Priority Substances List State of the Science Report for Aluminum Chloride, Aluminum Nitrate and Aluminum Sulfate

3.0 Risk Characterization

3.1 Environment

3.1.1 Gerneral Considerations

The aluminum salts considered in this report are produced and used in large quantities in Canada. In many applications, aluminum chloride, aluminum nitrate and aluminum sulfate become part of fabricated products and are not released to the environment. Water treatment uses a large amount of aluminum chloride and aluminum sulfate for the removal of fine suspended or soluble materials and is an important source of aluminum release to the aquatic and terrestrial environments (Germain et al., 1999).

Given its physical properties, most of the aluminum associated with the aluminum salts used in water treatment hydrolyses to aluminum hydroxide and becomes part of the floc structure, which settles in the form of sludge. A small amount of the aluminum added may stay in finished water or sludge in either colloidal particulate (Al(OH)₃) or soluble (e.g., Al(OH)₂⁺, Al(OH)₄^-) form and, in jurisdictions where this is permitted, be released to the aquatic environment when filters are backwashed, clarifiers purged or sedimentation basins emptied. Sludge may be released to sewers or spread on agricultural land. When released to aquatic environments, solid aluminum hydroxide present in process wastewater will remain in solid form. Aluminum present in soluble form may react with soluble materials present in water and precipitate, or it may stay in soluble form.

Based on information on the sources and fate of aluminum salts in the ambient environment, biota are expected to be exposed to aluminum resulting from the use of aluminum salts primarily in water and sediment receiving alum sludges discharged from DWTPs and in soil receiving applications of such sludges. The focus of the environmental risk characterization will therefore be sensitive aquatic and terrestrial organisms exposed to these sources of aluminum.

3.1.2 Environmental risk characterization

3.1.2.1 Aquatic organisms

3.1.2.1.1 Pelagic

Environmental exposure to aluminum in water is expected to be greatest in areas near direct releases of process wastewater to the aquatic environment by DWTPs. Unfortunately, measured values in receiving environments following direct releases are available only in the Ottawa River, when RMOC is backwashing filters. However, aluminum levels in effluents were calculated for the 30 DWTPs that reported direct wastewater releases to the receiving environment in a municipal survey (Germain et al., 1999). Table 5 presents estimated total aluminum concentrations in backwash waters for 13 DWTPs, including those with the highest estimated levels. The aluminum level in RMOC's DWTP's wastewater was quite similar to that measured at the beginning of the plume in the river (30 mg/L vs. 36 mg/L). P&P mills also use aluminum salts, but their releases are generally one order of magnitude lower than those of DWTPs (Germain et al., 1999). Table 5 includes total aluminum levels present in effluent from two typical Canadian mills.

Measurements of total concentrations of a metal can rarely be correlated directly with their biological effects. Metal in particulate form is generally not bioavailable, and the formation of complexes with inorganic (e.g., OH^-, SO₄^2-) or organic (e.g., fulvic acid) ligands can reduce the bioavailable fraction of the dissolved form of a metal. In order to estimate the amount of bioavailable aluminum present in rivers, we performed speciation with MINEQL+ and WHAM models. We calculated the level of dissolved inorganic monomeric form of aluminum following the releases of wastewater from selected DWTPs and two P&P plants, using the estimated aluminum level in effluents and assuming a 1:10 dilution. For the DWTPs considered, average concentrations of dissolved inorganic monomeric forms of aluminum (which are assumed to be the bioavailable forms) varied from 0.027 to 0.348 mg/L during backwash events, assuming that microcrystalline gibbsite is controlling the aluminum solubility. According to Hem and Robertson (1967), the precipitate formed when the pH of water is in the 7.5-9.5 range has a solubility similar to that of boehmite. This precipitate will evolve to bayerite, a more stable and insoluble form of aluminum hydroxide, within a week. If it is assumed that boehmite is controlling the solubility, dissolved aluminum levels would be lower, ranging from 0.005 to 0.059 mg/L (Table 5). For the two P&P mills considered, the dissolved aluminum values were among the lowest, whatever form is controlling the aluminum solubility.

The calculated dissolved aluminum concentration of 0.348 mg/L represents the saturation concentration, assuming that microcrystalline gibbsite controls solubility when aluminum salts are used to treat drinking water. This value was calculated for a location in the Canadian Prairies, where the pH of receiving waters (8.38) and solubility were the highest of all sites examined (Fortin and Campbell, 1999). As noted in Section 2.2.2.2, backwash events last for about 30 minutes and occur every 48-72 hours for each filter at a DWTP. If it is assumed that most DWTPs have about 20 filters (small DWTPs have fewer filters), it is estimated that concentrations in receiving waters near the point of discharge could be as high as 0.348 mg/L as much as 10% of the time. The rest of the time aluminum concentrations would approach background values, which, for locations on the Prairies, are likely on average to be about 0.022 mg/L as monomeric inorganic aluminum (see Section 2.3.2.2). The temporally weighted concentration of dissolved monomeric aluminum at this location averaged over a period of several days would therefore be about 0.055 mg/L. This concentration was taken as a conservative (reasonable worst-case) Estimated Exposure Value (EEV) for waters close to discharge points.

Because aluminum releases reported by DWTPs occur in circumneutral to neutral waters, two Critical Toxicity Values (CTV) corresponding to the pH of waters where releases occur could be chosen. The work of Neville (1985) provides a NOEC of 0.075 mg/L as inorganic monomeric aluminum, based on the absence of deleterious effects on ventilation and respiratory activity of rainbow trout at pH 6.5. This CTV is considered valid for the pH range 6.5-8.0. A second CTV for alkaline conditions (pH >8.0) is based on the work of Gundersen et al. (1994), who determined similar LC₅₀s (~0.6 mg dissolved Al/L) during several experiments in the pH range 8.0-8.6 and water hardness range 20-100 mg/L (as calcium carbonate). A NOEC for mortality of 0.06 mg dissolved Al/L can be derived for rainbow trout from data given for one of the 16-day exposures at 20 mg/L hardness and pH 8.0. The chemical concentrations in Gundersen et al. (1994) are expressed as "total" and "dissolved" aluminum; there was, unfortunately, no attempt to identify the forms of dissolved aluminum present. At the experimental pH, it is probable that a good proportion of the dissolved aluminum was the monomeric aluminate ion. Since the pH in waters for which the EEV was estimated is 8.38, the corresponding CTV is 0.06 mg/L as dissolved inorganic monomeric aluminum.

