Risk of neuropsychiatric disability among painters in the United States.

Scandinavian studies have suggested that working with solvents is associated with chronic neuropsychiatric disease. In the United States the Social Security Administration's records of white male recipients of disability compensation were used in a case-referent study on this topic. The cases were defined as men with any one of several neuropsychiatric diseases, and the referents as men with other disabling conditions. The men were considered exposed if they had worked as a painter prior to disability and unexposed if they had worked as a bricklayer. A job-exposure matrix verified the painters' potential exposure to solvents and the bricklayers' lack of potential exposure. The painters had a significant excess of neuropsychiatric disability [adjusted odds ratio (OR) 1.42, 95% confidence interval (95% CI) 1.04-1.94]. Construction painters had an excess of neuropsychiatric disability [OR 1.47 (95% CI 1.07-2.02)] in contrast to spray painters [OR 0.77 (95% CI 0.38-1.54)]. The limitations of the data are discussed, including potential diagnosis bias and exposure misclassification.

Clinical studies have reported electrophysiological abnormalities among painters (1) and paint-producing workers (2); these abnormalities can persist years after a person has been removed from exposure (3). Moreover, deficits in visual intelligence and memory have been associated with organic solvent exposure (3)(4)(5). These impairments among painters were also found two years after the initial diagnosis (6) or were present when diagnosed five or more years after the termination of the solvent exposure in question (7).
Some recent clinical studies have not found evidence for intellectual or behavioral deficits due to solvent ex-posure (13)(14)(15)(16)(17). Maizlish et al (15) found only mild sensory deficits among workers exposed to solvents from office furniture, automotive parts, and printing plants, and no evidence of a dose-response relationship. In addition deficits on verbal and nonverbal psychometric tests were not apparent for people previously exposed to solvents when preexposure intelligence and postexposure intelligence measures were equalized (13,14,16,17).
Epidemiologic studies have found an excess of chronic neuropsychiatric diseases among disabled men who had been employed in solvent-exposed occupations, including painting (8)(9)(10)(11)(12). Three of these studies were of a case-referent nature (8,10,11), and two were cohort studies that selected cases from union files (9,12). In addition, the cases were defined by the medical diagnosis of presenile dementia, cerebral atrophy, and other neuropsychiatric disorders that were not congenital or due to trauma. Moreover, exposure was quantified by years employed in painting, varnishing and carpetlaying (8), a job-exposure matrix (12), and an exposure history (10). Again, these studies used different definitions of nonexposure, which included skilled workers (8), bricklayers (9), unexposed occupations according to a job-exposure matrix (12), nonexposure to solvents according to questionnaire data (10), or occupations other than painters or carpetlayers in the construction industry (11).
Among these studies only Axelson et al (8) reported a significant odds ratio (OR) for all combined cases of neuropsychiatric diseases among painters. Three studies, however, reported a "dose-response" relationship between exposure and the estimated risk of dis-ability from neuropsychiatric disease (8,10,12). Specifically, the following exposed groups had the highest risk of disabling neuropsychiatric disease: painters, varnishers and carpetlayers who had been employed more than 30 years [age-adjusted OR 2.2] (8), carpenters and cabinet-makers who had worked more than 4000 h lacquering and gluing (OR 2.8) (10), and "always exposed " painters (OR 3.1) (12). In two studies , the exclusion of cases of alcoholism resulted in larger risk estimates (9,12).
Importantly, these epidemiologic studies did not uniformly define their cases. For instance, Axelson et al (8) included the diagnoses of neuroses, in contrast to their exclusion by Rasmussen et al (12). It may be critical to include neuroses cases because Lindstrom et al (13) reported a significant odds ratio of 5.5 for neuroses. In addition, Rasmussen et al (12) included questionable neuropsychiatric conditions in their case definition, such as alcoholic psychoses, cerebrovascular diseases of hypertensive encephalopathia, and ischemic cerebral atherosclerosis. Olsen & Sabroe (10) were also nonspecific in their case definition by using the mention of dementia in medical records to determine the inclusion of cases. These studies were also vague about the relevance of alcoholism in their case definitions. It was not clear whether disability due to alcoholism ought to be included in the case definition because it may be a misdiagnosis of chronic neuropsychiatric disability or it may be that painters are at a higher risk of alcoholism for reasons other than solvent exposure.
