Formaldehyde exposure and health status in households.

This report describes a case study concerned with acute and subacute health effects of formaldehyde in the indoor air, which is based on a large group of control houses and houses retroinsulated 4 to 5 years earlier with urea formaldehyde foam insulation (UFFI). Both groups underwent an environmental and health assessment on two occasions separated by an interval of 12 months, during which about one-half of the UFFI group performed remedial work on their houses. The results show that in the first survey of the study population, before remedial work, there was a moderate excess of many adverse health status indicators among the UFFI subset relative to the controls. This was associated with the presence of direct exposure-response relationships between formaldehyde levels in the UFFI houses and the prevalence of a number of symptoms. No comparable relationships were seen among the controls. At the second survey, performed following the removal of the UFFI, there was an appreciable reduction in the excess of most adverse health status indicators among the UFFI subjects. This improvement in health status among the UFFI removal subset was not associated with any significant diminution of formaldehyde exposures, although the previously observed exposure-response relationships had vanished. These observations imply that the findings obtained in the preremedial stage of the study cannot be explained by formaldehyde exposure alone.


Introduction
Formaldehyde has long been known to have acute surface irritant effects on humans, as indicated mainly by observations made in industry. Based on whether the exposed individual is highly sensitive or average in sensitivity, ocular irritation is observed at between 0.05 ppm and 1.0 ppm, nasal and throat effects at 0.1 to 1.0 ppm, and cough at 5 to 30 ppm. Less information is available relevant to the nonindustrial indoor setting (1-7).
The following report describes a case study concerned with acute and subacute health effects of formaldehyde in the indoor air, which is based on a large group ofcontrol houses and houses retroinsulated 4 to 5 years earlier with urea formaldehyde foam insulation (UFFI). The field work was done during the period 1983 to 1985, and the main publications appeared in 1988 (6)(7)(8).
The objective ofthe study was to compare the health status between the control and UFFI households and, ifdifferences were found, to examine the role of formaldehyde exposure. This study was conducted at a time when the Government ofCanada was providing subsidies for UFFI remedial work. Thus a longitudinal component was included in the study to allow the efficacy of the remedial measures to be assessed.
The data are presented here in a manner focusing on an unexpected finding encountered in this study, rather than on the adverse effects ofUFFI and the efficacy ofremedial work, which have been the focus of the earlier publications (6)(7)(8). This finding is of interest in this workshop because it exemplifies the complex nature ofthe adverse human responses which seem to be attributable to the indoor environment.

Methods
The methodology of this study has been described in detail previously (6). Our study was based on about 200 control and 600 UFFI households, each of which was examined on two occasions separated by an interval of 12 months. Between the two surveys, about one-halfofthe UFFI houses underwent remedial work. The UFFI households that were enrolled consisted ofthree subsets based on their intention either to have the UFFI removed, to do other remedial work such as sealing the UFFI-insulated wall cavities, or to remain unchanged. Houses were entered into the study in sets of four consisting of one control and one of each ofthe three UFFI subsets. These sets were matched by location, to be within about 1 mile ofeach other, and by time ofentry, to undergo their initial assessment within the same 4to 6-week period.
The houses were assessed using a questionnaire that collected information about a variety of features, including construction details, ventilation, and pollution sources. Formaldehyde levels were monitored with a pump and impingers centrally, in all bedrooms, and outside the houses, on 2 sequential days, during the same period as the occupants were examined. Assays were performed by the chromatropic acid method, using rigorous quality control procedures, which included frequent blank, doped, and side-by-side field controls; regular calibration curves; split samples tested on an on-going basis in an independent laboratory; and blinding ofour own technical staffas to the identity of the test samples (6). CO2 levels, temperature, and humidity were measured centrally and outside on the same occasions as formaldehyde. The temperature and humidity results are not important in the findings to be described and will not be mentioned further.
The occupants were assessed using a questionnaire that collected demographic details as well as other information, including the presence or absence ofa variety of symptoms, need for medical services, and history ofactive and passive smoking. Also, all subjects over the age of 16 underwent a series oftests including pulmonary spirometry, nasal airway resistance, sense of smell, nasal surface inflammatory cell and epithelial cytology, and patch test for formaldehyde allergy. The results ofthe objective tests are not relevant to the information being described here and will not be given below. However, these tests did assist in establishing the credibility of the data base by demonstrating a number of expected relationships, such as the influence of age, gender, and smoking on pulmonary function and the association of respiratory symptoms with changes in the tests of pulmonary function and nasal resistance (8).

