Three-year follow-up of serial nerve conduction among lead-exposed workers

Three-year follow-up of serial nerve conduction among lead-exposed workers. Objectives The purpose of this study was to investigate the serial nerve conduction parameters of a group of lead-exposed worlcers and determine their correlation with the serial blood lead results over a three-year period. A "nonresponse" level (defined as no significant changes in the nerve conduction in response to changes in the blood lead level of each exposed worker observed over the period of study) was also determined for blood lead in respect to the peripheral nerves. Methods Seventy-two male workers from a lead battery manufacturing factory were followed at six-month intervals for three years. At each follow-up, the blood lead level was determined and nerve conduction tests (ulnar and median nerves) were conducted. A group of 82 unexposed subjects served as referents. Results Significant differences were observed for some of the mean values of the median nerve conductio~l parameters between the exposed and reference groups. The 28 exposed workers who completed the follow-up were divided into the following two blood lead categories: <40 pg . 100 ml-' (< 1.93 pmol . I-') and 2 40 pg . 100 ml-I (2 1.93 pn~ol . I-'). In the latter, the median motor conduction velocity, median distal latency, median amplitude, ulnar motor conduction velocity, and ulnar amplitude were significantly correlated (adjusted for age and within-subject variation) with the blood lead levels, but not so in the former. to the three-year serial results, the "nonresponse" level for blood lead with respect to the peripheral nerves would be <40 pg . 100 ml-I (< 1.93 pmol . 1-I).

To our knowledge, only three epidemiologic studies (1)(2)(3) have been published that concern the potential neurologic effects of lead exposure and report repeated prospective testing of the same adults. However, there were shortcoinings in these studies. In the study by Spivey et al (1), the referents could have been exposed to lead and other neurotoxic metals in aluminum processing plants. The study of Araki et al (2) suffered from small numbers (19 gun inetal foundry workers) and the exposed workers were also exposed to elevated levels of zinc and copper, which may antagonize the effect of the lead (4). Similarly, Seppalainen & Hernberg (3) reported no significant differences between referents and 11 exposed worl~ers after four years of exposure. One possible explanation given by the authors was the small number of subjects left at the end of the study.
To date, there is no report of a sufficiently large cohort of lead-exposed workers for which serial measurement of the nerve conduction parameters and the blood lead level (B-Pb) have been studied. The objectives of our study were (i) to study the serial nerve conduction values of a group of lead-exposed workers and correlate them with the serial blood lead results over a three-year period and (ii) to determine a "nonresponse" level for blood lead in respect to the peripheral nerves. The term "nonresponse" used in this study is defined as no significant changes in the nerve conduction parameters in response to changes in the blood lead level for each exposed worker observed over the period of study.

Study population
ing, and the forming, assembling and welding of the battery connectors. The worlters are usually rotated through these various sections and had had considerable exposure throughout their worlt life.
The reference group consisted of 82 male factory worlters who were undergoing a preeinploytnent medical examination. None of the referents had any significant history of exposure to lead or any ltnown neurotoxic substances.
The exposed and reference subjects were selected according to the following criteria: (i) no history of any metabolic diseases, for example, diabetes tnellitus or thyrotoxicosis, (ii) clinically normal, and (iii) no alcohol consun~ption of more than 50 g of alcohol per day.
The 72 worlters were followed at six-month intervals, by occupational physicians as part of a compulsory medical exanlination for workers exposed to lead (5). At each examination, blood was taken for analysis. A nerve conduction test was carried out for each of the worl<ers by the same technician througl~out the study period of three years (1992)(1993)(1994). Although we started with 72 worlters, only 28 completed the three-year period. Table  1 gives a breakdown of the numbers of exposed subjects studied at six-month intervals and their corresponding geometric mean B-Pb levels.
The original study protocol was to follow the 82 referents throughout the three-year period. Unfortunately, as the study was strictly voluntary, at the end of the first year o~lly 26 subjects were still in the study. Only four subjects completed the three-year period. As such, the serial results of the referents have not been used in the subsequent analysis of the serial data.
The basic characteristics of the study population are presented in table 2. No marked differences were noted in the age and alcohol intake of the referents, all the exposed subjects and the exposed subjects who completed the three-year follow-up. There was a higher percentage of Malay and stnolters among the exposed workers. 4.4-19.8) pg . 100 1lil-'. There were no marked differences in the group of worlters who dropped out before the cnd of the three-year period.

