Professional Documents
Culture Documents
Background: The immune system and inflammation are statistically significantly higher risk of invasive breast can-
implicated in the pathogenesis of cancer. Prospective stud- cer (hazard ratio [HR], 1.15; 95% confidence interval [CI],
ies linking biomarkers of inflammation with cancer in- 1.04-1.26), colorectal cancer (HR, 1.19; 95% CI, 1.00-
cidence and mortality have been inconclusive. 1.41), endometrial cancer (HR, 1.42; 95% CI, 1.12-
1.79), and lung cancer (HR, 1.63; 95% CI, 1.35-1.97).
Methods: To determine whether there is an indepen- The findings were similar when cancers that occurred dur-
dent association of white blood cell (WBC) count with ing the first 2 years of follow-up were excluded. Statis-
incident cancer in postmenopausal women, a prospec- tically significant associations remained for invasive breast
tive cohort study was performed at 40 US clinical cen- cancer and endometrial cancer when the analyses were
ters involving 143 748 postmenopausal women aged 50 limited to nonsmokers. The WBC count was also statis-
to 79 years who were free of cancer at baseline and were tically significantly associated with breast cancer, lung
enrolled in the Womens Health Initiative. The main out-
cancer, and overall cancer mortality.
come measures were incident invasive breast, colorec-
tal, endometrial, and lung cancer.
Conclusion: Postmenopausal women with higher WBC
Results: In multivariate models, there was a graded as-
counts have a higher risk of incident invasive breast, co-
sociation of WBC count with incidence of all 4 types of lorectal, endometrial, and lung cancer, as well as a higher
cancer. Compared with the lowest quartile of WBC count risk of breast, lung, and overall cancer mortality.
(2.50-4.79109 cells/L), women with a WBC count
in the upper quartile (6.80-15.00109 cells/L) had a Arch Intern Med. 2007;167(17):1837-1844
A
CAUSAL LINK BETWEEN IN- Several prospective studies 6-8 have
flammation and cancer found that the white blood cell (WBC)
was first hypothesized in count is a predictor of cancer mortality,
1863 by Virchow, who ob- although a study9 among Korean men and
served leukocytes in neo- women found no association, and an-
plastic tissues.1 Well-known associations other study10 found the association to be
between cancer and certain infections (eg, specific to smoking-related cancers. The
chronic hepatitis B and hepatocellular car- WBC count and C-reactive protein level
cinoma) and inflammatory conditions (eg, have been shown to be associated with an
Author Affiliations: Crohn disease and colorectal carcinoma) increased risk of colorectal cancer in 4
HealthPartners Research have supported the notion that cancer studies,11-14 but the C-reactive protein level
Foundation, Minneapolis,
was unassociated with colorectal cancer in
Minnesota (Dr Margolis);
Fred Hutchinson Cancer For editorial comment another study15 that enrolled only women.
Research Center, Seattle, A meta-analysis16 of fibrinogen levels has
Washington (Ms Rodabough
see page 1822 also shown an association with increased
and Dr McTiernan); and risk of colorectal cancer mortality.
Department of Nutritional arises at sites of chronic inflammation.1 The
Sciences (Dr Thomson) and apparent protection from certain types of CME course available online
the Arizona Cancer Center cancer by aspirin and nonsteroidal anti- www.archinternmed.com
(Dr Lopez), University of inflammatory drugs provides an addi-
Arizona, Tuscon. tional line of evidence of a causal rela- The Womens Health Initiative (WHI)
Group Information: A full list
tion.2 Researchers have begun to elucidate is a multicenter prospective study of
of the members of the Womens
Health Initiative Research the molecular mechanisms by which in- 161 808 postmenopausal women com-
Group was published in fectious agents, inflammatory cells, and in- posed of diverse racial/ethnic and socio-
Arch Intern Med flammatory mediators may initiate and economic groups. At baseline, partici-
(2007;167[9]:901). promote cancer.3-5 pants in the WHI underwent a WBC count,
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1837
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1838
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1839
(continued )
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1840
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1841
Abbreviations: CI, confidence interval; HR, hazard ratio (computed from Cox proportional hazards models stratified by participation in the Womens Health
Initiative clinical trials or observational study); Q, quartile (Q1, 2.50-4.79 109 cells/L; Q2, 4.80-5.69 109 cells/L; Q3, 5.70-6.79 109 cells/L; and
Q4, 6.80-15.00 109 cells/L).
a Multivariate model is adjusted for age, race/ethnicity, smoking, alcohol intake, physical activity, use of aspirin or nonsteroidal anti-inflammatory drugs,
hormone therapy use, body mass index, height, age at menarche and menopause, parity, months of breastfeeding, bilateral oophorectomy, history of benign
breast disease, and family history of breast cancer.
b Multivariate model is adjusted for age, race/ethnicity, smoking, alcohol intake, physical activity, use of aspirin or nonsteroidal anti-inflammatory drugs,
hormone therapy use, body mass index, history of diabetes mellitus, and family history of colorectal cancer.
c Analyses limited to women with a uterus. Multivariate model is adjusted for age, race/ethnicity, smoking, physical activity, use of aspirin or nonsteroidal
anti-inflammatory drugs, hormone therapy use, body mass index, height, age at menarche and menopause, parity, and family history of endometrial cancer.
d Multivariate model is adjusted for age, race/ethnicity, pack-years of smoking, and current and past smoking.
