Aluminum sodium oxide

Epidemiological data

Administrative data

Endpoint:

epidemiological data

Type of information:

migrated information: read-across from supporting substance (structural analogue or surrogate)

Adequacy of study:

other information

Reliability:

2 (reliable with restrictions)

Rationale for reliability incl. deficiencies:

other: Acceptable, well documented publication which meets basic scientific principles.

Data source

Reference

Reference Type:

publication

Title:

Longitudinal study on potential neurotoxic effects of aluminium: I. Assessment of exposure and neurobehavioural performance of Al welders in the train and truck construction industry over 4 years

Author:

Kiesswetter, E. et al.

Year:

2007

Bibliographic source:

Int Arch Occup Environ Health 81:41–67

Materials and methods

Study type:

longitudinal study

Endpoint addressed:

neurotoxicity

Principles of method if other than guideline:

Longitudinal study comparing repeatedly measured exposure data and neurobehavioural data of exposed subjects with data of an age-matched control group of similar age, based on three examinations over a period of 4 years.

GLP compliance:

no

Test material

Test material information

Constituent 1

Reference substance name:

Aluminium

EC Number:

231-072-3

EC Name:

Aluminium

Cas Number:

7429-90-5

Molecular formula:

Al

IUPAC Name:

aluminum

Details on test material:

- Name of test material (as cited in study report): Aluminium
- Analytical purity: no data

Method

Type of population:

occupational

Ethical approval:

confirmed and informed consent free of coercion received

Details on study design:

HYPOTHESIS TESTED (if cohort or case control study): the aim of the study was to examine the relationship between aluminium exposure and long-term changes of neurobehavioural performance in aluminium welders as well as neurobehavioural performance changes of a non-exposed control group.


METHOD OF DATA COLLECTION
- Type: Medical interview / Questionnaire / Neuropsychological tests
- Details:
Medical examination
A standardized medical interview including occupational history was conducted prior to the first examination. In addition, a physical examination, including neurological status and pulmonary function tests, was carried out. In the following two examinations, relevant parts of the interview and physical examination were repeated.

Neurobehavioural methods

Schedule
The neurobehavioural methods comprised a questionnaire which served to record neurotoxic symptoms as well as psychological tests for the assessment of different functional domains including approaches to the modelling of premorbid intelligence.
The application scheme (schedule) of questionnaire and tests across the three examinations is shown in Table 1.
The Standard Progressive Matrices Test, a test of logic thinking, and a test of verbal intelligence (WST) were each conducted only once, in the first and second examination, respectively. In the second and third examinations, the computer-aided European Neurobehavioural evaluation system (EURO-NES) was applied. The symptom questionnaire Q16 and all other test methods were used in all three examinations.

Symptom questionnaire
A German version of the Swedish questionnaire Q16 was applied for symptom monitoring, with focuses on psychological and neuropsychological symptoms related to neurotoxic exposures.

Neuropsychological tests
The tests can be related to four neuropsychological domains: cognitive abilities, psychomotor performance, short-term memory and attention.

Cognitive abilities
Verbal intelligence was measured with the German multiple choice vocabulary test (Wortschatztest, WST). Deductive and inductive thinking as well as general intelligence were tested with the German version of the Standard Progressive Matrices Test (SPM). Visuospatial thinking and reasoning were assessed with the block design test (BDT). Intelligence and speed aspects were tested with the trail making test (TMT) by measuring the average time a subject needs to connect sequentially a series of numbers which are randomly distributed on a sheet of paper.

Psychomotor performance
Four computerised tests (‘steadiness’, ‘line tracing’, ‘aiming’, ‘tapping’) were used to assess psychomotor performance. The tests were carried out both the dominant and non-dominant hand separately. The computer-measures include accuracy and speed as parameters. Tremor can indirectly be measured with the steadiness-test, and to a lower degree the line tracing tests.

Short-term memory, working memory
The tests ‘recall of digits’ forward and backward of the Wechsler Adult Intelligence Scale (WAIS; in German: Hamburg-Wechsler-Intelligenztest für Erwachsene, HAWIE) were used for measuring short-term memory or working memory. Additionally, the computer test ‘digit span’ (DS) of the European Neurobehavioural Evaluation System (EURO-NES) was employed. The latter consists of an intensive adaptive test of short-term memory comprising steps of increasing/decreasing difficulty which approximate the individual reproduction ability. Furthermore, the symbol-digit substitution test (SDS) of the EURO-NES was included in this category.