Biological measurements
A blood sample was obtained by venepuncture with leadfree disposable syringes and stored in lead-free bottles at each examination. The B-Pb was deternlined using an atomic absorption spectrophotometer with a granite furnace. External quality control was carried out under a quality assessment scheme (NEQAS) in the United Icingdorn. Table 3 shows the accuracy in determining B-Pb since 1985.

Nerve conduction methods
The nerve conduction studies were performed for each subject by the same technician (throughout the study period) from a nemology laboratory. The measurements were made with a Medelec electrophysiological system (Model MS6). The room temperature was maintained constantly at 30-31°C and sltin itnpedance was kept below 10 1tQ for all the recordings.
The ~naxinluin rnotor conduction velocity and distal motor latency of the rnedian and ulnas nerves were determined in the dominant forearms. The median rnotor con- Table 1 . Mean blood lead levels of the exposed subjects at the time of the nerve conduction tests. (GSD = geometric standard deviation) Nerve N Blood lead concentrationb conduction testsa (r.lq . 100 m l l )

Geometric
GSDb mean The mean B-Pb for the reference group was 10.5 (range a The nerve conduct~on tests were conducted at six-month intervals  duction velocity and distal latency of the median nerve were obtained by stimulation of the median nerve at the wrist (13 c n~ from the base of the forefinger) and at the elbow. The ulnar motor conduction velocity and distal latency of the ulnar nerve were obtained by stimulation at the lateral aspect of the wrist (1 0 cm from the base of the little finger) and just below the medial epicondyle.
The muscle action potential of the thenar (for the median nerve) and hypothenar (for the ulnar nerve) muscles were recorded with surface disc electrodes. The median maximum sensory nerve conduction velocity and median sensory amplitude were measured by stimulating the forefinger with ring electrodes and recording orthodrotnically at the wrist (13 cm from the base of the forefinger) continuously for 10 s. Action potentials from two sites along the nerves were picked up, amplified, and averaged. Artifacts causing overloaded signals, for example, introduced by muscle contraction in the region of the recording electrodes, were automatically rejected. Similarly, the ulnar maximum sensory nerve conduction velocity and the ulnar sensory amplitude were measured by stimulating the little finger with ring electrodes and recording orthodromically at the lateral aspect of the wrist (10 cm from the base of the little finger).
In the base-line measurements made at the beginning of the study, there were significant differences in the median sensory nerve conductioll velocity, median motor conduction velocity, median nerve, and median sensory amplitude means between the exposed and reference groups, even after adjustment for age, ethnic group, smoking and drinking habits in the ANCOVA (table 4). The ulnar nerve conduction values were not significantly different for the exposed worlcers and the referents, except for the ulnar nerve.

Reproducibility of the serial nerve conduction parameters
To ensure the reproducibility of the nerve conductioil values, the nerve conduction values of seven healthy staff members of our department [with B-Pb levels of 2.6-5.0 pg . 100 ml-' (0.12-0.24 pmol . I-') and in the age range of 26-43 years] were tested by the same procedure throughout the study period. Each person repeated the tests seven times. The analysis of the results of the seven subjects did not show any significant within-subject variation (table 5).

Statistical analysis
The distribution of the nerve conduction values, and B-Pb levels were skewed. Logarithmic transformation was used to normalize the distribution. The mean nerve conduction values were compared between the exposed and reference subjects after adjustment for age, ethnic group, and smoking and drinking habits in an ANCOVA. A cluster sample is often used to investigate the statistical relationship between numeric variables. Typically cluster sampling is carried out in the following two stages: sampling of clusters and then sampling of measurements within clusters. Standard regression analysis is often employed to estimate the regression parameters, namely, the intercept (alpha) and linear regression coefficient of Y on X (beta), for data from a clustered sample, for which the unit of analysis is a measurement (X and Y) in which no consideratioll is given to the cluster identity of the measurements. This method is incorrect because standard regression stipulates that the Y measurements are statistically independent, whereas in a clus-  tered sample the Y measurements tend to be more alike within a cluster than between clusters. The method used in the analysis of the serial nerve conduction values and the B-Pb levels (which was a clustered sample) was based on the analytic method used by econometricians (6). The analysis of the within-cluster regression coefficient was done by the statistical analysis system package (7) together with SAS codes, packaged as a macro (8).
The exposed subjects who completed the three-year follow-up were thus selected for the analysis of the within-cluster regression coefficients of the serial nerve conduction and the B-Pb levels. To assess the "nonresponse" effect, the B-Pb levels of the 28 exposed subjects were divided into different categories. Each category was then tested using the "within-cluster" package to see if there was any relationship between the nerve conduction values and the B-Pb levels. any association, especially for the ulnar nerve parameters. Even for the median nerve conduction parameters, the regression correlation coefficients were of smaller magnitudes for the lower B-Pb category. Because or the small number of workers (28 left) and the three-year period and nawow B-Pb range, we could not divide the Table 6 shows the serial measurement data for the nerve B-Pb levels into more groups. Neither could we group conduction parameters and B-Pb levels. The mean B-Pb / the worl<ers by a certain maximum B-Pb. Therefore, Table 6. Serial neurophysiological data and blood lead levels of the exposed subjects at the six-month intervals over the three-year period. (

