Priority Substances List Assessment Report for Aluminum Chloride, Aluminum Nitrate, Aluminum Sulphate

2.0 Summary of Critical Information

2.1 Identity and physical/chemical properties

Aluminum chloride is also known as aluminum trichloride, aluminum chloride (1:3) and trichloroaluminum (ATSDR, 1992). It has the Chemical Abstracts Service (CAS) registry number 7446-70-0, and its chemical formula is AlCl₃. In its hydrated form, AlCl₃·6H₂O, it is called hexahydrated aluminum chloride (CAS No. 7784-13-6). Trade names include Aluwets, Anhydrol and Drichlor.

Synonyms for aluminum nitrate include aluminum trinitrate and aluminum(III) nitrate (1:3). Its CAS registry number is 13473-90-0, and its chemical formula is Al(NO₃)₃. The nonahydrate aluminum nitrate, Al(NO₃)₃·9H₂O (CAS No. 7784-27-2), is the stable form of this compound.

Aluminum sulfate can also be identified as alum, alumsulfate (2:3), aluminum trisulfate, dialuminum sulfate and dialuminum trisulfate. Its CAS registry number is 10043-01-3, and its hemical formula is Al₂(SO₄)₃. Alum is often represented as Al₂(SO₄)₃·14H₂O. It may be found in different hydrated forms. The commercial product, called cake alum or patent alum, is an octadecahydrate aluminum sulfate, Al₂(SO₄)₃·18H₂O.

In addition to these three compounds, aluminum polymers such as polyaluminum sulfate (PAS) and polyaluminum chloride (PAC) are used in water treatment. The general formula for PAS is Al_a(OH)_b(SO₄)_c, where b + 2c = 3a; for PAC, the general formula is Al_a(OH)_bCl_c, where b/a is usually about 2.5 (e.g., Al₂(OH)₅Cl). Mixed aluminum polymers may also be used; their general formula is Al_a(OH)_bCl_c(SO₄)_d, and b/a varies between 0.4 and 0.6.

At room temperature, aluminum chloride is a white, grey or yellow to greenish solid (Budavari et al., 1989). It fumes in air and explodes when in contact with water. Hexahydrated aluminum chloride is either a colourless crystal or a white or slightly yellow crystal or powder. Aluminum nitrate is a crystal that deliquesces rapidly in a humid atmosphere. Aluminum sulfate is a white, lustrous crystal or powder. When heated, aluminum chloride will sublimate at 181°C, and aluminum nitrate and aluminum sulfate will decompose at 135°C and 770°C, respectively (Budavari et al., 1989; Lewis, 1992). Hydrated aluminum chloride, aluminum nitrate and aluminum sulfate are very soluble in water. Physicochemical properties of the three aluminum salts are presented in Table 1.

2.2 Entry characterization

2.2.1 Production, importation, exportation and use

2.2.1.1 Aluminum chloride

Aluminum chloride is used in either anhydrous or hydrated form. In the anhydrous form, it is used as a catalyst, in Friedel-Crafts reactions, in the manufacture of rubber, in the cracking of petroleum and in the manufacture of lubricants. In its hydrated form, it is used by pharmaceutical industries and in the preparation of adhesive, paint pigment, resins, fertilizers, deodorants, antiperspirants and astringents (Germain et al., 1999). In water treatment, it is used mainly in polymeric forms as a coagulant.

In 1996, approximately 13 500 tonnes of aluminum chloride and PAC were produced in Canada, 1670 tonnes imported and 5900 tonnes exported, for an apparent consumption of 9270 tonnes (Table 2). According to the sales reported by the different suppliers, pulp and paper (P&P) industries and municipal water and wastewater treatment plants (MWWTPs) consumed, respectively, 30% and 45% of all aluminum chloride and PAC sold in Canada.

2.2.1.2 Aluminum nitrate

Aluminum nitrate can be used as a chemical reagent, in the leather tanning industry, as an antiperspirant, as a corrosion inhibitor, in uranium extraction and as a nitrating agent (Budavari et al., 1989). In Canada, only one user was identified in a survey done in 1997 by Environment Canada (1997b). This user reported that less than 400 kg of aluminum nitrate were included in fertilizers and exported to the United States in 1996 (Germain et al., 1999).

2.2.1.3 Aluminum sulfate

Aluminum sulfate is used primarily in water and wastewater treatment. There are other applications, however, in the leather industry, in the paper industry, as a mordant in dyeing, in fireproofing and waterproofing of textiles, in manufacturing of resins, in fertilizer preparation and in paint pigment preparation in Canada (Germain et al., 1999). Aluminum sulfate can also be used in waterproofing concrete, in decolorizing petrol, in antiperspirants and pesticides (Budavari et al., 1989) and in the treatment of eutrophic or mesotrophic lakes to reduce the amount of nutrients present in the water. It may also be found in food such as baking powder.

Approximately 252 000 tonnes of aluminum sulfate were produced in Canada in 1996, 15 500 tonnes imported and 7060 tonnes exported (Table 2). The apparent consumption was thus about 261 000 tonnes (Germain et al., 1999). The P&P industries and MWWTPs were the main users of aluminum sulfate with, respectively, 21% and 55% of the sales reported. Some MWWTPs reported that they used PAS. However, the quantities involved are small compared with those of aluminum sulfate or aluminum chloride.

2.2.2 Sources and releases

Aluminum sulfate minerals such as aluminite and alunite occur naturally in Canada in certain restricted geological environments. Aluminum chloride and aluminum nitrate do not occur naturally in the environment. Aluminum can be released from natural aluminum sulfate minerals; however, since aluminum is a common constituent of rocks, where it occurs dominantly in aluminosilicate minerals (e.g., kaolinite, boehmite, clay, gibbsite, feldspar, etc.), which weather and slowly release aluminum to the surface environment, releases of aluminum from aluminum sulfate cannot be distinguished from other natural aluminum releases. These releases would, however, be small compared with releases from aluminosilicate minerals.

Aluminum chloride, aluminum nitrate and aluminum sulfate are produced commercially and used in many applications in Canada. Although they are present in some commercial products, associated releases to the environment are expected to be small. They are used in large amounts in water treatment plants (industrial water, drinking water or wastewater). In this application, aluminum will react rapidly, producing sludge, usually in the form of aluminum hydroxide (Al(OH)₃). Sludge produced by MWWTPs or industries is sent to landfills or spread on land. Sludge purged from clarifiers or accumulated in sedimentation basins of drinking water treatment plants (DWTPs) cannot be released directly to the aquatic environment in many provinces. It is sent to sewers, spread on land or landfilled. Backwash waters (used to clean filters) are also regulated by some provinces; they cannot be discharged directly into an open body of water unless it is shown that there are no adverse effects on the receiving body of water.

Although the majority of the aluminum so released is in particulate hydroxide form, some will be dissolved. Since it is the monomeric aluminum present in the dissolved fraction that can adversely affect organisms, the following discussion considers releases of aluminum in general, focusing mostly on dissolved forms. This approach was necessary because very few studies reported monomeric aluminum levels in the environment or in anthropogenic releases.

2.2.2.1 Natural sources

Atmospheric deposition of aluminum on land or water is small compared with internal releases by weathering and erosion of rock, soil and sediment (Driscoll et al., 1994). Weathering and erosion of "alum"-containing rocks will release aluminum into soils and streams, in part as Al³⁺ and other dissolved anionic and cationic species, depending on pH and the availability of complexing ions (Garrett, 1998). These releases will be small, however, in relation to releases from weathering and erosion of aluminosilicate minerals.

There are no reliable estimates of the amounts of aluminum released to the environment by natural processes on a global scale, most of which comes from natural aluminosilicate minerals. Quantification of total or dissolved aluminum releases in Canada and elsewhere is very difficult and can provide only a rough estimate. Using Garrels et al.'s (1975) global stream flux of 2.05 g/m² per year, we estimated a total aluminum release (including particulate material) of approximately 20.45 million tonnes per year for Canada. Studies of weathering flux in selected

Canadian and U.S. catchments (e.g., Likens et al., 1977; Kirkwood and Nesbitt, 1991) yield similar or somewhat lower estimates (2-20 million tonnes per year) when extrapolated to the whole of Canada.

2.2.2.2 Anthropogenic sources

Very limited information is available on historical releases of the three aluminum salts identified by the Ministers' Expert Advisory Panel (1995). Accidental releases were reported to Environment Canada's National Analysis of Trends in Emergencies System (NATES) database. Between 1974 and 1991, 24 events released 316.2 tonnes of aluminum sulfate, mainly to land, and approximately 80% of the spilled material was recovered. Four accidental releases of aluminum chloride occurred in 1986 and 1987, and the product was not recovered on two occasions, resulting in a total release of 18.18 tonnes (Environment Canada, 1995).

Releases of aluminum salts or aluminum resulting from the industrial use of aluminum salts (excluding municipal water treatment operations) totalled approximately 8800 tonnes in 1996 (Table 3), with most (8124 tonnes) going to surface waters (Germain et al., 1999). Some releases were transferred to municipal wastewater treatment plants (MWTPs) (305 tonnes) or disposed of by landfarming (317 tonnes). Mean total aluminum levels measured or calculated in wastewater released into rivers by P&P industries ranged from 0.46 to 4.8 mg/L. Mean total aluminum levels measured for other types of industries ranged from 0.01 to 2.3 mg/L. Environmental Effects Monitoring (EEM) reports provided by the P&P industries mention the distance from the point of release needed to dilute the effluents to less than 1% in the receiving water body. In some cases, effluents were diluted to less than 1% within a few metres; in others, it took up to 300 km. In these cases, water inputs from other rivers were needed in order to reach the 1% dilution.

Municipal water treatment plants, either DWTPs or MWTPs, are the main users of aluminum chloride and aluminum sulfate. Aluminum salts are used as a flocculant or a coagulant to cause fine materials that are suspended, soluble or both to agglomerate, for subsequent removal via sedimentation and filtration. As part of this agglomeration or coagulation process, most of the aluminum associated with the added aluminum salt hydrolyses to aluminum hydroxide, which precipitates and becomes part of the floc structure. As such, it makes up a part of the sludge generated by the treatment process. A small amount of the aluminum added may stay with the finished water in either colloidal particulate (Al(OH)₃) or soluble form (e.g., Al(OH)₂+, Al(OH)₄-), dictated by the conditions of the treatment process.

Many of the provinces regulate the releases of process wastewater by DWTPs. In many cases, only backwash waters can be released to the receiving environment. In some cases, older plants are allowed to continue their practice if no major improvements are done.

Approximately half of the 240 DWTPs that answered the questionnaire sent by Environment Canada (Environment Canada, 1997b) reported use of aluminum chloride or aluminum sulfate to treat water. Approximately 30% of these release their sludges containing aluminum directly to rivers (in all cases, the receiving environment was circumneutral; Germain et al., 1999). These releases account for 58% of the 574 000 tonnes of sludges disposed of in Canada annually (Table 4). Using the available information, we estimated that aluminum released by DWTPs directly into Canadian rivers and on soil by landfarming practices represents approximately 3000 tonnes. Release to sewers with subsequent treatment by MWTPs is the other main disposal practice (32%). A few DWTPs use spreading on agricultural land or landfarming as a disposal practice; about 1700 tonnes are disposed of this way annually. Alberta, Ontario and Quebec have guidelines for the disposal of sewage sludge on agricultural land; spreading on agricultural land is permitted only when the pH is greater than 6.0 or when liming and fertilization (if necessary) are done.

In a study done with sludge from Calgary and Edmonton, AEC (1987) found that less than 0.02% of aluminum bound with sludge (containing 78 187 mg Al/kg d.w.) was released in water (i.e., 0.20-0.32 mg/L). Srinivasan et al. (1998) studied the speciation of aluminum at six different stages of water treatment at Calgary's DWTP. Total aluminum concentrations ranged from 0.038 to 5.760 mg/L, and dissolved inorganic aluminum concentrations varied from 0.002 to 0.013 mg/L. George et al. (1991) measured monomeric aluminum concentrations of less than 0.06 mg/L in alum sludge from 10 different DWTPs containing up to 2900 mg total Al/L; Calgary's DWTP was one of the plants studied.

Only Calgary's DWTP reported the aluminum content in the backwash water following the cleaning of its filters. Dissolved aluminum levels ranged from 0.07 to 0.44 mg/L, and total aluminum concentrations varied from 0.76 to 3.3 mg/L. The backwash waters from this DWTP were not released to the river but were treated and sold as fertilizer (Do, 1999).

