JOEM • Volume 48, Number 11, November 2006 1189

Cancer Risk Among Firefighters: A Review and

Meta-analysis of 32 Studies

D
Grace K. LeMasters, PhD uring the course of their work, fire-
Ash M. Genaidy, PhD fighters are exposed to harmful sub-
stances at the fire scene as well as at
Paul Succop, PhD the firehouse. At the fire scene, fire-
James Deddens, PhD fighters are potentially exposed to var-
Tarek Sobeih, MD, PhD ious mixtures of particulates, gases,
mists, fumes of an organic and/or in-
Heriberto Barriera-Viruet, PhD organic nature, and the resultant pyrol-
Kari Dunning, PhD ysis products.1,2 Specific potential
James Lockey, MD, MS exposures include metals such as lead,
antimony, cadmium, uranium, chemi-
Objective: The objective of this study was to review 32 studies on firefighters cal substances, including acrolein,
and to quantitatively and qualitatively determine the cancer risk using a benzene, methylene chloride, polyaro-
meta-analysis. Methods: A comprehensive search of computerized databases and matic hydrocarbons, perchlorethylene,
bibliographies from identified articles was performed. Three criteria used to assess toluene, trichloroethylene, trichloro-
the probable, possible, or unlikely risk for 21 cancers included pattern of phenol, xylene, formaldehydes, miner-
meta-relative risks, study type, and heterogeneity testing. Results: The findings als such as asbestos, crystalline, and
indicated that firefighters had a probable cancer risk for multiple myeloma with a noncrystalline silica, silicates, and var-
summary risk estimate (SRE) of 1.53 and 95% confidence interval (CI) of ious gases that may have acute, toxic
1.21–1.94, non-Hodgkin lymphoma (SRE ⫽ 1.51, 95% CI ⫽ 1.31–1.73), and
prostate (SRE ⫽ 1.28; 95% CI ⫽ 1.15–1.43). Testicular cancer was upgraded effects.1,2 In some situations, respira-
to probable because it had the highest summary risk estimate (SRE ⫽ 2.02; 95% tory protection equipment may be in-
CI ⫽ 1.30 –3.13). Eight additional cancers were listed as having a “possible” adequate or not felt to be needed
association with firefighting. Conclusions: Our results confirm previous findings resulting in unrecognized exposure.3
of an elevated metarelative risk for multiple myeloma among firefighters. In At the firehouse where firefighters
addition, a probable association with non-Hodgkin lymphoma, prostate, and spend long hours, exposures may oc-
testicular cancer was demonstrated. (J Occup Environ Med. 2006;48: cur to complex mixtures that comprise
1189 –1202) diesel exhaust, particularly if trucks are
run in closed houses without adequate
outside venting. In light of the World
From Epidemiology and Biostatistics, University of Cincinnati College of Medicine (Dr LeMasters, Trade Center disaster, concerns have
Dr Succop), Cincinnati, Ohio; Industrial and Manufacturing Engineering and Epidemiology and
reemerged and heightened related to
Biostatistics, University of Cincinnati College of Engineering and College of Medicine (Dr Genaidy),
Cincinnati, Ohio; the Department of Mathematical Sciences, University of Cincinnati College of Arts building debris particle exposures from
& Sciences (Dr Deddens), Cincinnati, Ohio; the Department of Industrial Medicine and Occupational pulverized cement and glass, fiberglass,
Diseases, Cairo University Faculty of Medicine (Dr Sobeih), Cairo, Egypt; the Department of Industrial asbestos, silica, heavy metals, soot,
Engineering, Interamerican University of Puerto Rico (Dr Barriera-Viruet), Bayamon, Puerto Rico; the and/or organic products of combustion.3
Department of Rehabilitation Sciences, University of Cincinnati Medical Center (Dr Dunning),
Cincinnati, Ohio; and Occupational and Environmental Medicine and Pulmonary Medicine, University
To date, only one meta-analysis
of Cincinnati College of Medicine (Dr Lockey), Cincinnati, Ohio. conducted by Howe and Burch in
This study was supported in part by a grant from the Ohio Bureau of Workers Compensation. 1990 examined the extent of cancer
Address correspondence to: Grace K. LeMasters, PhD, Department of Environmental Health, risk among firefighters in 11 mortal-
University of Cincinnati College of Medicine, Cincinnati, OH 45267-0056; E-mail: grace.lemasters@
ity studies.4 They reported that there
uc.edu.
Expert testimony was provided by Drs Grace LeMasters and James E. Lockey in pending workers was an increased association with the
compensation claim relating to non-Hodgkin’s lymphoma. occurrence of brain tumors, malig-
Copyright © 2006 by American College of Occupational and Environmental Medicine nant melanoma, and multiple my-
DOI: 10.1097/01.jom.0000246229.68697.90 eloma with the evidence in favor of
1190 Cancer Risk Among Firefighters • LeMasters et al

causality somewhat greater for brain data were extracted from each article (ie, SMR, PMR, RR, SIR, and OR).
tumors and multiple myeloma. Since by one reviewer and was verified by These criteria were used in a forward
then, there have been numerous mor- another. Discrepancies identified by approach as illustrated in Figure 1 in
tality and incidence studies. Hence, the second reviewer were resolved in which at each stage, a new criterion
the purpose of this study was two- a consensus meeting. was applied, and the probability of
fold. The first purpose was to update Likelihood of Cancer Risk. Statis- cancer risk was reassessed. The likeli-
the Howe and Burch findings by tically significant increases in cancer hood for cancer risk was given an
reviewing the methodologic charac- risks among firefighters were evalu- assignment of “probable,” “possible,”
teristics of these studies and deter- ated as the likelihood for cancer risk or “not likely” patterned after the In-
mining the probability of cancer by given a three-criteria assessment. The ternational Agency for Research on
assessing the weight of evidence, includ- three criteria included “pattern of Cancer (IARC) risk assessment of hu-
ing the calculated metarisk estimates. meta-relative risk association,” “study man carcinogenicity in terms of weight
The second purpose was to describe a type,” and “consistency” among stud- of the evidence.5
methodology for use in a meta-analysis ies. These criteria were particularly The “pattern of meta-relative risk
when diverse investigations are being important given the different method- associations” was the first criterion and
evaluated and summarized. ologies used for evaluating cancer risk included a two-step evaluation. For the

Materials and Methods

Search Strategy and
Inclusion Criteria
Standardized mortality ratio (SMR),
proportional mortality ratio (PMR),
relative risk (RR), standardized inci-
dence ratio (SIR), and case– control/
mortality odds ratio (OR) studies re-
lated to firefighters and cancer risk
were evaluated. For publication selec-
tion, at least 1 year in service as fire-
fighters was required except for those
studies basing employment on death
certificates. Publications were retrieved
by a search of computerized databases,
including Medline (1966 –December
2003), Health and Safety Science Ab-
stracts (since 1980–December 2003),
Cancerlit (1963–December 2003),
NIOSHTIC and NIOSHTIC2 (up to De-
cember 2003), BIOSIS Previews (1980–
December 2003), and PubMed (up to
December 2003) using the following key
words: firefighters, fire fighters, cancer.
In addition to the computerized search,
bibliographies in identified papers were
reviewed for additional studies.
The search was restricted to reports
published in English; abstracts and re-
views were not included. Studies were
excluded without basic data (eg, con-
fidence intervals) that are necessary in
the derivation of the meta-analysis
risk estimate. If there was more than
one article with the same or overlap-
ping population, preference was
given to the article providing more
comprehensive information. The Fig. 1. Likelihood of cancer risk.
JOEM • Volume 48, Number 11, November 2006 1191

first step, the strength of the meta- inconsistency. The level was down- effect size.8 The weighing factor Wi*
analysis by each study type (eg, SMR, graded if heterogeneity (inconsistency) in the DerSimonian and Laird random-
PMR) was assigned a score. The score testing among all combined studies effects model is
of “⫹⫹” was assigned if the meta- had an ␣ ⱕ0.10.
1