In determining Estimated No-Effects Values (ENEV) for aluminum, the nature of the biological response was considered, since some organisms respond to a narrow aluminum concentration range. This results in an abrupt "threshold" where an evident biological response occurs, with no observable effects at slightly lower concentrations (Hutchinson et al., 1987; Roy and Campbell, 1995). Consequently, since the CTV chosen is a NOEC, the application factor used to derive an ENEV from the CTV was 1.0. Aluminum being a natural element, it is also useful to consider whether the ENEV is within the range of natural background concentrations. Although based on limited data, on an overall basis, the 90th-percentile value for dissolved aluminum at sampling stations located upstream of points of discharge of aluminum salts is 0.06 mg/L. It should be noted that only a portion of this dissolved aluminum is in inorganic monomeric forms (corresponding to the ENEV). Thus, the 90th-percentile value for inorganic monomeric aluminum in uncontaminated water is expected to be less than 0.06 mg/L.

The reasonable worst-case quotient for receiving water can therefore be calculated as follows:

Since this conservative quotient is relatively close to 1, it is helpful to consider further the likelihood of biota being exposed to such concentrations in Canada.

It is likely that chemical equilibrium modelling overestimates inorganic forms of aluminum in solution, since it appears to overestimate dissolved aluminum. One reason for the overestimate is that a large fraction of the aluminum released from DWTPs during backwash events is probably in solid form, while calculations used to estimate the EEV assumed that all of the aluminum was in dissolved form (Germain et al., 1999). This "solid" aluminum would not tend to participate in solution reactions and would precipitate. Dissolved concentrations may also be overestimated because of the assumption that the solubility of aluminum is controlled by microcrystalline gibbsite. Based on limited data on concentrations of dissolved aluminum at different treatment steps at one Canadian DWTP, solubility may be controlled by less soluble forms of aluminum hydroxide, such as boehmite (Fortin and Campbell, 1999).

The possibility that modelled concentrations overestimate actual values is further supported by data for two sites on the North Saskatchewan River, where the dissolved inorganic aluminum concentrations predicted by modelling are 0.110 and 0.099 mg/L, while the measured concentrations at these sites are 0.005 and 0.010 mg/L (Roy, 1999b).

In a study done with sludge from Calgary and Edmonton, AEC (1987) concluded that all sludges tested were non-toxic using a microbial test and acutely and subacutely non-toxic to rainbow trout. In addition, as noted previously in Section 2.2.2.2, Srinivasan et al. (1998) studied the speciation of aluminum at six different stages of water treatment at Calgary's DWTP. The total aluminum concentration ranged from 0.038 to 5.760 mg/L, and the dissolved inorganic aluminum concentration varied from 0.002 to 0.013 mg/L. George et al. (1991) measured <0.06 mg monomeric Al/L in alum sludge from 10 different DWTPs containing up to 2900 mg total Al/L. These results show that the concentration of dissolved aluminum in process wastewaters is less than the ENEV.

3.1.2.1.2 Benthic

Toxic effects due to exposure to high concentrations of aluminum are unlikely, because of the solubility constraints in receiving waters discussed above. However, based on the limited effects data available, alum sludge released from DWTPs can deposit and form a blanket over sediments in rivers with slow water velocity, and macroinvertebrate populations may be stressed due to a lack of oxygen and carbon sources on which to feed (George et al., 1991). AEC (1984), for example, reported smothering effects related to settled sludge on sediments following disposal to rivers in Alberta. Usually, however, the influence of sludge deposit is limited in space, typically extending only a few hundred metres downstream. Consequently, population-level impacts are expected to be minimal. Furthermore, it is reported that when the sludge blanket is removed, the systems recover.

3.1.2.2 Terrestrial organisms

Terrestrial organisms are exposed to added aluminum when alum sludges from DWTPs are applied to agricultural soils.

The lowest level of dissolved aluminum reported to adversely affect terrestrial organisms is 0.135 mg/L, which can reduce root and seedling growth in sensitive grain and forage crops. This concentration was therefore selected as the CTV, assuming that most of the dissolved aluminum was in inorganic monomeric forms. Considering that this CTV was derived from experiments using solution cultures, the effects data on which the CTV is based could overestimate the sensitivity of crops grown in soils in the field. Because of that, the fact that many species were affected at the same low level and the fact that aluminum is naturally present in soil, an application factor of 1.0 was applied to the CTV to derive the ENEV. The conservative ENEV for soil-dwelling organisms is therefore 0.135 mg dissolved monomeric Al/L.

No data were identified on concentrations of dissolved aluminum in soils that have received applications of alum sludge. However, as was noted in Section 2.2.2.2, spreading on agricultural land is permitted in Canada only when the pH is greater than 6.0 or when liming and fertilization (if necessary) are done. Thus, the pH of receiving soils will likely be in the circumneutral range, where the solubility of aluminum is at a minimum. Based on results of equilibrium modelling, with the total dissolved aluminum concentrations being controlled by the precipitation of microcrystalline gibbsite, total dissolved aluminum concentrations would not exceed the ENEV unless soil pHs were less than about 5.1 (Belanger et al., 1999). Because it is very unlikely that the pHs of soils receiving alum sludge applications will be this low, it is very unlikely that the ENEV of 0.135 mg/L is exceeded in Canadian soils receiving such applications.

