Indoor personal factors related to the sick building syndrome.

J. Indoor air quality and personal factors related to the sick building syndrome. Scand J Work Environ Health 1990;16:121-8. The "sick building syndrome" in volves symptomssuch as eye, skin and upper airway irritation, headache, and fatigue. A multifactorial study was performed among personnel in consecutive casesof sickbuildingsto investigaterelationships between suchsymptoms, exposure to environmental factors,andpersonal factors.Thetotalindoorhydrocar bon concentration wassignificantlyrelated to symptoms.Other indoor exposuressuch as room tempera ture, air humidity, and formaldehyde or carbon dioxideconcentration did not correlate with the symp toms. Personal factors suchas reported hyperreactivity and sickleavedueto airway diseases werestrong ly related to the sick building syndrome. Other factors associatedwith the sick building syndrome were smoking, psychosocial factors, and experience of static electricity at work. Neither atopy, age, sex, nor outdoor exposurescorrelated significantlywith the number of symptoms. It was concludedthat the sick building syndromeis of multifactorial origin and related to both indoor hydrocarbon exposureand in dividual factors.

In the past, occupational medicine dealt mainly with severe diseases related to long-term or high-level exposure to specific chemicals in the work environment. However, during the last decade, concern has increased about the more subtle health effects of low-level exposure to complex mixtures of chemicals.
One such new area is the possible health effects related to poor indoor air quality in certain nonindustrial buildings, often described as "sick" buildings. As defined by a working group of the World Health Organization, the "sick building syndrome" involves various nonspecific symptoms such as eye, skin and upper airway irritation, headache, and fatigue (1). Although several investigators have reported a higher prevalence of symptoms among persons working in certain "sick buildings" than among workers in other buildings (2)(3)(4)(5)(6)(7)(8)(9) and many different chemicals have been quantified in indoor air (9, [10][11][12][13][14][15], only a few investigations have studied the influence of specific indoor exposures on the prevalence of the sick building syndrome (6). In addition, only few investigations have taken into consideration the influence of personal (6,16) and outdoor climatological (17,18) factors on the prevalence of the syndrome.
Earlier studies have indicated that room temperature, formaldehyde, carbon dioxide concentration, and building age should be taken into consideration in in- vestigations of the indoor climate. A high room temperature may enhance the prevalence of the sick building syndrome (19,20). A relation between indoor formaldehyde concentration and the sick building syndrome has earlier been demonstrated (21,22). Measurements of carbon dioxide can be used to indicate the degree of fresh air supply in relation to the number of individuals in a building, and a concentration of 0.08 vol% of carbon dioxide (800 ppm) has been suggested as a control limit value (10). A negative association between building age and the prevalence of symptoms has also been demonstrated (6). This study was performed in order to study the influence of indoor and outdoor exposures and personal factors on the occurrence of symptoms among personnel in presumedly "sick" buildings.

Methods
Since March 1984, the Department of Occupational Medicine at the University Hospital in Uppsala has offered a specific combination of a self-administered questionnaire and exposure measurements as a means of investigating workplaces in sick buildings. Consequently, occupational health care centers in the tricounty region (the counties of Gavleborg, Kopparberg and Uppsala) have spontaneously referred presumed cases of sick building syndrome to our department. This tricounty region contains 84 health care units serving 140000 employees in the public sector and 120 000 employees in the private sector.
The study was based on all consecutive cases of sick buildings (N = 11) with more than 10 employees re-ferred to the Department of Occupational Medicine during a three-year period (March 1984-April 1987. For statistical reasons, workplaces with less than 10 employees were excluded from the material. These 11 sick buildings were selected by the occupational health care centers to be investigated by us. In addition, information on those sick buildings known by the health care centers but not referred to us (N = 44) were obtained from a sick-building survey performed later, during 1988 (23). This information was used to compare the building characteristics of the referred sick buildings with those of sick buildings not referred to us to determine whether the studied sick buildings were representative for all known sick buildings in the region. The workplace characteristics of the referred and nonreferred sick buildings are presented in table 1. The Air condition ing (air cooling or humidification) 9 7 Wall-to-wall carpets in the workplaces 9 9 a Workplaces with at least ten employees known to be " sic k" by the occupational health cen ters in the tricounty region before 1987. b Workplaces with 10-19 employees. C Built after the so -called " energy cristo" (1974). d Primary schools or day care centers. majority of the buildings were new buildings in the public sector equipped with mechanical ventilation. Air cooling, air hum idification , and wall-to-wall carpets were rare.

