INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Firefighters are routinely exposed to a range of hazardous products of combustion. These products include acrolein, carbon monoxide, hydrogen chloride, hydrogen cyanide, oxides of nitrogen, sulfur dioxide, and particulates, as well as a range of other aldehydes and carcinogens (1). Firefighters generally wear personal protective equipment including a self-contained breathing apparatus (SCBA) during periods of smoke exposure. Positive-pressure SCBAs offer a high degree of respiratory protection, often exceeding the designated protection factor of 10,000, indicating a 10,000-fold reduction in concentration of contaminants inside the respirator facepiece as compared with the ambient environment (5, 6). However, respiratory protection may not be worn during all phases of firefighting, and smoke exposure during periods when it is not worn may result in adverse health effects (7).

Studies of firefighters have demonstrated both acute and chronic changes in their pulmonary function. Acute changes, including reductions in spirometric parameters and increased airway reactivity, have been described after firefighting, although they have generally been transient and have occurred when respiratory protection was not used during periods of smoke exposure (7). In longitudinal studies of firefighters in the 1960s and 1970s, Peters and coworkers and Sparrow and colleagues demonstrated accelerated declines in spirometric parameters in Boston firefighters, most likely as a result of exposure to products of combustion, (11, 12), and a similar trend was found in London firemen (13). Other surveillance studies, including more recent evaluations, have yielded mixed results, with a decreased rate of decline in spirometric parameters attributed to both selection factors and improved use of respiratory protection (14).

In Seattle firefighters, measurements of the single-breath diffusing capacity of carbon monoxide (DL_CO) have been used to complement routine measures of ventilatory capacity. In contrast to spirometry, DL_CO testing provides a measure of gas exchange. No prior longitudinal study of DL_CO measurements in firefighters has been reported in the literature. In this report we characterize our experience with longitudinal measurements of both spirometric parameters and DL_CO in Seattle firefighters over an 8-yr follow-up period.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study utilized a retrospective, repeat cross-sectional design, assessing ventilatory capacity in firefighters as measured annually, over an 8-yr period, through FVC, FEV₁, and DL_CO. Seattle firefighters have participated since 1988 in a medical surveillance program with the Seattle/King County Health Department and the University of Washington Occupational and Environmental Medicine Program. The surveillance program was designed to fulfill asbestos and hazardous materials surveillance and respiratory clearance requirements. Longitudinal study of ventilatory capacity and DL_CO were included to provide early indications of adverse trends. Components of the program included entry and annual questionnaires, annual physical examinations, spirometry, and DL_CO measurements, and, when indicated, chest radiographs and laboratory analyses. DL_CO monitoring was initiated in 1989, during the second year of the surveillance program. Participation was voluntary, and individual firefighters chose which components of the program they would complete, although efforts were made to encourage full participation in all components.

Questionnaires

A self-administered entry questionnaire submitted at baseline evaluation provided extensive information on previous firefighting activities and medical history. At annual follow-up examinations, another self-administered questionnaire was submitted, including information on interim firefighting activity. Firefighters were asked to estimate the number of actual fires (incidences in which smoke or fire was present) in which they had been involved in the previous year, with selections of 0 to 10, 11 to 20, 21 to 30, 31 to 40, 41 to 50, and more than 50. Questions about use of respiratory protection for different phases of fire response were included. The firefighters were also asked to estimate the average duration of their use of a respirator during the different phases of firefighting (extinguishment, entry/ventilation, rescue, support/standby, and overhaul) in the previous year, with selections of none, 1 to 25%, 26 to 50%, 51 to 75%, and 76% or more.

Pulmonary Function Testing

Spirometry was done with a Puritan-Bennett spirometer during 1988 and with a CyberMedic PFT system (Spinnaker TL Total Lung Analyzer System; CyberMedic Inc., Louisville, CO) for all subsequent years. Spirometric measurement data were collected at baseline and then annually, and measurements were made in accordance with testing criteria of the American Thoracic Society (ATS) (18) and corrected for temperature, pressure, and water saturation. A minimum of three acceptable tracings were recorded, with the subject in a sitting position. If the best FVC and FEV₁ measurements were within 5% of one another for each variable, the subject proceeded to DL_CO testing. Otherwise testing continued until the best measurements were within 5% or a maximum of six efforts had been made. The maximum acceptable spirometric measurement was reported and used for data analysis.

