Priority Substances List Assessment Report for Formaldehyde

3.0 Assessment of "Toxic" under CEPA 1999

3.1 CEPA 1999 64(a): Environment

The environmental risk assessment of a PSL substance is based on the procedures outlined in Environment Canada (1997a). Analysis of exposure pathways and subsequent identification of sensitive receptors are used to select environmental assessment endpoints (e.g., adverse reproductive effects on sensitive fish species in a community). For each endpoint, a conservative Estimated Exposure Value (EEV) is selected and an Estimated No-Effects Value (ENEV) is determined by dividing a Critical Toxicity Value (CTV) by an application factor. A conservative (or hyperconservative) quotient (EEV/ENEV) is calculated for each of the assessment endpoints in order to determine whether there is potential ecological risk in Canada. If these quotients are less than one, it can be concluded that the substance poses no significant risk to the environment, and the risk assessment is completed. If, however, the quotient is greater than one for a particular assessment endpoint, then the risk assessment for that endpoint proceeds to an analysis where more realistic assumptions are used and the probability and magnitude of effects are considered. This latter approach involves a more thorough consideration of sources of variability and uncertainty in the risk analysis.

3.1.1 Assessment endpoints

Formaldehyde enters the Canadian environment mainly from natural and anthropogenic combustion sources, from industrial on-site releases, from off-gassing of formaldehyde products, and through secondary formation as a result of oxidation of anthropogenic and natural organic compounds in air. Almost all releases and formation in the ambient environment are in air, with small amounts released to water.

Given its physical-chemical properties, formaldehyde is degraded by various processes in air, with very small amounts transferring into water. When released to water or soil, formaldehyde is expected to remain primarily in the original compartment of release, where it undergoes various biological and physical degradation processes. Formaldehyde is not bioaccumulative or persistent in any compartment of the environment.

Based on the sources and fate of formaldehyde in the ambient environment, biota are expected to be exposed to formaldehyde primarily in air and, to a lesser extent, in water. Little exposure of soil or benthic organisms is expected. While formaldehyde occurs naturally in plants and animals, it is readily metabolized and does not bioaccumulate in organisms. Therefore, the focus of the environmental risk characterization will be on terrestrial and aquatic organisms exposed directly to ambient formaldehyde in air and water.

3.1.1.1 Terrestrial

Data on terrestrial toxicity are available for a variety of microorganisms, plants and invertebrates (Section 2.4.1.1), as well as from mammalian toxicology studies (Section 2.4.3). The most sensitive identified endpoints include primarily effects on the growth and development of plants (Haagen-Smit et al., 1952; Barker and Shimabuku, 1992; Mutters et al., 1993).

Bacteria and fungi are ubiquitous in terrestrial ecosystems and, as saprophytes, are essential for nutrient cycling. Terrestrial plants are primary producers, provide food and cover for animals, and provide soil cover to reduce erosion and moisture loss. Invertebrates are an important component of the terrestrial ecosystem, consuming both plant and animal matter while serving as forage for other animals. Vertebrate wildlife are key consumers in most terrestrial ecosystems.

Therefore, although limited, the available toxicity studies cover an array of organisms from different taxa and ecological niches and are considered adequate for an assessment of risks to terrestrial biota. The single most sensitive response for all of these endpoints will be used as the CTV for the risk characterization for terrestrial effects.

3.1.1.2 Aquatic

Aquatic toxicity data are available for a variety of algae, microorganisms, invertebrates, fish and amphibians (Section 2.4.1.2). Identified sensitive endpoints include effects on the development and survival of algae and invertebrates (Bills et al., 1977; Bringmann and Kühn, 1980a; Burridge et al., 1995a,b), inhibition of cell multiplication in protozoa (Bringmann and Kühn, 1980a), immobilization of crustaceans (Bills et al., 1977) and mortality in fish (Reardon and Harrell, 1990).

Algae are primary producers in aquatic systems, forming the base of the aquatic food chain, while zooplankton, including protozoans and crustaceans, are consumed by many species of invertebrates and vertebrates. Fish are consumers in aquatic communities and themselves feed piscivorous fish, birds and mammals.

Therefore, although limited, the available studies cover an array of organisms from different taxa and ecological niches and are considered adequate for an assessment of risks to aquatic biota. The response for all of these endpoints that occurs at lowest concentration is the CTV for the risk characterization for aquatic effects.

3.1.2 Environmental risk characterization

3.1.2.1 Terrestrial organisms

Environmental exposure to formaldehyde in air is expected to be greatest near sites of continuous release or formation of formaldehyde, namely in urban centres and near industrial facilities releasing formaldehyde. Extensive recent data for concentrations in air are available for 27 sites, covering a range of industrial, urban, suburban, rural and remote locations in Canada.

3.1.2.1.1 Hyperconservative analysis

The highest reported concentration of formaldehyde in ambient air in Canada is 27.5 µg/m³. This value was obtained for a 24-hour urban sample collected in Toronto, Ontario, on August 8, 1995. The mean concentration for six measurements made at this site during a 30-day period encompassing this date (July 14 to August 12, 1995) is 22.15 µg/m³. This mean concentration will be used as the EEV in the hyperconservative analysis of the chronic exposure scenario for terrestrial organisms. A 1-month mean is selected for the EEV because it corresponds to a longer exposure period relative to the life span of test organisms for which data are available.

For the exposure of terrestrial organisms to formaldehyde in air, the CTV is 18 µg/m³, based on the corresponding amount in fog (9000 µg/L) that affects the growth and reproduction potential of rapeseed (Brassica rapa) exposed 4.5 hours per night, 3 nights per week, for 40 days (Barker and Shimabuku, 1992). This value is the lowest from a moderate data set composed of acute and chronic toxicity studies conducted on at least 18 species of terrestrial plants, microorganisms, invertebrates and mammals exposed to air and/or fog water.

The 40-day intermittent exposure of Brassica rapa can be considered as chronic exposure (covering a significant portion of a life stage of the organism). For the hyperconservative analysis, the ENEV for terrestrial organisms is derived by dividing the CTV by a factor of 10. This factor accounts for the uncertainty surrounding the conversion of the effect concentration to a no-effect value, the extrapolation from laboratory to field conditions, and interspecies and intraspecies variations in sensitivity. As a result, the ENEV is 1.8 µg/m³.

The hyperconservative quotient is calculated by div iding the EEV by the ENEV as follows:

Since the hyperconservative quotient is more than 1, there is a need to proceed to a more realistic analysis of whether formaldehyde emissions cause adverse effects on terrestrial organisms in Canada.

3.1.2.1.2 Conservative analysis

For a conservative analysis, a more realistic estimate of long-term terrestrial exposure would be the highest of 90th percentile values calculated for each monitored site. A highest 90th percentile value is still representative of high-end concentrations at the site of greatest concern, yet it also excludes unusually high measurements, some of which may have been caused by rare ambient conditions or undetected analytical error. Analysis of the abundant data available shows that only once in the last 10 years were such high air concentrations measured in Canada for as long a period (1 month) as that from which the mean was selected for the hyperconservative EEV. Based on these data, the highest 90th percentile value is 7.48 µg/m³, calculated from 354 measurements made in Toronto, Ontario, between December 6, 1989 and December 18, 1997. This value will be used as the EEV for the conservative analysis of the exposure scenario for terrestrial organisms. For comparison, the 90th percentile value calculated for all 3842 NAPS measurements available between 1997 and 1998 is 5.50 µg/m³. The overall mean and median are 2.95 and 2.45 µg/m³, respectively.

For a conservative analysis, a more realistic ENEV could be calculated by dividing the hyperconservative CTV of 18 µg/m³ (rapeseed) by a more refined application factor. According to Fletcher et al. (1990), there is remarkable agreement between field and laboratory EC₅₀ values for plant species. In a study of sensitivity to pesticides in a wide range of plants, only 3 of 20 field EC₅₀ values were 2- fold higher than laboratory EC₅₀ values, and only 3 of 20 laboratory EC₅₀ values were 2-fold higher than field EC₅₀ values. Therefore, no application factor may be necessary for laboratory to field extrapolations for plant effects. Furthermore, data indicated that extrapolations among plant species within a genus can be confidently made without uncertainty factors. When extrapolating from one genus to another within a family, an uncertainty factor of 2 captured 80% of the potential variability. Extrapolations across families within an order or across orders within a class should be discouraged, but, if necessary, factors of 15 and 300 should be used for intraorder and intraclass extrapolations, respectively, to capture 80% of the variability (Chapman et al., 1998). In the case of the Barker and Shimabuku (1992) study from which the CTV was selected, the four test species consisted of a deciduous tree (aspen), a coniferous tree (slash pine), a grain crop (wheat) and a seed crop (rapeseed), representing diverse growth forms and morphology from four orders and two classes (monocots and dicots). In two of these, there were no adverse effects at test concentrations, while in a third species (slash pine), there was an arguably adverse increase in top growth at the lowest concentration. Other studies indicate that other acute and chronic effects begin to occur only at airborne concentrations clearly higher than for the rapeseed in fog, even in developmental stages (e.g., lily pollen LOEC of 440 µg/m³). The rapeseed seedling therefore appears to be by far the most sensitive of a variety of species tested. Given the diversity of the data set, only a minimal application factor may be required for interspecies extrapolation. Regarding the extrapolation from effect concentration to no-effect concentration, it should be noted that Barker and Shimabuku (1992) used a relatively low threshold of statistical significance (a = 0.1), and effects on the rapeseed did not include any of the visual symptoms such as necrosis observed in other liquid- and gas-phase formaldehyde studies. This may therefore allow for a smaller application factor to be used on the CTV for rapeseed. Therefore, keeping a CTV of 18 µg/m³, the application factor of 10 used in the hyperconservative scenario can be reduced to 2 for the conservative assessment. As a result, the ENEV for the conservative analysis of the exposure scenario for terrestrial organisms will be 9 µg/m³.

