Health effects of aluminium exposure

Clinical manifestations of aluminium toxicity have been recognised for many years, and include dementia, bone disease and anaemia. These problems were initially identified in patients with renal impairment exposed to high concentrations of aluminium from dialysis solutions and phosphate binders, who developed so-called ‘dialysis dementia’. Up to 80% of patients with dialysis dementia exhibit motor impairment, with myoclonic jerks, ataxia and dyspraxia⁽Reference Burn and Bates²⁾. Cortical atrophy of the frontal lobes has been reported on brain scans⁽Reference Cusamo and Savazzi³⁾ and autopsy studies of patients have shown particularly high concentrations of aluminium in grey matter⁽Reference McDermott, Smith and Ward⁴⁾. Bolla et al. ⁽Reference Bolla, Briefel and Spector⁵⁾ performed detailed neurocognitive testing in adults undergoing dialysis, and reported that serum aluminium concentrations were predictive of visual memory. Associations between aluminium concentrations and tests of frontal lobe function and attention/concentration were also seen, although only in subjects with lower vocabulary scores. Similar neurological and cognitive problems were subsequently identified in adults receiving PN solutions, and also in aluminium smelting plant workers, especially those exposed before the new smoke hoods were introduced in 1972⁽Reference White, Longstreth and Rosenstock⁶⁾.

Adverse effects of aluminium on bone have been identified in adults with uraemia and low turnover osteomalacia⁽Reference Ott, Maloney and Klein⁷⁾ and in those with normal renal function undergoing long-tern PN⁽Reference Ott, Maloney and Coburn⁸⁾. Both groups of patients had low bone formation on iliac crest biopsy, with patchy osteomalacia. Histochemical staining of biopsy samples showed aluminium accumulation at the mineralisation front, which was quantified as surface-stainable aluminium. The latter correlated closely with quantitative measurements of aluminium in bone determined by atomic absorption spectroscopy. These findings were supported by multiple studies in several species of experimental animals including rats⁽Reference Goodman, Gilligan and Horst⁹⁾, piglets⁽Reference Sedman, Alfrey and Miller¹⁰⁾ and dogs⁽Reference Goodman, Henry and Horst¹¹⁾.

Aluminium exposure in preterm infants

Moreno et al. ⁽Reference Moreno, Dominguez and Ballabriga¹²⁾ calculated that PN solutions were the main source of aluminium exposure in neonates, accounting for 89% of total aluminium intake. In earlier PN solutions (prior to the mid-1990s), aluminium contamination occurred mostly from casein hydrolysates and trace element components, but the problem with more modern PN solutions relates mainly to small-volume acidic solutions stored in glass vials, notably calcium gluconate which was found to account for 81% of the contamination in one study⁽Reference Mouser, Wu and Herson¹³⁾. It has been recognised for some time that, when fed parenterally, infants retain >75% of the aluminium (compared to approximately 40% in adults), with high serum, urine and tissue levels⁽Reference Moreno, Dominguez and Ballabriga^12, Reference McGraw, Bishop and Jameson¹⁴⁾. For example, Sedman et al. ⁽Reference Sedman, Klein and Merritt¹⁵⁾ prospectively studied plasma and urinary aluminium concentrations in eighteen premature infants receiving intravenous therapy and in eight term infants receiving no intravenous therapy. They also measured bone aluminium concentrations in autopsy specimens from twenty-three infants, including six who had received at least 3 weeks of intravenous therapy. Preterm infants who received intravenous therapy had high plasma and urinary aluminium concentrations compared with normal controls. The bone aluminium concentration was also ten times higher in infants who had received at least 3 weeks of intravenous therapy than in those who had received limited intravenous therapy.

Although aluminium exposure and tissue aluminium accumulation were well documented in neonates two decades ago, it was unclear at the time whether this exposure had any health consequences. A causal relationship between early aluminium exposure and adverse health outcomes cannot be established in observational studies, particularly in preterm infants in whom the duration of PN (and hence, aluminium exposure) is very likely to act as a proxy for poor health, which is itself associated with adverse outcome. Thus, an experimental study was required, with preterm infants randomised to different aluminium exposure and follow-up to measure health outcomes. While it was not ethical to randomise a group of infants to receive ‘high’ aluminium exposure, it was ethical and feasible to randomise them to receive a lower aluminium exposure than they would receive in normal clinical practice.