Attention
Attentional processes were assessed in the ‘switching attention’ task of the EURONES and, partially, in simple reaction time (SRT) tasks. The SRT test includes the reaction time parameter and the movement time of the index finger between resting and target key. In the ‘switching attention’ task, the reaction time and error measures correspond to the increasing difficulty of the three subtests: block, arrow, mixed. Attention has to be directed to a few constant features (side, arrow) of the stimuli in the simple forms of the test, while switching between stimulus response rules is needed (mixed) in the complex form.


SETTING: occupational; aluminium welders of the train body and truck trailer construction industry. The enterprises belonged to the aluminium-processing industry from different regions of Germany (category: small or medium sized firms).


STUDY POPULATION
- Total population (Total no. of persons in cohort from which the subjects were drawn): 44
- Selection criteria: Workers having neurological diseases related to injuries or relevant metabolic illnesses as well as subjects with insufficient knowledge of German language were excluded. Only aluminium welders with at least 2 years of exposure to aluminium and no current or previous exposure to other potential neurotoxic substances, such as solvents or metals were included.
- Total number of subjects participating in study: 44
- Sex/age/race: male/43.3 ± 7.4 years
- Total number of subjects at end of study: 20
- Matching criteria: aluminium welders and assembly workers (controls) were comparable regarding gender, age, education, physical work environment and social environment (also refer to Table 2).
- Other: the study population was steady diminished due to economic problems, close down of one firm, and lay-offs in the others. The repeated-measurement analyses of the first two examinations were based on 33 subjects. The final analyses were based on 20 workers who had taken part in all three examinations.
The average values of carbohydrate deficient-transferrin (CDT, marker for alcohol) were below 5 U/L, indicating moderate alcohol consumption.


COMPARISON POPULATION
- Type: Control or reference group
- Details: The control population comprised assembly workers from the same enterprises as the exposed subjects.
- Total number of subjects participating in study: 37
- Sex/age/race: male/42.9 ± 5.7 years
- Total number of subjects at end of study: 12
- Matching criteria: aluminium welders and assembly workers (controls) were comparable regarding gender, age, education, physical work environment and social environment (also refer to Table 2).
- Other: the control population was steady diminished due to economic problems, close down of one firm, and lay-offs in the others. The repeated-measurement analyses of the first two examinations were based on 26 subjects. The final analyses were based on 12 workers who had taken part in all three examinations.
The average values of carbohydrate deficient-transferrin (CDT, marker for alcohol) were below 5 U/L, indicating moderate alcohol consumption.



HEALTH EFFECTS STUDIED
- Other health effects: neurobehavioural performance

Exposure assessment:

measured

Details on exposure:

TYPE OF EXPOSURE: inhalation during aluminium welding


TYPE OF EXPOSURE MEASUREMENT: Personal sampling / Biomonitoring (urine) / Biomonitoring blood


EXPOSURE LEVELS: refer to Table 3 and 4 under remarks on results.


EXPOSURE PERIOD: exposed and control groups worked both in a weekly rotated three-shift system (morning, afternoon, night shift; 8 h) which requires a weekly new adjustment of the circadian system of the shift-workers with special adaptation difficulties during and after the night shift week.


POSTEXPOSURE PERIOD: examinations were conducted standardized on the second day shift week between 8:00 and 13:00 h in order to avoid circadian and adjustment effects.

Statistical methods:

Exposure data
The interdependency, reliability, and stability (2-, 4-year intervals) of external and internal exposure measures were investigated using correlation and regression methods. In addition, ANOVA was used to test, whether aluminium biomonitoring data were also sensitive to acute exposure changes (pre-, post-shift differences).

Neurobehavioural data
The neurobehavioural data were analysed with multivariate models of covariance (MANCOVA) for repeated measurements. In some cases, different components of a test (such as speed and errors), tests within one domain and motor performance of both hands were analysed simultaneously in multivariate analysis models. Psychomotor data of the left and right hand were reorganized to data of dominant and non-dominant hand.
The MANCOVA models used included a grouping factor (welders vs. control), a repetition factor (examination, three levels), an interaction factor 'examination x exposure' and as covariates education index, age and the alcohol marker Carbohydrate Deficient Transferrin (CDT). The interaction term was used to test whether neurobehavioural performances of exposed and controls changed differently across the examinations. The variances explained by the respective model components (eta square) are given additionally to the error probability P. The association between exposure and neurobehavioural measures was tested by stepwise regression in a further statistical approach.
MANCOVA models were used to investigate moderating effects of different potential indicators of 'premorbid' intelligence.