DL_CO testing was initiated in 1989 with the same Cybermedic PFT system used for spirometry. During all years of the study, testing followed ATS standards (19). Data from two to four maneuvers were collected, with a minimum of 5 min between repeat measurements. Measurements were considered adequate if inspiration reached  90% of FVC within 2 s and the subject's breath was held for 10 s. Testing was continued until two test results were obtained that were within 10% of one another, or to a maximum of four tests. The acceptable maneuver yielding the highest test values was used for data analysis. All DL_CO test results were evaluated by a physician. The PFT system used for DL_CO measurement was calibrated twice daily, and underwent frequent maintenance and servicing, including replacement of the CO cell.

Data Analysis

For regression analyses, FEV₁ and FVC were expressed in liters, and DL_CO measurements were expressed in units of ml/min/mm Hg, which were not corrected for alveolar volume or hematocrit. Longitudinal spirometric data and DL_CO scores were modeled with a multivariate, linear, random-effects model (20). Random-effects models are most appropriate for longitudinal data gathered under circumstances in which each subject is not evaluated every year or at equally spaced intervals, and take into account overall effects, which are considered fixed effects, and intrasubject effects that are considered random effects and in which repeated measures may be correlated. Estimation was made by the restricted maximum likelihood method, using SAS PROC MIXED software (SAS Institute, Inc., Cary, NC). A model-building approach was used to determine the factors that best predicted spirometric and DL_CO function at entry, and the annual rate of decline in this function. As a first step, demographic variables (age, gender, race), physiologic measures (height and FVC), and prior exposure measures (pack-years of smoking) known to be associated with spirometric or DL_CO function were forced into the model. After controlling for these factors, we tested for potential predictors of longitudinal rate of decline and included them in the model if they reached significance at the 0.05 level. In this model, the effects of selected variables on longitudinal change are evaluated by the interaction of the selected variable with a time term. For example, the effect of average annual number of fires fought (AVEFIRE) on longitudinal changes in DL_CO is measured by the AVEFIRE · time variable.

For each firefighter, average annual number of fires fought was estimated from the questionnaire data, by using the midrange value of the ordinal exposure categories (i.e., five fires for the 0 to 10 selection, 15 fires for the 11 to 20 fire selection, etc.). Lack of use of respiratory protection was also calculated from annual questionnaire data, by using the midrange value of the ordinal exposure categories (i.e., 13% for the 1 to 25% selection, 38% for the 26 to 50% selection, etc.), expressed as percentages and not as actual units of time. These values were then averaged over the years in which spirometic and DL_CO measurements were obtained. Smoking as used in the regression models was defined as any cigarette use during the study period.

DL_CO Instrument Analysis

During 1995, repeat measurements were made with a second Cybermedic Spinnaker TL unit lent to us by the manufacturer, and with the same computer and software as originally used, in order to help verify the accuracy of the initial unit. During a 1-mo period, firefighters were tested on both machines, with a 10-min interval between tests. Dual testing was performed as time permitted on medical staff members and two or three firefighters per day, and the order of testing was reversed after testing of each firefighter. Comparison of the regular and borrowed DL_CO units was done with paired t tests. During 1996, all firefighters with a maximum DL_CO measurement of less than 70% predicted were offered retesting at a later time on a DL_CO unit (PF/DX 1085D; Medgraphics, St. Paul, MN) in a hospital-based pulmonary function laboratory. Firefighters with repeat DL_CO measurements that were below 70% predicted were offered full exercise testing.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

From 1989 to 1996, 1,104 firefighters underwent pulmonary function testing, of whom 1,076 had at least one annual set of DL_CO tests. In order to evaluate longitudinal changes in DL_CO, we restricted our analysis to 812 firefighters completing at least two sets of annual DL_CO measurements, for a total of 3,348 annual sets of DL_CO measurements. The characteristics of this cohort by examination year are presented in Table 1. Exact information on the total number of Seattle Fire Department firefighters on active duty for each year was not available, but it was estimated to be approximately 1,000. 