Randomised trial of aluminium exposure from parenteral nutrition in preterm infants

To investigate the short- and long-term health effects of neonatal intravenous aluminium exposure, a randomised double-blind trial was initiated in 1988, comparing standard PN solutions with solutions specially sourced for low aluminium content. Details of the original trial design are reported elsewhere but summarised briefly here⁽Reference Bishop, Morley and Day¹⁾. Two hundred and twenty-seven preterm infants with birthweight <1850 g were recruited from a neonatal unit in Cambridge, UK. Infants were eligible if a clinical decision was made to start PN, and were randomly assigned to receive either standard PN solution or a specially sourced low-aluminium solution. Details of the composition of the two solutions are shown in Table 1, together with the measured aluminium concentrations in each solution. The solutions were identical except that the aluminium-depleted solution contained calcium chloride instead of calcium gluconate. The use of a mixed sodium–potassium phosphate solution in place of potassium acid phosphate minimised the increase in chloride. By design, the total aluminium intake when the infant received 180 ml/kg per d differed markedly; 45 μg/kg per d for the standard solution compared with only 4–5 μg/kg per d for the aluminium-depleted solution. All decisions on infant feeding were made by the clinicians responsible for the care of the infant; the study team were not involved in this aspect. Data were collected on the clinical course of each infant, detailed records of exact intravenous and oral intake, daily blood samples for electrolytes, calcium and acid-base status, and weekly samples for plasma chloride.

Table 1. Composition and aluminium content of the standard and aluminium-depleted intravenous feeding solutions used in a randomised trial⁽Reference Sedman, Klein and Merritt¹⁵⁾

Vamin infant contained essential amino acids without added electrolytes. Intralipid 20% was a fat emulsion containing 20 g fatty acids/dl. Vitalipid contained fat-soluble vitamins and Solivito contained water-soluble vitamins. Neotrace was an in-house preparation containing Cu and Zn only. Vamin infant, intralipid 20%, vitalipid and solivito were manufactured by Kabi Vitrum.

Cognitive outcome

As at 18 months of age, no significant differences in cognitive outcome were found between randomised groups during adolescence. Non-randomised analyses were also performed to assess the impact of aluminium exposure, because an infant's actual exposure depended not only on the PN solution received but also on the duration of intravenous feeding, which varied considerably. The actual neonatal exposure to aluminium (shown in Fig. 1) covered a wide range, with considerable overlap between randomised groups. No significant differences were found when comparing subjects with total neonatal aluminium exposure above or below the median (55 μg/kg).

Fig. 1. Neonatal aluminium exposure for subjects studied at 13–15 years, according to randomised parenteral nutrition (PN) solution in a randomised trial⁽Reference Sedman, Klein and Merritt¹⁵⁾.

These findings must be considered in the context of the fact that the subjects seen at 13–15 years were a selected group, excluding those with known neuromotor impairment at 18 months or a Bayley MDI<85; they also had higher birthweight SDS than subjects who were not followed. Hence, it could be argued that the findings show no evidence of longer-term cognitive effects in this relatively lower risk cohort who had normal cognitive outcome at 18 months post-term, but cannot perhaps be generalised to smaller infants who already have evidence of neurocognitive impairment apparent in infancy: in effect, the follow-up protocol may have excluded the children already adversely affected by neonatal aluminium exposure. However, the sub-group of subjects studied at 13–15 years (particularly those subjects who had received more than 10 d of PN) showed the same trend towards higher Bayley MDI at 18 months in the aluminium-depleted group as observed in the larger cohort. Statistical significance was not reached for this comparison, possibly due to the small sample sizes available for the analyses; nevertheless these data suggest that the children followed-up were fairly representative of all subjects seen at 18 months in terms of the effects of aluminium exposure on cognitive outcome. One possibility is that those with Bayley MDI>85 in infancy were able to compensate subsequently for any adverse effect of early aluminium exposure.