Usually, backwash occurs every 48-72 hours for a given filter, and a complete filter backwash cycle extends over 30-35 minutes, with most of the aluminum being released in the first minutes of the cycle. Sludge containing aluminum is also released at the clarifier or at the sedimentation basin, depending on the equipment present at each plant. Some purge the sludge building up in the clarifier or sedimentation tank for 30 seconds every 10 minutes, for 1 minute every half-hour, or for 5-10 minutes every 2 days, and others "empty" their sedimentation basin once every 3-4 months. Hutchison et al. (1973) reported total aluminum levels in backwash waters, clarifiers and sedimentation tanks for approximately 105 DWTPs in Ontario; mean values were 6.2 mg/L, 133.5 mg/L and 1256 mg/L, respectively. On average, backwashes and "purges" use between 5% and 10% of treated water. The estimated average total aluminum concentrations in effluents are presented in Table 5 for some Canadian DWTPs. These DWTPs are those for which estimated average aluminum levels in effluent were highest among the 30 plants reporting direct releases to the environment (between 5.6 and 155 mg/L).

Not all the MWTPs use aluminum-based coagulants; a quarter of the 287 MWTPs that responded to the Environment Canada questionnaire (Environment Canada, 1997b) treat wastewater with one of the aluminum salts. MWTPs are not permitted to dispose of the sludge produced during water treatment in the rivers. Nearly 90% of the sludges produced are disposed of by agricultural spreading (222 000 tonnes) or landfill (91 000 tonnes) (Germain et al., 1999).

Mean total aluminum levels in the effluent of MWTPs using aluminum salts varied from 0.03 up to 0.84 mg/L, and the maximum value reported by one plant reached 1.8 mg/L. Some plants do not use aluminum-based coagulant but still reported aluminum levels in their effluents; their mean total aluminum levels ranged from 0.003 up to 0.90 mg/L (Germain et al., 1999). These figures are in the same order of magnitude as those reported by Orr et al. (1992) for 10 Ontario MWTPs and by MEF and Environnement Canada (1998) for 15 Quebec MWTPs.

P&P mills are the other major user of aluminum salts. In Ontario, MISA (1990) reported mean total aluminum levels ranging from 0.12 up to 6.1 mg/L for the different mills. After 1995, P&P mills had to respect the federal Pulp and Paper Effluent Regulations promulgated in 1992 under the Fisheries Act. In Quebec, this resulted in a 57% decrease in total aluminum levels in mill effluent, with reported levels ranging from 0.16 mg/L up to 1.89 mg/L (Germain et al., 1999). Total aluminum concentrations reported by two P&P mills and estimated dissolved inorganic aluminum concentrations in receiving water are presented in Table 5; levels are lower than those for most DWTPs.

2.3 Exposure characterization

2.3.1 Environmental fate

The sections below summarize the information available on the distribution and fate of aluminum and aluminum chloride, aluminum nitrate and aluminum sulfate. More detailed fate information is discussed in supporting documents prepared for this characterization (Bélanger et al., 1999; Germain et al., 1999; Roy, 1999a).

2.3.1.1 Air

In air, hydrated aluminum chloride will react with moisture and produce hydrochloric acid and aluminum oxide (Vasiloff, 1991). It is likely that aluminum nitrate and aluminum sulfate will react the same way and form nitric and sulfuric acids, respectively. Since these aluminum salts are usually not emitted to air, the amount of aluminum present in air that is related to the aluminum salts being considered here would be negligible compared with the amount coming from natural erosion of soil.

2.3.1.2 Water

Natural sources of aluminum release to aquatic systems include weathering of rocks, glacial deposits and soils and their derivative minerals and atmospheric deposition of dust particles. The most obvious increases in aluminum concentrations have consistently been associated with environmental acidification (Driscoll and Schecher, 1988; Nelson and Campbell, 1991). Soil minerals such as gibbsite and jurbanite are considered as the primary sources of aluminum release to the aqueous environment, especially in poorly buffered watersheds (Driscoll and Schecher, 1990; Campbell et al., 1992; Kram et al., 1995). In more buffered watersheds, a solid-phase humic sorbent in soil is involved in the release of aluminum (Cronan et al., 1986; Bertsch, 1990; Cronan and Schofield, 1990; Cronan et al., 1990; Seip et al., 1990; Taugbol and Seip, 1994; Lee et al., 1995; Rustad and Cronan, 1995).

The three aluminum salts considered in this report are highly soluble and form various dissolved species on contact with water. The fate and behaviour of aluminum in the aquatic environment are very complex. Aluminum speciation, which refers to its partitioning among different physical and chemical forms, and solubility are affected by a wide variety of environmental parameters, including pH, temperature, dissolved organic carbon (DOC) and numerous other ligands. Metals in solution may be present as dissolved complexes, as "free" or aquo ions, in association with particles, as colloids or as solids in the process of precipitating. The reactivity of aluminum, its geochemical behaviour in the aquatic medium as well as its bioavailability depend on its speciation (Neville et al., 1988).

There are two general types of ligands that can form strong complexes with aluminum in solution. Inorganic ligands include anions such as sulfate (SO₄^2-), fluoride (F-), phosphate (PO₄^3-), bicarbonate (HCO₃^-) and hydroxide (OH-), among others. Organic ligands include oxalic, humic and fulvic acids (Driscoll et al., 1980; Sparling and Lowe, 1996). The relative concentrations of the inorganic and organic ligands will basically determine which type of complex is formed in solution. Interactions with pH (Campbell and Stokes, 1985; Hutchinson and Sprague, 1987; Schindler, 1988; Driscoll and Postek, 1996) and DOC (Hutchinson and Sprague, 1987; Kullberg et al., 1993) are, however, of primary importance to the fate and behaviour of aluminum.

DOC will complex aluminum in water, forming aluminum-organic complexes and reducing concentrations of monomeric forms of aluminum (Farag et al., 1993; Parent et al., 1996). At pH 4.5, 1 mg DOC/L can complex approximately 0.025 mg Al/L, with its complexing capacity increasing as pH increases (Neville et al., 1988). Fractions of dissolved organic aluminum in various rivers in Canada were estimated with MINEQL+ and WHAM models; results varied between 1% and 94%, with the highest values at the lowest pH in the range of pH considered (6.5 -8.5) (Fortin and Campbell, 1999).

Aluminum is a strongly hydrolysing metal and is relatively insoluble in the neutral pH range (6.0-8.0) (Figure 1). In the presence of complexing ligands and under acidic (pH <6) and alkaline (pH >8) conditions, aluminum solubility is enhanced. At low pH values, dissolved aluminum is present mainly in the aquo form (Al³⁺). Hydrolysis occurs as pH rises, resulting in a series of less soluble hydroxide complexes (e.g., Al(OH)²⁺, Al(OH)₂⁺). Aluminum solubility is at a minimum near pH 6.5 at 20°C and then increases as the anion, Al(OH)₄-, begins to form at higher pH (Driscoll and Schecher, 1990; Witters et al., 1996). Thus, at 20°C and pH <5.7, aluminum is present primarily in the form Al³⁺ and Al(OH)²⁺. In the pH range 5.7-6.7, aluminum hydroxide species dominate, including Al(OH)²⁺ and Al(OH)₂⁺. In this range, aluminum solubility is low, and availability to aquatic biota should also be low. At pH >6.7, Al(OH)₄^- becomes the dominant species. Aluminum-hydroxide complexes predominate over aluminum-fluoride complexes under alkaline conditions. However, calculated aluminum speciation in some rivers in Canada indicated that only one river has a significant concentration (>1%) of aluminum-fluoride complexes; this river had a pH below 7 (Fortin and Campbell, 1999).

Aluminum mononuclear hydrolytic products combine to form polynuclear species in solution (Bertsch and Parker, 1996). Aluminum begins to polymerize when the pH of an acidic solution increases notably over 4.5:

2Al(OH)(H₂O)₅²⁺ Al₂(OH)₂(H₂O)₈⁴⁺ + 2H₂O

Polymerization gradually proceeds to larger structures, eventually leading to the formation of the Al₁₃ polycation (Parker and Bertsch, 1992a,b). In nature, conditions that favour the formation of polynuclear forms of aluminum can occur during the liming of acidic aluminum-rich watersheds (Weatherley et al., 1991; Lacroix, 1992; Rosseland et al., 1992) and possibly during the addition of alum to circumneutral waters (Neville et al., 1988; LaZerte et al., 1997).

When released into water, within DWTPs, for example, most of the aluminum associated with the aluminum salts considered in this report hydrolyses to form aluminum hydroxides (Hossain and Bache, 1991). Reactions between aluminum salts, water and associated "impurities" result in the formation of a floc or sludge, which separates from the water phase. A small fraction of the aluminum can stay in the water in either colloidal or truly dissolved form. Barnes (1985) describes the different reactions involved in the formation of aluminum hydroxide in aqueous solution; the overall reaction can be represented by the following equation:

Al₂(SO₄)₃ + 6H₂O 2Al(OH)₃⁰ + 3H₂SO₄

When used to treat sewage water, alum will also react with phosphate, and the reaction gives (Romano, 1971; Barnes, 1985):

Al₂(SO₄)₃ + 2PO₄^3- AlPO₄ + 3SO₄^2-

The aluminum hydroxide present in sludge is expected to remain mostly solid after its release to surface water. Ramamoorthy (1988) showed that less than 0.2% of the aluminum hydroxide present in sludge was released in supernatant water at pH 6 and less than 0.0013% at pH 7.65. In both cases, aluminum hydroxide was present mostly in particulate form. At these pHs, aluminum solubility is low, and kinetics favours the solid form of aluminum hydroxide.

2.3.1.3 Soil

Atmospheric deposition of aluminum is attributed mostly to the deposition of dust particles and is generally low (Driscoll et al., 1994). Aluminum is the third most abundant element in the earth's crust and makes up approximately 8% of its rocks and minerals (Skinner and Porter, 1989). Approximately 75% of Canada is covered by glacial till (Landry and Mercier, 1992); examples of aluminum-bearing minerals inherited from glacial till (i.e., primary minerals) are feldspars, micas, amphiboles and pyroxenes. On the other hand, transformation of primary minerals by chemical weathering reactions results in new solid phases (i.e., secondary minerals). Aluminum-bearing secondary minerals such as smectite, vermiculite and chlorite are often found in Canadian soils developed on glacial till.

Inputs of aluminum into soil solutions usually occur by mobilization of aluminum derived from the chemical weathering of soil minerals. The most important reaction in the chemical weathering of the common silicate minerals is hydrolysis. However, aluminum is not very soluble over the normal soil pH range; thus, it generally remains near its site of release to form clay minerals or precipitate as amorphous or crystalline oxides, hydroxides or hydrous oxides. Silica is much more soluble than aluminum at normal soil pH and is always in excess of the amount used to form most clay minerals, so that some is removed from the soil system in leachates (Birkeland, 1984). In some parts of the world, the extent of chemical transformation by chelation is believed to exceed that by hydrolysis alone. In forest soils of cold and humid regions, such as those of eastern Canada, aluminum is believed to be transported from upper to lower mineral soil horizons by organic acids leached from foliage and the slow decomposition of organic matter in the forest floor (Courchesne and Hendershot, 1997). The movement of aluminum-organic complexes stops when the soil solution becomes saturated (or when the aluminum-to-organic-carbon ratio reaches a critical value), thereby reducing their solubility. In pristine conditions, aluminum is normally retained within the B horizon of the soil. A third important reaction involving aluminum is the transformation of one mineral into another through the exchange of interlayer cations (Sposito, 1996).

Although the dissolution and precipitation reactions of aluminum-bearing minerals are often good indicators of the solubility of aluminum in soils, they are by no means the only pedogenic processes controlling the concentrations of aluminum in soil solutions. Many other processes may partly control the bioavailability of aluminum to plant and soil organisms. Aluminum may be 1) adsorbed on cation exchange sites, 2) incorporated into soil organic matter, 3) absorbed by vegetation or 4) leached out of the soil system (Ritchie, 1995).

In eastern Canada, the atmospheric deposition of strong acids, such as nitric acid and sulfuric acid, has accelerated the soil's natural acidification. The increased H⁺ activity (lower pH) in the soil solution creates a new equilibrium where more Al³⁺ is dissolved in the soil solution, cation nutrients (Ca²⁺, Mg²⁺ and K⁺) are replaced on the soil exchange complex by Al³⁺ and the base cations are eventually leached out of the soil.