The conservative quotient is calculated by dividing the EEV by the ENEV as follows:

Alternatively, for a conservative analysis, it may also be more realistic to use a CTV from a toxicity study involving exposure to formaldehyde in gas phase in air rather than back-calculating from exposure in fog. Reasons to do this include the exploratory nature of the fog study (Barker and Shimabuku, 1992) from which the hyperconservative CTV was selected. The conversion of fog water concentrations to expected air concentrations in the study could not be verified because variables (temperature, vapour pressure, water solubility, Henry's law constant) required for the conversion were not specified in the study. Reported exposure concentrations represented an estimated average based on the observed rate of degradation in the experimental system. Since formaldehyde in the fog water readily undergoes hydration and degradation, it is not certain how its properties may change its toxicity. Analysis of the terrestrial data set available indicates no other reports of studies on effects of fog or effects as sensitive as those in Barker and Shimabuku (1992). In addition, no data were found on concentrations of formaldehyde in fog in Canada or frequency of fog incidence in urban areas to be able to support an assumption that Canadian biota are being exposed to formaldehyde under such conditions as those used in the experiment. Also, the study did not seem to take into consideration potential exposure to gas-phase formaldehyde in between exposures to formaldehyde in fog. A study of chronic exposure to formaldehyde in gas phase in air may be more realistic.

For the conservative analysis of the exposure of terrestrial organisms to formaldehyde in air, the CTV is 78 µg/m³, based on the lowest average concentration in air that caused a slight imbalance in the growth of shoots and roots in the common bean (Phaseolus vulgaris) exposed for 7 hours per day, 3 days per week, for 4 weeks in air (day: 25°C, 40% humidity; night: 14°C, 60% humidity) (Mutters et al., 1993). This value was selected as the most sensitive endpoint from a moderate data set composed of acute and chronic toxicity studies conducted on at least 18 species of terrestrial plants, microorganisms, invertebrates and mammals exposed to air and/or fog water.

The 28-day intermittent exposure of the bean plant can be considered as chronic exposure (covering a significant portion of a life stage of the organism). Dividing the CTV by a factor of 10 to account for the unce rt ai nt y surrounding the conversion of the effect concentration to a no-effect value, the extrapolation from laboratory to field conditions, and interspecies and intraspecies variations in sensitivity, the resulting ENEV is 7.8 µg/m³. This yields the following conservative quotient:

This quotient is very close to one.

Given the arguments for reducing the application factor of the hyperconservative rapeseed CTV and the even milder effects observed for the common bean plant (Mutters et al. [1993] themselves did not conclude any ill effects from formaldehyde), the application factor can be reduced from 10 to 2 for a more realistic ENEV of 39 µg/m³. This results in a lower conservative quotient:

Since all three conservative quotients are less than 1, it is unlikely that formaldehyde in air causes adverse effects on terrestrial organisms in Canada.

In considering a weight-of-evidence approach, other data similarly do not indicate the likelihood of high risks associated with atmospheric exposure. It is uncertain what the potential ecological impacts could be for sensitive effects such as imbalance in growth of roots and shoots. Based on the toxicity data set available, it appears that plants are most sensitive during their early life stages. In Canada, sensitive early life stages of plants usually occur in the spring. Highest air concentrations of formaldehyde have generally been measured in late summer (August) (Environment Canada, 1999a), when atmospheric formaldehyde formation and photochemical smog formation are greatest. It would therefore appear that only the more tolerant adult plants would be exposed to the highest concentrations. In addition, in studies other than those used in the hyperconservative and conservative scenarios above, there has been considerably more tolerance to exposure to formaldehyde (e.g., no injury at concentrations below 840 µg/m³ for alfalfa; Haagen-Smit et al., 1952), with no effects on plants at concentrations of 44 mg/m³ (Wolverton et al., 1984).

A summary of the values used in the environmental risk analysis of formaldehyde in the terrestrial environment is presented in Table 7.

Table 7 Summary of the environmental risk analysis for terrestrial organisms

Terrestrial organisms	EEV (µg/m3)	CTV (µg/m3)	Application factor	ENEV (µg/m3)	Quotient
Tier 1: Hyperconservative
Highest urban mean, rapeseed in fog	22.15	18	10	1.8	12.3
Tier 2: Conservative
Highest 90th percentile, rapeseed in fog	7.48	18	2	9	0.83
Highest 90th percentile, bean plant in air	7.48	78	10	7.8	0.96
Highest 90th percentile, bean plant in air	7.48	78	2	39	0.19

3.1.2.2 Aquatic organisms

Environmental exposure to formaldehyde in water is expected to be greatest near areas of high atmospheric concentrations (where some formaldehyde can partition from air into water) and near spills or effluent outfalls. Measured concentrations are available in Canada for surface waters, effluents and groundwater. For surface water, data are available on limited sampling at four drinking water treatment plants in urban areas of Ontario and Alberta. Measured concentrations in effluent are available for one of the four industrial plants reporting releases of formaldehyde to water. Groundwater data are available for three industrial sites associated with spills or chronic contamination and six cemeteries in Ontario.

3.1.2.2.1 Hyperconservative analysis

A hyperconservative analysis has been conducted for aquatic organisms exposed to concentrations measured in surface water, effluents and groundwater.

The highest concentration of formaldehyde reported in surface water is 9.0 µg/L, obtained for a sample collected from the North Saskatchewan River near a treatment plant in Edmonton, Alberta (Huck et al., 1990). The highest 1-day concentration identified in an industrial effluent was 325 µg/L (Environment Canada, 1999a). In various groundwater samples, the highest concentration of formaldehyde was 690 000 µg/L at an industrial site (Environment Canada, 1997c). These values will be used as the EEVs in the hyperconservative analysis of aquatic organisms in surface water, effluent and groundwater, respectively. The effluent EEV is based on the conservative assumption that organisms could be living at the point of discharge. The groundwater EEV is based on the conservative assumption that the groundwater could recharge directly to surface water at its full concentration.

For exposure of aquatic animals to formaldehyde in water, the CTV is 100 µg/L, based on the concentration that causes 40-50% mortality after 96 hours in day-old zygotes of the marine alga, Phyllospora comosa (Burridge et al., 1995a). This value was selected as the most sensitive endpoint from a large data set composed of toxicity studies conducted on at least 36 species of freshwater and marine aquatic algae, microorganisms, inve rteb rate s, f ish an d amphibians.

The 96-hour exposure for Phyllospora comosa zygotes can be considered as chronic exposure (covering a significant portion of the lifetime of the organism). For a hyperconservative analysis, the ENEV is derived by dividing the CTV by a factor of 10. This factor accounts for the uncerta inty in the extrapolation from a chronic EC₅₀ to a chronic no-effects value, the extrapolation from laboratory to field conditions, and interspecies and intraspecies variations in sensitivity. The resulting ENEV is 10 µg/L.

The hyperconservative quotients are calculated by dividing the EEV by the ENEV as follows:

Surface water analysis

Since the hyperconservative quotient is less than 1, it is unlikely that formaldehyde causes adverse effects on aquatic organisms in ambient surface water in Canada, and more realistic exposure scenarios need not be considered.

Effluent analysis

Since the hyperconservative quotient is greater than 1, it is necessary to consider further the likelihood of biota being exposed to such concentrations in surface water near point sources in Canada.

Groundwater analysis

Since the hyperconservative quotient is greater than 1, it is necessary to consider further the likelihood of biota being exposed to such concentrations in Canada.

3.1.2.2.2 Conservative analysis

For a conservative analysis, more realistic estimates of aquatic exposure must be used. In the case of effluent, dilution can be considered. For a conservative analysis, the hyperconservative EEV of 325 µg/L can be divided by a generic and conservative dilution factor of 10 derived for all types of water bodies to estimate ambient concentrations of formaldehyde near outfalls. This results in a conservative effluent EEV of 32.5 µg/L.