The major objective of the present study was to evaluate the hypothesis that painters in the United States (US) have an excess risk of chronic neuropsychiatric disease. Information on persons receiving disability pensions from the US Social Security Administration was used for this purpose. We employed approximately the same coding categories of the International Classification of Diseases (eighth revision) as Axelson et al (8) for operationally defining chronic neuropsychiatric disease. Thus, cases were pension recipients who had been disabled by anyone of several chronic neuropsychiatric conditions. The referents were disabled with conditions other than chronic neuropsychiatric diseases. The exposure under study was employment as a painter (defined by occupation title) prior to the application for disability. We defined nonexposure as employment as a bricklayer (a skilled occupation with an education level similar to that of painters) prior to disability. In his study Mikkelsen (9) used bricklaying work as a definition of nonexposure, and Axelson et al (8) used other "skilled occupations" in his case-referent anal ysis.
We also used information from the National Occupational Hazard Survey to verify the probable presence of solvents in the workplace of painters and the lack of such exposure among bricklayers (18). This was exposure information in addition to the assessments made on the basis of occupation coding categories.

Data sources
Social security disability data. The cases and referents were selected from a longitudinal file, ie, the continuous disability history file, of the US Social Security Administration Disability Insurance Program. Specifically, this file was a 20 070 sample of individuals who qualified for disability benefits for a given year. In this study we used 94 889 white male recipients of such benefits for 1969 to 1973 and 1975 to 1976 who were at least 35 years old and who were employed as bluecollar workers [codes 600-900 of the Dictionary of Occupational Titles (19)] prior to being disabled. Data were not available for 1974 due to a change in the computer processing of the file.
Applicants were allowed benefits on the basis of the following administrative and medical or vocational criteria: (i) an unemployed applicant must have had an insured status in covered employment in 40 quarters (one quarter equaling three months) prior to application for benefits; (ii) the applicant then must have had a medically determinable physical or mental impairment, which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months. The characteristics of the Social Security disability benefit system have been reviewed by Popick (20).
Additional exposure information. We used a jobexposure matrix based on the 1972-1974 National Occupational Hazard Surve y (population different from that of the continuous disability history me) for characterizing solvent exposure among the painters and bricklayers (18). The Survey was a walk-through survey of 4656 establishments, which represented 639 codes of the Standard Industrial Classification. Exposure was defined as a chemical, physical, or biological agent that in one or more forms (vapor, liquid, or solid) was likely to contact the body of the employee. Either the actual chemical or their tradenames were noted. Occupation was coded according to the occupation codes of the 1970 census.
For the purposes of our analysis, the job-exposure matrix pro vided an estimated number of people exposed to specific hazards (numerator) and an estimated number of workers sur veyed in specific occupations and industries (denominator). The ratio of these two values provided an estimate of the percentage of workers in painting and bricklaying observed to be working in the presence of solvents (percentage observed).
We were only able to duplicate the lCD codes of Axelson et al (8) at the three-digit level so that our selection of cases was not as precise as th eirs (8), which used a five-digit ICD code. For instan ce, we could not select vertigo (ICD 780.50) or enc ephalopathy (ICD 781.70) cases, which are not differentiable from other disorders coded as ICD 780 or 781 (symptoms referable to the nervous system). We did no t select ner vousness (ICD 790) cases becau se we could not exclude debilit y (ICD 790.19). We did include the ma jor category of affective disorders (ICD 296 and 297), although Axel son et al (8) was able to exclude organic brain disorders such as manic depre ssion (ICD 296.1) and manic depression psychoses, circular type (ICD 296.3). On the other hand, whereas Axelson et al (8) did not differentiate between neurosis (lCD 300) and personality disorder (ICD 301), we separated these two disea se classifications.