The expectation that the solubility and hence bioavailability of aluminum in sludges applied to agricultural soils will be extremely limited is supported by data on aluminum levels in plants growing on such soils. For example, aluminum in yellow mustard seed (Sinapsis alba) and Durum wheat seed (Triticum turgidum var. durum) collected from plants grown in soil amended with alum sludge from Regina's DWTP were found to be not statistically different from those of seeds collected in control plots (Bergman and Boots, 1997).

Finally, although it has been noted that aluminum in the sludge can fix labile phosphorus by forming stable aluminum-phosphorus complexes and hence make it unavailable to plants, causing deficiencies (Jonasson, 1996; Cox et al., 1997), this is unlikely to occur when soil receiving sludge is also fertilized as required in Canada.

3.1.2.3 Uncertainty and Recommendations

There are a number of uncertainties in this risk characterization. Regarding effects of aluminum on pelagic organisms, there are only a few studies conducted at circumneutral pHs (6.5-8.0), conditions similar to those of aquatic environments receiving releases from DWTPs. There are also uncertainties associated with the decision to use an application factor of 1.0 to derive an ENEV for pelagic organisms - a choice that was made considering concentrations of aluminum in uncontaminated waters and the biological response of organisms to a narrow concentration range, resulting in an abrupt "threshold" where biological response occurs.

There are uncertainties associated with levels of aluminum released by DWTPs and with the levels and form of aluminum present in the aquatic environment. The use of the MINEQL+ and WHAM models provided aluminum results higher than those measured in the receiving environments when calculations were done assuming that aluminum solubility is controlled by microcrystalline gibbsite. When calculations were done with the boehmite form of aluminum hydroxide, levels were much lower than what was calculated with the microcrystalline form (Fortin and Campbell, 1999). Addition: Measurement and speciation of aluminum following the releases of DWTP's wastewaters would confirm the estimated levels and forms provided by MINEQL+ and WHAM models.

Other uncertainties exist relating to the impact of alum sludge releases on benthic organisms. There are some indications that sludge releases, whatever the flocculant used, may have a smothering effect on benthos. Given the potential for localized impacts on benthic organisms resulting from direct releases of sludge (not only alum sludge) from DWTPs to surface waters, it is recommended that this practice be discouraged even if the level of total aluminum measured in sediments in one site in Canada influenced by alum sludge is of the same order of magnitude as concentrations measured in unaffected areas elsewhere in Canada. Provisions to manage sludges should be incorporated in the planning stage before construction of new DWTPs and when an old facility is being upgraded.

Because of the increase in aluminum solubility in acidic water, releases of backwash waters, clarifier purge waters or waters from sedimentation basins should be discouraged in lakes and rivers where the pH can be less than 6.

In relation to terrestrial organisms, there are uncertainties associated with the limited data available for effects on soil-dwelling organisms other than plants. The lack of information on aluminum levels in pore waters of soils receiving applications of alum sludge is not considered critical, since these levels are constrained by theoretical limits on solubility that are below the ENEV for sensitive vegetation.

3.2 Environment upon which life depends

Based on available information on releases and their physical and chemical properties, aluminum chloride, aluminum nitrate and aluminum sulfate do not deplete stratospheric ozone, contribute to the formation of ozone in the troposphere or influence climate change.

3.3 Human health

3.3.1 Population exposure

3.3.1.1 Estimated population exposure

The average daily intake of aluminum has been estimated based on the levels in air, drinking water, soil and food and the amounts consumed by the various age groups of the general population of Canada (Table 9) (Environmental Health Directorate, 1998), although these intakes should be considered in the context of relative bioavailability from various routes of exposure (Section 3.3.1.2). The total average daily intake for all age groups of the general population of Canada from all sources is estimated to range from 113 to 598 mg/kg bw/ day. Based on the available data, the greatest source of exposure to aluminum for children, teens, adults and seniors of the general population of Canada is food. For infants and toddlers, the greatest source of exposure is the inadvertent ingestion of soil.Inhalation of airborne aluminum contributes only a small amount to the total daily intake, as estimated intakes range from <0.01 to 0.10 mg/kg-bw per day for ambient air and from 0.17 to 1.0 mg/kg-bw per day for indoor air.

For those who regularly use aluminum-containing over-the-counter oral therapeutic products, these products represent the major source of daily aluminum intake. Based on the manufacturers' maximum recommended daily doses, estimated daily intakes of aluminum from these products range from 1.4 to 21 mg/kg-bw per day (Table 10). The highest intakes are for children, teens, adults and seniors who use antidiarrheal agents and for adults and seniors who use antacids and adsorbents.

The very limited data available indicate that certain aluminum-containing over-the-counter oral therapeutic products may be used regularly by only a small fraction of the Canadian population. In the Statistics Canada National Population Health Survey, 0.2% of a sample population of 17 011 individuals over the age of 12 indicated that they had consumed one aluminum-containing antacid within 2 days of being surveyed. For the same population, 0.7% stated that they had taken one aluminum-containing salicylate (analgesic) within 2 days of being surveyed (Statistics Canada, 1995).

As shown in Table 11, estimated adult dermal exposure to aluminum through the use of aluminum-containing cosmetic products in Canada ranges from 0.01 to 1500 mg/kg-bw per day. The greatest cosmetic sources of daily dermal exposure to aluminum are skin cleansers, hair dye and hair conditioners.

3.3.1.2 Bioavailability

Data on the fractional absorption (bioavailability) of aluminum compounds via the inhalation, oral or dermal routes of exposure have been reviewed by Yokel and McNamara (2000). Estimates of bioavailability from various routes and media of exposure are necessarily crude owing to the limitations of the data on which they are based and the considerable number of influencing factors (Section 3.3.1.3). Generally, the bioavailability of aluminum via inhalation, ingestion and dermal exposure appears to be relatively low - in the range of tenths of a percent to a few percent.