Results and discussion

Results:

EXPOSURE
- Number of measurements: 6 (2 measurements per examination (pre- and post-shift))
- Median concentrations in the course of the longitudinal study: (refer also to Tables 3 and 4)

Exposed group
Total dust: 4.5-6.8 mg/m³
Aluminium in urine (pre-shift): 98-164 µg/L; 59-92 µg/g creatinine
Aluminium in urine (post-shift): 94-145 µg/L; 64-144 µg/g creatinine
Aluminium in plasma (pre-shift): 9.6-11.1 µg/L
Aluminium in plasma (post-shift): 11.6-14.3 µg/L

Control group
Aluminium in urine (pre-shift): 5.8-8.0 µg/L; 4.0-8.5 µg/g creatinine
Aluminium in plasma (pre-shift): 2.8-4.5 µg/L

- Date(s) of measurement(s): at 0, 2 and 4 years
- Other:

Exposure data of the subsample analysed neurobehaviourally: course and statistical analysis
The course of total dust load across examinations for the group of aluminium welders with neuropsychological measures had a minimum of 5.5 mg/m³ total dust at the second examination and a maximum of 8.1 mg/m³ at the last examination. The differences were statistically not significant (P = 0.44).
The course of the creatinine-related aluminium concentrations had a maximum of 140 µg/g creatinine at the second examination and decreased to ca 88 µg/g creatinine at the third examination. The aluminium in plasma values rose slightly from the first to the second examination and remained around ca. 16 µg/L (about 2 µg above the average values of the maximum sample as shown in Table 3). The creatinine-related aluminium concentrations showed a significant difference over time (P < 0.001). The changes of the aluminium concentrations in plasma were not significant (P = 0.22).
The comparison between pre-shift and post-shift for neurobehaviourally examined welders with data at all examinations revealed that the post-shift aluminium concentrations were higher than pre-shift concentrations by 30 µg/g creatinine in urine, and by 3.5 µg/L in plasma.
In order to test the sensitivity of the biomonitors to acute shift aluminium exposure, an ANOVA model for repeated measurements was used, applying the factors examination (three levels) and shift time (two levels). The creatinine-related aluminium concentrations in urine were significantly different for examination (P = 0.003) and shift time (P = 0.038) but there was no significant interaction examination x shift (P = 0.88). For data on aluminium in plasma, a borderline significance for shift time was observed (P = 0.062).
The pre-shift biomonitoring data of the welder and control samples used in the neurobehavioural analyses were significantly different with regard to creatinine-related aluminium concentrations in urine (P ≤ 0.001) and to aluminium concentrations in plasma (P ≤ 0.001).


Intercorrelation and temporal stability of biomonitoring data
The temporal stability of the biomonitoring data for neurobehaviourally investigated welders was analysed (Pearson correlation). The correlations of the individual measurements were examined for the two 2-year intervals (E1-E2, E2-E3) and for the 4-year intervals (E1-E3). A significant temporal stability of individual data was observed for the biomonitoring data. A moderate stability for 2-year intervals was seen for the dust data, but no stability over 4 years.
A high intercorrelation and a significant moderate correlation to the dust measures was observed for the biomonitoring parameters of welders averaged across the three examinations.
The linear regressions of the averaged dust values on the creatinine related post-shift aluminium concentrations and on the post-shift aluminium in plasma values suggested that the relation between dust- and creatinine-related aluminium concentrations in urine was slightly more systematic than between dust and aluminium in plasma values.


FINDINGS


Neurobehavioural results
Data for the three neurobehavioural examinations and related MANCOVA analyses are presented in Tables 5, 6, 7 and 8. The interaction between examination and exposure (examination x exposure) was the most important term, which indicates whether exposed and non-exposed subjects show diverging changes of the neurobehavioural parameters in the course of 4 years. Using explorative models, it was examined to what degree exposure or (and) individual cognitive abilities determine the general structure of neurobehavioural data.

Questionnaire Q16
In the symptom questionnaire Q16, welders and control subjects indicated only few symptoms (Table 5). A parallel reduction of the average symptom score was seen in the course of the examinations: from less than 3 to less than 2 in welders and from less than 2 to less than 1 in the control group. Both the group difference and the interaction term (examination x exposure) were not statistically significant under consideration of covariates (Table 6: P = 0.08 and P = 0.45, respectively).