In the case of groundwater, the very high concentrations at one contaminated site were related to a recognized historical contamination

that has since been contained and remediated (Environment Canada, 1999a). The next highest concentration reported for groundwater was for an industrial site in New Brunswick (maximum of 8200 µg/L). It is highly unlikely that the groundwater at a single sampling station would recharge directly to surface water. A more realistic representation of groundwater quality at the site could be achieved using the median concentration in groundwater at all sampling stations. The median was 100 µg/L for measurements taken at five wells at the contaminated site during 1996-1997. Assuming some degree of dilution similar to that of effluent in receiving water bodies, the median value can also be divided by the generic and conservative dilution factor of 10 to obtain a conservative estimate of possible concentrations in the event of surface recharge. As a result, the conservative EEV for groundwater is 10 µg/L.

For a conservative analysis, an endpoint should be selected that is more appropriate than that for the CTV used in the hyperconservative analysis, which was based on toxicity to a marine alga endemic to Australia. A more meaningful value can be derived by considering toxicity to the seed shrimp, Cypridopsis sp., a common freshwater ostracod, yielding a CTV of 360 µg/L, based on the 96-hour EC₅₀ (immobility) for this organism (Bills et al., 1977). This value was selected as the most sensitive endpoint from a large data set composed of toxicity studies conducted on at least 34 freshwater species of aquatic algae, microorganisms, invertebrates, fish and amphibians. For the conservative analysis, the ENEV is derived by dividing the CTV by a factor of 10. This factor accounts for the uncertainty surrounding the extrapolation from the EC₅₀ to a chronic no-effects value, the extrapolation from laboratory to field conditions, and interspecies and intraspecies variations in sensitivity. The resulting ENEV is 36 µg/L.

The conservative quotients are calculated by dividing the EEV by the ENEV as follows:

Effluent analysis

Since the conservative quotient is less than 1, it is unlikely that concentrations of formaldehyde in groundwater are causing adverse effects on populations of aquatic organisms in Canada.

A summary of the values used in the environmental risk analysis of formaldehyde in the aquatic environment is presented in Table 8.

Groundwater analysis

Since the conservative quotient is less than 1, it is unlikely that concentrations of formaldehyde in groundwater are causing adverse effects on
populations of aquatic organisms in Canada.

A summary of the values us ed in the environmental risk analysis of formaldehyde in the aquatic environment is presented in Table 8.

3.1.2.3 Discussion of uncertainty

There are a number of potential sources of uncertainty in this environmental risk assessment. Regarding effects of formaldehyde on terrestrial and aquatic organisms, uncertainty surrounds the extrapolation from available toxicity data to potential ecosystem effects. While the toxicity data set included studies on organisms from a variety of ecological niches and taxa, there are relatively few good chronic studies available. To account for these uncertainties, application factors were used in the environmental risk analysis to derive ENEVs.

Regarding environmental exposure, there could be concentrations of formaldehyde in Canada that are higher than those identified and used in this assessment.

Table 8 Summary of the environmental risk analysis for aquatic organisms

Aquatic organisms	EEV (µg/L)	CTV (µg/L)	Application factor	ENEV (µg/L)	Quotient
Tier 1: Hyperconservative
Surface water - marine algae	9.0	100	10	10	0.9
Effluent - marine algae	325	100	10	10	32.5
Groundwater - marine algae	690 000	100	10	10	69 000
Tier 2: Conservative
Effluent - seed shrimp	32.5	360	10	36	0.9
Groundwater - seed shrimp	10	360	10	36	0.28

For exposure in air, the measurements used in this assessment are considered acceptable because they were selected from an extensive set of recent air monitoring data of urban and other sites, including from sites at or near industrial facilities that use and release formaldehyde in Canada. These sites can also be associated with high concentrations of VOCs associated with secondary formation of formaldehyde. Thus, available data on atmospheric concentrations are considered representative of the highest concentrations likely to be encountered in air in Canada.

Only limited data are available for water, although concentrations of formaldehyde are expected to be low because of the limited releases to these media that have been identified and the limited partitioning of formaldehyde to these compartments from air. The available data on concentrations in groundwater include data from industrial sites of the users of formaldehyde. Since data are not available regarding surface recharge of the contaminated groundwater, the assessment very conservatively assumed that recharge occurred at concentrations equivalent to those measured in the groundwater with minimal dilution.

Despite some data gaps regarding the environmental effects and exposure of formaldehyde, the data available at this time are considered adequate for making a conclusion on the environmental risk of formaldehyde in Canada.

3.2 CEPA 1999 64(b): Environment upon which life depends

Formaldehyde does not deplete stratospheric ozone, and its potential for climate change is negligible. The photolysis of formaldehyde leads to the direct formation of radicals that are active in the formation of ground-level ozone (Carter et al., 1995). In addition, formaldehyde is more reactive with hydroxyl radicals (POCP of 105) than compounds such as ethene that are recognized as important in the formation of ground-level ozone (Bunce, 1996). Given its reactivity and concentrations measured in air in Canada, formaldehyde represented approximately 7.8% of the total volatile organic carbon reactivity, ranking it 4th among non-methane hydrocarbons and carbonyl compounds contributing to the formation of ground-level ozone (Dann and Summers, 1997). Formaldehyde is therefore important in the photochemical formation of ground-level ozone.

3.3 CEPA 1999 64(c): Human health

3.3.1 Estimated population exposure

Estimates of the total daily intake of formaldehyde by six age groups of the general population of Canada were developed primarily to determine the relative contributions from various media. These estimates indicate that the daily intake of formaldehyde via inhalation is consistently less than that estimated for the ingestion of foodstuffs. However, it should be noted that critical effects associated with exposure to formaldehyde occur primarily at the site of first contact (i.e., the respiratory tract following inhalation and the gastrointestinal tract following ingestion) and are related to the concentration of formaldehyde in media to which humans are exposed, rather than the total intake of this substance. For this reason, effects of exposure by inhalation and ingestion are addressed separately.

Due primarily to limitations of available data as a basis for characterization of exposure via ingestion, the principal focus of the assessment is airborne exposure. The less representative assessment for ingestion involves comparison of the concentration of formaldehyde in a limited number of food products with a Tolerable Concentration (ingestion).

The general population in Canada is exposed to low concentrations of formaldehyde in outdoor air and to generally higher concentrations in indoor air. A subset of data from the NAPS program was selected to represent the range and distribution of concentrations to which the general population of Canada is currently assumed to be exposed via inhalation of outdoor air. The selected data are from sites classified as suburban (n = 4) or urban (n = 4) and include all 24-hour concentrations of formaldehyde (n = 2818) measured at these sites between January 1990 and December 1998 (Health Canada, 2000). The distribution of concentrations is positively skewed, with median, arithmetic mean and upper-percentile concentrations as summarized in Table 9. The distribution of concentrations at one of the four urban sites (i.e., in Toronto) was selected as a reasonable worst case. The distribution of concentrations of formaldehyde at this site is also positively skewed, and statistical parameters of this distribution are summarized in Table 9.

Table 9 Concentrations of formaldehyde in outdoor air and residential indoor air in Canada

Medium of exposure	Number of samples	Mid-points of distributions (µg/m3)		Upper percentiles of distributions of concentrations (µg/m3)
		Median	Mean 5	75th	90th	95th	97.5th
Outdoor air - NAPS data 1	2818	2.8	3.3	4.1	6.0	7.3	9.1
Outdoor air - reasonable worst-case site 2	371	2.9	4.0	4.8	7.3	10.4	17.3
Indoor air - five studies 3	151	29.8	35.9	46.2	64.8	84.6	104.8
Indoor air - lognormal distribution 4	-	28.7	-	46.1	70.7	91.2	113.8

Pooled data (n = 151) from five studies in which concentrations of formaldehyde were measured in the indoor air of residences in Canada between 1989 and 1995 were the basis for the range and distribution of concentrations to which the general population of Canada is currently assumed to be exposed via inhalation of residential indoor air (Health Canada, 2000). Sampling duration was 24 hours in two of the studies selected (n = 47 samples). These samples were collected and analyzed by the same methodologies and by the same laboratory as for the NAPS data referred to above. Passive sampling for 7-day periods and different analytical methodology were employed in the remaining three studies (n = 104). The distributions of concentrations of formaldehyde from the 24-hour active and the 7-day passive samples were compared. These distributions were judged to be sufficiently similar to justify pooling the data from the five studies. Median, arithmetic mean and upper-percentile concentrations of the distribution of pooled concentrations are summarized in Table 9.

The distribution of pooled concentrations is positively skewed. When plotted in 10 µg/m³ bins, there is a good fit to a lognormal distribution characterized by the same geometric mean (i.e., 28.7 µg/m³) and standard deviation (2.92). Upper percentiles of this lognormal distribution were calculated and are shown for comparison in Table 9. The values of these percentiles are higher for the lognormal distribution than for the more limited data set. This is to be expected, since the lognormal distribution approaches the x-axis asymptotically.