冋 冉 冊册
relative risk was statistically signifi- W i* ⫽
cant and greater than 1.1. The score of Statistical Methods 1
D⫹
“⫹” was assigned if the meta-relative For all cancer outcomes having two wi
risk was not statistically significant, or more studies, the observed and ex- where Wi is the statistical weight for
but the point risk estimate was greater pected values from each study were a given study for the fixed-effect
than 1.1. The score of “⫺” was as- summed and a meta-relative risk esti- model and is equal to 1/SEi2 with SEi
signed if the meta-relative risk was not mate (mRR) was calculated. An mRR being the standard error for a given
statistically significant, and the point was calculated for each cancer by each study according to Chen and Seaton.9
risk estimate was equal to or less than study type, eg, SMR studies and as a

冘W
1.1. At the second step, these scores summary meta-relative risk across all n
were used to assign a probable, possi- study types. The mRR was defined as 关Q ⫺ (n ⫺ 1)兴 * i
ble, or unlikely designation for the the ratio of the total number of ob-

冉冘 冊
i⫽1
pattern of meta-relative risk associa- served deaths or incident cases to the D⫽ 2

冘W
n n
tion. A “probable” was assigned to the total number of expected deaths or Wi ⫺ i
2
cancer-specific site if one meta- incident cases as follows: i⫽1 i⫽1
relative risk (ie, mSMR, mPMR,

冘O
mSMR and PMR, mRR, mSIR, mOR) n It should be noted that D is set to 0
was statistically significant (score of i if Q ⬍ n ⫺ 1. The random-effects
⫹⫹) and at least another was greater i⫽1 model was validated against data
mRR ⫽ provided in Petitti,10 which after ap-
than 1.1 (score of ⫹). A “possible”
冘E
n
assignment was given if only one plication using our equations gave
i
meta-relative risk was available and i⫽1 identical results. For this study, an
was statistically significant (score of ␣ ⱕ10% or less for declaring heter-
⫹⫹) or if at least two meta-relative where Oi denotes observed deaths ogeneity was adopted.11
risks were greater than 1.1 but were (cases) in each individual study, Ei The SAS software was used to per-
not statistically significant (score of denotes expected deaths (cases), and n form the calculations and validated our
⫹). “Not likely” was assigned if the is the total number of studies.7 The program for the fixed-effect model
cancer-specific site did not meet the 95% confidence interval (CI) of mRR using data from different studies
probable or possible criteria. may be computed using the Poisson compiled by Howe and Burch4 on
The second criterion examined probability distribution as described by standardized mortality ratios and
the “study type” used to generate Breslow and Day.8 The standard error proportional mortality ratios among
meta-relative risks. If the meta-relative (SE) for the meta-relative risk is cal- firefighters. Where there were no
risk estimate reached statistical signif- 1 observed deaths or incident cases,
culated as SE⫽ where Wi is the
icance (score of ⫹⫹), based primarily 冑兺Wi the lower confidence interval for an
on PMR studies, the level was down- statistical weight for a given study individual study was set at 0.1 as
graded. PMR studies do not measure defined as 1/SEi2 and SEi is the stan- suggested in the method used by
the risk of death or death rates but dard error for a given study. Collins and Acquavella. 12 This
rather the relative frequency of that In the absence of heterogeneity, the method was compared with the data
particular cause among all causes of fixed-effect model was applied for de- excluding studies with a zero relative
death. Hence, the limitation of a PMR riving the meta-relative risk estimate; risk, and the results were similar.
study is that the estimate may be ab- otherwise, the random-effects model
normally low or high based on the was used. A test for heterogeneity for Results
overall increase or decrease in mortal- the fixed-effect approach is given by
ity and not due to the cause of interest.6 Q ⫽ 冘i⫽1n
Wi * {log(RRi) ⫺ log(mRR)}2
Identification and
Also, if the mSMR point risk estimate where RRi and mRR are the relative Characteristics of Studies
was not significant and ⱕ1.1 (⫺), the risk and the meta-relative risk, respec- The computerized literature search
level was downgraded. The third crite- tively. The hypothesis of homogeneity identified 21 U.S. and 14 non-U.S.
rion used for generating the likelihood among studies would be rejected if Q articles.13– 47 It was determined that
of cancer risk was an assessment of exceeds ␹n⫺1,␣
2
. Then the random- three studies were not eligible for the
“inconsistency” among studies. Heter- effects model was used with a different meta-analysis because of either insuf-
ogeneity testing as described in statis- study weight (Wi*) that further ac- ficient data,41 data were combined for
tical methods was used to evaluate counts for the interstudy variation in firefighters and other personnel,42 or
1192 Cancer Risk Among Firefighters • LeMasters et al

the text was not published in En- ageal cancer (2.03). Incidence stud- ness of the hazards or use of pro-
glish.43 In addition, four studies44 – 47 ies showed significant meta-SIR for tective equipment.
were excluded because of overlapping cancers of the stomach (1.58), pros- A final check on the three criteria
populations with other reports.18,30 For tate (1.29), and testis (1.83). assessment presented in Table 4 was
example, in 1992, Demers et al18 re- As shown in Table 3, only one made by calculating an overall sum-
ported more observed and expected cancer type, non-Hodgkin’s lym- mary of cancer risk across all studies
cancers than in the 1994 article.46 Four phoma, had two mortality OR anal- (ie, SMR, PMR, RR, SIR, OR).
additional studies48 –51 were identified yses, and both were significant. The There was agreement that cancer was
in the review by Howe and Burch4 and estimated mOR was essentially unlikely between the criteria assess-
used in the meta-analysis. These latter based on Ma et al14 due to the much ment and the not significant sum-
four studies are not presented in Table larger sample size of firefighters mary risk estimates for esophagus,
1. Hence, a total of 28 studies received (n ⫽ 4800) compared with 23 for liver, pancreas, larynx, lung, bladder,
a detailed review as shown in Table 1, Figgs et al.15 Odds ratios were sig- kidney, and Hodgkin’s disease and
which describes the study design char- nificantly higher for buccal cavity/ all cancers (Table 5). Differences
acteristics, exposure, and outcome def- pharynx (5.90) and Hodgkin’s dis- between the two approaches were
initions. Sixteen were U.S. studies and ease (2.4)14 as well as the single found for cancers of the buccal cav-
12 were non-U.S. investigations. Five incidence study related to bladder ity/pharynx and leukemia because
studies had an internal comparison cancer (2.11) and non-Hodgkin’s these were designated as possible by
group with the remaining using re- lymphoma (3.27).22 the criteria assessment but as not
gional or national comparison groups. The next step was to determine the significant in the summary risk esti-
Fourteen ascertained exposures from likelihood of cancer risk based on the mate. The remaining cancers were all
employment records and defined ex- three criteria assessment. Cancers re- rated as probable or possible and all
posure as a dichotomous (yes/no) vari- ceiving “probable” and “possible” had significant summary risk esti-
able. The majority of the studies relied designations are shown in Table 4. mates. Of note, testicular cancer
on death certificates for assessing a Based on evaluating the first crite- received the highest summary risk
cancer diagnosis. Of a total of 32 rion “pattern of meta-relative risk” estimate (OR ⫽ 2.02; 95% CI ⫽
articles, 26 are included in the meta- for the 20 cancer sites, eight were 1.30 –3.13) related to the SIR stud-
analysis as shown in Table 2. The six designated as “probable,” four as ies compared with the “possible”
additional articles are case– control/ “possible,” and eight as an unlikely designation by the three criteria
mortality odds ratio studies and pre- risk. Based on the second criteria assessment.
sented in Table 3 with one meta- “study type” stomach, rectum, skin
analysis for non-Hodgkin’s cancer, and malignant melanoma risk Discussion
lymphoma. were downgraded because of reli- The meta-analysis and criteria as-
ance on PMR studies for statistical sessment designate the likelihood of
Overview of Meta-analysis significance or the mSMR point risk cancer among firefighters as proba-
Table 2 summarizes the meta- estimate was not significant and ble for multiple myeloma and
analysis results by study type. Stud- ⱕ1.1. prostate cancer. Thus, the findings
ies were mostly mortality and were For the third criterion, “inconsis- related to multiple myeloma are in
analyzed using SMRs and PMRs. tency” among all studies caused a agreement with Howe and Burch.4
All-cause mortality had an SMR downgrading for only colon cancer The Philadelphia firefighter study13
10% less than general population to “possible.” This inconsistency was the largest cohort study reported
rates. Mortality from all cancers was may have been related to several to date investigating exposure–
similar to the general population us- factors, including study type and a response relationships. For Philadel-
ing SMR and RR indices, but PMR cohort effect. There were 14 SMR phia firefighters, the SMR results for
studies showed a 10% significantly and PMR colon cancer studies with multiple myeloma demonstrated an
higher rate (Table 2). For individual elevated meta-risk estimates of 1.34 increasing trend with duration of em-
cancers, there were statistically sig- and 1.25, respectively (Table 2). Of ployment as a firefighter: 0.73 (95%
nificant elevated meta-SMR esti- these 14 studies, there were 11 CI ⫽ 0.10 –5.17) for under 9 years,
mates for colon cancer (1.34) and (78.6%) with firefighters employed 1.50 (95% CI ⫽ 0.48 – 4.66) for 10 to
multiple myeloma (1.69). PMR stud- on or before 1950. In contrast, there 19 years, and 2.31 (95% CI ⫽ 1.04 –
ies demonstrated three significantly were six mRR and SIR studies with 5.16) with six observed deaths for
elevated meta-PMR values that in- meta-risk estimates of 0.91 and 0.90, greater than 20 years. Except for
cluded skin (1.69), malignant mela- respectively, with half employed race, there are essentially no known
noma (2.25), and multiple myeloma on or before 1950. It is possible risk factors for multiple myeloma
(1.42). There was one significantly that the older cohorts had higher other than occupational exposures
elevated meta-relative risk for esoph- exposures due to a lack of aware- (eg, paints, herbicides, insecticides,
TABLE 1
Characteristics of Studies From Electronic Search
JOEM
•

