Solvent exposure, alcohol consumption and liver injury in workers manufacturing paint.

Liver enzyme activity was examined in 89 South African paint makers currently exposed to a mixture of organic solvents at fairly low levels. However, the duration of exposure was substantial for many. Fifty-eight workers (65%) had at least one enzyme value above the upper reference limit. Gamma glutamyl transferase (gamma GT) activity was elevated in 46% of the workers and aspartate aminotransferase (ASAT) in 52%. In a comparison between workers with high and low solvent exposure, the gamma GT and ASAT values were higher in the most exposed group (eg, gamma GT mean 108 versus 69 U.l-1, P > 0.05). Adjustment for confounding by alcohol consumption and body mass index eliminated the differences due to exposure. It was concluded that the measures of liver injury used did not demonstrate solvent-induced hepatic damage but that excessive alcohol consumption was an important factor.

Case reports of liver injury caused by occupational exposure to organic solvents are well known. Consequently the hepatotoxic potential of some organic solvents, notably aliphatic chlorinated hydrocarbons, is not in doubt. Convincing epidemiologic evidence of liver injury in workers exposed to mixtures of solvents at usual workplace levels is, however, scant. Many studies reporting liver injury in exposed workers are based on mortality data (1) or case series often without an adequate comparison group or complete exposure data (2)(3)(4). Furthermore, negative studies are well represented in the literature (5)(6)(7), although their value is limited to some extent by the small number of workers with heavy solvent exposure.
In general, it can be said that the epidemiologic evidence supporting a causal relationship between chronic exposure to mixed solvents at low levels and clinically significant liver disease is weak. This statement does not necessarily imply however that the investigation of liver function in workers using these compounds is not warranted. Nonworkplace hepatotoxins such as alcohol and medicinal drugs may increase the risk of liver injury in these workers. Also of concern, particularly in underdeveloped countries, is the high incidence of infective liver disease (8) and the possible role of poor nutrition in increasing hepatic damage in these workers. The combined effect of occupational and nonoccupational factors on workers in developing countries may be important and was the motivation for this study of liver function in workers manufacturing paint in a South African factory.
Workers manufacturing paint are usually exposed to a mixture of aliphatic and aromatic organic solvents at relatively moderate intensity. To our knowledge no studies of the hepatic effects of this type of industrial exposure have been conducted in South Africa nor in a developing country. The major objectives of the cross-sectional study reported in this paper were (i) to determine the prevalence of abnormalities in routinely used biochemical indicators of liver injury in long-service workers making paint, (ii) to investigate the association between these indicators and solvent exposure, and (iii) to examine possible modifications in solvent effects caused by alcohol consumption and past exposure to hepatitis B virus.

Subjects
Two hundred and thirteen African men were employed at the paint factory. Limited resources prevented the inclusion of all 213. Consequently, only those workers with eight or more years of service and those with shorter service who were currently working in areas with relatively high solvent exposure (as identified by factory management and walkthrough inspections of the factory) were asked to participate in the study . Ninety workers satisfied the first criterion, and 12 the second. The response rates were 91% (82 workers) and 92% (11 workers), respectivelythus 93 worker s participated in the study. Reasons for nonresponse were prolonged sick leave for two workers (diagnosis not established), traumatic death for one, and refu sal to donate blood for the remainder.
Four of the 93 participants were excluded from the analy sis either becau se of missing data (three subjects had blood specimens damaged in transit and were reluctant to provide a repeat specimen) or positive HB s antigen (one subject).

The workplace
Paint manufacture started on the present site in the mid-1950s, and the product ion proce ss has remained fundamentally unchanged for many years. Improved paint technology has, however, resulted in alterations in the constituents of the paint s, and solvent usage has varied over the years. Recently the factory has predominantly manufactured industrial solvent-based paints , but water-based decorative paints have made up roughly a third of the 10 million liters of paint produced per year. The major inter vention to decrease levels of solvents in workplace air has been the covering of paint-filled pots with plastic sheeting to reduce the surface area from which solvent vapor could be generated. Long-service employees testify to the success of this strategy, and a reduction in the air levels of solvents with time is very likely . The current levels should be treated as the minimum to which the workers have been exposed.