These data are used to estimate the distribution of time-weighted 24-hour concentrations of formaldehyde to which the general population is exposed (Health Canada, 2000). This requires consideration of the proportion of the 24-hour day that is spent indoors versus the time spent outdoors. Recent deterministic (i.e., point) estimates (EHD, 1998) indicate that, in general, all age groups spend a daily average of 21 hours in indoor environments and 3 hours outdoors in Canada. Probabilistic estimates of the proportion of time spent indoors versus outdoors are more desirable, as these would provide an indication of the distributions of these average estimates, but these estimates were not available. Instead, a mean time spent outdoors of 3 hours is assumed based on the point estimates of time spent indoors and outdoors (EHD, 1998). The distribution of the time spent outdoors is arbitrarily assumed to be normal in shape with an arithmetic standard deviation of 2 hours. In the probabilistic simulation, this distribution is truncated at 0 hours and 9 hours. The time spent indoors is calculated as 24 hours minus the time spent outdoors.

Estimates of the distribution of time-weighted 24-hour concentrations of formaldehyde to which the general population is exposed were developed using simple random sampling with Crystal Ball^TM Version 4.0 (Decisioneering, Inc., 1996) and simulations of 10 000 trials. Each trial involves random sampling of the distribution of concentrations in outdoor air and multiplying this by a random sample of the time spent outdoors. This results in an estimate of the concentration- time product for formaldehyde (CO , in µg- hour/m³) resulting from exposure to outdoor air. The "time spent indoors" is then calculated as 24 hours minus "time spent outdoors." This "time spent indoors" is then multiplied by a random sample from the distribution of concentrations in indoor air and results in an estimate of the concentration-time product for formaldehyde (C1, in µg-hour/m³) resulting from exposure to indoor air. The average 24-hour time-weighted concentration of formaldehyde for each trial is then calculated as (1/24) x (C_O + C_I ) for exposure to outdoor and indoor air.

Two simulations were run. In both simulations, the distribution of concentrations of formaldehyde in outdoor air is represented by a frequency histogram of the data from the eight selected NAPS sites (n = 2818 samples). In the first simulation, the distribution of concentrations of formaldehyde in residential indoor air is represented by a frequency histogram of the pooled data from the five selected studies (n = 151 samples). In the second simulation, the distribution of concentrations of formaldehyde in residential indoor air is represented by an assumed lognormal distribution with the same geometric mean (28.7 µg/m³) and standard deviation (2.92) as for the pooled data. This assumed lognormal distribution is truncated at 150 µg/m³, the highest concentration measured among the five studies. It is assumed that the general population is exposed to similar distributions of concentrations in the indoor air of public places. Exposure to formaldehyde in the indoor air of workplaces is not addressed specifically; therefore, the general population is assumed to be exposed to similar concentrations of formald ehyde in t he indoor air of all w orkplaces. Estima tes of the median, arithmetic mean and upper percentiles of the distributions of 24-hour time-weighted average concentrations of formaldehyde determined from these probabilistic simulations are summarized in Table 10. The two simulations were each run five times. The relative standard deviations of the upper-percentile est imates of time-weighted average concentrations were calculated to determine the stability of these upper-percentile estimates. These relative standard deviations are also summarized in Table 10. Examples of the shapes of the distributions resulting from the two simulations are available in Health Canada (2000). Based on the assumptions underlying these probabilistic simulations, the estimates summarized in Table 10 indicate that one of every two persons would be exposed to 24-hour average concentrations of formaldehyde in air of 24-29 µg/m³ or greater (i.e., median concentrations). Similarly, 1 in 20 persons (i.e., 95th percentile) would be exposed to 24-hour average concentrations of formaldehyde in air of 80-94 µg/m³ or greater.

Table 10 Probabilistic estimates of 24-hour time-weighted average concentrations of formaldehyde in air

	Mid-points of distributions (µg/m3)		Upperpercentiles of distributions of concentrations (µg/m3) and relative standard deviations (%)
	Median	Mean3	75th	90th	95th	97.5th
Simulation 1 1	29	36	46 (± 0.5%)	62 (± 1.3%)	80 (± 1.9%)	97 (± 0.7%)
Simulation 2 2	24	33	45 (± 1.2%)	75 (± 1.2%)	94 (± 1.6%)	109 (± 1.3%)

1. In simulation 1, the distribution of concentrations of formaldehyde is represented by a frequency histogram of the pooled data from the five selected studies (n = 151 samples).
2. For simulation 2, a lognormal distribution of concentrations, truncated at 150 µg/m³, is assumed. This lognormal distribution has the same geometric mean (28.7 µg/m³) and standard deviation (2.92) as the distribution of concentrations for the pooled data from the five selected studies.
3. This is the arithmetic mean concentration.

Based on limited data from the United States, concentrations in drinking water may range up to approximately 10 µg/L, in the absence of specific contributions from the formation of formaldehyde by ozonation during water treatment or from leaching of formaldehyde from polyacetal plumbing fixtures. One-half this concentration (i.e., 5 µg/L) was judged to be a reasonable estimate of the average concentration of formaldehyde in Canadian drinking water, in the absence of other data. Concentrations approaching 100 µg/L were observed in a U.S. study assessing the leaching of formaldehyde from domestic polyacetal plumbing fixtures, and this concentration is assumed to be representative of a reasonable worst case.

Similarly, very few data are available with which to estimate the range and distribution of concentrations of formaldehyde in foods to which the general population in Canada is exposed. According to the limited available data, concentrations of formaldehyde in food are highly variable. In the few studies of the formaldehyde content of foods in Canada, the concentrations of formaldehyde were within the range from less than 0.03 to 14 mg/kg (Health Canada, 2000). However, the proportion of formaldehyde in foods that is bioavailable is unknown.

3.3.2 Hazard characterization

Inhalation, the likely principal route of exposure of the general population to formaldehyde, has been the focus of most studies on the effects of this substance in humans and laboratory animals. Available data on effects following ingestion or dermal exposure to formaldehyde are limited. Since formaldehyde is water soluble, highly reactive with biological macromolecules and rapidly metabolized, adverse effects resulting from exposure are observed primarily in those tissues or organs with which formaldehyde first comes into contact (i.e., the respiratory and gastrointestinal tracts following inhalation and ingestion, respectively).

Effects following inhalation that occur primarily at the site of contact are, therefore, the principal focus of this section.

3.3.2.1 Genotoxicity and carcinogenicity

3.3.2.1.1 Genotoxicity

Results of epidemiological studies in occupationally exposed populations are consistent with a pattern of weak positive responses for genotoxicity, with good evidence of an effect at site of contact (e.g., micronucleated buccal or nasal mucosal cells). Evidence for distal (i.e., systemic) effects is equivocal (chromosomal aberrations and sister chromatid exchanges in peripheral lymphocytes). The contribution of co-exposures to observed effects cannot be precluded.

The results of a large number of in vitro assays of a variety of endpoints indicate that formaldehyde is weakly genotoxic in both bacterial and mammalian cells. The spectrum of mutation induced by formaldehyde in vitro varies among cell types and concentrations to which cells were exposed but includes both point and large-scale changes. The results of in vivo studies in animals are similar to those in humans, with effects at site of contact being observed (e.g., modest increase in the proportion of pulmonary macrophages with chromosomal aberrations in rats following inhalation and cytogenetic alterations in the gastrointestinal epithelium of rats following oral exposure). Evidence of distal (systemic) effects is less convincing. Indeed, in the majority of studies of rats exposed to formaldehyde via inhalation, genetic effects within peripheral lymphocytes or bone marrow cells have not been observed.

Formaldehyde also induces the formation of DNA-protein crosslinks in a variety of human and rat cell types in vitro and in the epithelium of the nasal cavity of rats and respiratory tract of monkeys following inhalation, which may contribute to the carcinogenicity of the compound in the nasal cavity of rats through replication errors, resulting in mutation.

Overall, formaldehyde is weakly genotoxic, with effects most likely to be observed in vivo in cells from tissues or organs with which the aldehyde comes into first contact.

3.3.2.1.2 Carcinogenicity

Inhalation

In epidemiological studies of occupationally exposed populations, there has been little evidence of a causal association between exposure to formaldehyde and lung cancer. Indeed, results of studies in a rather extensive database of cohort and case-control studies do not fulfil traditional criteria of causality in this regard, such as consistency, strength and exposure-response. Increases in mortality or incidence have not been observed consistently, and, where examined, there has consistently been no evidence of exposure-response. The data for nasal and nasopharyngeal cancer are less clear. In case-control studies, there have been increases in cancers of the nasal or nasopharyngeal cavities that fulfil, at least in part, traditional criteria of causality, with tumours having been observed in workers with highest levels or duration of exposure. It should be noted, though, that measures of exposure in these population-based investigations are rather less reliable than those in the larger, most extensive cohort studies of occupationally exposed populations; moreover, methodological limitations complicate interpretation of several of the case-control studies. Excesses of cancers of the nasal or nasoph aryngeal cavit ies h ave not been observed cons istently in cohort st udies. Where there have been excesses, there has been little evidence of exposure-response, although the total number of observed tumours was small.