Study Number of Comparison Exposure Exposure Cancer

Reference Company Location Design/Analysis Period Workers Group Variable Source Source Cofactors
Baris, 200113 Philadelphia Cohort mortality (SMR) 1925–1986 7789 INT/NGP/NED 1, 3, 5 ER DC Age
Ma, 199814 24 US states Case– control (MOR) 1984 –1993 6607 INT 4 DC DC Age/race
Figgs, 199515 24 US states Case– control (MOR) 1984 –1989 23890 (cases) RGP 4 DC DC Age
119,450 (controls)
Burnett, 199416 27 US states PMR 1984 –1990 5744 INT 4 DC DC Age
Demers, 199317 4 US states Case– control (OR) 1977–1981 692 (cases) LGP 4 TRV TRV Age
1683 (controls)
Demers, 1992a18 Seattle, Tacoma (WA) Cohort mortality (SMR) 1944 –1979 4528 LGP 4 ER DCN, TRV Age
Incidence (SIR) INT/LW/NGP
Demers, 1992b19 Seattle, Tacoma, WA Cohort mortality (SMR) 1944 –1979 4546 INT/LW/NGP 2, 3 ER DCN Age
Portland
Beaumont, 199120 San Francisco Cohort mortality (RR) 1940 –1970 3066 NGP 3, 6 ER DCN Age/yr
Grimes, 199121 Honolulu PMR, RR 1969 –1988 205 RGP 3, 4 ER DC Race
Sama, 199022 Massachusetts Case– control (MOR) 1982–1986 315 LW/RGP 4, 7 TRV TR Age/smoke
Volume 48, Number 11, November 2006

Vena, 198723 Buffalo Cohort mortality (SMR) 1950 –1979 1867 NGP 3 ER DCN Age/yr
Feuer, 198624 New Jersey PMR 1974 –1980 263 LW/RGP/NGP 3, 8 ER DCN Age
Morton, 198425 Portland, Vancouver Incidence (SIR) 1963–1977 1678 RGP 4 TR TRV Age
Dubrow, 198326 British & USA Cohort mortality (SMR) 1950 –1977 — — 4 AR DC None
Musk, 197827 US Cohort mortality (SMR) 1915–1975 5655 RGP, NGP 4 ER DC Age
Berg 197528 US, Great Britain Cohort mortality (SMR) 1949 –1953 — NGP 4 DC DC Age
and PMR 1959 –1963
Stang, 200329 Germany Case– control OR) 1995–1997 269 (cases) RGP 4 ER MR Age
797 (controls)
Bates, 200130 New Zealand Cohort mortality (SMR) 1977–1995 4221 NGP 3 AR DC, TR Age/yr
Incidence (SIR)
Firth, 199631 New Zealand Incidence (SIR) 1972–1984 26207 NED 4 TR TR Age
Deschamps 199532 France Cohort mortality (SMR) 1977–1991 830 NGP 2 ER DCN Age
Delahunt, 199533 New Zealand Case– control (RR) 1978 –1986 710 (cases) NGP 4 TR TR Age/smoke
12,756 (controls)
Aronson, 199434 Canada Cohort mortality (SMR) 1950 –1989 5414 RGP 3, 6, 7 ER DCN Age/yr
Tornling, 199435 Sweden Cohort mortality (SMR) 1931–1983 1153 LGP 1, 3, 7 ER DC, TR Age/yr
Incidence (SIR)
Giles, 199336 Australia Incidence (SIR) 1980 –1989 2865 RGP 3, 6, 7 TRV TR Age
Guidotti, 199337 Canada Cohort mortality (SMR) 1927–1987 3328 RGP 2 ER DCN Age/yr
Hansen, 199038 Denmark Cohort mortality (SMR) 1970 –1980 886 NED 4 OTH DC Age
(Continued)
1193
1194 Cancer Risk Among Firefighters • LeMasters et al

engine exhausts, and organic sol-

Cofactors
vents).52–57 Benjamin et al58 re-

Age/yr

NED ⫽ national employment database

Age
ported that blacks compared with

RGP ⫽ regional general population

NGP ⫽ national general population
whites have at least double the risk

LGP ⫽ local general population

of being diagnosed with multiple
myeloma and twice the mortality
Cancer
Source
DC

DC
rate. Race may be ruled out as a
potential factor among firefighters,
LW ⫽ local workers
Comparison Group:

because cancer risk was investigated

Exposure

INT ⫽ internal

primarily for whites.

Source

DC
ER

The analyses for non-Hodgkin’s

lymphoma were consistent across a
diversity of study designs, including
Exposure

SMR, PMR, SIR, and OR incident/

Variable

mortality studies. All showed ele-

3

vated meta-risk or point estimates.

SMR, standardized mortality/morbidity ratio

The overall summary risk estimate

Comparison

was significantly elevated at 1.51

(95% CI ⫽ 1.31–1.73). Hence, non-
Group
RGP

RGP

SIR, standardized incidence ratio

Hodgkin’s lymphoma is considered a

PMR, proportional mortality ratio

probable cancer risk for firefighters.