The general production proce ss of the plant is as follow s. Dry raw materials, such as pigments, are ball-milled in a resin vehicle (milling section ). The milled material is then transferred to a mixer where it is extended with solvents and mixed (mixing section). The paint is then decanted into containers (filling secti on) or customized by the addition of special -function agents (special function sections I and 2) before the decanting. Paint is moved between sections in large pots, and these pots are cleaned with solvents in the pot cleaning section. The final steps are labeling (labeling section) and despatch (warehouse). Water-based paints are made in a separate polyvinyl acetate plant. As usual , in a paint factory , several other specialized processes are in operation. Small batches of special application paint are made in batches sections 1 and 2; solvents are decanted into small containers in the thinners section ; cans of spray paint are produced in the aerosol section; paint from damaged containers is decanted into new containers in the refilling section; and special colors are mixed in the small mix section. The production process is supported by workshops, stores , qualit y control, transport services, and administration offices. Exposure After written con sent was obtained, a questionnaire covering alcohol consumption, medication, previous liver disease, and full details of past and current occupational history was admini stered by a trained interviewer to each subj ect in his home language. Alcohol consumption was assessed by detailed inquiry of both the current and past peak weekly intake of beer, wine, spirits, and traditional alcoholic beverages. Total weekly intake was converted to grams per day for current and peak consumption separately.
The current exposure to solvents was determined from measurements of workplace air level s of the eight most extensively used solvents (table 1), and past exposure was assessed from data collected during routine environmental monitoring performed by the company's Occupational Health Service during 1987-1989. The current evaluation of solvent in air was conducted by an experienced industrial hygienist who divided the factory into solvent and nonsolvent sections according to walkthrough inspections and the routine monitoring data . The solvent sections were all those in which solvent expo sure was likely, and they included sections directly involved in paint making , decanting, or cleaning with solvents, plus sections with historical solvent levels in excess of 5% of the thre shold limit value (TLV) of the American Conference of Governmental Industrial Hygieni sts for solvent mixtures (9). Level s of solvent in air were determined from samples obtained from 41 solvent section workers wearing personal samplers comprised of an activated charcoal tube connected by tubing to a low-flow (20 ml . min:' calibrated orifice) Casella SPI5 pump (London, United Kingdom). Vapors collected on the activated charcoal were determined with gas chromatographic analyses (10). The selection of workers to wear personal sampler s was done by the industrial hygieni st. who inspected each solvent section and selected workers representative of all the important procedures or jobs (tasks) performed in that section . Limited resources prevented the use of a random selection method such as that proposed by the National Institute for Occupation al Safety and Health in the United State s (11). The measurements took place during a period of routine production. The proportion of workers who wore samplers was 0%,14%,25 %,27% and 50% in aerosol, batche s I, mixing , filling and special function 1, respect ively , and 100% in the remaining sections. Solvent levels were not measured in the aerosol section as workers were temporarily engaged in nonroutine activity.
The solvent levels in the air of the solvent-exposed section s are shown in table 1, along with the number of workers currently employed in each section. Eighty-three workers were currently working directly with solvents, and the solvent level in the breathing zone ranged from 50 mg . m-3 (special function section 2) to 434 mg . rrr? (pot cleaning section). The past and current levels were not always in agreement  (1987)(1988)(1989) levels of solvents in the solvent sections of the paint factory. Levels for the three most important solvents, the sum of all eight solvents measured (total solvent exposure), and the proportion of the thresholc limit value (TLV) for the m ixture of solvents (% TLV)" are presented. ( • TLV of mixture (C,fT,) +(e,tT,) + (c,tT3) ... where C, is the measured atmospheric concentration of a particular solvent and T, Is the threshold limit value. • Numberof workers currently employed in the section in parentheses. c Includes toluene, xylene, white spirit, ethyl acetate, methyl ethyl ketone, methyl-lsobutyl ketone, N butyl acetate, and cellulose acetate. (eg , for the pot cleaning section), probabl y because of the vari abilit y in daily work proc edures and factors such as ambient temp erature and natural ventilation. The co mbined 1987-1990 levels were used in the calculations of the cumul ative expo sure and the seco nd exposure index. Not show n in table I is that three samples were colle cted from three workshop workers who all had samples below the detection limits.
