Objective: A prospective study of civilian (nonfirefighter) smoke inhalation patients was carried out to test the hypotheses that: 1) absorption of carbon monoxide and hydrogen cyanide from smoke can be predicted by clinical examination and historical data; and, more specifically 2) a history of exposure to burning synthetic polymers is an important predictor of systemic cyanide levels.
Methods: The study was conducted over a three-year period at six urban hospitals. Patients with or without burns who were exposed to smoke within five hours of hospital arrival were sampled for carboxyhemoglobin, whole blood cyanide, urine cotinine and urine creatinine. Controls consisted of a smaller group of smoking status-matched, nonsmoke-exposed burn patients.
Analysis: Historical information was obtained on SMOKING status, FIRETYPE (structural vs other), MATERIAL burned (natural vs synthetic) and LAGTIME (from exposure to sampling). A smoke inhalation SCORE (0-10) was assigned to each case, based on physical examination findings and changes on chest X ray, and carboxyhemoglobin and cyanide levels were entered into various multivariate linear regression models.
Results: A total of 40 cases and 9 controls were recruited, ranging in age from 15 to 92 years. Thirty-four cases were discharged alive and six expired in-hospital. Observed carboxyhemoglobin levels ranged from 1.2% to 41.6% in cases (mean 8.6%), and from 0.5 to 7.3% in controls (mean 2.9%). Observed cyanide levels ranged from nondetectable (< 0.05 micrograms/mL) to 2.79 micrograms/mL in cases (mean 0.25 micrograms/mL), and from nondetectable to 0.11 micrograms/mL in controls (mean 0.03 micrograms/mL). Among cases, linear regression models explained up to 35% of the observed variance in carboxyhemoglobin levels (p < 0.001) and up to 48% of the variance in cyanide levels (p = 0.0001).
Conclusions: SCORE was the strongest predictor of both carboxyhemoglobin and cyanide levels; LAGTIME also explained significant variance for [log-transformed] carboxyhemoglobin. Historical factors, such as FIRETYPE, MATERIAL, and SMOKING status, did not explain significant variance in most of the statistical models employed.