There are no data available on the bioavailability to humans of aluminum from ambient or indoor air. An estimate of 1-2% bioavailability (Yokel and McNamara, 2000) can be derived from the occupational exposure studies of Gitelman et al. (1995) in which pre- and post-work shift serum and urinary aluminum levels were measured in aluminum industry workers exposed to respirable particles containing aluminum at a median concentration of 25 mg/m³ and Pierre et al. (1995) in which 24-hour urine samples were collected from workers occupationally exposed to approximately 200-500 mg soluble aluminum/m³. However, whether the airborne aluminum was absorbed from the lungs following inhalation, from the gastrointestinal tract following mucociliary clearance of the lungs or from the olfactory tract was not investigated in the identified occupational studies of populations for which exposure would be considerably greater than that expected in the general environment.

The oral bioavailability of aluminum in drinking water has been assessed using a variety of ²⁷Al- and ²⁶Al-radiolabelled aluminum salts (e.g., aluminum hydroxide, aluminum chloride, aluminum sulfate, aluminum citrate, etc.) in water administered to rats, rabbits and humans. Yokel and McNamara (2000) estimated a range of 0.1-0.5% based on the more recent studies in which 1) doses (0.13-3.2 mg/kg-bw per day) were similar in magnitude to estimated daily intakes from drinking water, 2) the form of aluminum used (aluminum hydroxide, aluminum chloride or aluminum in municipal tapwater) is likely to have speciation similar to that of aluminum in drinking water, and 3) fractional absorption was based on urinary excretion, serum levels or plasma aluminum versus time curves for oral versus intravenous administration (Hohl et al., 1994; Drüeke et al., 1997; Jouhanneau et al., 1997; Schönholzer et al., 1997; Priest et al., 1998; Stauber et al., 1999).

Estimates of bioavailability from the limited number of studies that have examined aluminum in food and beverages are within a range similar to that for drinking water (Yokel and McNamara, 2000). Priest (1993) estimated 0.1% bioavailability of aluminum from food based on previously published estimates of daily intake from food (15 mg), urinary excretion rate (25µg/day) and the percentage of aluminum retained in the body of a human volunteer (5%) following intravenous injection with ²⁶Al-citrate. More recently, Stauber et al. (1999) measured the 24-hour urinary aluminum excretion in 29 male and female volunteers who consumed low-aluminum water and meals and instant tea with known amounts of aluminum and estimated the bioavailability from food to be 0.53%.

Data on the oral bioavailability of aluminum from drugs are restricted mainly to studies of aluminum-containing compounds (e.g., sucralfate, aluminum lactate, Zeolite A^® and aluminum hydroxide) used as active ingredients in antacids, phosphate binders, toothpastes and other products (Yokel and McNamara, 2000). Estimates for sucralfate (basic aluminum sucrose sulfate), an antacid, range from 0.005% based on measurements of urinary aluminum in human volunteers (Haram et al., 1987) to 0.60% calculated from plasma aluminum versus time curves for oral versus intravenous administration in rabbits (Yokel and McNamara, 1988). For aluminum lactate, estimates of bioavailability range from 0.02% in rats (Wilhelm et al., 1992) to as high as 1.9% for rabbits administered high doses of the compound (540 mg/kg-bw) and fasted before administration (Yokel and McNamara, 1985). Cefali et al. (1996) reported the bioavailability of Zeolite A^®, an aluminum silicate inducer of osteoblast proliferation, as 0.023-0.032% based on plasma aluminum versus time curves for oral versus intravenous dosing in dogs. However, in the controls, there were large variations in plasma aluminum, while only small increments were observed following Zeolite A^® treatment. Moreover, the pharmacokinetic parameters for aluminum did not incorporate baseline adjustments (Yokel and McNamara, 2000). Although a wide range of estimates of bioavailability have been obtained for aluminum hydroxide, which is commonly used as an antacid and phosphate binder, they are generally less than those for other aluminum salts. Estimates range from 0.001%, determined from urinary aluminum measurements in humans administered 28 mg Al(OH)₃/kg-bw (Weberg and Berstad, 1986), to as high as 0.45% in fasted rabbits dosed with 270 mg/kg-bw of the compound (Yokel and McNamara, 1988).

There is no information available on the bioavailability of aluminum from the inadvertent ingestion of soil (Yokel and McNamara, 2000).

Limited data on the bioavailability of aluminum from dermal exposure are available from a study of an aluminum-containing underarm antiperspirant in male and female volunteers (Flarend et al., in press). Based on measurements of urinary ²⁶Al levels with corrections for the amount of aluminum retained on the skin and the duration of monitoring, the estimated bioavailability is 0.02%.

3.3.1.3 Factors influencing oral bioavailability

There is evidence that certain factors related to the speciation or forms of aluminum and their solubilities can influence the oral bioavailability of aluminum. A number of these factors have been reviewed by Yokel and McNamara (2000) and Greger and Sutherland (1997), including pH, citrate and other organic acids, silicates, phosphate and fluoride.

As discussed previously (Section 2.3.1.2), the forms or species of aluminum present in solutions and their solubilities vary considerably depending upon pH. Evidence that these pH-related changes in the form and solubility of aluminum can alter oral bioavailability includes observations of decreased aluminum absorption in humans administered ranitidine to increase gastric pH (Rodger et al., 1991) and a report of enhanced aluminum absorption at pH 4 compared with pH 7 in an in situ rat intestine model (van der Voet and de Wolff, 1986). In contrast, Beynon and Cassidy (1990) reported no difference in aluminum absorption between uremic patients with achlorhydria and normal subjects.