Thirty-four subje cts reported working with solvents prior to employment at the study factory. In general this work was for short periods only, mean 2.2 (SO 4) years. The distribution of this additional service relative to the cumulative exposure at the paint factory is shown in table 2.
The hepat otox ic potential of the various solvents was considere d to be equal. Therefore an additive model was used to calculate the TLV for mixtures of solvents. The TLV for the mixture of solvents was thus (e/T) + (C/T 2 ) + (C/T}) .. ., where C was 238 the measured atmo spheri c concentration and T the threshold limit value of a particular solvent.
Cumul ative exposure (CE) to solvents was ca lculated for each worker by multipl ying the number of years spent in a section by the level of mixed-solvent exposure, as the percentage of the TLV (%TLV), and then adding these products for each section in which the worker had worked: In the calculations of the cumulative exposure, factory sections which did not use solvents but which were directly adj acent to solvent sections were arbitrarily assigned a mixed-solvent exposure level of 5% of the TLV. The remaining nonsolv ent sections were assig ned an exposure level of 1%. Table 2 shows that the cumulative exposure to solvents was used to categorize workers i nto the t wo exposure groups with the cutoff for low exposure set at < 100 TLV% years. This cutoff was chosen to separate the low-and high-exposure groups into approximat ely equ al sizes, while ensuring that the group with low cumulative exposure did not includ e many worke rs with substantial current exposure. The distribution of the cumulative expo sure was positivel y skewed, with a median of 121 TLV % years. Using the median as the cutoff instead of 100 TL V% years altered the result s shown in table 2 only minim ally. As can he seen from the table , the average duration of service was substantial with a mean of 14 years. Solvent exposure at other workplaces was very similar for the groups, but the average duration of service at the paint factory was longer for the group with a high cumulative exposure ( 17 versus II years). The average cumulative exposure differed markedly, being 28 and 408 %TLV years for the low and high exposure groups, respectively. The current solvent expo sure levels were fairly low. The group with low cumulative exposure was significantly younger, but the body mass inde x did not differ between the groups (table 3).
A second exposure index was derived as for the cumulative exposure, except that the percentage of the TL V was replaced by total solvent exposure in milligrams per cubic meter.

Liver injury
Liver function was assessed through an analysis of serum bilirub in and the activities of the serum enzymes gamma glutamyl transferase ()'CiT), alkaline Scand J Work Environ Health 1993, vol 19, no 4 phosphatase (ALP), alanine aminotransferase (ALAT), and aspartate aminotransferase (ASAT) (table 4). These analyses were performed by the South African Institute for Medi cal Research -which is the regional reference laboratory -using standard methods on a Boehrin ger Mannheim Hitachi model 704 automatic analyzer (Tokyo, Japan) . Specimens were analyzed on the day of collection. The upper reference value for these determinations is shown in the legend of figure I . Workers with any value excee ding the reference were investigated for evidence of hepatitis A (HA) and hepatitis B (HB) virus infection [A: HA immunoglobulin M antibody; B: HB core (HB c) and HB surface (HB s) antibody and antigen], and a liver specialist was consulted regarding further management.
The workers had not been monitored for liver injury by the Occup ational Health Service prior to this study and the occup ational health staff had no recall of any worker being relocated from a worksite with high solvent expos ure to one with low exposure as a consequence of an abnormality in liver function.

Analysis
Multiple linear regression was done to investigate the relation ships between solvent exposure (cumulative exposure, second exposure index, duration of service in paint manufactur e, and current solvent exposure), Quetele t' s body mass index [weight (kilogramsj/height-(metersj], current alcohol consumption, age (years) , and the dependent variables )'CiT, ALP, ALAT, ASAT and the ALAT:ASAT ratio . An Table 3. Body mass inde x, age, and alcohol use of the workers accord ing to exposure group.  elevated ALAT:ASAT ratio has been used to differentiate solvent-induced liver injury from alcohol damage , which can cause a disproportionate increase in ASAT activity (12). The dependent variables were not normally distributed and were transformed (log 10) for the regression analyses, which were done with the help of PC-SAS (statistical analysis system for a personal computer) (13).