Of the 812 firefighters on whom the analysis was done, the distribution of number of annual DL_CO evaluations was as follows: two analyses in 187 (23%) firefighters, three in 186 (23%), four in 130 (16%), five in 99 (12%), six in 107 (13%), seven in 65 (8%), and eight in 38 (5%). For the 3,348 annual sets of DL_CO measurements, information was available on the second highest DL_CO measurement in 736 sets. In 682 (93%) of these 736 sets, the second highest measurement was within 10% of the highest DL_CO measurement. Annual evaluations of firefighter DL_CO were not excluded from analysis on the basis of these criteria. The intrasubject variability of DL_CO measurements was 9.77 ml/min/mm Hg. Intersubject variability was 44.74 ml/min/mm Hg, and declined with age.

The average annual spirometric measurement values (FEV₁ and FVC) and DL_CO values, expressed as % predicted (Crapo) (21) for the 812 firefighters completing at least two annual sets of DL_CO measurements, are presented in Figure 1. The average annual measurements are based on a slightly different set of firefighters each year because of variable participation and changes in the group due to retirement and new hires. The average % predicted DL_CO declined by 7.1%, from 94.4 ± 13.4% (95% confidence interval [CI] = 93.4% to 95.5%) in 1988 (n = 568 firefighters) to 87.3 ± 11.3% (95% CI: 86.2% to 88.3%) in 1996 (n = 381 firefighters), whereas the mean values of FVC and FEV₁ were statistically unchanged. In the larger cohort of 1,076 firefighters completing at least one set of annual DL_CO measurements, and in an "inception" cohort of 464 firefighters who had their initial DL_CO measurement in 1989 and at least one subsequent annual DL_CO measurement, decline in DL_CO without significant variation in spirometric parameters was consistent with the decline depicted in Figure 1.

View larger version (90K): [in this window] [in a new window]   Figure 1.   Pulmonary function by year for firefighters with two or more annual DLCO measurements (n = 812).

Regression models in which FEV₁ and FVC were respectively used as dependent variables (Table 2) did not show statistically significant longitudinal changes. Pack-years of smoking at the start of surveillance, smoking during the study, and change in number of female participants did not significantly contribute to the models.

Parameters and values for the regression model in which DL_CO was used as the dependent variable are shown in Table 3. This model used all 812 firefighters who had at least two annual sets of DL_CO measurements. Age, height, race, gender, FVC, and pack-years of smoking at entry were significantly associated with initial DL_CO. Age, female participation, FVC, smoking, and number of fires fought were all significantly associated with longitudinal decline in DL_CO measurements. Notably, the year of measurement (time) was strongly inversely associated with DL_CO. For a typical male firefighter (age: 36.7 yr; FVC: 5.64 L; average annual number of fires fought: 17.8), there was a substantial decrease in predicted diffusing capacity, of 1.02 ml/min/mm Hg per year (see also Figure 1). Interestingly, although the average number of fires fought was associated with a significantly higher baseline DL_CO of 0.05 ml/ min/mm Hg per fire/yr, this parameter was associated with a modest accelerated decline of 0.006 ml/min/mm Hg per fire/ yr in the follow-up period. Addition of the variables representing lack of use of respiratory protection during the overhaul phase (p = 0.934), annual FEV₁ measurements (p = 0.731), and minority firefighter participation (p = 0.669) did not contribute significantly to longitudinal change in DL_CO, and did not affect the contribution of number of fires fought.

Firefighters who had smoked at any time during the surveillance period (n = 142) started at a lower DL_CO than firefighters who had not smoked (n = 670), as shown by the contribution of the smoking variable. However, these same smoking firefighters actually showed a reduced rate of decline in DL_CO over time. This trend is graphically depicted in Figure 2, which compares decline in DL_CO in the smokers and nonsmokers. When regression analysis was performed on firefighters stratified by smoking status, the contribution of number of fires fought to longitudinal change in DL_CO was greater in nonsmokers (estimate = 0.009, p = 0.007) than in smokers (estimate = 0.004, p = 0.510). Similarly, when the cohort was stratified by age, the contribution of number of fires fought was greater in firefighters  35 yr of age (n = 361, estimate = 0.010, p = 0.0315) than in older firefighters (n = 451, estimate = 0.003, p = 0.408).

View larger version (108K): [in this window] [in a new window]   Figure 2.   DLCO by examination year for firefighters with two or more annual DLCO measurements (n = 812) stratified by smoking status.