There may be significant variation in Al³⁺ solubility with depth in a soil profile (Hendershot et al., 1995). In the surface horizons, the soil solutions tend to be undersaturated with respect to aluminum-bearing minerals; in the lower B and C horizons, aluminum in soil solutions can be expected to be near equilibrium with some aluminum solids. Although the equilibrium concentration is close to that which would be expected if gibbsite were controlling equilibrium, gibbsite has generally not been identified in Canadian soils. Other forms of aluminum, for example, hydroxy interlayered vermiculite, may control aluminum solubility at values close to those of gibbsite. Amorphous aluminum complexed with organic matter may also have a similar pH solubility curve that is a function of the pH-dependent variation in the number of binding sites.

The fluoride and hydroxide complexes are the two strongest groups of inorganic ion associations with aluminum in soil solutions (Nordstrom and May, 1995). In very acidic soils, aluminum in the soil solution is present mainly as free Al³⁺; as pH increases, free Al³⁺ hydrolyses to form complexes with OH^- ions (e.g., AlOH²⁺, Al(OH)₂⁺, Al(OH)₃⁰). Near pH 6.5, aluminum solubility is at a minimum, but it increases at neutral to alkaline conditions because of the formation of Al(OH)₄- (Driscoll and Postek, 1996). According to Lindsay et al. (1989), fluorine, the most electronegative and one of the most reactive elements, is released through the dissolution of fluoride-bearing minerals. In acidic soils (pH <5.5), low-ligand-number complexes such as AlF²⁺ are normally formed. In neutral to alkaline conditions, it is more difficult for F^- to compete with OH^- for aluminum in the soil solution because of the increased level of OH^-. Consequently, aluminum-hydroxide complexes predominate over aluminum-fluoride complexes in alkaline conditions.

The complexation of aluminum with sulfate is weaker than that with fluoride. However, in acidic soils where the sulfate concentration is high, aluminum may also form aluminum- sulfate complexes (Driscoll and Postek, 1996). At low sulfate concentrations, AlSO₄⁺ is the dominant aqueous form, whereas Al(SO₄)₂^- is predominant in soil solutions with higher sulfate concentrations. Brown and Driscoll (1992) showed that several aluminosilicate complexes, including AlSiO(OH)₃²⁺, are present in various regions of the eastern United States and Canada.

It has been shown that most dissolved aluminum in the soil solution of the forest floor is organically bound and that these aluminum-organic complexes become less abundant with increasing soil depth (Nilsson and Bergkvist, 1983; David and Driscoll, 1984; Driscoll et al., 1985). In the Adirondacks of New York, David and Driscoll (1984) found that 82% and 93% of the total dissolved aluminum in the organic horizons of conifer and hardwood sites, respectively, were organically complexed. The proportion of organic to inorganic aluminum decreased at both sites from the organic to the upper mineral horizons and from the upper to the lower mineral horizons. In the soil solutions of the mineral horizons, aluminum-organic complexes represented 67% and 58% of the total aluminum in the conifer and hardwood sites, respectively, which indicates the importance of aluminum-organic complexes in humus-rich forest soils of eastern North America.

The use of alum sludge as a soil amendment is the primary pathway by which aluminum present in the three compounds being considered in this report enters the terrestrial environment. However, the amount of aluminum added to soil through this practice is small in comparison with aluminum naturally present in soil. Moreover, since spreading on agricultural land is permitted only when the pH is greater than 6.0 or when liming and fertilization (if necessary) are done, the solubility (and hence bioavailability) of this aluminum is expected to be very limited.

2.3.1.4 Sediment

Sediments, where the metal is generally considered as biologically unavailable, are an important compartment for aluminum (Stumm and Morgan, 1981; Campbell et al., 1988; Tessier and Campbell, 1990). Aluminum occurs naturally in aluminosilicates, mainly as silt and clay particles, and it can be bound to organic matter (fulvic/humic acids) in sediments (Stumm and Morgan, 1981). At pH >5.0, dissolved organic matter (DOM) can co-precipitate with aluminum, thereby controlling its concentrations in lakes with elevated concentrations of DOM (Urban et al., 1990). DOM plays a similar role in peatlands (Bendell-Young and Pick, 1995). At pH <5.0, the cycling of aluminum in lakes is controlled by the solubility of mineral phases such as microcrystalline gibbsite (Urban et al., 1990). Lakes receiving drainage from acidified watersheds can act as a sink for aluminum (Troutman and Peters, 1982; Dillon et al., 1988; Dave, 1992).

Experimental acidification of lakes and limnocorrals has shown that aqueous aluminum concentrations rapidly increase in response to inputs of acid (Schindler et al., 1980; Santschi et al., 1986; Brezonick et al., 1990). Mass-balance studies have demonstrated that retention of aluminum by sediments decreases as pH decreases (Dillon et al., 1988; Nilsson, 1988). Under such conditions, sediments in acidified watersheds can provide a source of aluminum to the water column (Nriagu and Wong, 1986). Based on calculations of fluxes in acidic lakes, Wong et al. (1989) suggested that sediment is a source of aluminum to the overlying water column.

The release of aluminum hydroxide sludge from DWTPs directly to surface waters is the primary pathway by which aluminum from the three compounds being assessed enters sediment. If the water velocity is low at the point of discharge, much of the sludge released will settle onto the surface of local sediment. Since in Canada the waters receiving such discharges are typically circumneutral, the solubility (and hence bioavailability) of aluminum in the sludge will generally be minimal.

2.3.1.5 Biota

Few studies have examined the accumulation of aluminum by algae. While the algal bioassays conducted by Parent and Campbell (1994) were not specifically designed to determine the effect of pH on aluminum bioaccumulation, their data indicated that the accumulation of aluminum by Chlorella pyrenoidosa increased with the concentration of inorganic monomeric aluminum. In addition, the comparison of assays performed at the same concentration of aluminum but at different pH values showed that aluminum accumulation was suppressed at low pH (Parent and Campbell, 1994). Aquatic invertebrates can also accumulate substantial quantities of aluminum, yet there is evidence that most of the metal is adsorbed to external surfaces and is not internalized (Havas, 1985; Frick and Hermann, 1990). Using the results of Havas (1985), the bioconcentration factor (BCF) for Daphnia magna varied from 10 000 at pH 6.5 down to 0 at pH 4.5. Similar results, e.g., decreasing accumulation of aluminum with decreasing pH, were reported for crayfish (Malley et al., 1988), caddisfly (Otto and Svensson, 1983), unionoid clams (Servos et al., 1985) and a chironomid (Young and Harvey, 1991). Other studies with clams and benthic insects showed no relationship between water pH and tissue accumulation (Sadler and Lynam, 1985; Servos et al., 1985). Frick and Herrmann (1990) found that the largest portion (70%) of the aluminum was present in the exuvia of the mayfly, Heptagenia sulphurea, indicating that the metal was largely adsorbed and was not incorporated into the organism.

BCFs for fish were calculated to range from 400 to 1365 based on results presented in Roy (1999a). Numerous field and laboratory studies have demonstrated that fish accumulate aluminum in and on the gill. It has been suggested that the rate of transfer of aluminum into the body of fish is either slow or negligible under natural environmental conditions (Spry and Wiener, 1991). The initial uptake of aluminum by fish essentially takes place not on the gill surface but mainly on the gill mucous layer (Wilkinson and Campbell, 1993). Fish may rapidly eliminate mucus and the bound aluminum following the exposure episode. For example, Wilkinson and Campbell (1993) and Lacroix et al. (1993) found that depuration of aluminum from the gills of Atlantic salmon (Salmo salar) was extremely rapid once fish were transferred into clean water. The authors suggested that the rapid loss is due to expulsion of aluminum bound to mucus.

Far fewer studies have examined aluminum accumulation in benthic organisms. However, chironomids do not appear to accumulate aluminum to the same degree as other aquatic invertebrates. Krantzberg (1989) reported that the concentration of aluminum in chironomids was <0.3 nmol/g d.w. for the entire body and <0.1 nmol/g d.w. for the internal structures. Most aluminum either is adsorbed externally or is associated with the gut contents of chironomids (Krantzberg and Stokes, 1988; Bendell-Young et al., 1994).

BCFs for terrestrial plants were calculated based on data cited in the review by Bélanger et al. (1999). For both hardwood and coniferous species, the calculated BCF ranged from 5 to 1300 for foliage and from 20 to 79 600 for roots in studies done with aluminum solutions. For those conducted with soil, BCFs were lower for both foliage (0.03-1.3) and roots (325-3526). BCFs calculated for grain and forage crops ranged from 4 to 1260 in foliage and from 200 to 6000 in roots for experiments done with solutions. For soil experiments, the foliar BCF varied from 0.07 to 0.7.

2.3.2 Environmental concentrations

In order to identify aluminum levels in the Canadian environment, federal and provincial databases were consulted. Results obtained were expressed in different forms, according to media and analytical methods used. The methods used usually provide comparable results for a given form (e.g., total, extractable or dissolved) of aluminum. Data that were of questionable quality were not retained.

2.3.2.1 Ambient air

The aluminum content of particulate matter that is smaller than 10 µm in diameter (PM₁₀) in air varies depending upon the region and municipality where measurements are made (Germain et al., 1999). In 30 ambient air sites sampled over a 10-year period (1986-1995), mostly in urban locations across Canada, total aluminum concentrations measured in individual samples of PM₁₀ ranged from the limit of detection (approximately 0.001 µg/m³) to 24.94 µg/m³. The mean for each sampling site varied from 0.046 to 1.31 µg/m³. The lowest mean concentration was measured in Sutton, Quebec, a reference site, and the highest was measured in Vancouver, B.C. The Vancouver site also had the highest maximum value measured in Canada. In general, however, the aluminum salts considered in this characterization are unlikely to have contributed significantly to levels of aluminum measured in ambient air, because they are typically not released to air.

2.3.2.2 Indoor air

In a study conducted in Riverside, California, daytime and nighttime personal exposure PM₁₀samples were collected from 178 non-smokers; concurrently, PM₁₀ and PM_2.5 (particulate matter smaller than 2.5 µm in diameter) samples were collected from stationary monitors inside and outside their residences (Clayton et al., 1993; Thomas et al., 1993). Aluminum was measurable in more than 50% of daytime or nighttime personal exposure PM₁₀ samples but in less than 20% of fixed-location indoor and outdoor PM_2.5 samples. Levels greater than 0.55 mg/m³ were considered measurable. Daytime median concentrations of aluminum were 1.9, 2.5 and 3.4 mg/m³ for the PM₁₀ indoor, outdoor and personal exposure monitors, respectively; the corresponding nighttime concentrations were 0.99, 1.7 and 1.0 mg/m³ (Clayton et al., 1993). Based on a mass-balance model with multivariate analysis, outdoor air was estimated to be the major source of aluminum-containing particles in indoor air, and cooking and smoking were identified as major indoor sources of aluminum in air (Özkaynak et al., 1996).

2.3.2.3 Surface water

Aluminum is a naturally occurring element and is present in all the water bodies in Canada and elsewhere. Aluminum can be analysed under different forms, but historically results were reported mostly as total aluminum because of the low cost and ease of analysis. In many cases, results are also available for extractable or dissolved aluminum. Total aluminum represents all the aluminum present in a water sample, including the particulate fraction. Extractable aluminum includes both the "dissolved" fraction and weakly bound or sorbed aluminum on particles, and "dissolved" aluminum represents the fraction present in a sample filtered through a 0.45-µm membrane. All the bioavailable aluminum is considered to be present in this fraction, but not all the dissolved aluminum is bioavailable. Colloidal aluminum (0.01-0.1 µm) and organic aluminum (aluminum bound with soluble organic ligands) that are included in this fraction are not bioavailable.

At reference lake and river sites across Canada that have not been influenced by effluents from facilities using aluminum salts, mean total aluminum concentrations ranged from 0.05 to 0.47 mg/L, with a maximum value of 10.4 mg/L, measured in British Columbia. Mean extractable aluminum concentrations ranged from 0.004 to 0.18 mg/L, with a maximum value of 0.52 mg/L found in a lake from the Abitibi, Quebec, region. Mean dissolved aluminum concentrations varied from 0.01 to 0.08 mg/L; the highest dissolved aluminum value reported was 0.9 mg/L, in British Columbia.

Aluminum was measured in water taken both upstream and downstream of facilities using aluminum salts and releasing aluminum or aluminum salts, but sampling stations were typically not located close enough to sources to allow the local impact of the effluents to be assessed. Mean total aluminum levels generally varied from 0.002 to 2.15 mg/L, with a maximum value of 28.7 mg/L, measured in the Oldman River, 40 km downstream of Lethbridge, Alberta. Total aluminum levels are usually higher in the Prairies, in rivers with high total particulate matter content. Mean extractable aluminum concentrations ranged from 0.03 to 0.62 mg/L, and the maximum value of 7.23 mg/L was reached in the Red Deer River, at Drumheller, Alberta. Mean dissolved aluminum concentrations were much lower, ranging from 0.01 to 0.06 mg/L. In surface water, the maximum dissolved aluminum concentration (0.24 mg/L) was measured in the Peace River, Alberta (Germain et al., 1999). Concentrations in downstream locations were not consistently elevated in relation to concentrations in upstream locations, suggesting that the impacts of releases of aluminum salts are mostly local.