Discussion
The mean B-Pb for the exposed worlters was 36.9 yg . I00 ml-I (1.78 ymol . I-'). At this level, significant differences where observed for all the median nerve conduction parameters of the exposed workers when they were compared with the referents. However, there was no significant difference in the ulnar nerve conduction parameters of the exposed workers and the referents (except for distal latency). Therefore the median nerve appeared to be more susceptible to lead effects than the ulnar nerve. Davis & Svendsgaard, in a critical review and meta-analysis of nerve conduction studies and lead workers (9), also reported 011 this apparent susceptibility of the median nerve to lead. We will not comment on this phenomenon further as the susceptibility of the median nerve and the measurement of distal latency have been discussed in an earlier paper (10). However, we would like to stress that our finding is unliltely to be confounded by carpal tunnel syndrome among the exposed worlters. First, the exposed workers were not subjected to repetitive work since they are rotated through various sections of the factory (as mentioned earlier). Furthermore, none of the workers complained of paresthesia, numbness, or weakness of their fingers. Thus far, there have been no reports of worlters with work-related carpal tunnel syndrome in Singapore from the manufacturing of car batteries although it is a notifiable disease under the Workmen's Compensation Act of 1975 (1 1). Ehle (12) reviewed 15 articles on the correlation of B-Pb and nerve conduction velocity. He concluded: "In summary, it would appear from review oT all the studies of blood lead levels vcrsus nerve conduction velocity that the rnajority of studies do not support a coutinuous relationship between blood lead and nerve conduction velocity in adults at blood lead levels below 70 pg1100 ml (3.39 pmolll) [p 21 11" (12). The statement made by Ehlc involved studies correlating current B-Pb levels with nerve conduction \ielocities. Peripheral neuropathy caused by lead is usually associated with chronic lead exposure rather than with acute exposure, unless the exposure is very high (13). Indeed our result (table 6) reinforced the point that it takes time for the effects of lead on nerve conduction parameters to be seen. Although the mean B-Pb levels for the group pealted in the third test [48.3 pg . 100 inl-I (2.33 pmol . I-)'], the corresponding decrease in the median sensory nerve conduction velocity and ulnar sensory nerve conduction velocity was only observed in the fourth test (ie, six months after the third test was conducted) (table 6). As damage of the peripheral nerves takes time, it would be more objective to loolt at serial B-Pb levels rather than current B-Pb values to determine the effect of lead on the peripheral nervous system.
One of the drawbacks of this study on serial nerve conduction tests and B-Pb results is the lack of a reference group (with serial readings) for comparison. A reference group of 82 subjects was followed at the begillning of the study, but oiily four remained at the end of the three-year period. Although a reference group was not available, a srnall group of seven subjects were followed throughout the study period. No significant within-subject variation was observed for the nerve conduction parameters studied, and this finding demonstrates the repeatability of the nerve conduction tests collected throughout the study. Furthermore, it would have been highly unliltely for the reference group to have had a wide fluctuation in their B-Pb levels if they had been followed up, as they were not exposed to lead. Thus the objectives of this study could still be studied without a reference group to observe the serial results of the nerve conduction parameters and B-Pb levels.
Another concern is related to the high attrition rate of the exposed groupfrom an initial number of 72 to 28 worlters. The high attrition rate was unavoidable as the study was strictly voluntary and worlters who left the company, usually for a better paying job, were not keen to be followed further. The profiles of the worlters who left were fairly similar to those who con~pleted the study with respect to their mean B-Pb levels, exposure duration, and mean age (table 2). Thus these workers were not very different from those who completed the study.
One of the objectives of this research was to study the serial results of the nerve conduction parameters and the association with B-Pb levels. We have shown that certain nerve conduction parameters correlate significantly with B-Pb levels over a three-year period when adjusted for age and within-subject variation in nerve conduction (table 7). The correlation coefficients for the B-Pb level and the median nerve coilduction parameters were stronger than for the B-Pb levels and the ulnar nerve conduction parameters. Again, the serial results showed that the median nerve is more susceptible to the effects of lead than the ulnar nerve is.