A ssessment of symptoms and personal factors
The six-month prevalence of symptoms among the persons working in the sick buildings were recorded by mean s of a self-administered questionnaire. The questionnaire contained queries on individual factors such as smoking hab its, atopy, hyperreactivity, sick leave, work stress, work sa tisfaction, and climate of cooperation at work. The nonresponse rate was less than 5 010 in all of the sick buildings. Th e mean age of the employees was 41 years, and the mean number of reported da ys of sick leave due to respiratory illness was 5 d durin g the last six months. Other demographic and exposure data for the responding sick building personnel are presented in table 2. In order to avoid any influence of the results of the exposure measurements on the questionnaire responses, the questionnaire investigation was always completed before the exposures were measured . In all cases, the questionnaire investigations were performed during the heating season (October-April).
The questionnaire included "yes" I"no" questions on 16 different symptoms covering the last six months. No information on the severeness or the duration of the symptoms was gathered . Since the sick building syndrome comprises different symptoms, a single effect variable was created through the calculation of a symptom score ranging from 0 to 16 for each individual. In addition, the prevalence of the different symptoms was also calculated . Work stress, work satisfaction , and climate of cooperation during the last six months were mea sured by analogue rating scales (24). A psycho social index ranging from 0 to 100 % was calculated through the addition of the perceived degree of work dissa tisfaction , work stress, and lack of work cooperation . The symptom que stions and the analogue rating scales appear in the appendix. A ssessm ent of exposure Initially, information about the workplaces was obtained from the occupational health centers to which the workplaces were connected. This information included building age, type of work, type of employment, number of employees, type of floor covering, and type of ventilation, air humidification and air conditioning. The same type of information was gathered for both the referred cases and those cases of sick buildings known to the health care centers but not referred to us. In addition, chemical and climatological mea surements were performed in the referred sick buildings. Room temperature and relative humidity were recorded with an Assman psychrometer. Indoor formaldehyde concentration was measured with glass fiber filters im-a Range of ari thmeti c mean prevalence in each sick bUild ing. Table 3. Symptom prevalence during the last si x months in the total material of sick bUilding personnel (N = 261).

Results
Few individuals were a nnoyed by environmental tob acco smo ke, worked at a video display terminal , or had wall-to-wall carpets in their work pla ces. Th e effe cts of th ese rare expo sures were not analyzed fu rther.
Sympt om s from the eyes, nose , and throat, and also tiredness and head ache, were common in the referred sick buildings. Nau sea and dermal symptoms were less common (table 3).
The average indoor concentrati on for th e sum of compounds differed with one or two ord ers of magnitude bet ween the di ffer ent sick buildings (tabl e 4). Th e great est variat ion was found fo r terp enes and unident ified low-b oiling hydrocarbons. The largest 13 (27) were used in the calculation of the varia tion of different indoo r expo sures between workplaces. The multifactorial an alysis of the data was perform ed by mult iple linear regression, using the sympt om score as the dep end ent var iable. Logarithmic values of th e indoor volat ile hydrocarbon concent rations and building age were used in the regression mod el for a rea sonable symmetry of distribution of these variables.
Two different strat egies for ente ring var iable s into the regression model were used to ensure that the conclusion s were not influenced by the variable selectio n procedure. In the first strategy, con ventional stepwise regression (P <0.05) was performed , with no fo rcedin var iables. In the second, the anal ysis was performed in four steps. As a first step, all potential confounders were for ced into the model. Second, all significant confounders (P < 0.05) were maintained in the model. Third, the expo sure variables were forced into the model one by one. Fourth, all the significant expo sure variables were included in the model. Finally, all nonsignificant expo sur e variables were excluded. pregnated with 2,4-dinitrophenylhydra zine (25), the air sampling rate being 0.25 II min for 2 h. The filters were a na lyzed by liquid chro ma tography . The indoor carbon d ioxide concentra tion was measured by co lo rimetri c det ector tubes (Drager 0.01 % /a) .
The vo latile organi c compounds in the indoor and outdoor air were sampled on charcoa l sorbent tubes (SKC 226-0 1), th e air sampling rat e being I 11min for 2 h. In order to avoid moi sture interference with the ab sorption of the hydrocarbons, the measurements were performed dur ing October-April , when the relative hum idity of the indoor air in th is part of Sweden is low (10-40 %).
Th e charcoal tub es were kept at -20°C until an alysis. The charcoa l tubes were desorbed with I ml of ca rbo n disulfide prior to the analysis, which was per formed within a week of the sampling day with a gas chroma tograph (Hewlett Packard model 5880) equipped with a flame ionization detector and packed glas s co lumns . The injector temperature was 150°C, the detector temperature was 200°C , and th e flow of the carrier gas nitrogen was 30 mllmin . Fift een common hydr ocarbons were identified and qu antified through th e use of an .extem al standard technique. These hydro carb ons were iden tifi ed thr ough compa rison of the retention tim es of the samp les with the retentio n times of the standards on two di ffe rent columns. When the low-boiling un calibr at ed peak s (C3-CI 2) were quantified, the response factor of n-decan e was used . When the high-b oilin g unknown hydro carbons ( >CI2) were quantified , the response factor of a mixture of high boiling hydrocarbons (dodecyl ben zenes) was used. The to tal air concent ration of th e identified and unidentified hydrocarbons was calc ulated.
T he number of measurement s performed in eac h wor kplace depended on the size of the build ing. In larg er workplaces, the number of mea surements was greater to enable the calculation of a mo re representati ve average exposure for the workplace. For air temperature, air humidity, and carbon dioxide, two to eight sho rt-term measurements were taken . For the charcoal-sampled indoor hydrocarbon measurements, two to five individual 2-h average samples were taken. For formaldehyde one or two 2-h measurements were taken . For the outdoor hydrocarbon measur ement s, one individual 2-h average sample was ta ken . Arithmetic averag e exposures were calculated for each sick building workplace .
Data on aver age temperature, precipitation, and average atmos pheric pressure during the month when th e qu estionnaire was distributed to the different sick buildings were obtained from the Swedish Met eorological and Hydrolo gical Institute in Norrkoping.