Although information on the forms of dissolved aluminum present at these monitoring locations was not identified, results of equilibrium modelling suggest that most dissolved aluminum in waters with pH values of 8.0 and higher is in inorganic monomeric forms (Fortin and Campbell, 1999). For the 12 Prairie locations where dissolved and total aluminum levels were reported, pH levels were 8.0 or higher, and dissolved aluminum represented less than 3% of total aluminum (Roy, 1999b). The overall average concentration of dissolved aluminum at these sites was 0.022 mg/L, similar to levels of inorganic monomeric aluminum reported in comparatively pristine Adirondack surface waters (pH from ~5.8 to ~7.2), where most values were around 0.027 mg/L (Driscoll and Schecher, 1990).

Empirical data indicating an increase in aluminum levels in ambient water receiving inputs of aluminum salts were available for only a few locations. A total aluminum concentration of 36 mg/L was attained just downstream of the discharge pipe of the Regional Municipality of Ottawa-Carleton's (RMOC) DWTP in water samples taken during backwash in 1993; samples taken 200 m downstream of the discharge pipe showed a total aluminum level of 0.5 mg/L. In 1994, the total aluminum level reached 11.3 mg/L just downstream of discharge. In the Kaministiquia River, the increase in mean total aluminum noted from upstream to downstream stations corresponds approximately to the inputs from the P&P mill located in Thunder Bay, Ontario. The mean difference of 0.071 mg/L observed in total aluminum concentrations for samples taken on the same day at both stations for the 1990-1996 period is equivalent to the predicted aluminum increase of 0.069 mg/L calculated with the aluminum releases reported by the mill (Germain et al., 1999). For the Ottawa and Kaministiquia rivers, estimated dissolved monomeric aluminum levels were 0.027 mg/L and 0.040 mg/L, respectively. These values were obtained using MINEQL+ model and estimated concentrations in effluents, assuming solubility controlled by microcrystalline gibbsite (Fortin and Campbell, 1999). Using boehmite as the controlling phase provided lower dissolved inorganic aluminum levels (0.005 mg/L and 0.007 mg/L, respectively).

2.3.2.4 Drinking water

Aluminum is present in drinking water from natural sources and due to the use of salts such as alum or polyaluminum chloride as coagulants in DWTPs to remove organic compounds, microorganisms and particulate matter (Health Canada, 1998).

Data on concentrations of aluminum in drinking water were obtained from monitoring programs in five provinces and territories for the years 1990-1998. Samples were collected from municipal distribution systems for treatment plants with surface water sources and from municipal wells. The majority of the data are for an acid-leachable aluminum fraction that generally involves sample acidification in the field with nitric acid and filtration (using 0.4- to 0.45-mm filters) prior to analysis if solids are present. Because it does not require vigorous digestion, the acid-leachable fraction is less expensive and more practical for routine analysis; in the absence of significant amounts of particulate matter, it is assumed to approximate total aluminum.

The frequency of detection of aluminum at the provincial/territorial sites ranged from 35% for the Northwest Territories in 1990-1992 (18/52 sites) to 100% for Ontario in 1996-1997 (124/124 sites) (Facey and Smith, 1993; OMEE, 1998). Detection limits ranged from 0.1 mg/L in Alberta to 60 mg/L in the Northwest Territories (Facey and Smith, 1993; Alberta Environmental Protection, 1998). Arithmetic mean concentrations of aluminum ranged from 20 mg/L in New Brunswick for 1995-1998 to 208 mg/L in Alberta for 1998 (New Brunswick Department of the Environment, 1996, 1998a,b; Alberta Environmental Protection, 1998). The highest mean concentration of aluminum for an individual sampling site, 3300 µg/L, occurred in a distribution system site in Manitoba, and the lowest mean concentration, < 0.1 µg/L (detection limit), in several sites in Alberta (Manitoba Environment, 1996; Alberta Environmental Protection, 1998).

For New Brunswick, data on aluminum levels in a variety of different sources of drinking water, including municipal distribution systems, municipal reservoirs, municipal wells and domestic (private) wells, were obtained for the years 1995-1998. Mean concentrations of aluminum were 20, 47, 20 and 44 µg/L for the municipal distribution systems, municipal reservoirs, municipal wells and domestic wells, respectively. Among the municipal distributions systems, reservoirs and wells, the highest mean concentration of aluminum, 221 mg/L, occurred at municipal well sites, and the lowest, < 2 mg/L (detection limit), occurred in both the municipal wells and distribution systems. Each domestic well was sampled only once, and the highest and lowest aluminum levels recorded were 3500 µg/L and the detection limit, respectively.

2.3.2.5 Soil

According to data provided by the Geological Survey of Canada (Garrett, 1998), median total aluminum levels for soils and their associated glacial till parent materials lie in the range of 5.0-5.6% in the Prairies and 6.1-6.8% in southern Ontario. The Prairie values are slightly lower than those for southern Ontario, probably due to the greater proportion of Prairie surface materials derived from carbonate (limestone and dolomite) and sandstone bedrock sources than in the southern Ontario survey area, where aluminum-rich Precambrian Shield rocks are more abundant (Germain et al., 1999). The Ontario Ministry of the Environment has developed typical range values (OTR₉₈), which indicate the expected maximum concentrations for different elements or compounds. The aluminum OTR₉₈ value for old urban parkland is 2.7% and for rural parkland, 3.0%, based on a partial extraction with nitric acid (OMEE, 1994). In soil collected in a sugar maple (Acer saccharum Marsh.) stand located at St. Hippolyte in the Lower Laurentians, Quebec, oxalate-extractable aluminum levels ranged from 0.1% to 2.9% (Hendershot and Courchesne, 1991). It should be remembered that these total and strong acid-extractable aluminum data provide little indication of the amounts of aluminum that might be bioavailable.

In general, unless the soil pH falls below 4, levels of the Al³⁺ bioavailable form in the soil pore fluids are likely to be low. Hendershot and Courchesne (1991) measured aluminum in soil solution at St. Hippolyte, Quebec. The median total dissolved aluminum level was 0.570 mg/L, the median inorganic aluminum level 0.190 mg/L and the median Al³⁺ level 0.0003 mg/L in samples collected at a depth of 25 cm (pH = 5.5). Total dissolved aluminum was also measured in soil solution in the Niagara, Ontario, region; its level reached 1.214 mg/L (pH 4.2) in untreated soil. Following treatment with lime, aluminum was not detected in soil pore waters, and the pH increased to 4.8-5.5 prior to planting alfalfa (Medicago sativa L.). After three cuts of alfalfa, the pH was elevated to 6.0 in control plots and to 7.5-8.0 in limed plots; the mean total dissolved aluminum level was 0.335 mg/L in pore waters in the control plots and 0.016-0.397 mg/L in limed plots (Su and Evans, 1996).

Data relating to aluminum levels in soils treated with aluminum hydroxide sludges are limited. Near Regina, Saskatchewan, 1100 tonnes of alum sludge from a DWTP were spread on 16 ha of soil at a rate of 75 tonnes per hectare. There was no statistical difference in the mean acid-extractable aluminum level in both control (4.0%) and treated (4.1%) soil (Bergman and Boots, 1997). In a study done for the American Water Works Association, Novak et al. (1995) measured the aluminum content of soil before (pH 4.7 and 5.5 at two sites) and after application of water treatment residuals. The PAC residual contained 2330 mg Al/kg d.w., and the alum residual, 6350 mg/kg d.w. In cropland soil, relatively bioavailable aluminum (measured with a Mehlich III extraction procedure) varied between 405 and 543 mg/kg d.w. (or 0.04% and 0.05%) before the application of the water treatment residuals. Addition of PAC and alum residuals resulted in an increase of bioavailable aluminum to 770 mg/kg d.w. and 1115 mg/kg d.w., respectively. In another experiment, alum residual containing 150 000 mg Al/kg d.w. was applied to forest soil (pH 4.7). Soil analyses done 30 months later showed no differences between the control and the treatment plots for bioavailable and total aluminum.

2.3.2.6 Sediment

Based on limited data, total aluminum levels in Canadian sediments are of the same order of magnitude as those measured in soils, with levels varying between 0.9% and 12.8%. The highest levels were found in Lake St. Louis, Quebec. Of particular interest are aluminum levels measured in sediment of the Ottawa River less than 300 m downstream of a location where backwash water discharges from RMOC's DWTP. The total aluminum content of sediment from a control site was 1.7%, while the value closest to the outfall was 12.5% (Germain et al., 1999). However, based on the data available, the area with high aluminum levels extended no more than a few hundred metres downstream from the discharge point.

2.3.2.7 Vegetation

Aluminum concentrations in vegetation related to the production or use of the aluminum salts considered in this report were available for only a few locations in Canada. Vasiloff (1991, 1992) reported aluminum levels in bur oak (Quercus macrocarpa Michx.) foliage collected from trees near an aluminum chloride producer in Sarnia, Ontario. Total aluminum levels ranged from 25 to 170 mg/kg d.w. in 1989 and from 57 to 395 mg/kg d.w. in 1991. Levels were higher in the foliage of trees closer to the aluminum chloride plant. These levels were below the Ontario Rural Upper Limit of Normal for aluminum in tree foliage (Vasiloff, 1992).¹

Novak et al. (1995) measured aluminum levels in soils before (pH 4.7 and 5.5 at two sites) and after the application of water treatment residuals (PAC and alum sludge), as well as aluminum contents in tissues of corn (Zea mays), wheat (Triticum aestivum) and loblolly pine (Pinus taeda L.) in control and treated soils. Statistical differences in aluminum contents were noted only in corn tissues. Aluminum levels were lower (15.1 mg/kg d.w. vs. 18.6-19.6 mg/kg d.w.) in plants grown in soil treated with 2.5% of PAC water residual than in plants grown in soil treated with 1.34% alum or in controls; however, crop yields (kg/ha) were not lower. Aluminum levels in loblolly pine tissues were not statistically different in trees grown in control (270 mg/kg d.w.) and treated (152-170 mg/kg d.w.) soil.

2.3.2.8 Food

Levels of aluminum were determined in a 1993-1996 market basket survey of various foodstuffs collected from six Canadian cities: Halifax, Montréal, Ottawa, Toronto, Vancouver and Winnipeg (Dabeka et al., 1999). Data included concentrations of aluminum in 124 individual food items of the various food composite groups routinely consumed by the general population of Canada (Environmental Health Directorate, 1998). The fraction analysed was total aluminum based on nitric and perchloric acid digestion of the samples, and detection limits ranged from 0.005 to 7.6 µg/g (Dabeka, 1999). Mean concentrations of aluminum were 0.27 µg/g in eggs, 0.34-1.1 µg/g in mixed dishes and soups, 0.02-1.3 µg/g in vegetables, 0.07-1.5 µg/g in dairy products, 1.8-4.0 µg/g in nuts and seeds, 0.09-4.4 µg/g in beverages (including soft drinks and alcohol), 0.02-4.2 µg/g in fruit, 0.08-5.6 µg/g in foods containing primarily sugar, 0.43-7.0 µg/g in meat and poultry, 0.14-7.2 µg/g in fats, 0.53-12 µg/g in fish and 0.15-165 µg/g in cereal products (Dabeka et al., 1999).

Relatively high levels of aluminum can occur naturally in foods such as raisins (17 µg/g), shellfish (12 µg/g) and cucumbers (6 µg/g) (Dabeka et al., 1999). However, some of the highest mean concentrations that were determined, for cakes (165 µg/g), muffins (93 µg/g), pancakes (85 µg/g) and danishes and donuts (46 µg/g) are, at least in part, due to the use of aluminum-containing food additives in these products. The Food and Drug Regulations of the Food and Drugs Act permit the use of aluminum-containing additives (e.g., calcium aluminum silicate, sodium aluminum silicate, sodium aluminum phosphate, aluminum sulfate, etc.) as agents for anticaking, colouring, emulsifying, gelling, stabilizing, thickening and other uses in a variety of food products. The Food and Drug Regulations indicate that the maximum levels of most aluminum-containing food additives should be in accordance with good manufacturing practice. For other additives, specific maximum levels of use prescribed in the Regulations range from 0.036% (360 µg/g) for aluminum sulfate in egg products to 3.5% for sodium aluminum phosphate in creamed and processed cheese products (Health Canada, 2000).