Non-Hodgkin’s lymphoma is, how-
MOR, mortality odds ratio

ever, several cancer types with five

Number of

International Classification of Dis-

Workers
990

1039

ease (ICD) codes (200, 202.0, 202.1,

Design/Analysis

OR, odds ratio

RR, rate ratio

202.8, 202.9). Of importance is how

the definition of non-Hodgkin’s lym-
phoma by ICD code may contribute
to the variability in study findings.
1939 –1978

1921–1953
Period

For example, in a study by Demers et

Study

al19 comparing firefighters with po-

lice, the mortality incidence-density
TRV, tumor registry (occupation) with
TR, tumor registry with no validation

ratio (IDR) for “lymphosarcoma and

validation from external sources
Cohort mortality (SMR)

Cohort mortality (SMR)

DCN, death certificate nosologist

reticulosarcoma” (ICD 200) was not

Design/Analysis

elevated (0.81)19 but was (1.40) for

Exposure or Cancer Source

“other lymphatic/hematopoietic”
ER, employment records

AR, association records

(ICD 202, 203). Subsequent to the

and PMR

MR, medical records

DC, death certificate

time period covered in this review,

Ma et al59 examined Florida fire-
fighters but evaluated only one of
OTH, other

two cancers for ICD code 200, ie,

lymphosarcoma but not reticular sar-
Company Location

coma and found nonsignificance

Australia

(SMR ⫽ 0.94). Hence, these studies

Canada

demonstrate the importance of being

cognizant that differences in cancer
5. Company type engine, ladder
4. Occupation (based on death

6. Time since first employment

certificate or tumor registry)

risk estimates and interpretation of

1. Number of firefighter runs
2. Duration of “active” duty
3. Duration of employment

risk may be influenced by outcome

overall as a firefighter

definition.
8. Employment status
Mastromatteo, 195940

Results showing a probable asso-

Exposure Variables

ciation for prostate cancer is curious.

Reference
Eliopulos, 198439

7. Age-specific

Prostate cancer is the most common

malignancy affecting men and is the
Continued

second leading cause of cancer.60

TABLE 1

Risk of developing prostate cancer is

associated with advancing age, black
JOEM • Volume 48, Number 11, November 2006 1195

TABLE 2
Meta-relative Risk Estimates and Test for Inconsistency for Mortality and Incidence*
95%
Number of Meta-relative Confidence P Value
Disease Studies Reference Observed Expected Risk Interval Inconsistency
Mortality studies
Standardized mortality
ratio (SMR)
All causes (001–999) 12 13, 19, 23, 27, 30, 8384 9273.8 0.90 0.85– 0.97 ⬍0.00
32, 34
35, 37– 40
All cancers (140–209) 13 13, 19, 23, 27, 30, 1801 1799.9 1.00 0.93–1.08 0.02
32, 34
35, 37– 40, 51
Buccal cavity and 5 13, 19, 32, 34, 37 34 29.8 1.14 0.79 –1.60 0.84
pharynx (140 –149)
Esophagus (150) 4 13, 19, 23, 34 17 25.1 0.68 0.39 –1.08 0.62
Stomach (151) 7 13, 19, 23, 30, 34, 75 81.3 0.92 0.73–1.16 0.72
35, 37
Colon (153) 10 13, 19, 23, 26, 28, 252 188.3 1.34 1.01–1.79 ⬍0.00
30, 34, 35, 37, 51

Rectum (154) 6 13, 19, 23, 30, 34, 35 54 40.7 1.33 1.00 –1.73 0.43
Liver/gallbladder 5 13, 19, 23, 34, 35 22 21.9 1.00 0.63–1.52 0.92
(155–156)
Pancreas (157) 6 13, 19, 23, 34, 35, 37 63 64.2 0.98 0.75–1.26 0.58
Larynx (161) 3 13, 19, 34 8 13.7 0.58 0.25–1.15 0.82
Lung (162) 8 13, 19, 30, 34, 35, 37, 378 359.2 1.05 0.95–1.16 0.50
38, 51
Skin (173) 3 13, 19, 37 16 15.7 1.02 0.58 –1.66 0.68
Malignant melanoma 2 30, 34 4 5.9 0.67 0.18 –1.70 0.23
(172)
Prostate (185) 6 13, 19, 23, 34, 35, 37 104 91 1.14 0.93–1.39 0.67
Testis (186) 1 34 3 1.2 2.50 0.50 –7.30 —
Bladder (188) 6 13, 19, 23, 30, 34, 37 41 33.0 1.24 0.68 –2.26 0.03
Kidney (189) 6 13, 19, 23, 34, 35, 37 30 30.9 0.97 0.44 –2.13 0.01
Brain and nervous 8 13, 19, 23, 27, 30, 34, 64 46.1 1.39 0.94 –2.06 0.07
system (191–192) 35, 37
Non-Hodgkin’s 3 13, 19, 34 30 20.6 1.46 0.98 –2.08 0.92
lymphoma
(200, 202)
Hodgkin’s disease (201) 2 19, 34 4 5.1 0.78 0.21–2.01 0.59
Multiple myeloma (203) 4 13, 26, 34, 51 24 14.2 1.69 1.08 –2.51 0.15
Leukemia (204 –208) 2 13, 19 30 29.9 1.00 0.68 –1.43 0.27
Proportional mortality
ratio (PMR)
All cancers (140–209) 6 16, 24, 39, 48, 49, 50 2443 2215.7 1.10 1.06 –1.15 0.64
Buccal cavity and — — — — — —
pharynx (140 –149)
Esophagus (150) — — — — — —
Stomach (151) — — — — — —
Colon (153) 4 28, 48, 49, 50 99 79.2 1.25 0.90 –1.74 0.08
Rectum (154) 1 16 37 25 1.48 1.05–2.05 —
Liver/gallbladder — — — — — —
(155–156)
Pancreas (157) — — — — — —
Larynx (161) — — — — — —
Lung (162) 4 16, 48, 49, 50 773 742.1 1.04 0.88 –1.23 0.04
Skin (172–173) 2 16, 24 42 24.8 1.69 1.22–2.29 0.41
Malignant melanoma 2 48, 49 9 4 2.25 1.03– 4.27 0.49
(172)
Prostate (185) — — — — — —
(Continued)
1196 Cancer Risk Among Firefighters • LeMasters et al

TABLE 2
Continued
95%
Number of Meta-relative Confidence P Value
Disease Studies Reference Observed Expected Risk Interval Inconsistency
Testis (186) — — — — — —
Bladder (188) 1 16 37 37.4 0.99 0.70 –1.37 —
Kidney (189) 1 16 53 36.8 1.44 1.08 –1.89 —
Brain and nervous 4 16, 48, 49, 50 64 54.9 1.17 0.90 –1.49 0.27
system (191–192)
Non-Hodgkin’s 1 16 66 50 1.32 1.02–1.67 —
lymphoma
(200, 202)
Hodgkin’s disease — — — — — —
(201)
Multiple myeloma 4 16, 48, 49, 50 46 32.5 1.42 1.04 –1.89 0.88
(203)
Leukemia (204 –208) 2 16, 24 65 53.5 1.21 0.94 –1.55 0.47
Relative risk (RR)
All causes (001–999) — — — — — — —
All cancers (140–209) 2 20, 21 291 295.6 0.98 0.87–1.10 0.17
Buccal cavity and 1 20 11 7.7 1.43 0.71–2.57 —
Pharynx (140 –149)
Esophagus (150) 1 20 12 5.9 2.03 1.05–3.57 —
Stomach (151) 2 20, 21 25 20.6 1.21 0.80 –1.81 0.55
Colon (153) 2 20, 21 25 27.5 0.91 0.60 –1.36 0.92
Rectum (154) 1 20 13 9 1.44 0.77–2.49 —
Liver (155–156) — — — — — — —
Pancreas (157) 1 20 17 13.6 1.25 0.73–2.00 —
Larynx (161) 1 20 3 3.8 0.79 0.17–2.35 —
Lung (162) 1 20 60 71.4 0.84 0.64 –1.08 —
Skin (172–173) 1 20 7 4.1 1.71 0.68 –3.49 —
Malignant melanoma — — — — — — —
(172)
Prostate (185) 2 20, 21 19 24.3 0.78 0.13– 4.82 ⬍0.00
Testis (186) — — — — — — —
Bladder (188) — — — — — — —
Kidney (189) 1 20 4 5.9 0.68 0.19 –1.74 —
Brain and nervous 2 20, 21 9 7.1 1.26 0.55–2.34 0.14
system (191–192)
Non-Hodgkin’s — — — — — — —
lymphoma
(200, 202)
Hodgkin’s disease — — — — — — —
(201)
Multiple myeloma — — — — — — —
(203)
Leukemia (204 –208) 1 20 6 9.8 0.61 0.22–1.33 —
Incidence studies (SIR)
All cancers (140 –209) 3 30, 35, 36 367 366.6 1.00 0.90 –1.11 0.61
Buccal cavity and 2 18, 36 25 19.6 1.28 0.83–1.88 0.73
pharynx (140 –149)
Esophagus (150) 2 18, 30 10 7.6 1.32 0.63–2.42 0.51
Stomach (151) 3 18, 30, 35 38 24.1 1.58 1.12–2.16 0.33
Colon (153) 4 18, 30, 35, 36† 59 65.3 0.9 0.69 –1.17 0.37
Rectum (154) 3 18, 30, 35 41 36.1 1.14 0.81–1.54 0.4
Liver (155–156) 1 35 4 4.7 0.85 0.23–2.18 —
Pancreas (157) 4 18, 30, 35, 36 22 18.2 1.21 0.76 –1.83 0.83
Larynx (161) 2 18, 31 13 8.3 1.57 0.17–14.51 ⬍0.00
Lung (162) 4 18, 30, 35, 36 111 120.0 0.93 0.76 –1.11 0.83
Skin (172–173) 1 35 5 3.3 1.52 0.49 –3.54 —
Malignant melanoma 4 18, 30, 35, 36 60 47.9 1.25 0.96 –1.61 0.87
(172)
Prostate (185) 4 18, 30, 35, 36 147 114.1 1.29 1.09 –1.51 0.56
(Continued)
JOEM • Volume 48, Number 11, November 2006 1197