In the group of exposed workers with a B-Pb level of 2 40 pg . 100 ml-I, the age and within-cluster adjusted regression correlation coefficients of the median motor conduction velocity, median nerve, median sensory amplitude, and ulnar sensory nerve conduction velocity were significantly correlated with the B-Pb levels and of a larger magnitude than that of the group with a B-Pb of <40 pg . 100 ml-' (table 4). In addition, the group with a B-Pb of <40 pg . 100 ml-I, the adjusted regression correlation coefficients were not significant. Spivey et a1 (1) studied 55 lead-exposed worlters and 31 referents initially and after 12-18 months of lead exposure. There were no significant changes from the initial to the final values for the ulnar motor and sensory and the peroneal motor and sural sensory nerve conduction velocities. However, the referents' (aluminumprocessing worlters) mean B-Pb level was 22.1 pg . 100 ml-I (1.07 pmol . I-'), which subsequently illcreased to 27.9 pg . 100 ml-I (1.35 pmol . I-') in a year's time.
The referents could have been exposed to lead and other neurotoxic metals in the aluminum-processing and therefore the results could have been confounded. Although the B-Pb analyses were done on a monthly basis (for most worlters), no attempts were made to study the subjects' nerve conductioil serially.
Nineteen gun metal foundry workers were studied by Aralci et a1 over a two-year period (2). Motor and sensory conduction velocities of the distal radial and median nerves, together with the B-Pb level, were measured twice in a 12-month interval. They reported that yearly changes in the motor conduction velocity of the radial nerve and the sensory conduction velocity of the median nerve (forearm) were inversely related to the change in alcohol ingestion rather than to the lead levels. It must be noted that the exposed workers were also exposed to elevated levels of zinc and copper, which may antagonize the effect of lead (4). This possibility may explain the negative findings.
Seppalainen & Hernberg (3) studied both median and ulnar motor and sensory conduction in 23 lead workers after the initial year of lead exposure (3). Fifteen of these men were available for restudy after two years of exposure, and 11 were restudied after four years. At one year, the median lnotor distal latency and sensory distal conduction were significantly slower than in the matched referents. At two years both the median sensory distal conduction and forearm sensory conduction were significantly slower, but the median motor distal latency was no longer slower. No significant differences were present between the referents and the 11 exposed workers after four years of exposure. The inconclusive result of the study is probably a result of a lack of power, as there were only 1 1 exposed worlters at the end of the fourth year (15).
Hernberg et al, ill a summary article on the same study (15), reported that a slowing of the nerve conduction velocities occurred only for the lead workers whose maximal B-Pb levels were between 30 (1.45 pmol . 1-I) and 50 (2.42 pmol . 1-I) pg . 100 ml-I (15). Those with a B-Pb level below 30 pg . 100 ml-I for up to two years had no significant changes in their nerve functjons; their average nerve conductioil velocities remained at the initial level and also at the level of the referents.
Our findings are similar to those of Hernberg et a1 (15). In the category of B-Pb 2 40 pg . 100 ml-I, with a mean B-Pb of 49.7 pg . 100 ml-I (2.40 pmol . I-'), there were significant correlations of the median motor conduction velocity, median distal latency, and ulnar motor conduction velocity data with the B-Pb levels. This significant correlation was observed for up to a three-year period (adjusted for age and within-subject variation). The other nerve conduction parameters were not significantly conelated. This result is consistent with the linowledge that inotor nerves are more susceptible to the effects of lead than the sensory nerves are. It also shows that a B-Pb of 2 40 pg . 100 ml-I does affect the motor nerve conduction since there is slowing of both the median and ulnar motor conduction velocities with increases in the B-Pb levels.
In the category of B-Pb <40 pg . 100 ml-I, all the nerve conduction parameters correlated poorly with the B-Pb levels. The subjects' median and ulnar nerves (for those parameters measured) were not affected at B-Pb levels of <40 pg . 100 ml-I within the three-year period.
In light of the results of this study, we would lilte to suggest that the "nonresponse" level of blood lead could be considered to be <40 pg . 100 ml-I with respect to the peripheral nerves. More studies are needed before our suggestion can be confirmed.