Data from the six-city market basket survey also included levels of aluminum in infant formula. The mean concentrations of aluminum in ready-to-use milk- and soya-based formula were 0.06 and 0.85 µg/g, respectively (Dabeka et al., 1999). In an earlier study, Dabeka and McKenzie (1992) measured total aluminum in 282 samples of infant formula and evaporated milk sold in Canada. Mean concentrations for milk-based ready-to-use, concentrated liquid and powdered formula were 0.18, 0.27 and 1.4 µg/g, respectively, and the corresponding concentrations for soya-based formula were 1.6, 1.4 and 5.2 µg/g, respectively.² The mean concentration of aluminum in evaporated milk was 0.08 µg/g.

A wide range of concentrations of aluminum have been reported in breast milk from various countries, which may, in part, be due to the variation in methods of analysis and sample preparation used. Mean concentrations range from 9.2 µg/L (0.009 µg/g)³ determined by acid digestion and graphite furnace atomic absorption analysis (GF-AAS) of samples from 15 U.K. women to 380 µg/L (0.37 µg/g) estimated for 42 Croatian women using similar sample preparation and analysis methods (Hawkins et al., 1994; Mandic et al., 1995). Levels of aluminum in breast milk reported in two Canadian studies fall within the range of levels reported for other countries. Using a unique non-destructive rapid neutron activation method, Bergerious and Boisvert (1979) determined an average aluminum concentration of 350 µg/L (0.34 µg/g) for samples from 5 Quebec women, whereas Koo et al. (1988), using GF-AAS analysis, reported a median aluminum concentration of 14 µg/L (0.014 µg/g) for breast milk from 12 Alberta women.

The aluminum content of foods and beverages, especially salty, alkaline or acidic foods, may be increased following cooking in aluminum pots and pans or storage in aluminum containers (ATSDR, 1999). Lione (1983) reported that the aluminum content of tomatoes increased from 1.3 mg/100 g d.w. uncooked to 32 mg/100 g d.w. following 2 hours of cooking in an aluminum pot and to 53 mg/100 g d.w. after overnight storage in the same pot. Greger et al. (1985) noted that cooking in aluminum versus stainless steel cookware significantly increased the aluminum content of applesauce (7.1 vs. 0.12 µg/g w.w.), roast beef (0.85 vs. 0.21 µg/g w.w.), cabbage (3.6 vs. 0.20 µg/g w.w.) and eggs (1.6 vs. 0.13 µg/g w.w.). However, the aluminum content of foods such as green beans, cod, ham and rice was not significantly altered.

Concentrations of aluminum in beverages such as coffee and acidic soft drinks can also be increased following brewing in aluminum coffee percolators and storage in aluminum soft drink cans, respectively (Lione et al., 1984; Muller et al., 1993; Abercrombie and Fowler, 1997).

2.3.2.9 Consumer products

Aluminum compounds are used in a variety of over-the-counter medicinal products sold in Canada, such as antacids and adsorbents for the treatment of heartburn, gas and indigestion (e.g., aluminum hydroxide); analgesics or buffered aspirins (e.g., dihydroxyaluminum aminoacetate); local anti-infectives for cold and canker sores (e.g., aluminum chloride); and hemostatics to control bleeding from minor cuts (e.g., aluminum potassium sulfate). Concentrations of aluminum compounds in over-the-counter products sold in Canada were obtained from the Health Canada Drug Product Database. The Drug Product Database contains brand name, Drug Identification Number (DIN), ingredient and other information for approximately 20 000 drugs approved for use in Canada. Based on the concentrations of specific aluminum compounds, the elemental aluminum contents of orally administered over-the-counter products marketed in Canada are estimated to be 5900-90 000 ppm for antacids and adsorbents, 11 000 ppm for cathartics and laxatives, 16 000 ppm for analgesics, 21 000 ppm for local mucosal anesthetics (heartburn medication) and 5500-110 000 ppm for antidiarrheal agents (Health Canada, 1999a).⁴ For topically (dermally) administered products, concentrations range from 661 ppm for an anti-infective liquid for the prevention of swimmer's ear to 58 000 ppm in an astringent powder used to treat eczema, poison ivy, insect bites, etc. (Health Canada, 1999a).

Compounds such as aluminum chlorohydrate, aluminum hydroxide, aluminum starch octenylsuccinate, aluminum-based dyes and aluminum silicate are also used in antiwrinkle preparations, dentrifices, eye and face makeup, shampoo, lipstick, moisturizers and other cosmetic products sold in Canada. Concentrations of aluminum compounds in cosmetic products were obtained from Health Canada's Cosmetic Notification System, a mandatory system under which manufacturers must submit information including composition data on cosmetics prior to first sale in Canada. Based on available information on the aluminum content of active ingredients listed in the Cosmetic Notification System, estimated concentrations of elemental aluminum in cosmetic products range from 346 to 330 000 mg/g for antiwrinkle preparations, 235 to 3300 mg/g for barrier creams, 1600 to 10 000 mg/g for dentrifices, 2000 to 93 000 mg/g for deodorants, 40 to 1.0 × 10⁶ mg/g for eye makeup, 42 to 35 000 mg/g for face makeup, 210 to 700 mg/g for fragrances, 1600 to 16 000 for hair conditioner, 442 to 30 000 mg/g for hair dye, 158 to 52 000 mg/g for lipstick, 1000 to 100 000 mg/g for manicure preparations, 3500 to 13 000 mg/g for various powders, 78 to 100 000 mg/g for skin cleansers and 235-100 000 for skin moisturizers (Health Canada, 1999b).

2.4 Effects characterization

2.4.1 Ecotoxicology

Below, a brief summary of effects data for the most sensitive terrestrial and aquatic organisms is presented. More extensive descriptions of environmental effects are provided in several reviews (ATSDR, 1992, 1999; Bélanger et al., 1999; Roy, 1999a).

When aluminum salts are added to water, they hydrolyse, and monomeric aluminum can be formed in the dissolved fraction. It is this monomeric aluminum, not the salts, that can adversely affect organisms (Driscoll et al., 1980; Parker et al., 1989; Baker et al., 1990). The following summary focuses, therefore, on the effects of the dissolved (particularly monomeric) forms of aluminum that are produced when aluminum salts dissociate.

Many studies were published regarding aluminum toxicity, but only studies published since 1980 with good quality controls or procedures were retained. Studies with dissolved aluminum levels greater than the aluminum limit of solubility were discarded.

2.4.1.1 Aquatic organisms

Most of the research on the impact of aluminum on aquatic life has been related to the impacts of acid rain. In this report, we examined studies looking at the impact of aluminum at many pHs, but paid particular attention to those conducted in circumneutral conditions similar to those occurring where aluminum salts or aluminum resulting from the use of one of the salts considered are released. Because of this constraint, the most relevant effects data identified were for fish.

2.4.1.1.1 Pelagic

pH is known to have a significant effect on the toxicity of dissolved aluminum. Under acidic conditions, aluminum is most toxic in the pH range 5.0-5.5. At more acidic pH, its toxicity decreases, while at still lower pH, aluminum can offer transitory protection against the toxicity of H⁺ (Muniz and Leivestad, 1980; Baker, 1982; van Coillie et al., 1983; Roy and Campbell, 1995). Elevated concentrations of the cations Ca²⁺ and Mg²⁺ reduce the toxicity of metals (Pagenkopf, 1983; Campbell, 1995), yet there are relatively few results examining the effects of elevated calcium on aluminum toxicity. In fish exposed to aluminum at low pH, elevated calcium improves survival (Booth et al., 1988; Mount et al., 1988; Sadler and Lynam, 1988), reduces losses of plasma ions (Brown, 1981; Sadler and Lynam, 1988; McDonald et al., 1989) and reduces accumulation of aluminum on gills (Wood et al., 1988a,b).

The toxicity of dissolved aluminum is reduced in the presence of inorganic ligands, such as fluorides, sulfates and silicates, as well as organic ligands, such as fulvic and humic acids (Roy, 1999a). It is well established that DOM in particular influences the speciation and bioavailability of aluminum. In laboratory studies with fish, the toxicity of aluminum was reduced in the presence of organic acids, such as citric acid (Driscoll et al., 1980; Baker, 1982), salicylic or oxalic acid (Peterson et al., 1989), humic acid (van Coillie et al., 1983; Parkhurst et al., 1990) and fulvic acid (Neville, 1985; Lydersen et al., 1990; Witters et al., 1990; Roy and Campbell, 1997). In laboratory studies with amphibians (frog eggs and tadpoles), LC₅₀s for aluminum increased (i.e., toxicity was reduced) in the presence of DOM. However, in the field, the effects of DOM in attenuating aluminum toxicity are difficult to separate from the influences of pH and aluminum concentration (Clark and Hall, 1985; Freda, 1991).

Most aquatic toxicity studies involving aluminum have been conducted under conditions of low pH, and a number of these accounted for the solubility of the metal in the experimental design. The general conclusion of these studies is that aluminum toxicity is related to the concentration of dissolved inorganic monomeric aluminum (Roy, 1999a).

At pH <6.0, fish, the salmonids in particular, are among the most sensitive organisms to dissolved aluminum. In soft acidic waters, the LC₅₀ can be as low as 54 µg/L (for Atlantic salmon at pH 5.2), while in chronic studies, a Lowest-Observed-Effect Concentration (LOEC) of 27 µg/L was determined for growth (for brown trout [Salmo trutta] at pH 5.0). Some species of algae show a comparable sensitivity. Parent and Campbell (1994) determined a LOEC of 150 µg/L (as inorganic monomeric aluminum) at pH 5.0 with the alga Chlorella pyrenoidosa. While many invertebrates tolerate elevated levels of aluminum, Havens (1990) found that exposures to 200 µg Al/L at pH 5.0 were extremely toxic to Daphnia galeata mendotae and D. retrocurva. France and Stokes (1987) concluded that stress from aluminum exposure was secondary to the stress of low-pH exposure for survival of Hyalella azteca. Results of other studies also suggest that invertebrates are more sensitive to low pH than to aluminum. Amphibians show a similar sensitivity. Freda (1991) summarized her work by concluding that aluminum can be lethal to amphibians that inhabit soft acidic (pH 4-5) waters if concentrations exceed 200 µg inorganic Al/L.

At pH 6.0-6.5, there are few studies that provide effects estimates in terms of inorganic monomeric aluminum. At pH 6.0, a LOEC of 8 µg/L (inorganic monomeric aluminum) for growth of the alga C. pyrenoidosa can be estimated from the data of Parent and Campbell (1994). Growth of the alga was reduced at this single exposure concentration in media without phosphate. This LOEC is, however, well within the likely range of natural concentrations of inorganic monomeric aluminum in surface water. In comparison, Neville (1985) observed that 75 µg Al/L (as inorganic monomeric aluminum) caused physiological distress to rainbow trout (Oncorhynchus mykiss) at pH 6.1 but not at pH 6.5.

At pH 6.5-8.0, there are few effects data available. At neutral or near-neutral pH, aluminum has a tendency to precipitate, and the chemistry of these solutions is difficult to control. While the toxicity of alum in neutral-pH waters has been the subject of many studies, the results are unreliable, due to extreme variation between replicates of the same exposure concentration and between duplicate experiments (Lamb and Bailey, 1981; Dave, 1985; Hall and Hall, 1989; George et al., 1995; Mackie and Kilgour, 1995). However, a No-Observed-Effect Concentration (NOEC) for respiratory activity at pH 6.5 is provided by the results of the study by Neville (1985), who found that rainbow trout tolerated 75 µg Al/L (as inorganic monomeric aluminum) during exposures at this pH.

At pH >8.0, LOECs for survival of rainbow trout are =1.5 mg/L as total aluminum (Freeman and Everhart, 1971). In a more recent study, Gundersen et al. (1994) reported LC₅₀s for exposures of rainbow trout in the pH range 8.0-8.6. The LC₅₀s at all pHs were approximately the same value, ~0.6 mg/L (range: 0.36-0.79 mg/L) as dissolved aluminum (i.e., filterable through a 0.4-µm filter), and were similar in both acute (96-hour) and longer-term (16-day) exposures at hardness levels ranging from 20 to 100 mg/L (as calcium carbonate). A NOEC for mortality of 0.06 mg dissolved Al/L can be derived from data given for one of the 16-day exposures conducted at 20 mg/L hardness and pH 8.0. Although these concentrations were measured as dissolved aluminum, it is probable that most, if not all, of the dissolved aluminum at this pH was the monomeric aluminate ion, AlOH₄^-.