The reference diseases were all other diseases, excluding nonaffective psychoses (ICD 297-298) , sexual deviancy (ICD 302) , psychoses a ssociated with dru g dependence (ICD 304-309), mental retardation (lCD 310-31 5), inflammatory diseases of the ner vous system (lCD 320-324), hereditar y di seases of the nervo us system (ICD 330-333), multiple sclerosis (ICD 348), and peripheral ner vou s system disease s (lCD 350-358). These disord ers were excluded primarily becau se they ma y have had organ ic or trauma causation. Diseases of the peripheral nervous system were not included becau se our outcome o f interest was disorders of the central ner vou s system.
Def inition of expo sure and nonexp osure status. Exposure status was given to tho se recipients of disability pensions who had been em ploy ed as painters prior to application for disability. The painting occupations cho sen included brush painters (DOT 740), spray painters (DOT 741), construction painters (DOT 840), and transportation painters (DOT 845). [See the Dictionary of Occupational Titles (19) for a more detailed description of these occupations.] Nonexpo sure statu s was given to those beneficiaries employed as a br ickla yer (DOT 861) prior to application.
Def inition ofco variates. The case and reference gro ups were stratified into the age gro ups of < 40, 40-44 , 45-49, 50-54, 55-59, 60-64, and > 64 years of age at th e time the disability pension was awarded . Ed ucation was stratified into 1-8, 9-12, 15-16, and > 16 years of education. We also employed the time period as a third covariate because the 1969-1973 and 1975-1976 continuous disability histor y samples could be con sidered separate data files, with data for 1974 missing.
184 Statistical analysis. We used the Mantel-Haen szel method for calculating the summary odds rat ios and the Cochran-Ma ntel-H aenszel statist ic for eva luating the association between case-referent and exposure statu s (22). We could ha ve used unweighted or weighted ob servation s for calculating th e case-re fer ent odds rat ios. Th e weights represent the inverse o f th e samplin g frac tion for the indi vidual sta tes (of the United Stat es), which va ried bet ween 10 and 100 UJO',(23)' . We knew from pr evious wor k that weighted and unweighted methods produced very similar stat istical results; the odds ratios were rou ghly equ ivalent statistica lly, both in terms o f point estimate and significance level. Th erefo re, for simplicity, we chose to do our ana lysis using unweighted observations and thereby co mbined the states and ignored the sample information .
The Breslow-Day test for the homogeneity of odds ratios was used to evaluate the effe cts of age and educat ion (22).
The Mantel-H aenszel summary odds ratios, which were adjusted for age , edu cation, and time period, wer e co nsidered sign ificantly greater th an uni ty at the 5 % alph a level.
Inclusion of the j ob-exposure matrix. Th e jo b-exposure mat rix help ed to verify the relati ve so lvent expo sure for const ruc tion painting, spr ay painting, and bri cklaying occup ations for the relative presen ce (per centage of pot enti al expos ure ob served) of org ani c solvents in th e wor kplaces.
The organ ic solvent hazards used to charac terize the painters and bricklayers were chosen by two cri teria . First , the y had to be organic so lvent haza rds speci fica lly observed (in the National Occup at ion al Hazard Sur vey) in paint ing occupations, and , second, they had to be orga nic solvents that were rep orted in the literatur e (3). These included white spirit or petroleum spirit, tolu ene, xylene, ketone, acetone , bu tanol , and butan one -2.
Th e job-exposure matrix was used to characterize the exposed and unexposed occupation s (ie, constru ction painte rs, spray painters, and bricklayers) by ranking the se occup ations against all other occupation s accord ing to th e percentage o f time ob served working in the presen ce of the specific solvents . Th e ran k for each o f the so lvents ha s been reported in percentiles.