TABLE 2
Continued
95%
Number of Meta-relative Confidence P Value
Disease Studies Reference Observed Expected Risk Interval Inconsistency
Testis (186) 2 30, 36 21 11.5 1.83 1.13–2.79 0.15
Bladder (188) 2 18, 30 31 29.9 1.04 0.70 –1.47 0.67
Kidney (189) 3 18, 30, 35 11 18 0.61 0.30 –1.09 0.69
Brain and nervous 3 18, 30, 35 19 15.4 1.23 0.74 –1.93 0.84
system (191–192)
Non-Hodgkin’s 1 36 4 2.2 1.82 0.49 – 4.65 —
lymphoma
(200–202)
Hodgkin’s disease — — — — — —
(201)
Multiple myeloma — — — — — —
(203)
Leukemia (204 –208) 4 18, 25, 30, 36 18 12.9 1.4 0.82–2.21 0.36

Note. Codes of the International Classification of Causes of Death (9th Revision) in parentheses; published data for references 48 –50 in
Howe and Birch.4
*Meta analysis completed only for two or more studies.
†Reference 36 is a combination of colon and rectum cancers.

TABLE 3
Mortality and Incidence Studies for Case–Control/Mortality Odds Ratio Studies
95% Confidence
Outcome References Odds Ratio Interval
All cancers (140 –209) Mortality 14 1.10 1.10 –1.20
Buccal cavity and pharynx (140 –149) Mortality 14 5.90 1.90 –18.30
Esophagus (150) Mortality 14 0.90 0.70 –1.30
Stomach (151) Mortality 14 1.20 0.90 –1.60
Colon (153) Mortality 14 1.00 0.90 –1.20
Incidence 22* 1.04 0.59 –1.82
Rectum (154) Mortality 14 1.10 0.80 –1.60
Incidence 22* 0.97 0.50 –1.88
Liver/gallbladder (155–156) Mortality 14 1.20 0.90 –1.70
Pancrease (157) Mortality 14 1.20 1.00 –1.50
Incidence 22* 3.19 0.72–14.15
Larynx (161) Mortality 14 0.80 0.40 –1.30
Lung (162) Mortality 14 1.10 1.00 –1.20
Incidence 22* 1.30 0.84 –2.03
Skin (172–173) Mortality 14 1.00 0.50 –1.90
Malignant melanoma (172) Mortality 14 1.40 1.00 –1.90
Incidence 22* 1.38 0.60 –3.19
Prostate (185) Mortality 14 1.20 1.00 –1.30
Testis (186) Incidence 29 4.00 0.70 –27.40
Bladder (188) Mortality 14 1.20 0.90 –1.60
Incidence 22* 2.11 1.07– 4.14
Kidney (189) Mortality 14 1.30 1.00 –1.70
Incidence 33 4.89 2.47– 8.93
Brain and nervous system (191–192) Mortality 14 1.00 0.80 –1.40
Incidence 22* 1.52 0.39 –5.92
Non-Hodgkin’s lymphoma (200, 202) Mortality 14,15† 1.41 1.10 –1.70
Incidence 22* 3.27 1.19 – 8.98
Hodgkin’s disease (201) Mortality 14 2.40 1.40 – 4.10
Multiple myeloma (203) Mortality 14 1.10 0.80 –1.60
Incidence 17 1.90 0.50 –9.40
Leukemia (204 –208) Mortality 14 1.10 0.80 –1.40
Incidence 22* 2.67 0.62–11.54

*Two control groups available; police rather than state employees selected as most comparable. Significance difference only for malignant
melanoma when using state employees odds ratio and 95% confidence interval was 2.92 (1.70 –5.03).
†Mortality odds ratio (mOR) calculated only for non-Hodgkin lymphoma as only case– control study with at least two studies. mOR estimated
based primarily on larger sample in Ma et al.14
1198 Cancer Risk Among Firefighters • LeMasters et al

Pattern of meta-relative risk: “⫹⫹” meta-relative risk is significant at the 5% level and ⬎1.1; “⫹” meta-relative risk is not significant at the 5% level but ⬎1.1; “⫺” meta-relative risk
ethnicity, a positive family history,

Likelihood of
Cancer Risk
and may be influenced by diet. Al-

Probable

Probable

Probable
Possible
Possible
Possible
Possible
Possible
Possible

Possible
Possible

Possible
though the positive association with
prostate cancer may be due to some
of these factors, it is unlikely that
Criteria 3
Likelihood of Cancer Risk Among Firefighters After Employing Pattern of Meta-relative Risk Association, Study Type, and Inconsistency Among Studies

these entirely explain the findings;

most studies analyzed white men ad-
Inconsistency
No change
No change

No change
No change
No change

No change
No change
No change
No change

No change
No change
Down one justing for age. The summary risk
estimate was 1.28 (95% CI ⫽ 1.15–
1.43). The mSIR was significantly
elevated, and all individual studies
showed excess SIR values. Parent
and Siemiatycki,61 in a review arti-
Likelihood of
Cancer Risk

Probable

Probable

Probable

Probable
Possible
Possible

Possible
Possible
Possible

Possible
Possible

Possible
cle, concluded that there was sugges-
tive epidemiologic evidence for
prostate cancer associated with expo-
Criteria 2

sure to pesticides and herbicides, me-

Study type: down one level, the meta-relative risk (⫹⫹) is based primarily on mPMR studies and/or negative (⫺) mSMR studies.
tallic dusts, metal working fluids,
No change

No change

No change
No change
No change
No change

No change
No change
Down one

Down one
Down one
Down one

polycyclic aromatic hydrocarbon,

Study
Type

Inconsistency among studies: down one level, heterogeneity significant among all combined studies at the 10% level.
and diesel engine emissions. Cer-
tainly firefighters are exposed to
these latter two agents. Recently,
exposure to complex mixture in the
Likelihood of
Cancer Risk

NA indicates no available studies; NC, not able to calculate because only one study of that type available.

semiconductor industry also has

Probable
Probable
Probable
Probable
Probable

Probable

Probable

Probable
Possible

Possible
Possible

Possible

been associated with an increase in

prostate cancer.62 Thus, it is possi-
ble that some of the mixed expo-
sures experienced by firefighters
may be prostate carcinogens. Ross
mOR

⫹⫹

and Schottenfeld63 have cautioned,

⫺
⫺
⫺
⫺
⫺
⫺

⫺
⫺
⫺

⫺
⫺

however, against associating occu-

pational exposures with prostate
mSIR
Pattern of Meta-relative Risk Association

⫹⫹

⫹⫹
⫹⫹
NC

NC

NA
⫹

⫺
⫹

cancer.
Although there were only four stud-
Criteria 1

ies evaluating testicular cancer, we

mRR

NC

NC
NC

NC

NA

NA

NA

NA
⫹
⫺

propose upgrading the likelihood of

cancer risk from possible to probable.
This upgrade is suggested because
mSMR and

testicular cancer had the largest sum-

PMR

⫹⫹
⫹⫹
⫹⫹

⫹⫹

⫹⫹
NC
NC

NC
NC

mary point estimate (2.02, 95% CI ⫽

⫺

1.30 –3.13) as well as consistency

among the one SMR study, two in-
cidence studies, and one case–
mPMR

⫹⫹
⫹⫹

⫹⫹

is ⱕ1.1 and not significant at the 5% level.