Finally, in a study done with DWTP sludge from Calgary and Edmonton, Alberta, AEC (1987) concluded that all sludges tested were non-toxic using a microbial test and acutely and subacutely non-toxic to rainbow trout.

2.4.1.1.2 Benthic

Alum can be used to treat eutrophic lakes to reduce the amount of phosphorus present in water or prevent its release from sediment, but no reference to this kind of use was found for Canada. Lamb and Bailey (1981) concluded that a well-planned alum treatment would not result in significant mortality in benthic insect populations. Connor and Martin (1989) measured no detrimental effects on midge or alderly larvae following treatment of Kezar Lake, New Hampshire, sediment, and long-term effects on benthic invertebrates were minimal. Narf (1990) reported that benthic population diversities and numbers increased or remained the same following lake treatment with alum. Smeltzer (1990) observed a temporary impact on benthos after treatment of Lake Morey, Vermont, with an alum/sodium aluminate mixture. Benthos density, already low in the year prior to treatment, and richness were lower following treatment. However, changes were not significant, the benthic community recovered, and two new chironomids appeared the following year.

The Sludge Disposal Committee examined the impact of alum sludge discharge in aquatic environments and concluded that residue will tend to deposit near the point of discharge if the water velocity is low (Cornwell et al., 1987). It could have adverse effects, including development of anaerobic conditions. Roberts and Diaz (1985) related the reduction in phytoplanktonic productivity observed during alum discharge in a tidal stream in Newport News, Virginia, to the reduction in light intensity. Lin et al. (1984) and Lin (1989) found no buildup of sludge in pooled waters in the Vermillion and Mississippi rivers following sedimentation basin cleaning of DWTPs in St. Louis, Missouri. There were no significant differences in types and densities of macroinvertebrates in bottom sediments, and even higher density and diversity were found in some sites.

George et al. (1991) reported that macroinvertebrates located downstream of four DWTPs appeared to be stressed by alum discharges. In the Ohio River, effects seemed temporary and were limited in space. A water-sediment microcosm study done with bottom sediment from the receiving rivers over a 72-day period showed significantly lower oligochaete content in bottom sediment treated with alum sludge. Testing with bentonite gave the same results, and the authors concluded that the smothering effect from sludge may prove to be more important than aluminum content to aquatic organisms. In studies related to wastewater releases by DWTPs, AEC (1984) reported smothering effects related to settled sludge on sediments following their release to rivers in Alberta.

2.4.1.2 Terrestrial organisms

The following discussion focuses on the effects of aluminum on sensitive plant species, since toxicity data identified for other soil-dwelling organisms (e.g., microorganisms, fungi and invertebrates) were very limited. It should be noted, however, that the problem with alum sludge may be associated not only with the direct toxic effects of aluminum on plants, but also with indirect effects related to phosphorus deficiencies (Jonasson, 1996; Cox et al., 1997).

Aluminum's capacity to fix labile phosphorus by forming stable aluminum-phosphorus complexes and hence make it unavailable to plants can be responsible for the observed effects.

Aluminum present in solution, soil solution or soil itself resulted in a decrease in seedling growth, elongation or branching of roots of hardwood and coniferous species at varying levels (Horst et al., 1990; Bertrand et al., 1995; McCanny et al., 1995; Schier, 1996). The most sensitive species was honeylocust (Gleditsia triacanthos L.) (Thornton et al., 1986a,b). All measures of growth, except root elongation, consistently declined as solution aluminum increased, 0.05 mM or 1.35 mg/L being the critical value for a 50% general decrease (pH = 4.0). Since honeylocust is not an important species in Canadian forests and since the results obtained by Thornton et al. (1986b) contradict the results obtained for this species by other researchers, it was decided that the two next Lowest-Observed-Adverse-Effect Concentrations (LOAECs) are more relevant. Hybrid poplar (Populus hybrid) (Steiner et al., 1984) and red oak (Quercus rubra L.) (DeWald et al., 1990) showed a 50% decline in root elongation at an aluminum solution level of 0.11 mM (2.97 mg/L). The most sensitive coniferous species is pitch pine (Pinus rigida Mill.) (Cumming and Weinstein, 1990). Seedlings inoculated with mycorrhizal fungus, Pisolithus tinctorius (Pers.), showed increased tolerance to aluminum, whereas non-mycorrhizal seedlings exposed to 0.1 mM (2.7 mg/L) (pH 4.0) aluminum exhibited decreased root and shoot growth.

In an experiment done with scots pine (Pinus sylvestris L.), Ilvesniemi (1992) found that when nutrition was optimal, pines tolerated high levels of aluminum, but in nutrient-poor solution, their tolerance to aluminum was reduced 10-fold. Hutchinson et al. (1986) and McCormick and Steiner (1978) also observed that pines were tolerant of high levels of aluminum in optimal nutrient solution.

Grain crop and forage crop species were also affected by different levels of aluminum (Bélanger et al., 1999). Wheeler et al. (1992) found that two barley (Hordeum vulgare L.) cultivars and eight common wheat (Triticum aestivum L.) cultivars were particularly sensitive, growth being decreased by more than 50% at aluminum levels as low as 0.005 mM (0.135 mg/L) (pH 4.5). Wheeler and Dodd (1995) also showed a 50% decline in growth of clover species, Trifolium repens, T. subterraneum and T. pratense, at 0.005 mM (0.135 mg/L) aluminum (pH 4.7). In a solution culture study, Pintro et al. (1996) found that the root elongation rate of maize (Zea maize L. HS777 genotype) was also negatively affected at an aluminum level of 0.005 mM (0.135 mg/L) (pH 4.4). In a study done on barley, Hammond et al. (1995) found significant amelioration of the toxic effects of aluminum on root and shoot growth when silicon was added to the solution medium. Silicon amelioration of aluminum toxicity in maize has also been reported (Barcelo et al., 1993; Corrales et al., 1997). In the presence of silicon, aluminum uptake seems to be decreased because of the formation of aluminum-silicon complexes, thus leading to a decrease in aluminum bioavailability.

Wheeler and Dodd (1995) investigated the effect of aluminum on yield and nutrient uptake of some temperate legumes and forage crops using a low ionic strength solution. The solution aluminum levels at which top yield and root yield of 58 white clover cultivars were reduced by 50% ranged from approximately 0.005 to 0.02 mM (0.135 to 0.540 mg/L) (pH 4.5-4.7).

Although inorganic monomeric forms of dissolved aluminum (Al³⁺, Al(OH)₂⁺ and Al(OH)₂⁺) are believed to be the most bioavailable and responsible for most toxic effects (Alva et al., 1986; Noble et al., 1988), information on the concentrations of different dissolved aluminum complexes was not reported in the effects studies reviewed. However, since all of the studies indicating particular sensitivity were carried out in the laboratory in artificial solutions, it is likely that the majority of the aluminum present in these key studies was in inorganic monomeric forms. Considering that solution culture experiments gave lower LOEC values than did sand culture experiments in forest species studies, the effects data reviewed are considered to be conservative estimates of the effects levels for vegetation grown in natural soils.

2.4.2 Abiotic atmospheric effects

Based on available information on the physical and chemical properties of aluminum chloride, aluminum nitrate and aluminum sulfate, and the fact that releases of these substances to the atmosphere in Canada are negligible (Section 2.2.2), these aluminum salts are not considered to be involved in the depletion of stratospheric ozone, tropospheric ozone formation or climate change.

2.4.3 Experimental animals and in vitro

Identified information on neurological effects in laboratory animals of aluminum salts administered orally, dermally or by inhalation is briefly summarized in this section in the context of the limited scope and objectives of this report, which builds on previous initiatives by Health Canada. The results of the majority of these studies have been reviewed elsewhere (Domingo, 1995; Golub and Domingo, 1996; WHO, 1997; ATSDR, 1999). Tables 6 and 7 contain brief descriptions of the species considered, exposure levels, duration and principal neurological results for each study.

Altered performance in a variety of neurobehavioural tests and pathological and biochemical changes to the brain have been observed in studies of the oral administration (i.e., drinking water, diet, gavage) of aluminum salts to mice, rats and monkeys for varying periods of time as adults or during gestation, weaning and/or post-weaning. Interpretation of the results of a number of these studies is limited by designs that focus on testing specific hypotheses rather than examination of a range of neurotoxicity endpoints, the administration of single doses or a lack of an observed dose-response, lack of information on concentrations of aluminum or bioavailability from basal diets, the use of specific ligands to enhance accumulation of aluminum and small group sizes. Indeed, there have been no studies in which a broad range of neurological endpoints (biochemical, behavioural and histopathological) have been investigated in a protocol including multiple dose groups. In general, there has been no significant evidence of Alzheimer's disease-like neuropathology in studies of the oral administration of aluminum to mice and rats (WHO, 1997; ATSDR, 1999).

Neurobehavioural alterations in adult mice and rats following the administration of aluminum salts in drinking water, in diet or by gavage for periods ranging from 1 month to 1 year include decreased locomotor activity (Commissaris et al., 1982; Golub et al., 1989, 1992a; Lal et al., 1993), motor coordination (rota-rod treadmill performance) (Bowdler et al., 1979; Sahin et al., 1995) and grip strength (Golub et al., 1992a; Oteiza et al., 1993), learning and memory deficits (impaired maze performance, active and passive conditioned avoidance responses) (Commissaris et al., 1982; Fleming and Joshi, 1987; Connor et al., 1988, 1989; Bilkei-Gorzó, 1993; Lal et al., 1993; Zheng and Liang, 1998), increased sensitivity to flicker (Bowdler et al., 1979) and increased (Oteiza et al., 1993) and decreased (Golub et al., 1992a) startle responsiveness (Table 6).

Histopathological changes to the brain have been reported in several studies in which rats consumed drinking water or diets supplemented with aluminum salts for periods ranging from 21 days to 6 months (Table 6). In some cases, effects such as cytoplasmic vacuolization and swelling of astrocytic processes and neuronal nuclear vacuolization and inclusions were scattered throughout the brain parenchyma (Florence et al., 1994). In other studies, multifocal neuronal degeneration, abnormal and damaged neurons and reduced neuronal density were identified in specific brain regions (e.g., cerebral cortex, subcortical region, hippocampus, base of brain) (Roy et al., 1991; Varner et al., 1993, 1998; Somova et al., 1997). Somova et al. (1997) observed deformity and vacuolization of nuclei and neurofibrillary degeneration localized in the hippocampus after administering aluminum chloride to rats in drinking water for 6 months, but the neurofibrillary degeneration was distinct from the neurofibrillary tangles (NFTs) of Alzheimer's disease. It has been postulated that the neuropathological effects observed in some of these studies (Varner et al., 1993, 1998; Florence et al., 1994) may have been the result of increased bioavailability of aluminum due to the administration of the citrate and fluoride salts (ATSDR, 1999).

Some of the biochemical changes to the brains of adult mice, rats and monkeys resulting from oral administration of aluminum salts for varying periods of time include alterations to second messenger systems (Johnson and Jope, 1987; Johnson et al., 1992; Hermenegildo et al., 1999), evidence of oxidative damage (lipid peroxidation) and effects on antioxidant systems (Fraga et al., 1990; Lal et al., 1993; Gupta and Shukla, 1995; Katyal et al., 1997; Abd El-Fattah et al., 1998; Sarin et al., 1998). Alterations to membrane lipid content and membrane-bound enzyme activities (Lal et al., 1993; Sarin et al., 1998), effects on cholinergic enzyme activities (Bilkei-Gorzó, 1993; Kumar, 1998), decreased levels and increased phosphorylation of microtubule-associated and neurofilament proteins (Johnson et al., 1992; Jope and Johnson, 1992) and alterations to catecholamine levels (Flora et al., 1991) have also been observed.

Gestational, lactational and/or post-weaning exposure of rats and mice to aluminum salts through the diet or by gavage has produced neurobehavioural effects in offspring; in some cases, these effects have persisted into adulthood (Table 7). Some of the most commonly observed effects include decreased grip strength (Golub et al., 1992b, 1995), reduced temperature sensitivity (Donald et al., 1989; Golub et al., 1992b), reduced auditory startle responsiveness (Misawa and Shigeta, 1993; Golub et al., 1994) and impaired negative geotaxis response (Bernuzzi et al., 1986, 1989a; Muller et al., 1990; Golub et al., 1992b). Decreased activity levels (Cherroret et al., 1992; Misawa and Shigeta, 1993) and locomotor coordination (Bernuzzi et al., 1989a,b; Muller et al., 1990) and impaired righting reflex (Bernuzzi et al., 1986, 1989b) have also been observed.