General characteristics
The referents were on the average 2.5 yea rs old er th an the cases, but had fewer years of edu cation . See table I for a summa ry of the age and edu cation covariates for the cases and referents.
In addit ion, those who were empl oyed as painters (N =4291) prior to being disabled had slightly less education than all the other disabled persons who had been  Mean SO Median Mean SO employed as blue-collar workers (N = 86 672) , but the sa me number of yea rs of education as bricklayer s (N = 1641) . Table 2 show s that neuroses, alco ho lism, and other brain diseases accounted for 81 % of th e cases . Presenile dementia was much less co m m o n in our study than in that of Axelson et al (8) , whereas the prominen ce of neuroses and alcoholism di sability was comparable.
As would be expected , disabilitie s due to circulatory disease and musculoskeletal disease were th e mo st common diagnoses among the referents ( There was no apparent interac tio n bet ween age and educat ion with exposure status, as the odds ratios were homogen eou s across both the age and edu cat ion strata. Spe cifi call y, for age the Breslow-Day test for the homogeneity of odds ratios wa s not signi ficant [X' = 3.1 5, degrees of freedom (df) 6, P =0.789]. Nor was it signi ficant for education (X' = 3.08 , df 4, P = 0.545).
We al so calculated the odds ratios , as adjusted for age, education, and time period, for spe cific exposure occupations. Construction painters, which accounted for about 88 % of the exposed painters, had a significant odds ra tio of 1.47. Spray paint ers, who comprised a Code of eighth revis ion of the InternationalClassification of Diseases in parentheses.  about 8 % of the exposed men, formed the only group that had odds ratios that were less than unity (table 6). mentia in the sample of disabled persons in the United States limited our ability to draw conclusions. On the other hand, we obtained a positive association between painting occupations and each of the other specific diagnostic categories, including alcoholism.
We also found that brush painters, construction painters, and transportation equipment painters had a significant excess of disabling neuropsychiatric diseases as compared with spray painters, who had an odds ratio that was less than unity. Previous epidemiologic studies did not report odds ratios for specific painting occupations, albeit spray painters and construction painters, so that our finding is singular. Our result is provocative in reference to the use of protective equipment used by spray painters (25). However, further investigation of whether there is a difference in amount or type of solvent exposure between spray painters and construction painters would need to be done before we can fully understand this finding.
Previous findings in regard to the relationship between alcoholism and solvent exposure are inconsistent. Mikkelsen (9) found that, when alcoholism cases were excluded from their case definition, the odds ratios were increased. On the other hand, Cherry et al (24) obtained an odds ratio of 5.7 for disabled alcoholic cases who had been previously employed as painters. In a separate analysis, we obtained a significant odds ratio, adjusted for age, education, and time period, of 1.51 (95 070 CI1.l3-2.02) for alcoholism, when nonexposure was defined as work in blue-collar occupations other than painting.
A number of mechanisms may be contributing to the association between disabling alcoholism and employment as a painter. First, the diagnosis of alcoholism may in fact be a diagnostic misclassification for other neuropsychiatric diseases, particularly affective and neurotic disorders. On the other hand, nonartifactual mechanisms may be occurring. For instance, painters consume more alcohol and, thus, are more likely to become alcoholic. However, at least one study reported that painters did not consume more alcohol than other types of workers in the construction industry (10). It may be possible that solvent-induced neuropsychiatric diseases are an intermediate step to alcoholism, such that people with other personality disorders are at a higher risk for developing alcoholism or alcoholics are at a higher risk for solvent-induced neuropsychiatric disability. Thus diagnosed alcoholism, which was the disabling condition for some of the cases, may be one of several neuropsychiatric disorders these cases possess. Finally, exposure to solvents may directly induce alcoholism.