Statistical methods
The differen ces in the proportion s were calculat ed by th e two-t ailed chi-square test or by Fisher 's exact test for 2x2 co ntingency table s (26). Geometric mean s and a Sum of toluene, m-xylene, o-xylene, p-xylene , and ethyl benzene . b Sum of n-octane, n-decane, n-decane, and n-undecane. C Sum of alpha-pinene, delta-carene, and limonene. d Sum of n-butanot and iso-butanol, e Sum of un identified hydrocarbon with a retent ion time below n-dodecane.
, Sum of un identified hydrocarbons w ith a retent ion time equal to or above n-dodecane.  The multiple regression analysis revealed that five individual factors were significantly related to the symptom score. These factors were nonspecific hyperreactivity (P<O.OOI), sick leave due to airway illness (P<O.OOI), psychosocial dissatisfaction (P<O.OI), own tobacco smoking (P<O.OI), and reported exposure to static electricity (P < 0.01). Among the measured exposures, only total indoor hydrocarbon concentration correlated significantly with the number of symptoms (P<O.OI). Room temperature, air humidity, and carbon dioxide or formaldehyde concentration did not correlate significantly with the symptoms. Neither did building age or any outdoor climatologicalor exposure variables correlate significantly with the symptom score. The two different strategies used for selecting variables into the analysis resulted in the same final regression model (table 7).
A further multiple regression analysis of the relation between different groups of compounds (table 4) and the number of symptoms, including the five significant confounders in the regression model, was also performed . In this analysis, only unidentified lowboiling hydrocarbons correlated significantly with the symptoms (P < 0.05). No identified groups of compounds correlated significantly with the number of symptoms.
The relation between the sick building syndrome and the total hydrocarbon concentration could also be demonstrated in a regression analysis of the average number of symptoms in the different sick building groups as a function of the logarithmic value of the  group of compounds was made up of unidentified lowboiling hydrocarbons. The chromatograms revealed that most of these unidentified compounds were highly volatile compounds with retention times shorter than the retention time of benzene. The indoor and outdoor  The relation between the indoor hydrocarbon concentration and different symptoms was also studied in comparisons of the geometric mean exposure among persons with and persons without a particular symptom (table 8). For 10 out of 16 symptoms, the mean hydrocarbon exposure was significantly higher among persons with the symptom as compared with persons without the symptom (P < 0.05). a The following 12 variables were not signifi cant (P>0.05) predi ctors of symptom score: age, sex, atopy, building age, indoor room temperature, indoor air humidity, indoor carbon dio xide concentration , outdoor volatile hydrocarbon concentration , outdoor temperature, air pressu re, and precipitation. b Difference in number of symptoms between individuals an' swering " yes" and those answering " no " On this question. c Two -tailed P <0.001. d Number of symptoms per yearly week of sick leave due to airway illness. e Difference in number of symptoms among individuals with 100 % On the psychosocial dissatisfaction index as compared to those with 0 % .

