Incidence of Lung Cancer and Melanoma Among Respondents to the Wade Fire
Provided to the Philadelphia Inquirer by
James D. Lewis, MD, MSCE
Brian L. Strom, MD, MPH
Revised October 4, 1999
How the experts reached their conclusions
The Philadelphia Inquirer asked Drs. Lewis and Strom to assess whether the incidence rate of melanoma and lung cancer in personnel involved in the Wade fire was greater than that of the general population. Of importance, however, the authors of this report were not involved in any of the data collection and have made no effort to validate the data provided by the Inquirer staff.
The following definitions are used in this report:
- Incidence rate – the number of new cases of cancer occurring per person per year
- Standardized incidence ratio (SIR) – the ratio of the incidence rate observed in the study population compared to the incidence rate in the general population. A SIR of 100 suggests that the incidence rate of cancer in the study population is neither increased nor decreased compared to the expected incidence rate for the general population. When the SIR is greater than 100, this suggests an increased incidence of cancer in the study population.
- 95% confidence interval – this is an approximate range of values one would obtain for the estimate of the SIR if the experiment were able to be repeated many times, each with the same sample size. Stated more simply, if the Wade fire was studied 100 times, in 95 of these studies, the calculated SIR for each of these experiments would fall within the range of the 95% confidence interval. When the 95% confidence interval excludes 100, the likelihood is less than 5% that the difference observed between the incidence rate of cancer in the Wade fire personnel and the expected rate for the general population is due to chance. This is the conventional criterion for statistical significance.
Follow-up duration was calculated from the date of the fire to the date of last known follow-up or the date of diagnosis of the cancer of interest, whichever was shorter. Total years of follow-up were stratified by race, sex, and 5-year age groups. Published age-, sex-, and race-specific incidence rates for melanoma and lung cancer were obtained from the SEER Program of the National Cancer Institute. This program collects cancer incidence data from population-based registries that include approximately 14% of the United States population. These incidence rates were multiplied by the number of patient-years of each stratum in the cohort to calculate an expected number of cases of each cancer. The total number of cases expected for the whole cohort was calculated by summing the expected number of cases from each sub-group of the cohort. This was then compared to the observed number of cases to calculate the SIR (SIR = 100 * observed number of cases / expected number of cases). Exact 95% confidence intervals for the SIR were calculated assuming a Poisson distribution for the incidence of the cancers. As explained above, a SIR greater than 100 with a 95% confidence interval that does not include 100 is consistent with an increased risk of cancer in the cohort.
Stratified analyses were then performed calculating the SIR for subjects closest to the fire compared to those farther away, for subjects with a duration of exposure greater than or equal to the median duration (8 hours), and for subjects with direct contact to the chemicals.
Sensitivity analyses were performed to account for the possible influence of missing data. In these analyses, we assumed that the missing subjects had comparable age, sex, and racial distribution as the cohort and that the follow-up duration was comparable. Using these assumptions, we calculated the number of subjects who would need to have been exposed to the fire and did not develop melanoma but for whom no information was available, such that the observed number of cases of melanoma would be equal to the expected number of cases. A similar analysis was performed to determine the number of subjects with missing data required such that the 95% confidence interval would include 100, thus implying that the observed increased incidence rate of melanoma might be due to a chance observation.
Data were provided from the Philadelphia Inquirer on 199 subjects. For 18 of these there was inadequate information on age, date of last follow-up, and the incidence of cancer to include in the analysis. Of the remaining 181 subjects, 179 were male. The role of the personnel at the fire included firefighters (n=127), other support roles (n=53), and unknown (n=1). The racial distribution was 156 Caucasian, 18 blacks, and 3 other (treated as black for all further analyses since SEER data are stratified as white, black, and all races combined). The median age at the time of the fire was 30.3 years (range 14.6 to 70.1).
There were 4 cases of melanoma. The expected number of cases was 0.65. Thus, the SIR was 612 (95% confidence intervals 167 - 1568). The incidence of melanoma was significantly elevated among the 122 subjects closest to the fire (observed 3 cases, expected 0.41, SIR 730, 95% confidence interval 150 - 2132). Among the 59 subjects further away from the fire, there was 1 case of melanoma compared to an expected 0.24 cases (SIR 413, 95% confidence interval 10 – 2298). There were no cases of melanoma among the 17 subjects the furthest away from the fire.
The median duration of exposure was 8 hours. The SIR for melanoma among the 115 subjects with 8 or more hours of exposure (including 4 subjects of unknown duration) was 487 (95% confidence interval 59 – 1758). [The results did not change when those 4 subjects were omitted.] The SIR for subjects with less than 8 hours of exposure (SIR= 825, 95% confidence intervals 100 – 2978). All 4 cases of melanoma occurred in the 109 subjects with direct contact with the chemicals (SIR=1078, 95% confidence interval 293 – 2760).