Five carcinogenicity bioassays have provided consistent evidence that formaldehyde is carcinogenic in rats exposed via inhalation (Kerns et al., 1983; Sellakumar et al., 1985; Tobe et al., 1985; Monticello et al., 1996; Kamata et al., 1997). The incidence of nasal tumours was not significantly increased in mice exposed to formaldehyde by inhalation (Kerns et al., 1983). This has been attributed, at least in part, to the greater reduction in minute volume in mice than in rats exposed to formaldehyde (Chang et al., 1981; Barrow et al., 1983), resulting in lower exposures in mice than in rats (Barrow et al., 1983).

Observation of tumours at the site of contact is consistent with toxicokinetic considerations. Formaldehyde is a highly water-soluble, highly reactive gas that is absorbed quickly at the site of contact. It is also rapidly metabolized, such that exposure to even high concentrations of atmospheric formaldehyde does not result in an increase in blood concentrations.

As described in Section 2.4.3.7, the mechanisms by which formaldehyde induces nasal tumours in rats are not fully understood. However, it has been hypothesized that a sustained increase in epithelial cell regenerative proliferation resulting from cytotoxicity is a requisite precursor in the mode of induction of tumours. Mutation, for which the formation of DNA-protein crosslinks serves as a marker of potential, may also contribute to the carcinogenicity of the compound in the nasal cavity of rats. Studies relevant to assessment of the mode of action include a cancer bioassay (Monticello et al., 1996) in which intermediate endpoints (proliferative response in various regions of the nasal epithelium) have been investigated. The relevant database also includes numerous shorter-term studies in which proliferative response and the formation of DNA-protein crosslinks in the nasal epithelium of rats and other species have been examined following exposure via regimens often similar to those in the cancer bioassays (Swenberg et al., 1983; Casanova and Heck, 1987; Heck and Casanova, 1987; Casanova et al., 1989, 1991, 1994; Monticello et al., 1989, 1991). It should be noted, though, that due to the limited data on intermediate endpoints in most of the cancer bioassays, information available as a basis for direct comparison of the incidence of intermediate lesions (i.e., proliferative response as a measure of cytotoxicity and DPX) and tumours is limited to that presented in Table 4.

In all cases where examined, without exception, sustained cytotoxicity and cellular proliferation were observed in the nasal cavities of the same strain of rats exposed in a similar manner in short-term studies to concentrations or doses that induced nasal tumours in the cancer bioassays (Monticello et al., 1991, 1996). However, the converse is not always true. Similarly, tumours have been observed only at concentrations at which increases in DNA-protein crosslinks have been observed in shorter-term studies in the same strain (Casanova and Heck, 1987; Heck and Casanova, 1987; Casanova et al., 1989, 1994).

In addition, where proliferative response (Monticello et al., 1991, 1996) and DPX (Casanova et al., 1994) have been examined in various regions of the nasal passages, sites at which there are increases are similar to those where tumours have been observed. The concentration-response relationships for DPX, cytotoxicity, proliferative response and tumours are highly non-linear, with significant increases in all endpoints being observed at concentrations of 4 ppm (4.8 mg/m³) and above (Table 4). This correlates well with the concentration at which mucociliary clearance is inhibited and glutathione-mediated metabolism saturated (i.e., 4 ppm [4.8 mg/m³]). Histological changes, increased epithelial cell proliferation and DPX are all more closely related to the exposure concentration than to the total cumulative intake or dose of formaldehyde (Swenberg et al., 1983; Casanova et al., 1994).

While the respective roles of DPX, mutation and cellular proliferation in the induction of tumours in the rat nose are not fully delineated, the hypothesized mode of carcinogenesis is in keeping with the growing body of evidence supporting the biological plausibility that prolonged regenerative cell proliferation can be a causal mechanism in chemical carcinogenesis. Regenerative cell proliferation following formaldehyde-induced cytotoxicity increases the number of DNA replications and thus increases the probability of a DNA-protein crosslink initiating a DNA replication error, resulting in a mutation. This proposed mode of action is consistent with the observed inhibition of DNA replication in the rat nose at elevated concentrations (Heck and Casanova, 1994) and point mutations in the p53 tumour suppressor gene in tumours from the noses of rats exposed to formaldehyde (Recio et al., 1992).

The hypothesized mode of induction of formaldehyde-induced tumours that satisfies several criteria for weight of evidence, including consistency, concordance of exposure-response relationships across intermediate endpoints and biological plausibility and coherence of the database, is likely relevant to humans, at least qualitatively. Increased cell proliferation (Monticello et al., 1989) and DNA-protein crosslink formation (Casanova et al. 1991) within epithelia of the upper respiratory tract have been observed in monkeys exposed to formaldehyde vapour. Although not sufficient in itself as a basis for inferring causality, direct evidence on histopathological lesions in the nose of humans exposed primarily to formaldehyde in the occupational environment is consistent with a qualitatively similar response of the upper respiratory tract in humans and experimental animals to formaldehyde. Increased human epithelial cell proliferation following in situ exposure to formaldehyde has also been observed in a model system in which rat trachea populated with human tracheobronchial epithelial cells were xenotransplanted into athymic mice (Ura et al., 1989).

Because formaldehyde is highly reactive at the site of contact, dosimetry is of critical importance when extrapolating across species that have significantly different anatomical features of the nasal and respiratory passages and patterns of flow of inhaled air. Since humans as well as other primates are oronasal breathers, compared with rats, which are obligate nose breathers, effects associated with the inhalation of formaldehyde are likely to be observed in a wider area deeper within the respiratory tract. Indeed, in rats exposed to moderate levels of formaldehyde, histopathological changes, increased epithelial cell proliferation as well as DNA-protein crosslink formation are restricted to the nasal cavity; in formaldehyde-exposed monkeys (as surrogates for humans), on the other hand, these effects have been observed further along within the upper respiratory tract. While the epidemiological studies taken as a whole do not provide strong evidence for a causal association between formaldehyde exposure and human cancer, the possibility of increased risk of respiratory cancers, particularly those of the upper respiratory tract, cannot be excluded on the basis of available data.

Based primarily upon data derived from laboratory studies, therefore, the inhalation of formaldehyde under conditions that induce cytotoxicity and sustained regenerative proliferation is considered to present a carcinogenic hazard to humans.

Oral exposure

Epidemiological studies of potential carcinogenic hazards associated with the ingestion of formaldehyde were not identified. Currently, there is no definitive evidence to indicate that formaldehyde is carcinogenic when administered orally to laboratory animals. However, consistent with the known reactivity of this substance with biological macromolecules in the tissue or organ of first contact, histopathological and cytogenetic changes within the gastrointestinal tract have been observed in rats administered formaldehyde orally. These observations and additional consideration of the mode of induction of tumours by formaldehyde lead to the conclusion that under certain conditions of exposure, potential carcinogenic hazard associated with the ingestion of formaldehyde cannot be eliminated.

3.3.2.2 Non-neoplastic effects

Sensory irritation of the eyes and resp iratory tract by formaldehyde has bee n observed consi stently in cli nical studi es and epidemiological (primarily cross-sectional) surveys in occupational and residential environments. The pattern of effects is consistent with increases in symptoms being reported at lowest concentrations, with the eye generally being most sensitive.

At concentrations higher than those generally associated with sensory irritation, generally small, reversible effects on lung function have been noted, although ev idence of cumu lative decrement in pulmonary function is limited.

Results are consistent with the increased prevalence of histological changes in the nasal epithelium in cross-sectional studies of workers being attributable to formaldehyde (Edling et al., 1988; Holmström et al., 1989c; Boysen et al., 1990; Ballarin et al., 1992). The criterion of biological plausibility for weight of evidence of causality is also satisfied by the convincing evidence in monkeys (Rusch et al., 1983) and rodents of histopathological alterations (degenerative changes consistent with cytotoxicity) within the upper respiratory tract. Other than damage to the gastric epithelium observed following the acute ingestion of large amounts of formaldehyde (K_ochhar et al., 1986; Nishi et al., 1988; WHO, 1989), studies on potential changes within the gastrointestinal tract in humans following the long-term ingestion of this substance were not identified. However, histological changes within the surface epithelium of the gastrointestinal tract of rats (e.g., erosions and/or ulcers, hyperkeratosis, hyperplasia, gastritis) have been observed following chronic oral exposure to formaldehyde administered in drinking water, at high concentrations (Til et al., 1989; Tobe et al., 1989).

Formaldehyde is not likely to affect reproduction or development at levels of exposure lower than those associated with adverse health effects at the site of contact. Based upon recent epidemiological studies of occupationally exposed individuals, there is no clear evidence indicating that either maternal or paternal inhalation exposure to formaldehyde is associated with an increased risk of spontaneous abortion (Hemminki et al., 1985; Lindbohm et al., 1991; John et al., 1994; Taskinen et al., 1994). In studies of laboratory animals exposed via inhalation (Saillenfait et al., 1989; Martin, 1990) or oral administration (Seidenberg and Becker, 1987; Wickramaratne, 1987), formaldehyde had no effect on reproduction or fetal development, at levels of exposure less than those causing notable adverse health effects at the site of contact.