NC

NC
NA
NA

NA
NA

control study showing elevated risk

⫹

estimates between 1.15 and 4.30.

Testicular cancer is the most com-
mSMR

mon malignancy between the ages of

⫹⫹

⫹⫹
NC
⫹
⫺

⫹
⫺
⫺

⫹
⫹

20 and 34. Except for cryptorchism,

no risk factor has been clearly dem-
onstrated.64 Because testicular can-
Multiple myeloma

cer occurs among younger men with

Non⫺Hodgkin’s
Cancer Site

high survival, mortality studies are

lymphoma
melanoma

less germane. Bates et al30 showed

Malignant

Leukemia
Stomach

Prostate
Rectum
Buccal

an increase in the incident cases of

TABLE 4

Colon

Testis
Brain
Skin

testicular cancer with firefighter ex-

posure duration as follows: 10 years:
JOEM • Volume 48, Number 11, November 2006 1199

TABLE 5
Summary of Likelihood of Cancer Risk and Summary Risk Estimate (95% CI) Across All Types of Studies for All Cancers
Likelihood of Cancer Summary Risk
Cancer Site Risk by Criteria Estimate (95% CI) Comments
Multiple Probable 1.53 (1.21–1.94) Consistent with mSMR and PMR (1.50, 95% CI ⫽ 1.17–1.89)
myeloma Based on 10 analyses
Heterogeneity—not significant at the 10% level
Non-Hodgkin Probable 1.51 (1.31–1.73) Only two SMR and another PMR studies
lymphoma Slightly higher than mSMR and PMR (1.36, 95% CI ⫽ 1.10 –1.67)
Based on eight analyses
Heterogeneity—not significant at the 10% level
Prostate Probable 1.28 (1.15–1.43) Consistent with mSIR (1.29, 95% CI ⫽ 1.09 –1.51)
Based on 13 analyses
Heterogeneity—not significant at the 10% level
Testis Possible 2.02 (1.30 –3.13) Slightly higher than mSIR (1.83, 95% CI ⫽ 1.13–2.79)
Based on four analyses
Heterogeneity—not significant at the 10% level
Skin Possible 1.39 (1.10 –1.73) Slightly lower than mSMR and PMR (1.44, 95% CI ⫽ 1.10 –1.87) – derived
on basis of PMR studies
Based on eight analyses
Heterogeneity—not significant at the 10% level
Malignant Possible 1.32 (1.10 –1.57) Slightly higher than mSMR and PMR (1.29, 95% CI ⫽ 0.68 –2.20)
melanoma Based on 10 analyses
Heterogeneity—not significant at the 10% level
Brain Possible 1.32 (1.12–1.54) Slightly higher than mSMR and PMR (1.27, 95% CI ⫽ 0.98 –1.63)
Based on 19 analyses
Heterogeneity—not significant at the 10% level; there was
heterogeneity among SMR studies
Rectum Possible 1.29 (1.10 –1.51) Slightly lower than mSMR and PMR (1.39, 95% CI ⫽ 1.12–1.70)
Based on 13 analyses
Heterogeneity—not significant at the 10% level
Buccal cavity Possible 1.23 (0.96 –1.55) Slightly higher than mSMR (1.18, 95% CI ⫽ 0.81–1.66)
and pharynx Based on nine analyses
Heterogeneity—not significant at the 10% level
Stomach Possible 1.22 (1.04 –1.44) Lower than mSIR (1.58, 95% CI ⫽ 1.12–2.16)
Based on 13 analyses
Heterogeneity—not significant at the 10% level
Colon Possible 1.21 (1.03–1.41) Slightly lower than mSMR and PMR (1.31, 95% CI ⫽ 1.08 –1.59)
Based on 25 analyses
Heterogeneity—significant at the 10% level; there was
heterogeneity among SMR and PMR studies
Leukemia Possible 1.14 (0.98 –1.31) Similar to mSMR and PMR (1.14, 95% CI ⫽ 0.92–1.39)
Based on eight analyses
Heterogeneity—not significant at the 10% level
Larynx Unlikely 1.22 (0.87–1.70) Higher than mSMR (0.58, 95% CI ⫽ 0.25–1.15)
Based on seven analyses
Heterogeneity—not significant at the 10% level
Bladder Unlikely 1.20 (0.97–1.48) Similar to mSMR and PMR (1.24, 95% CI ⫽ 0.83,1.49)
Based on 11 analyses
Heterogeneity—significant at the 10% level; there was
heterogeneity among SMR studies
Esophagus Unlikely 1.16 (0.86 –1.57) Higher than mSMR (0.68, 95% CI ⫽ 0.39 –1.08)
Based on eight analyses
Heterogeneity—not significant at the 10% level
Pancreas Unlikely 1.10 (0.91–1.34) Slightly higher than mSMR (0.98, 95% CI ⫽ 0.75–1.26)
Based on 13 analyses
Heterogeneity—not significant at the 10% level
Kidney Unlikely 1.07 (0.78 –1.46) Similar to mSMR and PMR (1.23, 95% CI ⫽ 0.94 –1.59)
Based on 12 analyses
Heterogeneity—significant at the 10% level; there was
heterogeneity among SMR studies
(Continued)
1200 Cancer Risk Among Firefighters • LeMasters et al

TABLE 5
Continued
Likelihood of Cancer Summary Risk
Cancer Site Risk by Criteria Estimate (95% CI) Comments
Hodgkin’s Unlikely 1.07 (0.59 –1.92) Higher than mSMR (0.78, 95% CI ⫽ 0.21–2.01)
disease Based on three analyses
Heterogeneity—not significant at the 10% level
Liver Unlikely 1.04 (0.72–1.49) Similar to mSMR (1.00, 95% CI ⫽ 0.63–1.52)
Based on seven analyses
Heterogeneity—not significant at the 10% level
Lung Unlikely 1.03 (0.97–1.08) Similar to mSMR and PMR (1.05, 95% CI ⫽ 0.96 –1.14)
Based on 19 analyses
Heterogeneity—not significant at the 10% level; there was
heterogeneity among PMR studies
All cancers Unlikely 1.05 (1.00 –1.09) Similar to mSMR and PMR (1.06, 95% CI ⫽ 1.02–1.10
Based on 25 analyses
Heterogeneity—significant at the 10% level; there was
heterogeneity among SMR studies

CI indicates confidence interval; SMR, standardized mortality ratio; PMR, proportional mortality ratio; SIR, standardized incidence ratio.