Since average DL_CO measurements varied by year, models including year of evaluation (1989 to 1996) were constructed. Although year of evaluation was significantly associated with initially measured DL_CO values, its inclusion in the model did not affect the association between number of fires fought and decline in DL_CO (Table 4). When models were run with data from all 1,076 firefighters (one or more annual sets of DL_CO measurements), the effect of average annual number of fires fought was essentially unchanged (estimate = 0.006, p = 0.043). Analysis restricted to the inception cohort (n = 464) revealed a similar association between fires fought and decline in DL_CO (estimate = 0.005), but with a smaller sample size this association lost statistical significance (p = 0.150).

Comparison of results with the regular and borrowed DL_CO units with paired t tests (n = 22 test/retest pairs) did not reveal a statistically significant difference. The measurements with the regular unit averaged 31.4 ± 11.4 ml/min/mm Hg, and those with the borrowed unit averaged 32.2 ± 11.5 ml/min/mm Hg. For the most recent (1996) DL_CO measurements, 18 (5%) of the firefighters had a maximum measurement below 70% of the Crapo predicted value. Repeat testing of 11 of these 18 firefighters at the hospital-based pulmonary function laboratory revealed an average increase in maximum DL_CO of 0.22 ml/min/mm Hg (95% CI: 1.61 to +2.04). Exercise testing was done on four of the 11 firefighters; in three cases the test results were considered normal, and one test indicated impaired gas exchange.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The finding of stable ventilatory capacity over time in Seattle firefighters is reassuring, suggesting that the accelerated ventilatory loss previously described in firefighter populations (11- 13) has not occurred in Seattle. However, the observed changes in diffusing capacity raise a number of questions and concerns. In addition to the expected changes in DL_CO associated with age, gender, ventilatory capacity, and smoking, we observed two temporal phenomena with regard to diffusing capacity, the first being a dramatic decline in DL_CO of 1.02 ml/ min/mm Hg/yr, associated with year of measurement, and the second being a smaller but statistically significant longitudinal decline of 0.006 ml/min/mm Hg associated with reported number of fires fought per year.

The large magnitude of the changes in DL_CO associated with year of measurement, with a substantial decrease in DL_CO in the latter years of the surveillance program, despite stability in the reported number of fires fought and use of personal protective equipment, suggests that diffusing capacity may lack sufficient reliability as a screening tool for respiratory function. However, in models adjusting for year of examination and other potential confounders, we continued to observe a modest accelerated loss in DL_CO associated with reported numbers of fires fought. This finding raises the alternative, and perhaps complementary, possibility that DL_CO provides increased sensitivity in assessing early preclinical effects of smoke inhalation, that cannot be not detected through changes in ventilatory capacity.

DL_CO measurements can be affected by a number of factors, including lung volume, airflow obstruction, ventilation/ perfusion mismatch, and loss of alveolar capillary membrane, as occurs in emphysema (22, 23). Age, height, gender, and race are included in predictive equations such as those developed by Crapo (21), and were associated with measured DL_CO values in our regression model. Smoking is also known to affect DL_CO measurements (24), and was associated with diminished diffusing capacity in our study population.

Despite the clear temporal association of baseline DL_CO with examination year, repeated confirmatory testing of the DL_CO measurement instrument in later years of the surveillance period did not show any evidence of machine error. Specifically, paired testing of the DL_CO instrument against a second, similar unit did not show a significant difference in average measurements, although there was wide variation in individual measurements. Comparison of the readings provided by the DL_CO unit used for the study with measurements made at a later time in a hospital-based pulmonary function laboratory did show generally higher, though not statistically significantly different, measurements with the hospital-based unit. This comparison was based on firefighters selected on the basis of low DL_CO measurements in annual screening examinations, and is therefore prone to regression toward the mean.

Although there is no clear indication of machine error or technical variation in diffusing capacity in our study, the explanation for such large temporal differences that we found in DL_CO, after adjusting for demographic variation, remains unclear. The prominent change of 1.02 ml/min/mm Hg associated with year of measurement, as compared with the relatively modest changes associated with reported numbers of fires fought, in the absence of any change in ventilatory capacity, suggests that variation unrelated to respiratory status may have occurred.