In most instances the exposure and outcome (serum enzyme activity) data were not normally distributed, and nonparametric tests of statistical signif icance were used unless otherwi se indicated.

Results
Fifty-eight of the 89 workers (65%) had a biochemical index of liver function above the upper reference value . Figure I shows the extent of liver function abnormality for each indicator tested. The )GT activity was abnormal for 41 workers (46%) , as was the ASAT for 46 workers (52%), while bilirubin was mildly elevated (22 umol . I-I) in one worker only.
Medication was unlikel y to be an important factor in accounting for these abnormalities. Although 34 workers (38%) had used a drug within three months of the study, drug use was not associated with solvent exposure (cumulative exposure) nor with biochemical abnormalities of liver function . In addition, other than hypertension (6 workers, 7%), con-ditions associated with the use of potentially hepatotoxic agents were not reported. For example, no worker had been exposed to an anesthetic recently or treatment for tuberculos is, epilepsy, diabetes mellitus, or a psychological condition within the past five years.
An important feature was the higher current alcohol consumption reported by the group with higher solvent exposure.
Considering the indicators of liver function , it can be seen that the mean )GT was well above the upper reference value (50 U . 1-1) for the whole group and for both exposure categories, the high-exposure group having the higher mean "(GT (108 versus 69 U . 1-1). The )GT values were, however, not normally distributed and the median values for the cumulative exposure groups were less different at 50.0 and 42.5 U· I-I. The ASAT pattern was similar, but the values were relatively less elevated above the reference value of 30 U . I-I. These difference s were not statistically significant at the 5% level for either the untransformed or transformed data . The ALP and ALAT activity and the ALAT:ASAT% ratio were not associated with solvent use.
The group with low cumulative exposure included two worke rs with current exposure at 59% of the TLV for mixtures and four workers with a corresponding value of 35%. Repeat ing the comparisons in table 4 with these six currently exposed workers excluded from the low-exposure group left the enzyme activities essentially unaltered -no change in enzyme value exceeded 2 U . I-I. Thus the relatively high current solvent levels for a small number of subjects with low cumulati ve exposure did not account for the lack of significant differ ences between the workers with high and low cumulati ve exposure.
To compare enzyme activity in groups with a greater contrast in solvent exposure, we analyzed the first and fourth cumulati ve exposure quarters. The mean cumulative exposures for these two group s were 12.7 (range 9-18) and 707.9 (range 386-1323) TLV% years . The differences in the )GT and ASAT values were large -"(GT 70.1 versus 132 U . I-I and ASAT 31.9 versus 42.2 U . I-I for the first and fourth quarters respectively -but did not reach statistical significance in either case (P>O. I) . This analysis was repeated for the first and fourth current solvent exposure quarters. The current mean solvent exposure was I % of the TLV (below detection limits) for the first quarter and 44% of the TLV (or 196 mgm") for total solvents in the fourth quarter, but the median "(GT, ASAT, ALAT, and ALP values were all slightly higher for the low current exposure quarter. To control for the effects of alcohol consumption and assess the contribution of paint manufacture to the effect on liver function, we stratified workers into low, medium, and high alcohol consumption groups so that exposure effects could be compared in groups with similar alcohol usage. alcohol effect on )'CiT and ASAT in both the low and high cumulative expo sure group s. A possible exposure effect was present in that the high cumulative expo sure worker s who drank littl e (0-19 g . d-' ) or moderatel y (20-79 g. d-I ) had)'CiT and ASAT values above the reference levels (50 and 30 U . 1-1 , respectively), and their values were also higher than the corresponding low cumulative exposure workers. This difference was unlikely to be due to obesity as the number of workers classified as obese (body mass index >29.9) was similar for the low-and high-exposure alcohol group s. The highest mean )'CiT and ASAT activities were measured for workers with high alcohol consumption in the low-exposure group, probably because two were particularly heavy drinkers.