Exposure of mice and rats to aluminum salts during gestation, lactation and post-weaning also produced some evidence of disturbances in brain biochemistry, such as alterations in brain lipid contents and increased lipid peroxidation (Verstraeten et al., 1998), delayed expression of a phosphorylated neurofilament protein (Poulos et al., 1996), differential effects on choline acetyltransferase activity in various brain regions (Clayton et al., 1992), decreased concentrations of manganese in brain (Golub et al., 1992b, 1993), alterations to signal transduction pathways associated with glutamate receptors and decreased expression of proteins of the neuronal glutamate-nitric oxide-cGMP pathway (Llansola et al., 1999).

2.4.4 Humans

In this section, information on the neurological effects for which associations with aluminum have been reported in humans following oral, dermal or inhalation exposure is briefly summarized in the context of the limited scope and objectives of this report, which builds on previous initiatives by Health Canada. Most of this information has been reviewed elsewhere (Nieboer et al., 1995; WHO, 1997; ATSDR, 1999; Smargiassi, 1999). Initially, a brief summary of the results of available epidemiological studies of associations between aluminum in the general environment and Alzheimer's disease or cognitive dysfunction is presented; more detailed descriptions of the relevant individual analytical epidemiological studies of aluminum in drinking water are provided in Table 8. Information on neurobehavioural effects in workers occupationally exposed to aluminum is also briefly summarized based primarily on the results of previous reviews (Neiboer et al., 1995; WHO, 1997; ATSDR, 1999; Smargiassi, 1999).

An association between Alzheimer's disease or cognitive dysfunction and exposure to aluminum in drinking water has been examined in analytical epidemiological studies (i.e., cohort, case-control and cross-sectional) conducted in populations from Ontario, Quebec, England, France and Switzerland (Table 8). Positive results have been reported for the studies from Ontario (four studies) (Neri and Hewitt, 1991; Neri et al., 1992; Forbes et al., 1992; 1994; 1995a; 1995b; Forbes and Agwani, 1994; Forbes and McLachlan, 1996; McLachlan et al., 1996), Quebec (one study) (Gauthier et al., in press) and France (two studies) (Michel et al., 1991; Jacqmin et al., 1994; Jacqmin-Gadda et al., 1996).

One of the studies from Ontario consisted of a series of analyses of the relationships between cognitive impairment, aluminum and physical-chemical parameters in drinking water conducted on the same sample population from the Ontario Longitudinal Study of Aging (LSA) (Forbes et al., 1992; 1994; 1995a; Forbes and Agwani, 1994). A second study from Ontario (Forbes et al., 1995b; Forbes and McLachlan, 1996) included an analysis of the association between aluminum in drinking water and Alzheimer's disease or presenile dementia recorded on death certificates and a follow-up analysis based on the same dataset but with a different age grouping and exposure categories, and adjusting for different water quality parameters. Neri and Hewitt (1991) examined the relationship between Alzheimer's disease or presenile dementia and aluminum using hospital discharge records from Ontario and Neri et al. (1992) later updated this analysis. The fourth study from Ontario was a case-control analysis of Alzheimer's disease in brain tissue bank donors and concentrations of aluminum in residential drinking water (McLachlan et al., 1996). The degree of overlap between the cohorts from these four studies is unclear. The single positive study from Quebec was a case-control analysis of Alzheimer's disease and exposure to various aluminum species in residential drinking water (Gauthier et al., in press).

Both of the positive studies from France were based on the Principle Lifetime Occupation and Cognitive Impairment in a French Elderley Cohort (PAQUID). In the first study (Michel et al., 1991) a positive association between aluminum in drinking water an Alzheimer's disease was reported, but the results have been discounted because of a reliance on outdated information on concentrations in drinking water (Jacqmin et al., 1994; Smith, 1995; WHO, 1997). The second French study included an initial report of the effect of pH on the association between aluminum and cognitive impairment (Jacqmin et al., 1994) and a follow-up report in which the analysis was expanded to include the effect of silica (Jacqmin-Gadda et al., 1996).

Studies in which no association was observed included three conducted on populations from England (Forster et al., 1995; Martyn et al., 1997; Wood et al., 1988) and one from Switzerland (Wettstein et al., 1991). The studies from England consisted of two case-control studies of Alzeimer's type presenile dementia and Alzheimer's disease in populations from a number of different regions across the country (Forster et al., 1995; Martyn et al., 1997) and a cross-sectional study of the relationship between dementia in patients from northern England and aluminum in drinking water (Wood et al., 1988). The degree of overlap between the populations analysed in these studies is not clear. The negative study from Switzerland was a cross-sectional examination of dementia in octogenerians from Zurich and aluminum in drinking water (Wettstein et al., 1991).

Although not as inherently sensitive as the analytical studies, in four ecological epidemiological studies conducted over the past 15 years, there has been a positive association between the occurrence of Alzheimer's disease and aluminum in drinking water (Vogt et al., 1986; Martyn et al., 1989, Flaten, 1990; Frecker, 1991).

In the analytical epidemiological studies conducted in France and Ontario, physical-chemical parameters in water such as pH, fluoride, and silica were reported to be cofactors for the association between aluminum in drinking water and Alzheimer's disease or cognitive dysfunction. For instance, the association between aluminum and cognitive impairment observed by Jacqmin et al. (1994) was positive for pHs £ 7.3 and negative for pHs > 7.3, and in the follow-up analysis (Jacqmin-Gadda et al.,1996), there was a significant positive association only when both pH and silica were low (7.35 and 10.4 mg/L, respectively). In their examination of the association between cognitive impairment and aluminum, fluoride and pH, Forbes et al. (1992; 1994) observed the lowest risks when levels of aluminum were relatively low (< 84.7 mg/L), levels of fluoride were relatively high ( ≥ 880 mg/L) and pHs were neutral (7.85-8.05). Based on these results, the authors suggested that more bioavailable forms of aluminum may be present at acidic or basic pHs and the effects of these forms of aluminum are decreased in the presence of fluoride (Forbes et al., 1994). Forbes et al. (1995a) also reported variations in associations between aluminum and cognitive impairment with low (< 790 mg/L) and high levels (≥ 790 mg/L) of silica. They observed the lowest risks when levels of aluminum and silica were both low or when levels of both substances were high. These results are consistent with those from the Ontario study of death certificates by Forbes et al. (1995b) in which some of the highest risks for Alzheimer's disease were observed when concentrations of aluminum of ≥ 336 mg Al/L were combined with high pHs (≥ 7.95), low levels of fluoride (< 300 mg/L) or low levels of silica (< 1500 mg/L). However, in the study from England by Martyn et al. (1997) in which no association was observed between exposure to aluminum and Alzheimer's disease, restricting the analysis to subjects exposed to low levels of silica in drinking water (< 6 mg/L) did not alter the overall results.

There have been no or only very weak associations between exposures to aluminum in antacids and Alzheimer's disease in a number of analytical epidemiological studies (Heyman et al., 1984; Graves et al., 1990; Flaten et al., 1991; CSHA, 1994; Forster et al., 1995). Associations between Alzheimer's disease and the use of aluminum containing antiperspirants were reported in two case-control studies, but the interpretation of the results is difficult due to methodological limitations of the studies (e.g., missing data, misclassification due to varying brands and subtypes of antiperspirant with varying aluminum contents, etc.) (Graves et al., 1990; CSHA, 1994). In a recent pilot study, there was an association between Alzheimer's disease and the consumption of foods containing high levels of aluminum food additives; however, the sample size was very small, and the association was significant only for selected categories of food containing additives (Rogers and Simon, 1999).

Subclinical neurological effects have been observed in a number of studies of workers (aluminum potroom and foundry, welders and miners) chronically exposed to aluminum. Many of these studies involved small numbers of workers and involved the assessment of exposure based on occupation rather than urinary, serum or airborne aluminum concentrations, and most involved mixed exposures to various dusts and chemicals. Endpoints examined in different studies varied and for those that were similar, results were not always consistent. The types of adverse neurological effects observed included impaired motor function (Hošovski et al., 1990; Sjögren et al., 1996; Kilburn, 1998), decreased performance on cognitive tests (attention-memory visuospatial function) (Hošovski et al., 1990; Rifat et al., 1990; Bast-Pettersen et al., 1994; Kilburn, 1998; Akila et al., 1999), reports of subjective neuropsychiatric symptoms (Sjögren et al., 1990; White et al., 1992; Sim et al., 1997) and quantitative electroencephalographic changes (Hänninen et al., 1994).

In one case-control study from England (Salib and Hillier, 1996) and two from the United States (Gun et al., 1997; Graves et al., 1998), the relationship between the occurrence of Alzheimer's disease and occupational exposure to aluminum has been investigated. In each study, disease status was defined by standard criteria (e.g., NINCDS-ADRDA and/or DSM),⁵ and exposure to airborne aluminum (e.g., welding fumes, dusts and flakes) was assessed through occupational history questionnaires administered to informants. In none of these studies was there a significant association between occupational exposure to airborne aluminum and Alzheimer's disease.

2.4.5 Supporting data - Aluminum and Alzheimer's disease

In addition to the association between aluminum in drinking water and the occurrence of Alzheimer's disease observed in some epidemiological studies (Section 2.4.4), a number of other lines of evidence have been considered by some as support for a role for aluminum in the development of the disease. For example, the induction of Alzheimer's-like neuropathology in the brains of certain species of experimental animals following administration of aluminum and the detection of elevated levels of aluminum in bulk brain tissue and in NFTs and neuritic plaques from Alzheimer's patients have been suggested as implicating aluminum in the etiology of the disease. Additional cited evidence includes observations of interactions between aluminum and b-amyloid protein (Ab), the principal protein component of neuritic plaques, and the positive results observed following the trial use of an aluminum chelator, desferrioxamine, for the treatment of Alzheimer's disease. Finally, evidence of a role for aluminum in the etiology of two other adverse neurological conditions in humans, dialysis encephalopathy and a Western Pacific variant of amyotrophic lateral sclerosis, has also been considered by some to provide support for the plausibility of the association between aluminum and Alzheimer's disease.

Information on these lines of evidence has been reviewed previously (WHO, 1997; Savory, 2000) and is summarized below.

As discussed previously (Section 2.4.3), the oral administration of aluminum to mice and rats has not produced significant evidence of Alzheimer's-type neuropathology (e.g., Alzheimer's-type NFTs or neuritic plaques). In contrast, administration of aluminum by non-traditional routes of exposure (e.g., intrathecal, intracerebral, subcutaneous, etc.) to certain species (e.g., rabbit, cat, guinea pig, ferret, etc.) can produce a progressive encephalopathy with extensive neurofibrillary pathology (e.g., neurofilament aggregates) (WHO, 1997; ATSDR, 1999). Although this aluminum-induced neurofibrillary pathology has some similarity with that seen in Alzheimer's disease, there are significant ultrastructural and biochemical differences that remain to be resolved. Aluminum-induced neurofilament aggregates are composed of 10-nm straight neurofilaments, whereas Alzheimer's disease NFTs are made up primarily of 20- to 24-nm paired helical filaments (PHFs). At a finer ultrastructural level, the protofilaments in the NFTs from Alzheimer's disease patients are larger than those from aluminum-induced neurofilament aggregates (3.2 vs. 2.0 nm). The staining characteristics of NFTs from Alzheimer's brains also differ from those of neurofilament aggregates in experimental animals (fluorescence with thioflavin S and birefringence with Congo red). Alzheimer's disease NFTs have a different peptide composition from aluminum-induced neurofilament aggregates (hyperphosphorylated tau vs. triplets of neurofilament protein) and different immunoreactivity (anti-tau and anti-ubiquitin vs. anti-neurofilament protein) (Wisniewski and Wen, 1992). However, in a recent study by Huang et al. (1997), aluminum-induced neurofilamentous aggregates reacted with a variety of monoclonal antibodies that recognize the phosphorylated and non-phosphorylated forms of tau, and Shin et al. (1995) reported that aluminum can bind, aggregate and stabilize tau in vitro and in vivo. While neuritic plaques are not a feature of aluminum-induced encephalopathy in animals, immunoreactivity to Ab and its parent molecule, amyloid precursor protein, has been observed in neurons from both rabbits and rats treated with aluminum (Shigematsu and McGeer, 1992; Huang et al., 1997).