Discussion
Selection bias can occur as a result of both a low response rate and incorrect study design. The participation rate was high in our study, and therefore the probability of selection bias due to loss of individuals from the studied sick building groups was minimized . The II sick building groups were referred to us by the occupational health centers in the region. This procedure might have introduced selection bias since we had no control of the criteria for referring the sick building workplace. However, the prevalence of building or workplace characteristics was similar in the studied sick buildings and sick buildings not referred to our clinic. Thus selection bias coupled to the referring process was less probable. places, the cross-sectional design of the study would underestimate the true effect of indoor air pollution. Another problem regarding the validity is the possible response bias due to awareness of the exposure. This study was performed in presumedly sick buildings where at least some of the individuals believed that their symptoms were related to their workplace. However, the measurements were performed after the questionnaire study was completed, and thus the results of the measurements could not have influenced the questionnaire responses. In addition, only specific ex-posures and specific individual factors correlated with the number of symptoms. Finally, in multivariate modeling of epidemiologic data, the strategy of entering the variables into the model may influence the result (28). However, in this study, the two different strategies resulted in the same regression model. In conclusion, it was not probable that the results of our study were due to response or selection bias or to the selection of a particular regression model.
In our study, sick leave due to respiratory illness and symptoms of hyperreactivity in situations outside the sick buildings were the most strongly correlated with the number of symptoms. None of these potential confounders has earlier been studied in other investigations on the sick building syndrome. However, some of the symptoms included in the sick building syndrome are similar to symptoms occurring in common airway infections. In addition, common airway infections may induce temporary bronchial hyperreactivity among normal subjects (29). Very probably, such temporary hyperreactivity also makes the individual more sensitive to irritants at work or in the home environment.
Tobacco smoking correlated with the sick building syndrome. An association between smoking and such symptoms has earlier been demonstrated (6). A relationship between the degree of psychosocial dissatisfaction and the sick building syndrome was also found. This result is in agreement with the findings of two other studies in which both climate of cooperation at work (30) and work stress (16) were found to correlate with symptoms associated with the sick building syndrome.
A relation between self-reported exposure to static electricity and symptoms was also demonstrated. Concerning the possible relation between static electricity and the sick building syndrome, there is conflicting information in the literature. In a recent Swedish study, individuals who often experienced electrostatic shocks in hospital environments without carpeting reported an enhanced prevalence of fatigue (30). In a British study, no such association could be demonstrated (2). There are other studies in which the degree of electrostatic charge was measured among office workers. One study revealed a significant association between the degree of charge and the prevalence of symptoms associated with the sick building syndrome (20). In another study, no such association could be demonstrated (3 I).
Although chemical stimulation of the trigeminal or olfactory nerves has been suggested as an explanation of the sick building syndrome (32), there are few epidemiologic studies that test this hypothesis. Formaldehyde concentrations one order of magnitude higher than the levelsmeasured in the present study have been shown to correlate with symptoms related to the sick building syndrome (21,22). In the "Town Hall Study." a positive association was found between total 126 hydrocarbon concentration and sick building syndrome in that the highest prevalence of symptoms was found for the building with the highest concentration of total hydrocarbons (6). Besides this study, no epidemiologic studies on the effect of volatile hydrocarbons other than formaldehyde on the sick building syndrome are found in the literature. In our study, the total indoor hydrocarbon concentration, ranging from 0.05 to 1.38 rng/rn ', was the exposure measure best correlated with the number of symptoms. Since this study demonstrated an effect of the hydrocarbon concentration on many different symptoms, it is justified to use a symptom score as a simple measure of the sick building syndrome. In a Danish exposure chamber study, the effect of 2-h exposures to a mixture of 20 common indoor hydrocarbons was studied (33). In that investigation, it was shown that, at a total hydrocarbon concentration of 5 mg/m', the hydrocarbon mixture could induce irritation of the eyes, nose, and throat. The effect was acute and showed no signs of adaptation. In addition, a digit-span performance test also showed decreased scores during exposure. Other human experimental data on the irritative effects of hydrocarbon mixtures similar to indoor air at lower exposure levels than 5 mg/m' are not available.
The sick buildings in our study were generally wellventilated above the existing ventilation standard of 5-6 I of outdoor air per second and person (10), and no significant association between carbon dioxide concentration and symptoms could be demonstrated. Although carbon dioxide is a good measure of the fresh air supply per individual, it is a reliable indicator of air quality only if humans or human activities are the dominating source of indoor air pollution. As demonstrated by Fanger (34), other nonhuman factors such as the buildings or the ventilation systems could be dominating sources of air poIIution in a modern indoor environment. This possibility could explain why no association between symptoms and carbon dioxide concentration could be demonstrated in our study.
We conclude that the sick building syndrome is of multifactorial origin, depending on both personal factors and environmental factors such as the indoor hydrocarbon concentration. Although some types of hydrocarbons might contribute to the irritative symptoms more than other compounds, the sum of hydrocarbons is a simple and convenient measure, which was also found to be better related to the sick building syndrome than the concentration of single groups of compounds. This finding is weIIin agreement with the hypothesis that the syndrome is best understood as a concurrent effect of sensory irritation from the complex mixture of hydrocarbons occurring in the modern indoor environment. However, since certain personal factors are also correlated with the prevalence of the syndrome, these factors must be carefuIIy controIIed in epidemiologic investigations on the sick building syndrome.