An important difference between our study and the Scandinavian investigations was that the diagnosis of presenile dementia was the least common among the disabled population in the United States (2.5 %) compared with, eg, 16 % in the Axelson et al study (8) Table 6. Summary odds ratio (OR) adjusted for age, education, and time period for specific exposed occupations. (95 % CI = 95 % confidence interval)  (8) reported an overall odds ratio of 1.8. However, unlike other epidemiologic studies, we did not obtain an odds ratio greater than unity for presenile dementia. The small number of cases of presenile de-neurop sychiatric disease amon g the disabled US workers, followed by alcoholism, other brain diseases, and affective psychoses. A difference in the distribution of disorde rs between the Scandinavian studies and ours may not reflect a difference in th e prevalence of disablin g diseases in different countries, but rathe r a difference in the diagnostic nom enclature or guidelines in medical or psychiatric diagnosis. Non etheless, neurotic and psychotic disorders have increased irritability, inappropriate affect, compulsive or perseveration behavior , and intellectual deficienc y as symptoms (26). The psychological profiles of the individual s exposed to solvents were also found to include some o f these cognitive, emotional, and behavioral changes, part icularl y tho se associated with memory and emot ional reactivity leg, 0rbrek et al (16)]. Moreover, the categories for the "clinical evaluation of patients" who had been exposed to solvents includ es "sustained personality or mood changes" (type 2A) and " impairment of intellectual function" (type 2B) (27). These two categories encompass symptoms for the diseases used in our case definition for operationally defining chronic neuropsychiatric disease.
There ar e several possible limitations when one generalizes from a disabled population to the entire popul at ion at risk. One is that the disabled population is different from the general population. For insta nce, people who appl y for disab ility may have other risk factors that increa se their vulnera bility to neuropsychiatric disord ers. We would have to assume that those factors are distributed evenly across occupational groups. A second limitation is that the prob abilit y for appli cation or award for disability is influenced by factors other than medical criteria or work histor y eligibility, part icularly the occupation prior to being disa bled . A recen t study , however, showed that occupation was not related to an awareness of government disab ility benefits or the decision to apply for disability (28). Th ird, our odds ratio, which is really a " disability odds ratio," would be affected by the proportion of other diseases. For example, bricklayers may be mor e disabled by musculoskeletal disorders than painters, neuropsychiatric disab ility being therefore proportionately reduced among bricklayers.
Another disadvantage of our case-referent study is that we did not have access to medical record s to verify the diagnoses, and we relied on a decentralized data collection system that was meant primarily for administrat ive evaluation of the Social Security disab ility benefit progr am . We agree with Baker & Fine (29) that using a case definition , such as ours, for chronic neuropsychiatric disease may not adequately capture neurop sychiatri c deficits that result from long-term solvent exposure. Ho wever, no one has precisely elucidat ed the exact nature of the effect of solvents on the central nervous system. Moreover , without kno wing the type o f solvents actuall y involved in the exposure, it would probably have been inappropriate to restrict our case definition . Given our lack of contro l of the data collection and coding, the optimum appro ach was designing a study that was most likely to select disability cases that would have resulted from solvent exposure.
We also had to rely on job titles for defining expo sure . Oth er epidemiologic studies , such as those of Olsen & Sabroe (10) and Shalat et al (30) were able to collect a work history to establish previous solvent exposure. In a discussion of the issue of miscIassification , Rasmu ssen et al (12) concluded that basing exposure classificat ion on job titles probably biases the estimated odds ratio towards unit y. Regardless, our job-exposure matrix indirectly supported the relat ive potential solvent exposure of exposed painters and lack thereof for unexposed bricklayers.
The result s of our anal ysis pro vide additional evidence that there may be a relationship between occupational exposure to or ganic solvents and chronic disabling neuropsychiatric diseases. However, our analysis does not stand alone, given the inherent biases and data quality issues associated with an existing administrative data source. The strength of our findings using disability dat a of th e Social Security Administration is the internal consistenc y of the results and the agreement with the European studies. Nonetheless, the interpretat ion of these results may not only appl y to the risk for painters per se for neurological damage from long-term solvent exposure, but also for an y worker who is exposed to solvents in the wor kplace.