The presence of aluminum at the target site has also been investigated in the context of its possible role in the pathogenesis of Alzheimer's disease. However, results of available studies are inconsistent. Elevated levels of aluminum in bulk brain tissue from Alzheimer's disease patients have been observed in several studies (Crapper et al., 1973; Trapp et al., 1978; Yoshimasu et al., 1980; Ward and Mason, 1987; Xu et al., 1992). However, other researchers have reported no such increases (McDermott et al., 1977; Markesbery et al., 1981; Jacobs et al., 1989) or mixed results (Traub et al., 1981). Methods of analysis based on neutron activation analysis and electrothermal atomic absorption spectrometry were employed both in studies where elevated aluminum levels were reported and in studies where no elevations were observed. In the most extensive bulk tissue analysis, there were no differences between the aluminum content of frontal and temporal cortex specimens from 92 Alzheimer's disease patients compared with controls (Bjertness et al., 1996). To refine the analyses beyond bulk brain tissue, microanalytical techniques have been employed to detect aluminum in NFTs. However, the results of these analyses have been somewhat mixed, as Perl and Brody (1980) measured aluminum in neurons containing NFTs using scanning electron microscopy and energy dispersive X-ray spectrometry, but this finding was not confirmed by two other groups using the same technique (Jacobs et al., 1989; Chafi et al., 1991). Good et al. (1992) used a more sensitive technique, laser microprobe mass spectrometry (LMMS), and reported an accumulation of aluminum in neurons containing NFTs from Alzheimer's brains. This result has been questioned due to possible contamination of fixatives and stains (Makjanic et al., 1998) and the contradictory results of an analysis by Lovell et al. (1993) in which similar amounts of aluminum were determined in neurons containing NFTs and NFT-free neurons from Alzheimer's disease brains using LMMS. However, there was no evidence of aluminum contamination upon later analysis of the stains and instrument settings differed between the two studies (Good and Perl, 1993; Lovell et al., 1993; Savory, 2000). Hence, the presence of aluminum in NFT-bearing neurons as identified by LMMS cannot be ruled out. The cores of neuritic plaques have been analysed for the presence of aluminosilicates, with Candy et al. (1986) reporting positive results with energy dispersive X-ray spectrometry, but Landsberg et al. (1993) unable to detect aluminum in plaque cores using particle-induced X-ray emission. Hence, there is little consistency in the results of the most refined studies; consequently, the weight of evidence of elevated aluminum levels in bulk brain tissue, NFTs and neuritic plaques from Alzheimer's patients is unconvincing, at present.

Aβ is a major component of the neuritic plaques that are characteristic of Alzheimer's disease, and there are reports that aluminum is capable of interacting with this protein both in vivo and in vitro . For instance, aluminum enhanced the aggregation of human Aβ into insoluble plaques in vitro (Mantyh et al., 1993). However, the aggregation of Aβ is also promoted by other metal ions, and Bush et al. (1994) observed that a much greater proportion of synthetic human Aβ_1-40 peptide aggregated in the presence of zinc than in the presence of aluminum on an equimolar basis. Aluminum also promotes a change in conformation of Aβ from α -helical to β-turn and random coil structures, which may, in turn, be the mechanism for aluminum-induced blockage of calcium channels formed in membranes by Aβ (Arispe et al., 1993; Exley et al., 1993). In addition, using Aβ_25-35 peptide, Exley et al. (1995) observed that in the presence of glucose, physiological concentrations of aluminum enhanced the aggregation of Aβ fibrils into structures similar to the PHFs in NFTs.

Dialysis encephalopathy is a well-recognized condition associated with aluminum intoxication. It is characterized by disordered speech, problems with swallowing (dysphagia), myoclonic jerks, epileptic seizures, dementia, spasms, inhalation pneumonia and death due to the dysphagia (Nieboer et al., 1995). Dialysis encephalopathy can occur in adults and children with chronic renal failure due to long-term exposure to dialysis fluids and parenteral solutions containing aluminum or orally administered aluminum containing phosphate binders for the control of secondary hyperparathyroidism (Wills and Savory, 1989; Nieboer et al., 1995; WHO, 1997). Although dialysis encephalopathy occurs in a specific subpopulation due to iatrogenic aluminum exposure, Aβ-reactive plaques and NFTs are present in brains from both dialysis encephalopathy and Alzheimer's disease patients (Brun and Dictor, 1981; Scholtz et al., 1987; Candy et al., 1992; Harrington et al., 1994). However, the characteristic PHFs that make up Alzheimer's disease NFTs have not been identified in dialysis encephalopathy brains, and, unlike Alzheimer's disease, aluminum in dialysis encephalopathy appears to be located in the glial cells and walls of blood vessels (Good and Perl, 1988; Wisniewski and Wen, 1992). In addition, it has been noted that the Aβ-reactive plaques in the study of Candy et al. (1992) were diffuse plaques, not associated with clinical disease (Wisniewski et al., 1997), and, in the study of Harrington et al. (1994), the level of Aβ in dialysis encephalopathy brain tissue was not correlated with the amount of aluminum present. Finally, the initial clinical presentation of dialysis encephalopathy is very different from the initial signs of memory loss and difficulties with space and time orientation that are characteristic of Alzheimer's disease (Wisniewski and Rabe, 1992; Nieboer et al., 1995).

Desferrioxamine is an aluminum chelator used in the treatment of dialysis encephalopathy. The results of clinical trials in which this compound was used to treat Alzheimer's disease have been cited as supporting a role for aluminum in the etiology of the disease. In a 2-year single-blind clinical trial, the progression of dementia was decreased in a desferrioxamine treatment group (intramuscular administration) compared with patients given oral placebos (Crapper McLachlan et al., 1991, 1993). However, it is difficult to attribute these results to chelation of aluminum per se, because desferrioxamine chelates other trivalent ions (e.g., Fe³⁺) and has anti-inflammatory properties (Hirschelmann and Bekemeir, 1986), which may have beneficial effects on Alzheimer's disease unrelated to the chelation of aluminum (McGeer and Rogers, 1992). Also, there were important differences between the treatment and placebo groups (treatment: intramuscular administration and unblinded vs. placebo: oral administration and blinded), and the difference between the groups was judged by a videotaped home behavioural assessment, whereas baseline intelligence, memory and speech ability did not differ (Crapper McLachlan et al., 1991, 1993; Nieboer et al., 1995).

Amyotrophic lateral sclerosis is a disease characterized by progressive muscular wasting and weakness, spasticity and hyper-reflexia, with little or no effect on the intellect or oculomotor and sensory functions. A form of this disease unique to the Western Pacific region (e.g., Guam, Kii Peninsula of Japan, West New Guinea) has been linked to aluminum exposure. Although amyotrophic lateral sclerosis is primarily a motor neuron disease, hippocampal neurons from individuals with the Western Pacific form and individuals from the same region with parkinsonism-dementia contain diffuse NFTs that are ultrastructurally and biochemically similar to those of Alzheimer's disease. However, the characteristic neuritic plaques of Alzheimer's disease appear to be absent in these individuals (Strong, 1995). Elevated levels of aluminum and calcium have been measured in NFT-bearing neurons from amyotrophic lateral sclerosis-parkinsonism-dementia patients (Perl et al., 1982; Garruto et al., 1984); within the neurons, aluminum and calcium are co-localized in the perikarya and dendritic processes (Gajdusek, 1985; Garruto and Yase, 1986). Soil and drinking water from the Western Pacific region contain low levels of calcium and magnesium and elevated levels of aluminum (Gajdusek and Salazar, 1982). It has been postulated that in affected individuals, a defect in mineral metabolism combined with chronic dietary deficiencies of calcium and magnesium produce a form of secondary hyperparathyroidism associated with greatly increased gastrointestinal absorption of aluminum (Strong, 1995).

2.4.6 Mode of action

There is evidence that ingested and absorbed aluminum are distributed to the potential target site, the brain, from studies in both laboratory animals and humans. For example, in a number of the studies of neurological effects of orally administered aluminum described in Tables 6 and 7 the accumulation of aluminum in the brains of mice, rats and monkeys from the treatment groups was reported. More sensitive testing in which of the ²⁶Al radioisotope were administered to rats resulted in elevated levels of aluminum in the brain measured from 48 hours to 30 days after dosing (Fink et al., 1994; Walton et al., 1995; Drüeke et al., 1997; Jouhanneau et al., 1997). In humans, the most consistent evidence for the accumulation of aluminum in the brain is that from a series of studies of patients who died of renal failure and were treated with aluminum-containing phosphate binders and dialysis or no dialysis. In general, the highest levels of aluminum were measured in brains from renal failure patients with dialysis encephalopathy, followed by dialysis patients without the condition, non-dialysed patients and controls (Alfrey et al., 1976; Flendrig et al., 1976; McDermott et al., 1978; Arieff et al., 1979; Alfrey, 1980).

Information related to potential modes of action by which aluminum affects the nervous system has been discussed in a number of recent reviews (Strong et al., 1996; WHO, 1997; ATSDR, 1999, Savory, 2000). Consequently, the presentation of this information is limited to a brief summary consistent with the limited scope and objectives of this report which builds on previous initiatives by Health Canada. Most of this information has been reviewed elsewhere.

As discussed previously (Section 2.4.3), the administration of aluminum by non-traditional routes of exposure (e.g., intrathecal, intracerebral, etc.) to certain species (e.g., rabbit, cat, guinea pig, etc.) can result in neuronal cytoskeletal pathology manifested as hyperphosphorylated neurofilamentous aggregates. These neurofilamentous aggregates may arise from aluminum-induced alterations to neurofilament protein gene expression (e.g., reductions in transcribable DNA, effects on DNA repair processes and suppression of mRNA levels for specific neurofilament proteins) and/or aluminum-induced post-translational modifications to neurofilament proteins (e.g., hyperphosphorylation, inhibition of dephosphorylation, increased resistance to proteolysis and cross-linking of neurofilament protein subunits).

Observations of alterations to the activities of choline acetyltransferase, acetylcholinesterase and the neuronal uptake of choline following the administration of aluminum to immature and adult rodents have led to the hypothesis that the neurotoxicity of aluminum may involve an effect on cholinergic neurotransmission.

An additional potential mechanism of aluminum-induced neurotoxicity involves alterations to second messenger systems in the brain. Decreased levels of inositol triphosphate and cAMP have been recorded in rats consuming aluminum in drinking water. in vitro , aluminum decreases agonist-stimulated inositol phosphate accumulation in brain slices. Aluminum may also affect calcium-dependent signalling and processes by inhibiting voltage-sensitive channels for the entry of calcium into neurons and the extrusion of calcium from the neuronal cytosol by Mg²⁺-ATPase.

Administration of aluminum to rodents can result in enhanced lipid peroxidation in the brain. Rather than a direct effect of aluminum, this appears to occur via an acceleration of iron-induced peroxidation. In addition, aluminum can suppress antioxidant systems and alter membrane lipid content and composition in the brain.

Aluminum can inhibit the activities of two enzymes that are important to glucose metabolism in the brain, namely hexokinase and glucose-6-phosphate dehydrogenase. The observation of inhibition of these two enzymes by aluminum coupled with that of reduced glucose metabolism in the brains of rats chronically administered aluminum and Alzheimer's patients have led to the hypothesis that aluminum-induced neurodegeneration may involve selective effects on glucose metabolism.

Alterations to the permeability of the blood-brain barrier may also be relevant in the induction of aluminum-induced neurotoxicity. Increases in the permeability of the blood-brain barrier to sucrose, thyroxine, cortisol, prolactin, growth hormone and luteinizing hormone have resulted from the intravenous or intraperitoneal administration of various aluminum salts to rats and mice. Studies in mice administered aluminum chloride indicate that the increases in permeability may be due to selective effects on specific transport systems.

While there is evidence, therefore, for the interaction with and impact of aluminum on different components of the neurological system, available data are inadequate to serve as a basis for a hypothesized mode of action of aluminum in inducing specific neurological disorders such as Alzheimer's disease.

---

¹Fugitive emissions of aluminum chloride and subsequent hydrolysis, resulting in the formation of hydrochloric acid, were responsible for the damage to trees, including death, that was observed at one location. The company ceased its operations in the mid-1990s. No such damage was reported near aluminum sulfate plants.
²Values for milk- and soya-based concentrated liquid and powdered formula were for undiluted products (Dabeka and McKenzie, 1992).
³Breast milk density assumed to be 1.03 g/mL (Environmental Health Directorate, 1998).
⁴Concentrations expressed in ppm (parts per million) are equivalent to µg/g for products in tablet or powder form or mg/L for products in liquid, suspension or jelly form.
⁵NINCDS is the National Institute of Neurological and Communicative Disorders and Stroke. ADRDA is the Alzheimer's Disease and Related Disorders Associations. DSM is the Diagnostic and Statistical Manual of Mental Disorders, American Psychiatric Association.