Results and Discussion
The first survey in this study was based on the 2446 occupants of 802 houses, all located within about 60 miles ofcentral Toronto. About 90% participated in the second survey (8). The findings ofthis study were mainly explained by differences between the controls and the UFFI subset that underwent removal oftheir insulation. Accordingly, the results to be described will be simplified by dealing only with the 605 controls and 699 UFFI removal subjects, all of whom participated in both the first and second surveys.
The general characteristics of the control and UFFI removal subjects were quite similar as seen at the first survey, before remedial work ( Table 1). The two groups did not differ significantly in racial origin, height, gender distribution, age, broad occupational categories, and in hours spent in the house per week. There was a nonsignificant trend for more of the UFFI removal subset to have been lifetime nonsmokers. Among those who were current or ex-smokers, the extent of smoking expressed as pack years was similar in both groups.
The prevalence and means ofadverse health status indicators at the first survey, before any remedial work, was moderately higher for a large number ofvariables among the UFFI removal subset relative to the controls ( Table 2). The risk ratio ranged from a low of 1.  thirst and throat discomfort. For the purpose of simplification only selected indicators will appear in subsequent tables, but this will preserve a representative picture of the findings. At the second survey, after remedial work had been performed by the UFFI removal subset, many ofthe statistically significant differences seen at the first survey had vanished, and those that remained were borderline (Table 3). There was a marked decrease in the risk ratio of some variables, such as throat discomfort and increased thirst; a moderate decrease in others, such as eye irritation and tiring easily; and no discernible change for a few, such as cough and sputum.
The average time interval between remedial work and survey 2 was 8 to 10 months, but the range was broad and in 20% the interval was 3 months or less. We accordingly examined the relationship between change in symptom prevalence from survey 1 and survey 2 and the numberofdays between remedial workbeing done and survey 2 ( Table 4). This demonstrated a weak relationship foronly the two variables nasal problem and headache, suggesting that the changes in health status that followed remedial work were generally independentoftimebeforethe second survey. The house variables demonstrated that the UFFI removal homes were significantly older than the controls (p = 0.0001), which was expected since the controls would have been weighted with newer houses insulated with other materials at the time of being built and would not need to be retroinsulated (Table 5). 'This slope indicates that there was a 0.3 percentage point decrease in the prevalence of nasal problem for each 10 days after remedial work was done. Both groups of houses were similar in socioeconomic appearance. The day of the year of being assessed was similar within each group between the two surveys, but was about 6 weeks earlier in the year fbr the UFFI removal subset relative to the controls (p = 0.0001). This difference in time ofyear between the two groups was not associated with any appreciable differences in the ambient temperature at the time ofbeing examined. There was an equally small decrease in the occurrence of smoking between surveys within both groups of houses. The indoor formaldehyde levels were about 20% higher in the UFFI removal subset than in the controls (p = 0.0001) and did not change between the two surveys ( Table 5). The outdoor for-maldehyde levels were similarly low in both groups. The indoor and outdoor CO2 levels were in the expected ranges and did not differ significantly either within or between the groups. Our inability to detect a decrease in formaldehyde in the removal subset at the second survey may have been related to the recent installation ofpressed wood items, as well as some possible diminution in fresh air infiltration, which is suggested by the trend for the CO2 level to increase in the UFFI removal subset at survey 2.

FORMALDEHYDE EXPOSURE AND HEALTH STATUS IN HOUSEHOLDS
The frequency distribution of the indoor formaldehyde levels of the UFFI removal subset at the first survey was shifted rightward to a small degree relative to the controls, as would be expected from the difference between their means. This did not change appreciably at the second survey. There were five houses among the UFFI removal subset having mean indoor formaldehyde levels in excess of0.112 ppm, whereas all in the control group were at this level or lower.
Exposure-response relationships were examined between the mean household formaldehyde level in the first survey of each house and the status ofthe various health indicators for each occupant. The UFFI removal subset at survey 1 showed a number ofdirect, significantly positive exposure-response relationships ( Table 6). For example, for each 0.01 ppm of formaldehyde exposure, there was a 2.47 percentage point increase in the occurrence of nasal problems, a 1.15 percentage point increase in thirst, and a 1.13 point increase in cough. No comparable significantly positive relationships were seen among the controls.
Because the UFFI removal subset had a number of formaldehyde exposures that were higher than any in the control group, the same analysis was repeated after deleting those subjects with the higher formaldehyde exposures (Table 7). This showed a persistence but weakening of the exposure-response relationships when the 29 people were deleted whose exposure was greater than 0.112 ppm, and even when those 74 were dropped who were exposed to higher than 0.08 ppm, which overlapped the upper end ofthe range ofexposure in the controls. When the exposure-response relationships were reexamined based on the second survey data for formaldehyde and health status indicators, no significant responses were observed ( Table 8).
The findings ofthis study therefore indicate that health status indicators have improved and exposure-response relationships with formaldehyde have disappeared at the second survey, in the absence ofany appreciable decrease in formaldehyde exposure relative to the first survey. Since these observations cannot be explained by a problem with our formaldehyde measurements, we  aThis analysis was performed in the same manner described for Table 6, using survey 2 data for the same population.
are forced to conclude thatthe preremedial findings demonstrated at the first survey were not due to formaldehyde alone. These observations could be explained alternatively on the basis that formaldehyde interacted with some other chemical or psychological factors associated with UFFI, or was a proxy for them, and that they were removed by remedial work. If due to an interaction or proxy effect between formaldehyde and other UFFIrelated chemicals, one might have expected to see a reduction in formaldehyde levels if removal of the UFFI was associated with a reduced exposure to the hypothetical other chemicals. Nevertheless, our results indicated a continuing source ofthe small excess of formaldehyde even after the removal of the UFFI, and thus this could continue to be associated with other related chemicals. Accordingly, we are unable to clearly distinguish whether this proposed interaction or proxy effect is between formaldehyde and other chemicals or psychological factors.

Conclusions
The results described show that in the first survey ofthe study population, before remedial work, there was a moderate excess of many adverse health status indicators among the UFFI subset relative to the controls. This was associated with the presence of direct exposure-response relationships between formaldehyde levels in the UFFI houses and the prevalence of a number of symptoms. No comparable exposure-response relationships were seen among the controls. At the second survey, performed following the removal of the UFFI, there was an appreciable reduction in the excess of most adverse health status indicators among the UFFI subjects relative to the controls. This improvement in health status among the UFFI removal subset was not associated with any significant diminution of formaldehyde exposures, although the previously observed exposure-response relationships had vanished. These observations are indicative of the complexities that may arise in assessing and understanding health risks in individual case studies related to chemicals in indoor air.