Based upon the available although limited data, exposure to formaldehyde is unlikely to be associated with suppression of the immune response. Indeed, the dermal hypersensitivity of some individuals to formaldehyde as well as the results of studies in animals indicate heightened immune responses linked to formaldehyde exposure. Information from epidemiological studies on suppression of the immune response associated with exposure to formaldehyde was not identified. Adverse effects on either cell- or humoral-mediated immune responses have not been consistently observed in studies conducted in laboratory animals (Dean et al., 1984; Adams et al., 1987; Holmström et al., 1989b; Jakab, 1992; Vargová et al., 1993). Although suggested in case reports for some individuals, no clear evidence that formaldehyde-induced asthma was attributable to immunological mechanisms has been identified. However, studies with laboratory animals have revealed that formaldehyde may enhance their sensitization to inhaled allergens (Tarkowski and Gorski, 1995; Riedel et al., 1996).

For the general population, dermal exposure to concentrations of formaldehyde in the vicinity of 1-2% (10 000-20 000 ppm) is likely to cause skin irritation; however, in hypersensitive individuals, contact dermatitis can occur following exposure to formaldehyde at concentrations as low as 0.003% (30 ppm). In North America, less than 10% of patients presenting with contact dermatitis may be immunologically hypersensitive to formaldehyde.

3.3.3 Exposure-response analysis

Cancer and non-neoplastic effects are addressed separately here. However, the weight of evidence indicates that formaldehyde is carcinogenic only at concentrations that induce the obligatory precursor lesion of proliferative regenerative response associated with cytotoxicity, although interaction with DNA must also be taken into account. For consistency with other assessments and for ease of presentation, cancer and non-cancer effects are considered separately here, although, based on consideration of mode of action, they are inextricably linked.

Emphasis in the dose-response analyses for cancer presented below is on a biologically motivated case-specific model that incorporates a two-stage clonal growth model. This model is supported by dosimetry calculations from computational fluid dynamics (CFD) modelling of formaldehyde flux in various regions of the nose and a single-path model for the lower respiratory tract. While this model entails simplification of cancer biology, which requires selection of a number of parameters and use of simplifying assumptions, it is considered to offer improvement over default methodology due to incorporation of as many biological data as possible.

There has been no sensitivity analysis conducted to determine which of the model parameters has greatest impact on risk estimates or to identify which parameters are known with the highest degree of certainty. However, output of the model is considered adequate as a basis to ensure that measures taken to prevent sensory irritation 1 in human populations are sufficiently protective with respect to carcinogenic potential.

3.3.3.1 Inhalation

3.3.3.1.1 Carcinogenicity

There is indisputable evidence that formaldehyde is carcinogenic in rats following inhalation, with the carcinogenic response being limited to the site of contact (e.g., the nasal passages of rodents). While the mechanism of action is not well understood, based primarily upon data derived from laboratory studies, regenerative proliferation associated with cytotoxicity appears to be an obligatory intermediate step in the induction of cancer by formaldehyde. Interaction with genetic material, the potential for which is indicated by DPX, likely also contributes, although the probability of mutation resulting from DPX is unknown.

Available data are also consistent with the hypothesis that humans would respond qualitatively similarly to experimental animals in this regard. However, since formaldehyde is highly reactive at the site of contact, dosimetry is of critical importance in predicting interspecies variations in response, as a function of flux to the tissue and regional tissue susceptibility, due to the significantly different anatomical features of the nasal and respiratory passages between experimental animals and humans.

The approach to dose-response modelling emphasized here, therefore, is biologically based, reflecting the non-linearity in concentration-response relationships for formaldehyde-induced nasal cancer and associated intermediate endpoints and incorporating, to the extent possible, mechanistic data and state-of-the-art analyses for species-specific dosimetry. It incorporates regenerative cell proliferation as a required step in the induction of tumours by formaldehyde and a contribution from mutagenicity (not defined specifically by DPX) that has greatest impact at low exposures through modelling of complex functional relationships for cancer due to actions of formaldehyde on mutation, cell replication and exponential clonal expansion. Species variations in dosimetry are taken into account by CFD modelling of formaldehyde flux in various regions of the nose and a single-path model for the lower respiratory tract of humans.

The outcome is compared with that derived based on empirical default methodology for estimation of tumorigenic concentrations in the experimental range for Priority Substances (Health Canada, 1994). However, it is the biologically motivated case-specific model that is considered to provide the most defensible estimates of cancer risk, on the basis that it encompasses more of the available biological data, thereby offering considerable improvement over default (Health Canada, 1998). Moreover, in view of the clear emphasis herein and preference for the biologically motivated case-specific model, there has been no attempt to incorporate more of the biological data in the calculation of tumorigenic concentrations by default methodology (e.g., dose and time dependence to der ive an empirical d ose metric for rats).

Figure 2 Two-stage clonal growth model (reproduced from CIIT, 1999)

Biologically motivated case-specific model

Derivation of the dose-response model and selection of various parameters are presented in great er detail in CIIT (1999); only a brief sum mary is provided here. The biologically based, two-stage clonal growth model developed (Figure 2) is identical in biological structure to other such models (also known as MVK models), incorporating information on normal growth, cell cycle time and cells at risk (in various regions of the respiratory tract).

Formaldehyde is assumed to act as a direct mutagen, with the effect considered proportional to the estimated tissue concentration of DNA-protein crosslinks. The dose-response curve for DNA-protein crosslink formation is linear at low exposure concentrations and increases in a greater than linear manner at high concentrations, similar to those administered in the rodent carcinogenicity bioassays. The second mode of carcinogenic action considers cytotoxicity and the subsequent regenerative cellular proliferation associated with exposure to formaldehyde, incorporating a "hockey stick" dose-response curve (i.e., dose threshold curve) within the model. Values for parameters related to the effects of formaldehyde exposure upon the mutagenic (i.e., DNA-protein crosslink formation) and proliferative response (i.e., regenerative cell proliferation resulting from formaldehyde-induced cytotoxicity) were derived from a two-stage clonal growth model developed for rats (Figure 3), which describes the formation of nasal tumours in animals exposed to formaldehyde.

Species-specific dosimetry within various regions of the respiratory tract in laboratory animals and humans was also incorporated. Regional dose is a function of the amount of formaldehyde delivered by inhaled air and the absorption characteristics of the lining within various regions of the respiratory tract. The amount of formaldehyde delivered by inhaled air depends upon major airflow patterns, air-phase diffusion and absorption at the air-lining interface.

Figure 3 Roadmap for the rat clonal growth model. SCC = squamous cell carcinoma (reproduced from CIIT, 1999)

The "dose" (flux) of formaldehyde to cells depends upon the amount absorbed at the air-lining interface, mucus/tissue-phase diffusion, chemical interactions such as reactions and solubility, and clearance rates. Species differences in these factors influence the site-specific distribution of lesions.

The F344 rat and rhesus monkey nasal surface for one side of the nose and the nasal surface for both sides of the human nose were mapped at high resolution to develop three-dimensional, anatomically accurate CFD models of rat, primate and human nasal airflow and inhaled gas uptake (Kimbell et al., 1997; Kepler et al., 1998; Subramaniam et al., 1998). The approximate locations of squamous epithelium and the portion of squamous epithelium coated with mucus were mapped onto the reconstructed nasal geometry of the CFD models. These CFD models provide a means for estimating the amount of inhaled gas reaching any site along the nasal passage walls and allow the direct extrapolation of exposures associated with tissue damage from animals to humans via regional nasal uptake. Although development of the biologically based, two-stage clonal growth model for rats required analysis of only the nasal cavity, for humans, carcinogenic risks were based on estimates of formaldehyde dose to regions (i.e., regional flux) along the entire respiratory tract.

The exposure-response model developed for humans (see Figure 4) predicts the additional risk of formaldehyde-induced cancer within the respiratory tract under various exposure scenarios.

Two of the parameters in the human clonal growth model - the probability of mutation per cell division and the growth advantage for preneoplastic cells, both in the absence of formaldehyde exposure, were estimated statistically by fitting the model to human 5-year age group lung cancer incidence data for non-smokers. 2 The parameter representing the time for a malignant cell to expand clonally into a clinically detectable tumour was set at 3.5 years.

Figure 4 Roadmap for the human clonal growth model (reproduced from CIIT, 1999)

In addition to the human nasal CFD model, a typical-path, one-dimensional model of formaldehyde uptake was developed for the lower respiratory tract. The latter model consisted of the tracheobronchial and pulmonary regions in which uptake was simulated for four ventilatory states, based on an ICRP (1994) activity pattern for a heavy-working adult male. Nasal uptake in the lower respiratory model was calibrated to match overall nasal uptake predicted by the human CFD model. While rodents are obligate nasal breathers, humans switch to oronasal breathing when the level of activity requires a minute ventilation of about 35 L/minute. Thus, two anatomical models for the upper respiratory tract encompassing oral and nasal breathing were developed, each of which consisted basically of a tubular geometry. For the mouth cavity, the choice of tubular geometry was consistent with Fredberg et al. (1980). The rationale for using the simple tubular geometry for the nasal airway was based primarily upon the need to remove formaldehyde from the inhaled air at the same rate as in a corresponding three-dimensional CFD simulation. However, in calculations of carcinogenic risk, the nasal airway fluxes predicted by the CFD simulations, and not those predicted by the single-path model, were used to determine upper respiratory tract fluxes.