Interpretation and practical implications of the findings

Data from this clinical trial suggest that neonatal aluminium exposure from PN in the high-risk preterm infant may have adverse effects on later bone health, as well as short-term cognitive outcome. Although there was no strong evidence for effects on later cognitive outcome, the group of subjects followed at 13–15 years were a selected population with normal development at 18 months; it is unclear whether persistent or additional adverse effects would be apparent in the subjects who already demonstrated sub-optimal development at 18 months of age. The observed effects are plausible given the known toxicity of aluminium for brain and bone seen in adults and in animal models. This is the only experimental study to systematically examine the health effects of aluminium in any population with high exposure and, despite its limitations, it seems unlikely that it will be repeated.

It is important to consider the likely practical significance of the observed effect of aluminium exposure on bone mass. This is difficult to quantify since there are no data directly relating bone mass at age 13–15 years to later fracture risk. Hip bone mass was 7·6% lower in subjects with neonatal aluminium exposure above the median, while the difference in lumbar spine bone mineral content was approximately 0·7 sd between groups (representing approximately 14% of population variance assuming a normal distribution), and the difference in lumbar spine bone mineral density was 0·36 sd (representing approximately 7% of population variation). These figures can be considered in the context of the study of Hernandez⁽Reference Hernandez²⁸⁾, who estimated that the peak bone mass was a better predicter of osteoporosis risk than either the age at menopause or the rate of age-related bone loss later in life, and calculated that a 10% increase in peak bone mass would delay the onset of osteoporosis by 10 years. Given the sizeable number of contemporary infants undergoing intensive care and still exposed to aluminium via PN, these findings have contemporary relevance. PN is also used more aggressively in modern neonatal units, starting earlier and with more rapid advancement than was typical at the time of the clinical trial; hence neonatal aluminium exposure may be greater. Furthermore, recommended mineral intakes for preterm infants are now higher than what it was 20 years ago, so preterm infants are exposed to greater volumes of calcium gluconate, the main offender in terms of aluminium intake.

Regulatory aspects of aluminium exposure

Potential methods for lowering aluminium exposure from PN solutions are given in Table 2. Despite widespread recognition of the problem, until recently little progress had been made on reducing exposure. Following a review of the literature, the Food and Drug Administration recommended that daily aluminium intakes should not exceed 5 μg/kg per d in vulnerable patients⁽²⁹⁾, including preterm infants. Manufacturers were required to ensure that large volume parenterals do not contain more than 25 μg/l of aluminium and to label them as such. No restrictions were placed on the aluminium content of small volume parenterals, but manufacturers were required to label them with the estimated aluminium content at expiry. While this represented an advance, Poole et al. ⁽Reference Poole, Hintz and Mackenzie³⁰⁾ calculated that meeting the Food and Drug Administration recommendations was currently impossible in patients <50 kg using available products. Furthermore, calculated aluminium intake in patients <3 kg in their study was 30–60 μg/kg per d – higher than in the randomised trial discussed in this paper. In a more recent study, the same investigators measured the actual aluminium content of PN solutions being administered to forty preterm infants and found that intakes were still three to five times the recommended Food and Drug Administration limit, although significantly less than the intake calculated using manufacturers’ values on product labels⁽Reference Poole, Schiff and Hintz³¹⁾. Most recently (2010), following a review of available data, the UK Medicines and Healthcare Regulatory Authority recommended that calcium gluconate in small volume glass containers should not be used for repeated or prolonged treatment in children <18 years, including preparation of PN⁽³²⁾. Manufacturers should now be required to address the issue at least for this particular component of PN solutions, and this represents an important step towards addressing the problem of aluminium exposure and toxicity in vulnerable infants and children.

Table 2. How can neonatal aluminium exposure from parenteral nutrition (PN) solutions be minimised using currently available solutions? Potential methods and barriers to implementation