SIR ⫽ 1.39, 95% CI ⫽ 0.2–5.0; 11 noted in Table 4, however, there titative summary risk estimates
to 20 years: SIR ⫽ 4.03, 95% CI ⫽ were elevated, but not significant, shown in Table 5, 10 cancers, or half,
1.3–9.4. In those exposed greater risk estimates across all studies, ie, were significantly associated with
than 20 years, the risk estimate re- mSMR, mPMR, mRR, and mSIR. firefighting. Three cancers were des-
mained elevated but declined (SIR ⫽ This consistency is all the more re- ignated as a probable risk based on
2.65, 95% CI ⫽ 0.3–9.6), possibly markable given the diversity of rare the quantitative meta-risk estimates
because testicular cancer generally cancers included in the category and our three criteria assessment.
occurs at a younger age. Bates et al30 “brain and nervous system.” Further- These cancers included multiple my-
argued that, although the reason for more, there was a 2003 study by eloma, non-Hodgkin’s lymphoma,
the excess risk of testicular cancer Krishnan et al65 published after our and prostate. A recommendation is
remained obscure, the possibility that search that examined adult gliomas also made, however, for upgrading
this is a chance finding was low in the San Francisco Bay area of men testicular cancer to “probable” based
because incident studies are likely in 35 occupational groups. This on the twofold excess summary risk
the most appropriate methodology study showed that male firefighters estimate and the consistency among
for a cancer that can be successfully (six cases and one control) had the the studies. Thus, firefighter risk for
treated. highest risk with an odds ratio of these four cancers may be related to
The 1990 findings of Howe and 5.93, although the confidence inter- the direct effect associated with ex-
Burch4 showing a positive associa- vals were wide and not significant. In posures to complex mixtures, the
tion with brain cancer and malignant addition, malignant melanoma was routes of delivery to target organs,
melanoma are compatible with our also initially scored as probable but and the indirect effects associated
results because both had significant was downgraded to “possible” due to with modulation of biochemical or
summary risk estimates. Brain can- study type. This study downgrade physiologic pathways. In anecdotal
cers were initially scored as probable was related to the negative SMR (⫺) conversations with firefighters, they
but then downgraded to possible (Ta- and reliance primarily on a PMR report that their skin, including the
ble 5). There was inconsistency study. Thus, in conclusion, our study groin area, is frequently covered with
among the SMR studies, which re- supports a probable risk for multiple “black soot.” It is noteworthy that
sulted in the use of the random- myeloma, similar to Howe and testicular cancer had the highest
effects model, yielding confidence Burch’s4 findings, and a possible summary risk estimate (2.02) and
limits that were not significant association with malignant mela- skin cancer had a summary risk esti-
(SMR ⫽ 1.39, 95% CI ⫽ 0.94 –2.06) noma and brain cancer. mate (1.39) higher than prostate
(Table 2). This inconsistency primar- (1.28). Certainly, Edelman et al3 at
ily resulted from the Baris et al Summary the World Trade Center, although
study,13 a 61-year follow up of 7789 We implemented a qualitative under extreme conditions, revealed
firefighters demonstrating a marked three-criteria assessment in addition the hazards that firefighters may en-
reduction in brain cancer (SMR ⫽ to the quantitative meta-analyses. counter only because air monitoring
0.61, 95% CI ⫽ 0.31–1.22). As Based on the more traditional quan- was performed.
JOEM • Volume 48, Number 11, November 2006 1201

As noted in Table 1, approxi- 4. Howe GR, Burch JD. Fire fighters and tional cohort studies in the United States.
mately half of the studies used local, risk of cancer: an assessment and over- Am J Epidemiol. 1992;136:1232–1240.
view of the epidemiologic evidence. 19. Demers PA, Heyer NJ, Rosenstock L.
regional, or national general popula-
Am J Epidemiol. 1990;132:1039 –1050. Mortality among firefighters from three
tion rates as the comparison group. 5. Overall evaluations of carcinogenicity: northwestern United States cities. Br J
These general population compari- an updating of IARC Monographs, vols Ind Med. 1992;49:664 – 670.
son groups raise concern that the 1– 42, suppl 7. IARC Monographs on the 20. Beaumont JJ, Chu GS, Jones JR, et al. An
actual risk of cancer may be under- Evaluation of Carcinogenic Risks to Hu- epidemiologic study of cancer and other
estimated due to the healthy worker mans. Lyon: International Agency for causes of mortality in San Francisco fire-
effect related to the strict physical Research on Cancer; 1989. fighters. Am J Ind Med. 1991;19:357–372.
6. Decoufle P, Thomas TL, Pickle LW. 21. Grimes G, Hirsch D, Borgeson D. Risk of
entry requirements, maintenance of
Comparison of the proportionate mortal- death among Honolulu fire fighters Ha-
better physical fitness, and good ity ratio and standardized mortality ratio waii. Med J. 1991;50:82– 85.
health benefits. The healthy worker risk measures. Am J Epidemiol. 1980; 22. Sama SR, Martin TR, Davis LK, et al.
bias may be less pronounced, how- 111:263–269. Cancer incidence among Massachusetts
ever, for cancer than for conditions 7. Wong O, Raabe GK. Application of firefighters, 1982–1986. Am J Ind Med.
such as coronary heart disease. Fur- meta-analysis in reviewing occupational 1990;18:47–54.
thermore, tobacco is unlikely a con- cohort studies. Occup Environ Med. 23. Vena JE, Fiedler RC. Mortality of a
1996;53:793– 800. municipal-worker cohort: IV. Fire fight-
tributing factor because cancers
8. Breslow NE, Day NE. Statistical methods ers. Am J Ind Med. 1987;11:671– 684.
known to be associated with smok- in cancer research. Volume II—the de- 24. Feuer E, Rosenman K. Mortality in po-
ing such as lung, bladder, and larynx sign and analysis of cohort studies. IARC lice and firefighters in New Jersey. Am J
were designated as unlikely and cor- Sci Publ. 1987;82:1– 406. Ind Med. 1986;9:517–527.
responding summary risk estimates 9. Chen R, Seaton A. A meta-analysis of 25. Morton W, Marjanovic D. Leukemia in-
were not statistically significant. mortality among workers exposed to or- cidence by occupation in the Portland–
These findings of an association of ganic solvents. Occup Med (Lond). 1996; Vancouver metropolitan area. Am J Ind
46:337–344. Med. 1984;6:185–205.
firefighting with significant increased
10. Petitti DB. Decision Analysis in Meta 26. Dubrow R, Wegman DH. Setting priori-
risk for specific types of cancer raise Analysis, Decision Analysis and Cost- ties for occupational cancer research and
red flags and should encourage further Effectiveness Analysis: Methods for control: synthesis of the results of occu-
development of innovative comfort- Quantitative Synthesis in Medicine, 2nd pational disease surveillance studies.
able protective equipment allowing ed. New York: Oxford University Press; J Natl Cancer Inst. 1983;71:1123–1142.
firefighters to do their jobs without 2000:102–118. 27. Musk AW, Monson RR, Peters JM, et al.
compromising their health. Studies are 11. Greenland S. Quantitative methods in the Mortality among Boston firefighters,
review of epidemiologic literature. Epi- 1915–1975. Br J Ind Med. 1978;35:104 –
especially needed that better character-
demiol Rev. 1987;9:1–30. 108.
ize the type and extent of exposures to 12. Collins JJ, Acquavella JF. Review and 28. Berg JW, Howell MA. Occupation and
firefighters. meta-analysis of studies of acrylonitrile bowel cancer. J Toxicol Environ Health.
workers. Scand J Work Environ Health. 1975;1:75– 89.
Acknowledgments 1998;24(suppl 2):71– 80. 29. Stang A, Jockel KH, Baumgardt-Elms C,
The authors thank members of the Orange 13. Baris D, Garrity TJ, Telles JL, et al. et al. Firefighting and risk of testicular
County Fire Authority, Battalion 4, Station Cohort mortality study of Philadelphia cancer: results from a German popula-
22, for their insights into cancer risk. The firefighters. Am J Ind Med. 2001;39:463– tion-based case– control study. Am J Ind
authors also thank Connie Thrasher and Mi- 476. Med. 2003;43:291–294.
chael Kuhlman for their assistance in prepa- 14. Ma F, Lee DJ, Fleming LE, et al. Race- 30. Bates MN, Fawcett J, Garrett N, et al. Is
ration and Drs Gary Marsh, Leslie Stayner, specific cancer mortality in US firefight- testicular cancer an occupational disease
and Sheila Zahm for their expert review and ers: 1984 –1993. J Occup Environ Med. of fire fighters? Am J Ind Med. 2001;40:
input. 1998;40:1134 –1138. 263–270.
15. Figgs LW, Dosemeci M, Blair A. United 31. Firth HM, Cooke KR, Herbison GP.
States non-Hodgkin’s lymphoma surveil- Male cancer incidence by occupation:
References lance by occupation 1984 –1989: a twen- New Zealand, 1972–1984. Int J Epide-
1. Brandt-Rauf PW, Fallon LF Jr, Tarantini ty-four state death certificate study. Am J miol. 1996;25:14 –21.
T, et al. Health hazards of fire fighters: Ind Med. 1995;27:817– 835. 32. Deschamps S, Momas I, Festy B. Mor-
exposure assessment. Br J Ind Med. 16. Burnett CA, Halperin WE, Lalich NR, et tality amongst Paris fire-fighters. Eur J
1988;45:606 – 612. al. Mortality among fire fighters: a 27 Epidemiol. 1995;11:643– 646.
2. Golden AL, Markowitz SB, Landrigan state survey. Am J Ind Med. 1994;26: 33. Delahunt B, Bethwaite PB, Nacey JN.
PJ. The risk of cancer in firefighters. 831– 833. Occupational risk for renal cell carci-
Occup Med. 1995;10:803– 820. 17. Demers PA, Vaughan TL, Koepsell TD, noma. A case– control study based on the
3. Edelman P, Osterloh J, Pirkle J, et al. et al. A case– control study of multiple New Zealand Cancer Registry. Br J Urol.
Biomonitoring of chemical exposure myeloma and occupation. Am J Ind Med. 1995;75:578 –582.
among New York City firefighters re- 1993;23:629 – 639. 34. Aronson KJ, Tomlinson GA, Smith L.
sponding to the World Trade Center fire 18. Demers PA, Vaughan TL, Checkoway H, Mortality among fire fighters in metro-
and collapse. Environ Health Perspect. et al. Cancer identification using a tumor politan Toronto. Am J Ind Med. 1994;26:
2003;111:1906 –1911. registry versus death certificates in occupa- 89 –101.
1202 Cancer Risk Among Firefighters • LeMasters et al