The possible exposure effect was exam ined throu gh a perform ance of multipl e linear regre ssion analyses using cumulative exposure, age, body mass index, and current alcohol con sumption (g' d:' ) as explanatory variables and oyGT, ALP, ALAT, ASAT (transformed log 10) and ALAT :ASAT as the dependent variables. Tabl e 6 presents the results of these analyses. The table does not include curnula-tive exposure as it did not contribute significantly to the model for any of the dependent variables. Alcohol con sumption was strongly associated with the activities of oyGT, ASAT, and ALAT , while body mass index was a significant factor for the )'CiT and ALAT levels and the ALAT:ASAT ratio. Age was associated with the ALP levels. Possible exposure effects were sought with the use of the second exposure index, duration of service in the paint factory, duration of solvent exposure (paint factory + addition al solvent exposure years), or current solvent levels in place of cumulati ve exposure in the model. None of these measures of exposure was significantly associat ed with enzyme activity.
The workers with an abnormality in liver function were investigated for viral hepatitis A and B infection. No worker was positive for HA immun oglobulin M antibody and only one worker was HB s antigen positive. Twenty workers (48% of those tested) had immunological evidence of past exposure to hepatitis B virus (HB c antibody and HB s antibody positive). Tabl e 7 shows that these 20 workers drank more alcohol than their serologically negati ve coworkers and they had a higher cumulative exposure. Table 5. Liver enzyme acti vity for the low and high solvent-exposure groups according to current alcohol co nsumpt ion.
(yGT =gamma glutamyl transferase , ALP =alkaline phosphatase, ALAT =alanine aminotransferase, ASAT =aspartate aminot ransferase , TLV =threshold limit value)   Although the "(GT and ASAT levels were higher in the serologically positive group, the differences were not statistically significant. Adding past exposure to hepatiti s B virus to the mult iple linear regression model failed to show an independent associ ation between past viral expo sure and the liver function variables tested.

Discussion
One objective of this study, the first in South Africa and probabl y the first reported from a developing country, was to determine the prevale nce of abnormal liver function among workers with long service in paint making. Five biochemical variables were tested, and well over half of the worke rs (65%) had at least one variable above the local laboratory refere nce limit. Comp aring our findings with those reported for groups of other paint industry workers is problematic, as factors such as alcohol usage, obesity, exposure intensity, and reference values can vary among the groups, but it is nevertheless interesting that this group had a much greater prevalence of values outside the reference range and far higher mean enzyme activities than has been reported in other studies of paint industry workers (4)(5)(6)14). Two factors should, however, be considered when these findings are interpreted. The first is that our reference values may have been inappropriate for the study population. Althou gh they were in general use by local physicians and pathology laboratories, they were determined with the use of the 95 percentile of values obtained during a survey of predomin antly white, middle-cl ass subjects -a group clearly different from the paint workers. Appropriate reference values for purposes of monitoring specific groups of workers is important , and their selection should form part of a monitoring program ( 15). The second factor is that screening with multiple laboratory tests increases the probability of abnormal results arising by chance. Th is well-known problem in using a battery of tests to monitor workers using hepatotoxins has been discussed by other authors ( 15,16). Although both of these factors are important, neither is likely 242 to account substantially for the high rate of abnormal "(GT and ASAT values, given the strong association shown between enzyme activiti es and alcohol use and the fact that many of the abnormal values were well above the reference level (figure I ).
Whatever the reason , the high prevalence of "abnormal" results found in this study are of more than passing interest. If these results are gene ralizab le to other South Africa n work forces, then strategies using liver enzymes to monitor worke rs will have to be designed to take account of large numbers of false positives (abnormal enzy me levels without chemical hepatitis).
Workers with liver function abnormality were tested for hepatitis B infection. Only one worker (2% of those tested) was positive for HB s antige n. Thi s number is lower than the 10% repor ted for black South African miners (8), but the difference may be due partly to the relatively small number of workers tested and to the urban setting.
