably provide the most reliable data from which to derive exposure-response relationships that form the basis for air quality standards. However, clinical studies are restricted for ethical reasons to the examination of mild, temporary effects of short-term exposures in a limited number of subjects. As such, they are most suitable for developing short-term exposure limits.

A good clinical study should control for extraneous variables and experimental bias. In order to satisfy this need, most clinical studies employ a control group. Comparison with this group provides an indication of the effects that the experimental conditions exert on exposed individuals. The inclusion of such a group is very important if the effect of the experimental intervention is to be isolated.

In order to further reduce experimental bias in clinical studies, the allocation of subjects into experimental and control groups should be a random process. That is, all subjects should have an equal chance to be assigned to either the experimental group or the control group. Through this process the study will have a good chance of obtaining comparable groups, such that the measured and unknown characteristics of the subjects at the time of group allocation will be, on the average, evenly balanced between groups. Also, since most statistical procedures are based on normal distributions, the randomization procedure is necessary in order to meet the assumptions of the statistical design.

In order to reduce experimental bias to a minimum, some studies employ designs in which the subjects, and sometimes the investigators, are not aware of the particular intervention to which a subject has been assigned. The advantage of these "blind" designs is that they reduce the possibility that the subjects or investigators will favour certain outcomes based on group status.

Only through the employment of blind designs can the investigator be reasonably assured that extraneous variables and experimental bias have been controlled as much as possible. However, it is sometimes impossible to employ such a design for practical reasons. If, for example, the investigators are testing the pulmonary effects of ozone, and the subjects and investigators can detect the treatment condition by the smell of ozone, it is clearly not practical to attempt a blind study. This is a limitation of many clinical studies of airborne pollutants.

3.1.3 Animal Studies

Although numerous studies of the effect of airborne pollutants in animal species have been conducted, levels of exposure have, in general, been much higher than those in ambient air. In addition, extrapolation of the results to ambient levels is complicated by the distinct anatomical differences between the respiratory tracts of animals and man. Also, studies are frequently confined to unusually high concentrations of no more than one or two pollutants, rather than to the low concentrations and mixtures of substances generally found in the home. However, the results of such studies are useful in the identification of target organs and systems, in the clarification of mechanisms of toxicity and in the assessment of carcinogenicity.

The reliability of carcinogenesis bioassays in animal species is evaluated on the basis of several features of the design and the results of the study. These features include the size of the experiment (i.e., the numbers of exposed and control animals); the influence of environmental factors (e.g., diet); the route and method of exposure; the doses administered; the species, strain and sex of the animals; the types, site, incidence and time for the development of tumours; and the nature of the exposure-response relationship. Information concerning the kinetics, metabolism and mechanism of action and the results of epidemiological studies in human populations are also considered when the relevance of the results of carcinogenesis bioassays for man is assessed.

3.2 Approach Used in Deriving Exposure Guidelines

In order to establish accurate, defensible exposure guidelines it is essential to determine the quantitative relation between a given pollutant and its effects. The terms "exposure-response" and "exposure-effect" denote this quantitative relationship. Owing to ethical considerations, such quantitative relationships are difficult to determine with precision in human populations. Nevertheless, clinical and epidemiological studies combined with laboratory animal studies can provide a substantial amount of quantitative information concerning the effects of exposure to a given pollutant.

3.2.1 Non-carcinogenic Substances

Regulatory agencies have traditionally attempted to determine a level of exposure below which there are no apparent detrimental effects. This so-called "threshold level" is closely related to the lowest level at which minimal, or reversible, effects can be observed the "lowest-observable-adverse-effect level" (LOAEL). A safety factor may be incorporated into the derivation of a regulatory standard or guideline depending upon the number and quality of studies upon which the LOAEL is based. This approach has been used to establish guidelines for a number of indoor air pollutants considered in this document.

The size of the safety factor depends to a large degree on whether human rather than animal data are available, whether studies have been conducted directly on those segments of the population believed to be at high risk and the quality of the studies themselves. Ultimately the choice is based on a consensus decision by experts, but strictly has no scientifically defensible basis.

Owing to uncertainty concerning data obtained in observational studies, the World Health Organization has used a safety factor of two in recommending guidelines for daily and annual exposure to air pollutants; this value has been adopted in deriving some of the guidelines specified in this document.