Cognitive abilities
Verbal IQ test (WST): The vocabulary test was conducted at the second examination and was used to assess premorbid intelligence in workers with German as their mother language. Related to the total number of subjects at the end of the study, 75% (n = 15) of the exposed and 58% (n = 7) of the controls could be examined (Table 7). An “average” intelligence was indicated by the IQ values of both groups (estimated from the WST).
The mean verbal IQ score was lower for welders than for controls (95.3 and 98.7, respectively). No significant difference could be stated considering a potential influence of age (Table 8, P = 0.263). A significant influence of CDT was seen in this model (P = 0.027).

Standard progressive matrices test (SPM)
This verbal-free test of logic thinking was conducted only at the first examination. The exposed group showed weaker performances than control group without statistical significance (Table 7). There was no significant group difference in the covariance analysis (Table 8, P = 0.299). A significant influence was seen with the covariate age (P < 0.001).

Block design
The block design test (HAWIE) is a measure of visuospatial thinking and general intelligence. In this test, the exposed group showed significantly lower scores than the control group (Table 8, P = 0.033). The covariate age was significantly related to the outcome variable (P = 0.026).
The time courses of group performances across examinations revealed, however, a tendency to convergence. There was no significant interaction between examination and group factor (Table 8, P = 0.51).

Trail making test
This test is also used for estimation of general intelligence. It cannot be compared to the methods mentioned above, since the speed factor is of high importance. Processing and faults were analysed together multivariately. A not significant tendency to poorer performance was seen in the exposed group (Table 8, P = 0.09) and a borderline influence of age (P = 0.075). The interaction factor 'examination x group' was highly significant (P ≤ 0.001). However, the trends were converging.

Psychomotor performance
Data for the four psychomotor performance tests are presented in Tables 5 and 6. Time for processing as well as faults of the dominant and non-dominant hand were analysed simultaneously in multivariate statistical approaches.

Steadiness, line tracing, aiming, tapping
No significant group differences were found in the four subtests of the motor performance series (Table 6). In the 'line tracing' test, the exposed group showed a borderline tendency to higher error numbers and longer error times (multivariate, P = 0.066). The age factor showed a significant influence on the 'line tracing' performance (P = 0.002). No significant interaction term 'examination x exposure' was seen in the psychomotor performance tests (Table 6; steadiness P = 0.746, line tracing P = 0.772, aiming P = 0.671, tapping P = 0.452). This indicated diverging courses of exposed subjects and controls.

Digit span, working memory Recall of digits (HAWIE)
There were no significant difference between the exposed and control group regarding the short-term memory performance in forward and backward reproduction of short number sequences (Tables 7, 8). The multivariate analysis of the interaction of forward and backward recall showed no statistical significance (P = 0.661).

Digit span, symbol-digit substitution (EURO-NES)
No significant difference between the exposed and control groups (Table 7; 8, P = 0.38) and no significant interaction (P = 0.417) in regard to the temporal trends was observed in the joint multivariate covariance analysis of the EURO-NES computer tests 'digit span' and 'symbol-digit substitution'.

Attention Simple reaction time
There were neither a significant group difference (Table 6, P = 0.894) nor significant interaction 'examination x exposure' (Table 6, P = 0.107).

Switching attention
Neither group differences nor interaction effects were indicated by the common, multivariate analyses of reaction times and errors in the three subtests of the 'switching attention' test (side, arrow, mixed) (Tables 7, 8).


INCIDENCE / CASES
- Incidence/ Number of cases for each disease / parameter under consideration:
- Other:


STATISTICAL RESULTS
- SMR (Standard mortality ratio):
- RR (Relative risk):
- OR (Odds ratio):
- Other:


OTHER OBSERVATIONS:

Explorative modelling: a priori differences and premorbid differences
The group difference in the block design test (P = 0.026) was the significant difference observed. In the trail making (P = 0.09) and line tracing (P = 0.066) tests only tendencies to differences were seen. The following analyses were conducted to investigate structural relationships between performance and exposure. Intelligence performance was examined either as dependent variable in relation to exposure measures, or as independent variable and possible moderator in relation to neurobehavioural performances in different explorative models.

Stepwise regression: exposure and general intelligence as main predictors of neurobehavioural performances
In a first step it was investigated, whether exposure parameters (diverse biomonitoring measures, dust, and exposure duration) explain variance in 'block design' performance. The 'block design' performance was not dependent on exposure parameters as shown in the stepwise regression (P > 0.05).
In a second model variables potentially showing power to predict performances in the several other neurobehavioural tests were tested. Block design was integrated in these explorative models as an independent variable predicting performance, assuming that it represents a general factor of intelligence. Several exposure parameters and exposure duration, education, age, and CDT were used as potential predictors besides block design. The significant predictors of neurobehavioural performance were determined by stepwise regression for each of the 32 neurobehavioural test parameters, whose number arose from 'examination x test x parameter’. These models were restricted to exposed subjects.
The regressions showed that education and exposure parameters had no meaning in all 32 models, in accordance to the results of multivariate analyses of covariance. The variable 'block design' was significantly associated with neurobehavioural test performance, explaining in 11 different regressions 13-48% of the variance of the respective dependent neurobehavioural parameters.