The finding that the solvent worke rs apparently drank more alcohol than their co-workers is interesting and may be due to a migration of heavy drinkers to more polluted and less desirabl e jobs.
The major objective of this study was to determine whether long-term solvent exposure in paint manufacturing is associated with liver injury, as measured by standard biochemical indicators. No convi ncing evidence of such an association was found. Although the more heavily ex posed workers genera lly had higher enzyme levels, even when the effects of alcohol were accou nted for by stratification (table 3), regressio n analyses did not confirm an indepe ndent exposure-effect relationship. The relative ly sma ll number of subjects tested may be one explanation.
Other explanations for the absence of evidence of a causal relationship between exposure and disease are possible. First, the current solvent levels were generally not high; thu s the lack of disease may be due to the fact that the workers were not sufficie ntly exposed. Many of the workers did have long service, however, as shown by the mean of 17 years for the high-exposure gro up, and it seems reaso nable to assume higher solvent levels in the past. Further-more, wor kplace air levels are likely to underestim ate ex pos ure, as they do not take into account skin absorption. The nature of the solve nts may have pla yed a role also, since alipha tic chlo rinated hydro carb ons were not co nstitue nts of the paints and these agents arc ge nera lly co nsi dered the most hepatotoxic.
Inaccurate expos ure cla ssifi cation of workers is another important po ssibility. Care was taken to ge t to kn ow the factory pri or to th e start of the study, and non solvent work ers were confidently ide ntifie d. As the y were predomin antl y warehou se, rail yard and e lec trical works hop wo rke rs, office cleaners, and drivers, the lo w so lve nt classifi cation is likely to be accurate. Significant skin absor ption wa s not a factor for the low-exposu re group.
Third, the de ter minations of solvents in ai r were limited in number and collected on rel ativ ely few wo rkdays . Therefore the levels obt ained may not be repr esentative of usual levels. Neverth ele ss, their utility in grouping worker s into expos ure cla sses appears j ustified.
A co mpe lling reaso n to accept the findi ng th at these paint manu facturers did not have ev ide nce of liver injury du e to work is the ag reement with most other st udies of paint industry workers . Increased )GT activity in the most heavily ex posed of 180 paint manu facturers and sprayers wa s attributed to solvent ex pos ure by the auth ors of a study co nducted in Tai 7) all failed to show an association betw een solvents and liver injury as assess ed by routine biochem ical mea sures. Neverth eless, these negative findings should not be interpreted as an absence of an hep atotoxic effec t. Serum bile acids (14) and transferr in ( 18) may be more se nsitive measu res of possibl e hepatotoxicit y than enzy me activ ity. In addition , fatty liver disease, which may be pre sent despite normal li ver function tests, has be en lin ked to solve nt ex pos ure during hou se painting (3) and a variet y of ot he r occ upa tio ns (1 9) . It is possibl e, therefore, that chro nic liver damage wa s occurring that was not reflected by enzy me abnorma lities.
Th e indu cti on of e nzy mes and cy totoxicity due to an exces sive use of alc ohol appeared to be a cause for con cern in thi s grou p of workers. Th e yGT and ASAT activities were strongly associated with repo rted alcoh ol con sumption, and strateg ies to reduce drinking may be the most important meth od of preventing liver disease. It can be co nc luded from th is study that standard tes ts of " live r fun ction " are unlikel y to be a co st-efficient me ans of mon itoring wo rkers exposed to rel ativel y low so lve nt le vels unless the results are link ed to a pro gram to decrease alcohol con sumpti on.
In co ncl usion, this study fo und that current so lvent ex pos ure during paint manufacture was generally we ll belo w the TLV for ind ividual age nts and for Scand J Work Environ Health 1993, vol 19, no 4 so lve nt mi xtures. Thi s was an une xpe cted finding give n the locati on in a developing country but may be ex plained by factors intrinsic to the produ ct ion pro ces s. A high prevalence of rai sed serum activ ity of liver enzymes was detected , but an associ ation bet ween so lve nt ex pos ure and enzy me ac tivity was not shown when fact ors such as alcoho l consumption and bod y mass were co ntro lled for .