To account for oronasal breathing, there were two simulations. In one simulation, the nasal airway model represented the proximal upper respiratory tract, while for the other simulation, the mouth cavity model was used for this region. In both simulations, the fractional airflow rate in the mouth cavity or in the nasal airway was taken into account. For each segment distal to the proximal upper respiratory tract, the doses (fluxes) of formaldehyde from both simulations were added to obtain the estimated dose for oronasal breathing. The site-specific deposition of formaldehyde along the human respiratory tract coupled with data on effects upon regional DPX and cell proliferation (derived from studies in animals) (Casanova et al., 1994; Monticello et al., 1996) were reflected in calculations of carcinogenic risks associated with the inhalation of formaldehyde in humans.

Estimates of carcinogenic risks using the human two-stage clonal growth model were developed for typical environmental exposures (i.e., continuous exposure throughout an 80-year lifetime to concentrations of formaldehyde ranging from 0.001 to 0.1 ppm [0.0012 to 0.12 mg/m³]). The human clonal growth model predicted non-zero additional risks throughout the exposure ranges examined. The two-stage model describes a low-dose, linear carcinogenic response for humans exposed to levels of formaldehyde of ≤0.1 ppm (0.12 mg/m³), where cytotoxicity and sustained cellular regenerative proliferation do not appear to play a role in tumour induction. Indeed, the effect of formaldehyde upon regenerative cellular proliferation did not have a significant impact upon the predicted carcinogenic risks at exposures between 0.001 and 0.1 ppm (0.0012 and 0.12 mg/m³). Based upon the two-stage clonal growth model, the predicted additional risks of upper respiratory tract cancer for non-smokers, associated with an 80-year continuous exposure to levels of formaldehyde between 0.001 and 0.1 ppm (1.2 and 120 µg/m³), range from 2.3 x 10^-10 to 2.7 x 10^-8, respectively (CIIT, 1999).

No excess risk was predicted by the human clonal growth model in a cohort exposed to formaldehyde at a specific plant examined in two epidemiological studies (Blair et al., 1986; Marsh et al., 1996). This was consistent with the observed number of cases of respirato ry tract cancer (113 observed deaths; 120 expected) in the cohort. Th us, the outcome of the model was consistent with the results of the epidemiological studies.

Default modelling

For comparison, based upon the approach typically employed in the assessment of Priority Substances, a Tumorigenic Concentration₀₅ (TC₀₅) (i.e., the concentration associated with a 5% increase in tumour incidence over background) of 7.9 ppm (9.5 mg/m³) (95% lower confidence limit [LCL] = 6.6 ppm [7.9 m g/m³]) formaldehyde was derived from data on the incidence of nasal squamous tumours in rats exposed to this substance in the single study (i.e., Monticello et al., 1996) in which exposure-response was best characterized. 3 The TC₀₅ is calculated by first fitting a multistage model to the exposure-response data. The multistage model is given by

P(d) = 1 -e ^{-q ₀ -q ₁ d-...-q k d k}

where d is dose, k is the number of dose groups in the study minus one, P(d) is the probability of the animal developing a tumour at dose d and q_i > 0, i = 1, ..., k are parameters to be estimated.

The model was fit using GLOBAL82 (Howe and Crump, 1982), and the TC₀₅ was calculated as the concentration C that satisfies

A chi-square lack of fit test was performed for each of the three model fits. The degrees of freedom for this test are equal to k minus the number of q_i s whose estimates are non-zero. A p-value less than 0.05 indicates a significant lack of fit. In this case, chi-square = 3.7, df = 4 and p = 0.45.

3.3.3.1.2 Non-neoplastic effects

There are considered to be sufficient data from clinical studies and cross-sectional surveys of human populations, as well as supporting observations from experimental studies conducted with laboratory animals, to indicate that the irritant effects of formaldehyde on the eyes, nose and throat occur at lowest concentration. Although individual sensitivity and exposure conditions such as temperature, humidity, duration and co-exposure to other irritants are likely to influence response levels, in well-conducted studies, only a very small proportion of the population experiences symptoms of irritation following exposure to ≤0.1 ppm (0.12 mg/m³) formaldehyde. This is less than the levels that reduce mucociliary clearance in the anterior portion of the nasal cavity in available clinical studies in human volunteers (0.3 mg/m³) and induce histopathological effects in the nasal epithelium in cross-sectional studies of formaldehyde-exposed workers (0.3 mg/m³). Additional investigation of preliminary indication of effects on pulmonary function in children in the residential environment associated with lower concentrations of formaldehyde (40-60 ppb [48-72 µg/m³]) (Krzyzanowski et al., 1990) is warranted.

3.3.3.2 Oral exposure

Lack of evidence for the potential carcinogenicity of ingested formaldehyde precludes an analysis of exposure-response for this effect.

Data on non-neoplastic effects associated with the ingestion of formaldehyde are much more limited than for inhalation. Owing to its high reactivity, non-neoplastic effects in the tissue of first contact following ingestion (i.e., the gastrointestinal tract) are more likely related to the concentration of the formaldehyde consumed, rather than to its cumulative (total) intake. Information from studies on humans is inadequate to identify putative exposure-response relationships with respect to toxicological effects associated with the long-term ingestion of formaldehyde. However, a Tolerable Concentration (TC) for formaldehyde in ingested products may be derived on the basis of the NOEL for the development of histological changes in the gastrointestinal tract of rats as follows:

where:

• 260 mg/L is the NOEL for effects (i.e., histopathological changes) in the gastrointestinal tract of rats administered formaldehyde in drinking water for 2 years in the most comprehensive study conducted (Til et al., 1989), and
• 100 is the uncertainty factor (´10 for interspecies variation, x 10 for intraspecies variation).

3.3.4 Human health risk characterization

Characterization of human health risks associated with exposure to formaldehyde is based upon analysis of the concentrations of this substance in air and some food products, rather than estimates of total daily intake per se, since effects are observed primarily in the tissue of first contact and are related to the level of exposure rather than to total systemic intake.

Emphasis for the characterization of health risks associated with the inhalation of formaldehyde in the environment in Canada is on non-neoplastic effects that occur at lowest concentrations (i.e., sensory irritation). The adequacy of this approach to protect for potential carcinogenicity is considered in the context of the biologically motivated case-specific model described above.

In humans (as well as laboratory animals), signs of ocular and upper respiratory tract sensory irritation have been observed at exposures typically greater than 0.1 ppm [120 µg/m³]). The estimated median and mean 24-hour time-weighted average exposures to formaldehyde in air in Canada are, at most, one-third of this value. This value is also greater than the estimated time-weighted average exposure to which 95% of the population is exposed. In some indoor locations, however, concentrations may approach the level associated with signs of eye and respiratory tract sensory irritation in humans.

The risks of upper respiratory tract cancer predicted by the biologically motivated case-specific model to be associated with exposure to the median, mean and 95th percentile concentrations of formaldehyde in air in Canada are also exceedingly low (i.e., <2.7 x 10^-8). Based on this estimate of risk, priority for investigation of options to reduce exposure in relation to the carcinogenicity of formaldehyde is low.

Available information is considered insufficient to fully characterize the exposure of individuals in Canada to formaldehyde in foodstuffs. However, based upon limited information, the levels of formaldehyde in drinking water appear to be more than 2 orders of magnitude less than the Tolerable Concentration (2.6 mg/L). Although the concentration of formaldehyde in some food products would appear to exceed the Tolerable Concentration, the extent of its bioavailability therein is unknown.

3.3.5 Uncertainties and degree of confidence in human health risk characterization

There is a moderate degree of confidence in the characterization of the principal source of exposure of the general population (i.e., residential indoor air). In the two studies where there was active sampling for a 24-hour duration, the analytical and sampling methodologies were optimum, all of the samples were analyzed by a single specialized laboratory, and the effects of diurnal variation were minimized by the 24-hour sampling duration. The data are also reasonably current (i.e., 1991-1993) and the measured values consistent with those determined in surveys in other countries. While some uncertainty is introduced by pooling of these data with those from the remaining three studies, which involved passive sampling, the ranges and distributions of concentrations in these subsets of data were similar. Some uncertainty is introduced by the limited size a nd representation of t he data set ( n = 151 homes in Win dsor, Hamilton, Trois- Rivières, Québec , Saskatoon and various locations in the Northwest Territories), lack of random sampling of the homes and involvement of volunteers.