Based on the 181 subjects with adequate data to include in the analysis, the average number of expected cases of melanoma per subject was 0.0036 (0.6532/181). Assuming that missing subjects had similar age, sex, and racial distribution as the identified subjects, there would need to be more than 900 additional exposed subjects in whom there were no cases of melanoma for the observed number of cases to equal the expected number. Similarly, there would need to be an additional 120 subjects not identified and in whom there were no additional cases of melanoma for the 95% confidence interval to include 100.
In summary, there appears to be a significantly increased incidence of melanoma among those personnel responding to the Wade fire. The magnitude of this risk may be greater among those subjects closest to the fire and with direct contact to the chemicals, although the data are not definitive on these points. There did not appear to be an association between increased duration of exposure and risk of melanoma. Among subjects classified as having direct contact to the chemicals, the risk of melanoma was more than 10 times that expected in the general population.
There were 10 cases of lung cancer among the cohort. The expected number of cases was 2.09. Thus, the SIR was 478 (95% confidence interval 230 – 880). There were 8 cases among the 121 subjects closest to the fire (SIR 612, 95% confidence interval 264 – 1205) and 2 cases among subjects farther away (SIR 255, 95% confidence interval 31 – 922). The SIR for subjects with 8 or more hours of exposure was 564 (95% confidence interval 243 – 1111). For subjects with shorter exposure time the SIR was 297 (95% confidence interval 36 – 1073). The SIR in subjects with direct contact was 673 (95% confidence interval 290 – 1325). In subjects without direct contact the SIR was 267 (95% confidence interval 32 – 964).
In summary, it appears that the respondents to the Wade fire had a higher incidence of lung cancer than would be expected from the general public. This was particularly evident in subjects closest to the fire, with 8 or more hours of exposure, and with direct contact to the chemicals.
These data suggest that the personnel responding to the Wade fire experienced an increased incidence of both lung cancer and melanoma compared to what would be expected in the general population. The magnitude of this increased risk is such that the observed difference is unlikely to be due to chance (less than 5% likelihood that this was a chance finding). However, these data do not prove causality between the exposure and the subsequent cancer risk.
The data on lung cancer are difficult to interpret due to the absence of complete data on smoking, a known risk factor for lung cancer. While most studies do not suggest an increased incidence of lung cancer among firefighters, in at least one previous study, firefighters have been identified as having an increased risk of lung. In a meta-analysis, Howe and Burch estimate the relative risk of death from lung cancer in firefighters to be 0.92 using cohort studies and 1.08 using proportional mortality studies (Howe and Burch, 1990).
The elevated risk of melanoma is striking. The occurrence of 2 of the 4 melanomas on the feet is somewhat unusual (see figure 1). Additionally, our sensitivity analyses suggest that even though some data are missing, this is unlikely to explain the observed increased incidence of this cancer. However, it is important to note that previous studies have suggested an increased risk of melanoma among firefighters. These studies are summarized by Landrigan et al. (Landrigan, 1994). Most of these studies suggest that firefighters have an increased risk of death from melanoma 1.2 to 2.9 times greater than the general population. In the meta-analysis of Howe and Burch, the pooled estimate of risk of death from melanoma was 1.73 (95% confidence intervals 1.03 – 2.74).
Importantly, for both lung cancer and melanoma, there appears to be a potential dose response relationship. In both cases, those subjects with the greatest exposure, estimated by duration of exposure (only of lung cancer), proximity to the fire, and direct contact with the chemicals appear to have a greater risk of developing cancer. It should be noted, however, that these are only qualitative differences in risk and that formal statistical comparisons between the groups were not performed. One should also recognize that those subjects with the greatest exposure in the Wade fire might also have had greater additional exposures to carcinogens over their lifetime. As such, one must be cautious in implicating the Wade fire exposure as specifically causing these cancers.
Note: Figure1 shows the anatomic site distribution of melanoma by sex in 731 males and females. This figure is copied from Langley RGB, Barnhill RL, Mihm MC, Fitzpatrick TB, Sober AJ. Neoplasms: Cutaneous Melanoma. In Dermatology in General Medicine. Freedberg IM, Eisen AZ, Wolff K, Austen KF, Goldsmith LA, Katz SI, Fitzpatrick TB, eds. McGraw-Hill, New York 1999.
Ries LAG, Kosary CL, Hankey BF, Miller BA, Edwards BK (eds.) SEER Cancer Statistics Review, 1973-1995, National Cancer Institute. Bethesda, MD, 1998
Landrigan PJ, Golden AL, Markowitz SB. Occupational cancer in New York City firefighters. Mount Sinai. 1994
Howe GR, Burch JD. Fire fighters and risk of cancer: an assessment and overview of the epidemiological evidence. Am J Epidemiol 1990;132:1039-1050.