Citrate, commonly found in fruit juices and other foods, can enhance the oral bioavailability of aluminum through the formation of an aluminum citrate complex. The absorption of aluminum hydroxide has been shown to be increased in the presence of citric acid in humans (Slanina et al., 1986; Weberg and Berstad, 1986; Nolan et al., 1990; Walker et al., 1990; Coburn et al., 1991; Rudy et al., 1991; Lindberg et al., 1993; Gomez et al., 1994; Nestel et al., 1994; Priest et al., 1996) and in animals (van der Voet et al., 1989; Partridge et al., 1989, 1992; Schönholzer et al., 1997). Similarly, aluminum citrate was absorbed to a greater degree than other forms of aluminum when administered to rats or rabbits (Yokel and McNamara, 1988; Froment et al., 1989a; Schönholzer et al., 1997). In contrast, Jouhanneau et al. (1993, 1997) reported no difference in the amount of aluminum absorbed by rats in the presence of added citrate; in human volunteers, Stauber et al. (1999) found no significant difference in the bioavailability of aluminum from alum-treated tapwater with or without citrate added. It has been proposed that citrate increases aluminum absorption by promoting the opening of tight junctions between gastrointestinal mucosal cells (Froment et al., 1989b; Taylor et al., 1998). Other hypotheses include increased solubility of aluminum at low pHs in the presence of citrate and chelation and transport of aluminum into gut mucosal cells by citrate (Greger and Sutherland, 1997).

Other organic acids present in food (ascorbic, gluconic, lactic, malic, oxalic and tartaric acids) can also increase the solubility and tissue retention of aluminum in rats (Partridge et al., 1989; Domingo et al., 1991, 1994).

Silicates can reduce the oral bioavailability of aluminum by the formation of hydroxyaluminosilicates (Harris et al., 1996). Plasma ²⁶Al concentrations were reduced by 85% in five men who consumed ²⁶Al and silicon in orange juice compared with orange juice and ²⁶Al without silicon (Edwardson et al., 1993). Silicic acid administered prior to and during the dosing of rats with aluminum citrate resulted in reduced tissue aluminum accumulation (Quarterly et al., 1993). However, the bioavailability of aluminum was not significantly altered in rats that were administered both aluminum and silicon after eating (Drüeke et al., 1997).

Aluminum hydroxide has been used as a phosphate binder to treat uremic patients with hyperphosphatemia. The basis for this treatment is the ability of aluminum to form insoluble complexes with phosphates in the gastrointestinal tract, thus preventing phosphorus absorption. Greger and Sutherland (1997) proposed that in the presence of sufficient quantities of phosphate, the formation of insoluble aluminum phosphate complexes could produce a similar effect on aluminum absorption. It has been suggested that phosphate-containing substances in the diet (e.g., phytate and casein) may reduce the absorption of aluminum (Glynn et al., 1995).

Aluminum salts can reduce the absorption of fluoride from the gastrointestinal tract (Spencer et al., 1981; Greger and Sutherland, 1997), and, similar to phosphate, it has been hypothesized that sufficient concentrations of fluoride should reduce the absorption of aluminum (Greger and Sutherland, 1997). However, Allain et al. (1996) observed increased plasma aluminum levels in rats administered aluminum fluoride versus aluminum chloride.

It has been suggested that certain species or forms of aluminum in drinking water may have a greater relative bioavailability than the forms present in food and consequently make a greater contribution to daily intake and potential adverse effects (Martyn et al., 1989). This is based in part on the increased levels of low molecular weight, dissolved, labile and presumably bioavailable species of aluminum observed after treatment of drinking water with aluminum-containing coagulants (Driscoll and Letterman, 1988; Van Benschoten and Edzwald, 1990; Gardner and Gunn, 1995). A comparison of aluminum levels in raw water and drinking water treated with alum from four provincial sites revealed that total recoverable aluminum levels decreased, while levels of total dissolved and dissolved extractable aluminum increased. The extractable aluminum is a labile fraction that includes free aluminum and all inorganic and organic forms exchangeable with a chelating resin (Bérubé and Brulé, 1996, 1999).

There is limited evidence from studies with animals and humans that other factors that are not directly related to the form of ingested aluminum may enhance aluminum absorption, including iron deficiency (Cannata et al., 1991, 1993; Brown and Schwartz, 1992; Florence et al., 1994), dietary calcium deficiency (Taneda, 1984; Provan and Yokel, 1990), vitamin D (Adler and Berlyne, 1985; Ittel et al., 1988; Long et al., 1991, 1994) and uremia (Ittel et al., 1987, 1988, 1991; Olaizola et al., 1989).

Administration of aluminum hydroxide and citrate resulted in a significantly greater increase in aluminum levels in blood in Alzheimer's patients aged 65-76, but not in those aged 77-89, compared with normal controls (Taylor et al., 1992). However, within the control groups, there was a significant correlation between age and increase in blood aluminum. Day et al. (1994) reported that patients with Down's syndrome, a disease that may be genetically linked to Alzheimer's disease, absorbed significantly greater amounts of aluminum (5-6 times) than controls when both groups were given ²⁶Al in the presence of citrate (i.e., orange juice).

3.3.2 Hazard characterization

As described in Section 2.4.4 and Table 8, in epidemiological studies in a range of populations the hypothesis that aluminum in the general environment (primarily drinking water) is a risk factor in the development or acceleration of Alzheimer's disease or impaired cognitive function in the elderly has been investigated. It is this potential association with Alzheimer's disease or impaired cognitive function in the elderly that has greatest implications for public health resulting from exposure in the general environment. Hence, the weight of evidence for these associations primarily from the studies of greatest inherent sensitivity - i.e., analytical epidemiological investigations - is considered here in the context of traditional criteria for causality, principally as a basis for characterizing additional relevant study.

The criteria against which the weight of evidence of causality for such associations are judged are outlined in Health Canada (1994) and subsequent updates included on the Environmental Substances Division website (www.hc-sc.gc.ca/ehp/ehd/bch/env_contaminants/psap/psap.htm). They include consistency and specificity, strength, dose-response, temporality, biological plausibility and coherence of the observed association.