Although it contributes less to total exposure, there is a high degree of confidence in the characterization of the concentrations of formaldehyde in ambient air in Canada, due to the magnitude and sensitivity of the relevant monitoring data. Analytical and sampling methodologies were optimum , all of the samples were analyzed by a single specialized laboratory, and the effects of diurnal variation were minimized by the 24-hour sampling duration. The data set is large (n = 2819) and reasonably current (i.e., 1990-1998), and the concentrations of formaldehyde are consistent with those reported for outdoor air in other Canadian and international studies. However, the locations of NAPS sites were not determined by a random sampling scheme, and a subset of only eight NAPS sites was selected. The data may also not be strictly representative of population exposure, since the air is sampled at elevations higher than the breathing zone at some sites and may be remote from populated areas. However, samples from Canada's three major urban centres (i.e., Montréal with two sites, Toronto and Vancouver) account for 54% of the 2819 samples, and samples from two sites in Windsor, Ontario, account for an additional 21% of the samples in this data set.

Uncertainty concerning the time spent indoors by Canadians is judged to be low, since the estimate is based on the most current Canadian data, the time-activity data were obtained based on a random sampling scheme, and analysis of the data involved population weighting. However, the same mean time spent outdoors is assumed for Canadians of all age groups and in all regions of the country, a normal distribution is assumed for the hours per day spent outdoors, and the variance of the assumed normal distribution is also assumed (i.e., standard deviation of 2).

The degree of uncertainty concerning the formaldehyde content of food currently consumed by Canadians is sufficiently high so as to preclude meaningful estimation of exposure from this source, except as a basis for determining potential relative proportions of total intake from various media. Identified data on concentrations in this medium are restricted to a small number of food samples collected in other countries, sometimes in early studies for which there is some suspicion of production of formaldehyde due to the relatively high temperatures and acidic reagents employed.

There are no indications that food items were selected on a random basis and often no indication whether the reported concentrations reflect formaldehyde content in the food as consumed. Due to its high volatility, the formaldehyde content would be expected to be reduced during processing and cooking. Formaldehyde is not expected to partition into the fatty compartments of foods, and direct contact of formaldehyde in food applications is very limited. Also, while there is some suggestion that formaldehyde is present in food in bound (unavailable) form, data to substantiate this contention were not identified.

There is a moderate degree of certainty that consumption of drinking water does not contribute significantly to the daily intake of formaldehyde by Canadians, since formaldehyde is relatively unstable in water. However, no data concerning the range and distribution of concentrations of formaldehyde in Canadian drinking water were identified.

With respect to toxicity, the degree of confidence that critical effects are well characterized is high. A relatively extensive database in both humans and animals indicates that critical effects occur at the initial site of exposure to this substance. The database in humans is also sufficiently robust to serve as a basis for confident conclusion concerning the consistently lowest levels at which effects (i.e., sensory irritation) occur, although additional investigation of an unconfirmed report of effects on respiratory function in children exposed to lower levels of formaldehyde is desirable.

The degree of confidence in the database that supports an obligatory role of regenerative proliferation in the induction of nasal tumours in rats is moderate to high, although the mechanism of carcinogenicity of formaldehyde is unclear. Although the biologically motivated case-specific model for estimation of cancer risks is clearly preferred due to incorporation of as many biological data as possible, there are a number of uncertainties described in more detail in CIIT (1999) and summarized briefly here, although sensitivity analyses were not conducted. For dosimetry, sources of uncertainty for which sensitivity analyses would have been appropriate include the use of individual rat, primate and human nasal anatomies as representative of the general population, the use of a typical-path human lung structure to represent people with compromised lungs, the sizes of specific airways, the use of a symmetric Weibel model for the lung, the estimation of the location and extent of squamous and olfactory epithelium and of mucus- and non-mucus-coated nasal regions in the human, and the values of mass transfer and dispersion coefficients. The lack of human data on formaldehyde-related changes in the values of key parameters of the clonal growth model accounts for much of its uncertainty.

In order to better define the mode of action of induction of tumours, elaboration of the quantitative relationship between DPX and mutation and the time course of loss of DNA-protein crosslinks is desirable. Additional characterization of the shape of the concentration-response relationship for regenerative proliferative response would also be informative.

For Priority Substances where the induction of cancer through direct interaction with genetic material cannot be ruled out and available data are inadequate as a basis for development of biologically motivated case-specific models, cancer potency is estimated based on empirical modelling of experimental data within or close to the experimental range, as described above (Section 3.3.3). Estimates of exposure are then compared with these quantitative estimates of carcinogenic potency (Exposure Potency Index) to characterize risk and provide guidance in establishing priorities for further action (i.e., analysis of options to reduce exposure) (Health Canada, 1994) under CEPA 1999. While the biologically motivated case-specific model is clearly preferred as a basis for characterization of exposure-response for cancer for formaldehyde due to its incorporation of as many of the biological data as possible, the priority for investigation of options to reduce exposure based on default methodology is presented here for comparison.

Utilization of this default approach in the case of formaldehyde would indicate that probabilistic estimates of the 24-hour median, mean and 97.5th percentile concentrations of formaldehyde in air in Canada (generally and for a worst-case site) would be approximately 327-, 263- and 98-fold lower, respectively, than the maximum likelihood estimate of the carcinogenic potency (i.e., TC₀₅ = 9.5 mg/m³)4 derived from a carcinogenesis bioassay in rats (Monticello et al., 1996). Overall, based upon these Exposure Potency Indices (ranging from 3 x 10^-3 to 1.0 x 10^-2), the priority for the investigation of options to reduce exposure to formaldehyde in air would have been considered to be high.

3.4 Conclusions

CEPA 1999 64(a): Based on analyses of worst-case situations that are likely to be encountered in Canada, risk quotients for water and air are less than 1. The environmental risks associated with concentrations of formaldehyde likely to be found in Canada therefore appear to be low. Therefore, available data indicate that it is unlikely that formaldehyde is entering or may enter the environment in a quantity or concentration or under conditions that have or may have an immediate or long-term harmful effect on the environment or its biological diversity, and it is not considered to be "toxic" as defined in CEPA 1999 Paragraph 64(a).

CEPA 1999 64(b): Formaldehyde is not involved in depletion of stratospheric ozone and likely does not contribute significantly to climate change. Because of its reactivity and abundance in air, formaldehyde contributes, along with other reactive volatile organic chemicals, to the formation of tropospheric ozone. Therefore, based on available data, formaldehyde is entering the environment in a quantity or concentration or under conditions that constitute or may constitute a danger to the environment on which life depends, and it is considered to be "toxic" as defined in CEPA 1999 Para graph 64(b).

CEPA 1999 64(c): Alth ough other factors (such as sustained cellular proliferati on) play an important role, there is likely a genetic component (i.e., mutation, for which DNA-protein crosslinks serve as a marker for potential) in the induction of tumours following the inhalation of formaldehyde. Therefore, formaldehyde is considered to be "toxic" as defined in Paragraph 64(c) of CEPA 1999. For compounds where the induction of cancer through direct interaction with genetic material cannot be ruled out, this approach is consistent with the objective that exposure be reduced wherever possible and obviates the need to establ ish an arbitrary "de minimis" level of risk for the determination of "toxic" under CEPA 1999. However, based on comparison of risks of cancer estimated on the basis of a biologically motivated case-specific model with calculated exposure in air of the general population in Canada, the priority for investigation of options to reduce exposure on the basis of carcinogenicity is considered to be low. While the majority of the population is exposed to concentrations of formaldehyde less than those associated with sensory irritation, continued investigation of options to reduce exposure to formaldehyde in indoor air is recommended as part of an overall program to reduce exposure to other aldehydes considered to be "toxic" under Paragraph 64(c) of CEPA 1999.

Overall conclusion: Based on critical assessment of relevant information, formaldehyde is considered to be "toxic" as defined in Section 64 of CEPA 1999.

3.5 Considerations for follow-up (further action)

Formaldehyde contributes to the photochemical formation of ground-level ozone. It is recommended that key sources of formaldehyde be addressed, therefore, as part of management plans for volatile organic chemicals that contribute to the formation of ground-level ozone. While indications are that concentrations currently in air and water are not causing environmental harm to biota, continued and improved monitoring at sites likely to release formaldehyde are desirable, notably with regards to industrial uses for resins and for fertilizers as well as releases from pulp and paper mills.

Although the priority for investigation of options to reduce exposure in the general environment is generally considered to be low, in relation to carcinogenic potential, in some indoor locations, concentrations are only slightly lower than, and may even approach, the level associated with signs of eye and respiratory tract sensory irritation in humans. Therefore, it is recommended that continued investigation of options to reduce exposure to formaldehyde in indoor air be considered under the authority of acts other than CEPA 1999 as part of an overall program to reduce exposure to other aldehydes (e.g., acrolein, acetaldehyde) in indoor air deemed to be "toxic" under Paragraph 64(c) of CEPA 1999. Where the control of any identified sources falls within the authority of an Act other than CEPA 1999, the results of these investigations should be forwarded to the appropriate authority for further consideration.