35. Tornling G, Gustavsson P, Hogstedt C. 46. Demers PA, Checkoway H, Vaughan TL, sures and cancer. Environ Health Per-
Mortality and cancer incidence in Stock- et al. Cancer incidence among firefight- spect. 1995;103(suppl 8):205–208.
holm fire fighters. Am J Ind Med. 1994; ers in Seattle and Tacoma, Washington 56. Davis DL, Blair A, Hoel DG. Agricultural
25:219 –228. (United States). Cancer Causes Control. exposures and cancer trends in developed
36. Giles G, Staples M, Berry J. Cancer 1994;5:129 –135. countries. Environ Health Perspect. 1993;
incidence in Melbourne Metropolitan 47. Bates MN, Lane L. Testicular cancer in 100:39 – 44.
Fire Brigade members, 1980 –1989. fire fighters: a cluster investigation. N Z 57. Sonoda T, Nagata Y, Mori M, et al.
Health Rep. 1993;5:33–38. Med J. 1995;108:334 –337. Meta-analysis of multiple myeloma and
37. Guidotti TL. Mortality of urban firefight- 48. Petersen GR, Milham S. Occupational benzene exposure. J Epidemiol. 2001;11:
ers in Alberta, 1927–1987. Am J Ind Mortality in the State of California, 1959 – 249 –254.
Med. 1993;23:921–940. 1961. Cincinnati: National Institute for Oc- 58. Benjamin M, Reddy S, Brawley OW.
38. Hansen ES. A cohort study on the mor- cupational Safety and Health; 1980. Myeloma and race: a review of the liter-
tality of firefighters. Br J Ind Med. 1990; 49. Milham S. Occupational Mortality in ature. Cancer Metastasis Rev. 2003;22:
47:805– 809. Washington State, 1950 –1971, vol I. 87–93.
39. Eliopulos E, Armstrong BK, Spickett JT, et Cincinnati: National Institute for Occupa- 59. Ma F, Fleming LE, Lee DJ, et al. Mor-
al. Mortality of fire fighters in Western Aus- tional Safety and Health; 1976. tality in Florida professional firefighters,
tralia. Br J Ind Med. 1984;41:183–187.
50. Gallagher R, Threfall WJ, Band PR, et al. 1972 to 1999. Am J Ind Med. 2005;47:
40. Mastromatteo E. Mortality in city fire-
Occupational Mortality in British Colum- 509 –517.
men. II. A study of mortality in firemen
bia 1950 –1984. Richmond, British Co- 60. Crawford ED. Epidemiology of prostate
of a city fire department. AMA Arch Ind
lumbia, Canada: Worker’s Compensation cancer. Urology. 2003;62(suppl 1):3–12.
Health. 1959;20:227–233.
Board Press; 1989. 61. Parent ME, Siemiatycki J. Occupation and
41. Muscat JE, Wynder EL. Diesel exhaust,
51. Howe GR, Lindsay JP. A follow-up study of prostate cancer. Epidemiol Rev. 2001;23:
diesel fumes, and laryngeal cancer. Oto-
a ten-percent sample of the Canadian labor 138 –143.
laryngol Head Neck Surg. 1995;112:
force. I. Cancer mortality in males, 1965–73. 62. Beall C, Bender TJ, Cheng H, et al.
437– 440.
42. Zahm SH, Brownson RC, Chang JC, et J Natl Cancer Inst. 1983;70:37–44. Mortality among semiconductor and stor-
al. Study of lung cancer histologic types, 52. Durie BG. The epidemiology of multiple age device-manufacturing workers. J Oc-
occupation, and smoking in Missouri. myeloma. Semin Hematol. 2001;38(suppl cup Environ Med. 2005;47:996 –1014.
Am J Ind Med. 1989;15:565–578. 3):1–5. 63. Ross RK, Schottenfeld D. Prostate can-
43. Elci OC, Akpinar-Elci M, Alavanja M, et al. 53. Costantini AS, Miligi L, Vineis P. An cer. In: Schottenfeld D, Fraumeni JF, eds.
Occupation and the risk of lung cancer by Italian multicenter case– control study on Cancer Epidemiology and Prevention,
histologic types and morphologic distribu- malignant neoplasms of the hematolym- 2nd ed. New York: Oxford University
tion: a case–control study in Turkey. Monaldi phopoietic system. Hypothesis and pre- Press; 1996:1180 –1206.
Arch Chest Dis. 2003;59:183–188. liminary results on work-related risks. 64. Huyghe E, Matsuda T, Thonneau P. In-
44. Rosenstock L, Demers P, Heyer NJ, et al. WILL (Working Group on Hematolym- creasing incidence of testicular cancer
Respiratory mortality among firefighters. phopoietic Malignancies in Italy). Med worldwide: a review. J Urol. 2003;170:
Br J Ind Med. 1990;47:462– 465. Lav. 1998;89:164 –176. 5–11.
45. Heyer N, Weiss NS, Demers P, et al. 54. Burmeister LF. Cancer in Iowa farmers: 65. Krishnan G, Felini M, Carozza SE, et al.
Cohort mortality study of Seattle fire recent results. Am J Ind Med. 1990;18: Occupation and adult gliomas in the San
fighters: 1945–1983. Am J Ind Med. 295–301. Francisco Bay area. J Occup Environ
1990;17:493–504. 55. Blair A, Zahm SH. Agricultural expo- Med. 2003;45:639 – 647.