As discussed previously, the majority of epidemiological studies of the potential association between aluminum in the general environment and Alzheimer's disease or impaired cognitive function have focused on drinking water as the source of exposure. Among these are studies of the inherently more sensitive analytical type (i.e., cohort, case-control and cross-sectional) conducted in populations from Ontario, Quebec, France, England and Switzerland (Table 8). However, it should be noted that variations in exposure of individuals in these analytical studies have been assessed to only a very limited extent, with information on concentrations of aluminum in drinking water being distinguished only by location of residence. In general, the quality of the studies has been sufficient in terms of assessment of outcomes and consideration of confounders to make them relevant to an assessment of causality.

A direct association between Alzheimer's disease or Alzheimer's-type presenile dementia and exposure to aluminum in drinking water was investigated in studies on populations from Ontario, France, England and Quebec (Michel et al., 1991; Neri and Hewitt, 1991; Neri et al., 1992; Forbes et al., 1995b; Forster et al., 1995; Forbes and McLachlan, 1996; McLachlan et al., 1996; Martyn et al., 1997; Gauthier et al., in press). The potential association between dementia or cognitive impairment and aluminum in drinking water was investigated in studies from Ontario, France, England and Switzerland (Wood et al., 1988c; Wettstein et al., 1991; Forbes et al., 1992; 1994; 1995a; Forbes and Agwani, 1994; Jacqmin et al., 1994; Jacqmin-Gadda et al., 1996). A positive association was reported in all of the studies from Ontario, France and Quebec (Michel et al., 1991; Neri and Hewitt, 1991; Neri et al., 1992; Forbes et al., 1992; 1994; 1995a; 1995b; Forbes and Agwani, 1994; Jacqmin et al., 1994; Jacqmin-Gadda et al., 1996; Forbes and McLachlan, 1996; McLachlan et al., 1996; Gauthier et al., in press), and no association was observed in those from England and Switzerland (Wood et al., 1988c; Wettstein et al., 1991; Forster et al., 1995; Martyn et al., 1997). The results of one of the French studies (Michel et al., 1991) have been discounted due to a reliance on outdated information on aluminum concentrations in drinking water (Jacqmin et al., 1994; Smith, 1995; WHO, 1997).

With respect to the assessment of outcomes, the diagnostic criteria for selection of cases were generally more stringent and specific to Alzheimer's disease among the studies and populations in which there was a positive outcome. In the positive study from Quebec, cases were ascertained based on a three step process that included a modified mini-mental state exam (MMS) and standardized diagnostic criteria (DSM, NINCDS-ADRDA and ICD⁶) (Gauthier et al., in press). ICD criteria were also used in two positive studies from Ontario (Neri and Hewitt, 1991; Neri et al., 1992; Forbes et al., 1995b; Forbes and McLachlan, 1996). In an additional positive study from Ontario, McLachlan et al. (1996) obtained autopsy-confirmed cases of Alzheimer's disease with clinical histories of dementia from brains donated to the Canadian Brain Tissue Bank. In the fourth study from Ontario, Forbes and colleagues (Forbes and Agwani, 1994; Forbes et al.,1992; 1994; 1995a) assessed impaired mental functioning by a questionnaire that included a modified mental status test. Cognitive impairment was determined by MMS score in the positive study from France conducted by Jacqmin-Gadda and colleagues (Jacqmin et al.,1994; Jacqmin-Gadda et al.,1996). Among the studies from England in which no association was observed, only Forster et al. (1995) used a diagnostic algorithm incorporating NINCDS-ADRDA and DSM criteria, and an MMS examination to identify cases. Martyn et al. (1997) relied on a diagnosis of Alzheimer's disease from hosptial notes or computerized tomographic scans to classify demented patients as cases, but standardized diagnostic criteria were not specified in the study. In the third negative study from England, Wood et al. (1988c) identified subjects with dementia using an unspecified mental test. Finally, in the negative study from Switzerland (Wettstein et al., 1991) dementia was diagnosed using the MMS.

There is somewhat less consistency in the control of potential confounders and other factors among the studies from Ontario, Quebec and France in which a positive association was reported compared to the studies on populations from England and Switzerland in which no association was observed. The study from Quebec (Gauthier et al., in press) and selected analyses from the Ontario study of cognitive impairment (Forbes et al., 1994; 1995a) involved control of a range of factors (e.g., age, sex, education, family history, genotype, occupation, etc.), and age and sex were accounted for in the Ontario study by Neri and colleagues (Neri and Hewitt, 1991; Neri et al., 1992). However, these factors were generally not addressed in the Ontario death certificate study (Forbes et al., 1995b; Forbes and McLachlan, 1996), in the study by McLachlan et al.(1996) or in some of analyses included in the Ontario study of cognitive impairment (Forbes et al., 1992; Forbes and Agwani, 1994). In the positive study from France by Jacqmin and colleagues (Jacqmin et al.,1994; Jacqmin-Gadda et al.,1996) analyses of cognitive impairment included control for age, sex, education and occupation. All of the studies from England had some form of adjustment for age. Forster et al. (1995) also controlled for a range of other factors including family history, disease history, head injuries, physical activity and smoking and the study of Martyn et al. (1997) included control for diagnostic centre and distance from residence to diagnostic centre. The other study from England, Wood et al. (1988), controlled for both age and sex, while the Swiss study, Wettstein et al. (1991) accounted for age, education and socioeconomic status. There is also some evidence that other characteristics of water quality such as pH are co-factors for the purported association, consistent with what might be expected, if the association was causal (Forbes et al., 1992; 1994; 1995a; 1995b; Jacqmin et al., 1994; Jacqmin-Gadda et al., 1996).