In instances where there are sufficient data from reliable clinical studies of transient changes in groups at risk (for example, changes in pulmonary function in exercising asthmatics), no safety factor is incorporated in the derivation of a short-term exposure guideline.

Because of the wide variation in individual susceptibility to irritants, notably aldehydes, short-term exposure guidelines have been derived by applying a factor of five to the lowest value reported to cause a significant increase in symptoms of irritation.

It has been suggested that occupational hygiene limits could be adapted for the residential indoor environment by applying a safety factor to accommodate differences such as exposure times, pollutant mixtures and population sensitivities. Without a thorough knowledge of the scientific basis for the occupational limits, however, the Working Group considered such an approach to be scientifically indefensible.

3.2.2 Carcinogenic Substances

There is evidence that for most carcinogenic substances threshold levels may not exist. It follows in such cases that there is no level of exposure at which a hazard does not exist, although at very low concentrations the health risks may be so small as to be undetectable. For most carcinogens, the derivation of acceptable exposure limits using experimentally derived LOAELs and safety factors is considered inappropriate.

Ideally, exposure to known or suspected carcinogens should be avoided. However, it is not possible to eliminate certain carcinogens from the environment. The maximum concentrations listed in this document for formaldehyde, a possible human carcinogen, and radon, a documented human carcinogen, are the lowest levels that are practical to achieve and at which there appears to be no undue public health risk.

To ascertain the level of risk or probability of adverse response at the low concentrations likely to be encountered in the environment, several statistical procedures have been developed to predict the shape of the dose-response curve at doses below those administered in experimental studies. The shape of the extrapolated dose-response curve can influence considerably the value of the exposure concentrations at which there appears to be a negligible human health risk, with estimates sometimes ranging over several orders of magnitude, depending on the mathematical model employed.

The long-term exposure guideline estimated for formaldehyde was based on mathematical extrapolation of tumour incidence in an animal bioassay, taking into account as much as possible information concerning the mechanisms of action. Nevertheless, it must be recognized that the calculated risk at low levels of exposure is probably overestimated owing to the conservative assumptions upon which the mathematical model is based.

Estimated cancer risks associated with exposure to low levels of radon were based on extrapolation of data on cancer mortality observed in workers exposed occupationally to much higher airborne concentrations. It should be recognized, however, that there are several sources of uncertainty in such estimates, which can serve only as very rough approximations.

3.3 Monitoring Procedures

Methods for monitoring indoor air quality have not yet been standardized. Many studies have been carried out with a combination of older methods and instruments (used also for ambient air monitoring) together with those that have only recently been developed. There is now a trend toward developing smaller portable instruments that can serve both as personal monitors and as stationary (area) monitors. Passive and active personal monitors that measure particulate matter (filters) and gases (absorbers and adsorbers) have been developed. The small size and quiet operation of this newer equipment render it suitable for use in indoor environments. Such monitors also offer the possibility of determining pollutant levels near the breathing zone of individuals, so that direct estimates of personal exposure may be made.

Monitoring to ascertain compliance with the short-term exposure guidelines should be conducted when "worst case" situations are anticipated, and procedures should be designed to provide an accurate assessment of occupants' actual exposure. It is recommended that samples be taken when and where maximum concentrations are expected and that the averaging times specified along with the ASTER be used.

In the case of the long-term exposure it is more difficult to specify appropriate monitoring procedures. Sampling should be conducted over periods long enough to encompass any diurnal, seasonal or other temporal fluctuations. The magnitude and frequency of these fluctuations may vary considerably from contaminant to contaminant and be a function of dwelling type, location and the activity of the occupants. In many cases annual averages have served as a basis for formulating long-term air quality standards. Frequently, annual averages are based on random sampling of 24-hour average readings. Monthly or weekly averages are typically determined through daily averages, although passive monitors now available may have collection periods of from 7 to 90 days.

It is recommended that a minimum sampling period of 24 hours be used and that samples be taken when concentrations are expected to be at their highest. Should concentrations be outside the specified long-term exposure range, additional samples should be taken to ascertain whether levels are likely to remain elevated, or whether fluctuations will result in the long-term (often annual) average concentrations falling within the guideline value.

If contaminant levels are found to be outside any of the recommended exposure ranges, the source of the problem should be identified. All possible corrective measures should be taken or advice sought from health authorities.