Explorative models analysing concepts of premorbid intelligence: education versus block design versus verbal IQ
Education, considered as a surrogate of 'premorbid intelligence', was not effective as covariate both in the regression analyses and multivariate analyses of covariance. The results of this study showed that 'block design' is a stable variable, having trait character in the examination course and not being associated to exposure. Therefore, block design was used as covariate in the two covariance analyses instead of school education, which had shown borderline differences between groups. For the purpose of comparison, additional covariance models with 'verbal IQ' were tested using the reduced German-speaking samples.
The error probability for group differences in the 'trail making' test was reduced from P = 0.09 to a new adjusted probability of 0.42 by the MANCOVA with age, CDT, and 'block design' as covariate. The same approach with 'verbal IQ' resulted in an error probability of 0.21.
A MANCOVA of 'line tracing' with 'block design' as covariate reduced the error probability from P = 0.066 to the new adjusted error probability of 0.20. With 'verbal IQ' instead of block design, the same approach resulted in a similar increase of error probability to P = 0.22.

Confounding factors:

Education, intellectual ability level, age, and objective indicators of alcohol consumption (CDT) were taken into account in statistical and interpretative judgements.

Strengths and weaknesses:

On average, the 10th to the 15th year of working life was examined for every exposed subject. Only cautious backward extrapolation and long-term forecasting are allowed by this short period allows. Although the neurobehavioural parameters did not correlate with exposure during the study period, neither early nor late aluminium exposure effects can be fully excluded.
Limiting factors are the small sample size, which reduces the statistical power, and possible selection tendencies. However, by using repeated measurement analyses, the longitudinal approach can compensate for the small sample size. Controlling the variability due to differences between workers and separating it from exposure effects and error variance are the advantages of repeated measurement models. In addition, the precision of the study was increased by introducing measurements on covariates.

Any other information on results incl. tables

Refer to attached background material.

Applicant's summary and conclusion

Conclusions:

In this longitudinal study, aluminium exposure and neurobehavioural data of aluminium welders in the train and truck construction industry were compared with the data of a non-exposed control group from the same enterprises. The cohort consisted of 20 (initially 44) male subjects in the exposed group and 12 (initially 37) subjects in the control group, matched for gendre, age, education, physical work environment and social environment. Three examinations were carried out over a period of 4 years at 2 years-intervals. Exposure was assessed as total dust in air as well as aluminium concentrations in pre-and post-shift urine and plasma. The exposed and control groups were examined for neurobehavioural performance with a battery of neurobehavioural tests comprising symptoms, verbal intelligence, logic thinking, psychomotor behaviour, memory, and attention. Computer-aided tests from the Motor Performance Series (MLS) and the European Neurobehavioural Evaluation System (EURO-NES) were employed. Biomonitoring data and the relationship to neurobehavioural data were examined with correlation and regression analysis methods.
Neurobehavioural data were analysed with multivariate covariance-analytical methods (MANCOVA) considering the covariates age, indicators of 'a priori' intelligence differences (education or 'premorbid' intelligence), and alcohol consumption (carbohydrate-deficient transferrin in plasma, CDT).
The mean total dust concentration during welding, near to the routinely worn ventilated helmets, ranged from 5 to 8 mg/m³. In the course of the study, aluminium concentrations in urine and plasma were in the ranges of 88-140 µg/g creatinine (pre-shift) and 13-16 µg /L, respectively, showing a high long-term stability but also sensitivity to acute shift dependent exposure changes.
No significantly increased symptom levels were observed in the exposed group (subjects who had been working as aluminium welders at an average of 15 years) when compared to the control group.
No correlation was observed between biomonitoring and neurobehavioural performance variables using explorative regression and covariance analyses. Likewise, there was no significant difference between the aluminium-exposed and control groups in the neurobehavioural performance courses during the 4 years period. Explorative modelling indicated that possible indicators of intellectual 'a priori' (premorbid) differences between subjects but not by their exposure information could determine the structure of neurobehavioural outcomes.