While there is some evidence of exposure-response in the individual available studies for the reported association between aluminum and Alzheimer's disease, there is little consistency in results among the different investigations in this respect, at least based on the limited extent of comparison permitted by the available data. This lack of consistency may be a function in part, of variations in sensitivity between the studies. For example, in the study from Quebec (Gauthier et al., in press), the classification of subjects was restricted to exposed (upper quartile of subjects) versus unexposed due to the limited number of cases and controls. There is evidence for a statistically significant increasing risk of Alzheimer's disease with increasing concentration of aluminum in drinking water (i.e., dose-response) only in the Ontario study by Neri and colleagues (Neri and Hewitt, 1991; Neri et al., 1992). In another Ontario study (McLachlan et al., 1996), the risks of Alzheimer's disease increased (3.6-7.6) as concentrations of aluminum increased from 125 to 175 µg/L, but the confidence intervals for the higher risk estimates were wide and included 1.

There are also inconsistencies in the dose-response trend across the studies from Ontario, Quebec and France in which positive outcomes were reported. Among the studies conducted in Ontario, Neri and Hewitt (1991) and Neri et al. (1992) determined relative risks (RRs) of 1.3-1.5 for concentrations of 100 to >200 µg Al/L, whereas McLachlan et al. (1996) reported odds ratios (ORs) of 1.7-2.5 for the composite exposure level of ≥100 µg Al/L and ORs of 3.6-7.6 for individual exposure levels from 125 to 175 µg Al/L. Using a cut point of 84.7 mg/L to define low versus high concentrations of aluminum, Forbes et al. (1992, 1994, 1995a) and Forbes and Agwani (1994) reported ORs for cognitive impairment ranging from 0.67 to 4.0 for exposure to high concentrations of aluminum in the presence of varying levels of pH, fluoride and/or silica. In general, the higher ORs corresponded to exposure scenarios when pH levels were high or fluoride and silica levels were low. When more complex multivariate analyses were conducted on the same dataset with control for a range of factors, the ORs ranged from 1.72 to 2.35 (Forbes and Agwani, 1994; Forbes et al., 1994; 1995a). In a death certificate study from Ontario, Forbes et al. (199b) analysed the risks of Alzheimer's disease from exposure to concentrations of aluminum ≥ 336 mg/L combined with high or low levels of pH, fluoride or silica and estimated RRs⁷ ranging from 0.88 to 4.0 with the higher RRs generally observed when pH was high or fluoride and silica were low. In the same study, the RRs for exposure to lower concentrations of aluminum (68-200 mg/L) were £ 1.0. The follow-up analysis by Forbes and McLachlan (1996) produced even higher RRs (4.8-10) for concentrations above 250 µg Al/L. Based on these results, Forbes and McLachlan (1996) hypothesized a J-shaped dose-response curve for the relationship between Alzheimer's disease and aluminum in drinking water with a minimum around 100 mg Al/L. In the positive study from Quebec, an OR of 2.67 was observed for exposure to 12 mg/L of monomeric organic aluminum in drinking water with lower ORs reported for exposures to other aluminum fractions in drinking water. In their study on a French population, Jacqmin et al. (1994) initially reported odds ratios (ORs) for cognitive impairment of 1.3-1.4 for 50 to 100 mg Al/L at pH 7, while in their follow-up analysis an OR of 4.0 was estimated for exposure to concentrations of aluminum higher than the first quartile of the distribution (≥ 3.5 mg/L) combined with low pH and silica levels. ORs for higher quartiles of the aluminum distribution (≥ 9.0 to ≥ 16.0 mg/L) were less than 1.2 (Jacqmin-Gadda et al., 1996). Adding to these inconsistencies is the fact that the highest concentrations of aluminum reported in negative studies from England and Switzerland (98-250 µg/L) were similar to those in the positive studies (Wood et al., 1988c; Wettstein et al., 1991; Forster et al., 1995; Martyn et al., 1997). There is additional evidence (albeit less reliable) for a dose-response trend in two of the four ecological epidemiological studies in which a positive association between Alzheimer's disease and aluminum in drinking water was observed (i.e., Martyn et al., 1989: RR = 1.4-1.7 for 20 to >110 µg Al/L; and Flaten, 1990: RR = 1.2-1.3 for 50 to >200 µg Al/L).

For the analytical epidemiological studies from Ontario, Quebec and France in which there was a positive association between Alzheimer's disease and exposure to aluminum, the strength of the observed association was generally moderate, with statistically significant RRs or ORs as high as 1.5-4.0 reported for the highest exposure groups (Neri and Hewitt, 1991; Neri et al., 1992; Forbes et al., 1992; 1994; 1995a; 1995b; Forbes and Agwani, 1994; Jacqmin et al., 1994; Jacqmin-Gadda et al., 1996; McLachlan et al., 1996; Gauthier et al., in press). An exception is the Ontario study of Forbes and McLachlan (1996), in which RRs for Alzheimer's disease in subjects 85 years of age or older exposed to aluminum concentrations of >250 µg/L disease ranged from 4.8 (p < 0.05) with no control for water quality parameters to 10 (p < 0.05) with control for water source, pH, turbidity and concentrations of silica, iron and fluoride.

The analytical epidemiological studies in which there was a positive association provide only limited evidence to satisfy the criterion of temporality of exposure and disease, since in few of these studies have the effects of the duration of exposure on the occurrence of Alzheimer's disease or cognitive dysfunction been assessed. In one of the studies conducted in Ontario (McLachlan et al., 1996) an OR of 1.7 was reported for exposure to ≥100 µg Al/L in drinking water, increasing to 2.5 when a 10-year weighted residential history prior to onset of the disease was taken into account. In their study of cognitive impairment, Forbes et al. (1994) indicated that they observed greater ORs when their analyses were restricted to subjects who had lived at their current addresses for more than 5 years. In the death certificate study of Alzheimer's disease from Ontario, Forbes and McLachlan (1996) noted that the death certificates for their cases did not state how long subjects had resided in the areas where their deaths occurred thus limiting the assessment of exposure duration. However, the risk of Alzheimer's disease and/or presenile dementia was greater for individuals > 75 compared to < 75 years of age and > 85 versus > 75 years of age (Forbes et al., 1995b; Forbes and McLachlan, 1996). According to the authors, these results may be attributed to a greater cumulative exposure to aluminum over a lifetime for individuals in the older age groupings. In the study from Quebec (Gauthier et al., in press) individual exposures were weighted for duration of residence at a given location. However, a significant relationship between organic monomeric aluminum in drinking water and Alzheimer's disease was reported only for the onset period of the disease. The authors suggested that the onset period may have actually represented long-term exposure, as the cases lived an average of 44 years at their residences prior to onset of the disease. In contrast to this limited evidence of increasing risk with increasing duration of exposure, the exclusion of patients greater than 65 years of age strengthened the relationship between Alzheimer's disease and aluminum in the ecological study of Martyn et al. (1989). However, the authors attributed this result to greater case ascertainment for the younger age group due to more aggressive clinical investigation.