Although we are not aware of other longitudinal studies of DL_CO in firefighters, similar studies in other working populations have also suggested temporal changes in diffusing capacity independent of exposure. In a 5-yr follow-up study of steelworkers, a 15% decline in predicted DL_CO values was observed; however, a more dramatic 20% decline occurred in an unexposed control group of similar age, weight, height, and smoking status (25). In a 5-yr prospective study of toluene diisocyanate (TDI) fabrication workers, an overall annual decline in DL_CO of 0.72 ml/min/mm Hg was observed; however, exposure was inversely associated with rate of decline (26). Although longitudinal studies in the general population have generally indicated stable diffusing capacity over a 5- to 10-yr period in nonsmokers (27, 28), it is difficult to draw any consistent conclusion of trends in DL_CO among working populations.

In the setting of potential variation in DL_CO unrelated to respiratory status, the observed decline in diffusing capacity associated with firefighting activity, predicted by our model to consist of an annual loss of 0.12 ml/min/mm Hg for the average firefighter reporting 20 fires fought per year, must be interpreted with caution. In addition to the large decline associated with year of measurement, three other observations limit our ability to attribute decline in DL_CO to smoke inhalation. First, if such exposure was associated with a decline in DL_CO, one would expect reported failure to use respiratory protection to be associated with such a decline, which we did not observe. Second, the average annual number of fires fought was actually associated with a higher baseline diffusing capacity, in contrast to the accelerated decline in the follow-up period. Third, one might have anticipated a positive interaction between cigarette smoking and fire-smoke exposure, with additive or synergistic biologic effects on respiratory function; however, only nonsmokers in our study experienced a significant longitudinal decline in DL_CO in relation to fires fought. Potential explanations for the latter two findings include: (1) a decline in DL_CO may have already occurred in smokers and older firefighters prior to the surveillance period, with less subsequent sensitivity to smoke exposure; and (2) younger and proportionally more nonsmoking firefighters may have worked at fire stations with more fire responses. However, we were not able to test these hypotheses.

A limitation of this study, as in most respiratory surveillance programs, is that the actual magnitude of smoke exposure could not be directly or quantitatively determined. Occupational smoke exposure in firefighters would be expected to have two potential sources: (1) improperly fitted or poorly functioning SCBA; and (2) inconsistent use of respiratory protection during firefighting activities. Information about SCBA function during active firefighting is limited. Although studies have shown that positive-pressure SCBAs provide a high degree of protection (5, 6), our data clearly indicate that such personal protective equipment is used less often in support, standby, and overhaul/clean-up phases of firefighting.

Despite these limitations, our findings raise the possibility that deleterious subclinical effects, detectable through subtle changes in diffusing capacity, result from firefighting activities. Although the dramatic effects of severe smoke inhalation injury have been well described, including bronchospasm, tracheobronchitis, and pulmonary edema (29), subclinical effects of chronic low-level smoke inhalation have not been well characterized. Increased serum levels of Clara cell protein (30), as well as increased pulmonary clearance of ^99mTc-diethenetriamine pentacetate (^99mTc-DTPA) (31) in firefighters following smoke exposure, have provided evidence of acute cellular injury, inflammatory changes, and increased permeability of the bronchoalveolar-capillary barrier. In the case of the study in which the last-named finding was made, changes in ^99mTc-DTPA seen in firefighter instructors occurred without detectable changes in spirometry (31). The changes in DL_CO associated with firefighting in our study suggest that chronic repeated episodes of inflammation caused by smoke exposure could result in loss of the alveolar-capillary interface, manifested by changes in diffusing capacity. DL_CO may also provide a more sensitive method for detecting these early changes than do conventional assessments of ventilatory capacity.

In conclusion, our study of Seattle firefighters demonstrated a marked decline in average measured DL_CO over an 8-yr follow-up period, without concomitant changes in ventilatory capacity as determined by FVC and FEV₁. These findings raise important questions about the efficacy of DL_CO in the screening of workers exposed to potential respiratory toxicants. The prominent temporal variation in DL_CO appears to substantially limit the reliability, and hence the clinical utility, of this respiratory parameter. Nonetheless, after controlling for year of measurement and other potential confounders such as demographic factors and smoking, we found that the average annual number of fires fought was associated with a modest but statistically significantly accelerated decline in DL_CO.

The alternative explanation, that DL_CO may provide a more sensitive method for detecting early subclinical effects of smoke exposure, merits further exploration. Our findings indicate the need for further studies utilizing sensitive markers of early respiratory injury, as well as improved measures of smoke exposure, to confirm the clinical significance of our results. Pending such investigations, our findings underscore the need for assiduous use of respiratory protection during all phases of firefighting.