The evidence for biological plausibility of the association between exposure to aluminum in drinking water and Alzheimer's disease is, at the very most, limited. Indeed, it is restricted primarily to observations, based on the limited available data, that effects observed consistently at lowest doses in experimental animals are those on the neurological system, the induction of Alzheimer's-like neuropathology in the brains of certain species of experimental animals following aluminum administration, and the observation of aluminum-induced neurological disorders in humans, such as dialysis encephalopathy (Section 2.4.5).

In relation to biological plausibility, four essential criteria that must be met in order to assign a role for aluminum as a definitive factor in the pathogenesis of Alzheimer's disease are commonly cited (Kruck and McLachlan, 1989). These are as follows:

1. It must be located at the site of toxic action without exception and be significantly diminished or absent in controls.
2. It must be present in and available from the environment.
3. There must be a demonstrable route of access to the site of toxic action.
4. Removal of aluminum from this site must be followed by retardation or arrest of the induced process.

In relation to these criteria, evidence of the presence of elevated levels of aluminum in bulk brain tissue and in NFTs and neuritic plaques from Alzheimer's patients is conflicting. There is, however, evidence that aluminum is accessible to the site of toxic action (i.e., central nervous system), and, based on data on speciation, more bioavailable forms of aluminum may be present in the media for which associations with Alzheimer's disease have been observed (i.e., drinking water). Variations in results are also consistent with what might be expected, when co-factors that influence the bioavailability of aluminum are taken into account.

However, these criteria fall far short of those considered appropriate for assessment of weight of evidence for mode of action as a basis for assessment of biological plausibility for risk assessment. Indeed, as outlined in Section 2.4.6, no plausible pathway for induction of Alzheimer's disease by aluminum has been proposed with measurable key events, for which sufficient investigation has been conducted to assess weight of evidence against traditional criteria of causality as outlined below.

Overall, then, the weight of evidence for causality for the observed associations between aluminum and Alzheimer's disease is weak, at best. There is only limited consistency in the results of the analytical epidemiological studies. While the criteria for diagnosis were generally more stringent in the studies in which there was a positive outcome, there was more consistent control of potential confounding factors in the studies in which no associations were reported. Moreover, while there is some evidence of exposure-response in the individual available studies for the reported association between aluminum and Alzheimer's disease, there is little consistency in results among the different investigations in this respect, at least based on the limited extent of comparison permitted by the available data. There are also limited data to serve as a basis of the extent to which the observed association between aluminum and Alzheimer's disease meets the criterion of temporality. Most limiting, however, in the assessment of the weight of evidence for causality of the observed association is the lack of relevant data on biological plausibility; indeed, there is no hypothesized plausible pathway from exposure to effect with measurable key events, for which sufficient investigation has been conducted to assess weight of evidence against traditional criteria of causality, such as consistency, strength, specificity, dose-response, temporal patterns, biological plausibility and coherence.

3.4 Uncertainties and recommendations

While the evidence of an association between exposure to aluminum and Alzheimer's disease is weak, it cannot be dismissed completely in view of the consistency of some results with several lines of circumstantial evidence and the paucity of data to serve as a basis for consideration of biological plausibility. These include reported associations being for the medium for which bioavailability may be the greatest and the documented accumulation and interaction of aluminum with the central nervous system. In view of the potentially significant public health implications if the association were causal, this area is considered a priority for research.

Health Canada initiatives related to development of appropriate research protocols for investigation in this area are described in the Introduction of this report. Specific recommendations of an international workshop regarding the design and conduct of the study included the use of mice (wild-type and transgenic for human genes associated with Alzheimer's disease) and rabbits as the appropriate species; exposure commencing in utero and continuing throughout the lifetime of the animals; use of an organo-aluminum compound; administration of a purified, low aluminum content diet; assessment of behavioural endpoints in the U.S. EPA neurotoxicity testing guidelines; appropriate brain histopathology, biochemical and hematological analyses; kinetic studies of aluminum accumulation in the brain; control of aluminum contamination; and Good Laboratory Practice.

As follow-up, an Expert Steering Committee developed an RFP for a study or studies based on the recommendations from the international workshop. The Steering Committee developed designs for a 2-year two-generation study in wild-type mice and a 1-year two-generation study in transgenic mice carrying copies of human genes rendering them predisposed to Alzheimer's-type neuritic plaques. Because of their susceptibility to Alzheimer's-type pathology, the transgenic mice could be a population of animals highly sensitive to any potential effect of aluminum on the development of the disease. Both studies were to involve exposure to aluminum maltolate in drinking water, a purified low-aluminum diet, a repeated battery of behavioural tests, and biochemical and morphological tests at sacrifice.

The Steering Committee also discussed the option of conducting a full-scale case-control epidemiological study of aluminum in drinking water and Alzheimer's disease and recommended that a subcommittee be struck at a later date to design such a study.

It was determined that any RFPs developed for either the mouse or epidemiological studies should not proceed without secured funding, which has not yet been acquired.

---

⁶International Classification of Diseases, World Health Organization.
⁷In the studies of Forbes et al. (1995b) and Forbes and McLachlan (1996) the relative risks (RRs) were estimated by rate ratios.