Adjustment of multivariate models for known risk fac- count and lung cancer by smoking is a possibility, al-
tors, exclusion of incident cases early in the follow-up though we adjusted for current smoking, past smoking,
period, and limitation of the analyses to nonsmokers di- and pack-years of smoking in the multivariate analysis.
minish the chance that our findings are a result of con- In addition to the epidemiological evidence, exten-
founding or reverse causality, especially for breast and sive basic research supports a role of the immune sys-
endometrial cancer. We also tested models limited to tem and inflammation in the pathogenesis of cancer. Most
women who had previously undergone breast cancer tumors are heavily infiltrated by inflammatory cells, which
screening (verified mammography reports) and colorec- in turn produce cytokines and chemokines that regu-
tal cancer screening (self-report only). Because these late the growth, migration, and differentiation of tumor
analyses included only one-third to one-half as many can- and stromal cells.1,2,24,25 Several genetic polymorphisms
cer cases, power was much diminished, and there were in cytokine genes have been associated with an in-
no statistically significant associations of WBC count with creased risk for colorectal adenomas, a precursor of co-
these cancers in this analysis. Because we did not dem- lorectal cancer.26 Neutrophils, eosinophils, and mono-
onstrate that an elevated WBC count is present before nuclear cells produce reactive oxygen and nitrogen species
breast and colorectal cancer is detectable by screening, that further damage cellular DNA, RNA, lipids, and pro-
definitive evidence of a causal relation is not present. teins.27 Proteases, matrix metalloproteinases, angio-
The weaker and statistically nonsignificant relation of genic factors, and other soluble mediators produced by
WBC count with colorectal cancer in nonsmokers could leukocytes affect cell survival and tissue remodeling.28
be because of chance or because the overall association Of the cytokines, tumor necrosis factor may play a cen-
is caused by elevated WBC counts in smokers. Smoking tral role as a tumor promoter, as well as in the interac-
is a risk factor for colorectal adenomas,23 and our data tion between tumor and inflammatory cells, in part
show that current smoking is also strongly related to WBC through induction of nuclear factor Bsignaling path-
count. Similarly, confounding of the relation of WBC ways.29 Cancer cells may use the same adhesion mol-
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1842
Abbreviations: CI, confidence interval; HR, hazard ratio (computed from Cox proportional hazards models stratified by participation in the Womens Health
Initiative clinical trials or observational study); Q, quartile (Q1, 2.50-4.79 109 cells/L; Q2, 4.80-5.69 109 cells/L; Q3, 5.70-6.79 109 cells/L; and
Q4, 6.80-15.00 109 cells/L).
a Multivariate model is adjusted for age, race/ethnicity, smoking, alcohol intake, physical activity, use of aspirin or nonsteroidal anti-inflammatory drugs,
hormone therapy use, body mass index, height, age at menarche and menopause, parity, months of breastfeeding, bilateral oophorectomy, history of benign
breast disease, and family history of breast cancer.
b Multivariate model is adjusted for age, race/ethnicity, smoking, alcohol intake, physical activity, use of aspirin or nonsteroidal anti-inflammatory drugs,
hormone therapy use, body mass index, history of diabetes mellitus, and family history of colorectal cancer.
c Analyses limited to women with a uterus. Multivariate model is adjusted for age, race/ethnicity, smoking, physical activity, use of aspirin or nonsteroidal
anti-inflammatory drugs, hormone therapy use, body mass index, height, age at menarche and menopause, parity, and family history of endometrial cancer.
d Multivariate model is adjusted for age, race/ethnicity, pack-years of smoking, and current and past smoking.
e Multivariate model is adjusted for age; race/ethnicity; smoking; alcohol intake; physical activity; use of aspirin or nonsteroidal anti-inflammatory drugs;
hormone therapy use; body mass index; height; waist to hip ratio; age at menarche and menopause; parity; bilateral oophorectomy; family history of breast,
colorectal, ovarian, and endometrial cancer; and history of diabetes mellitus, hypertension, and benign breast disease.
ecules and trophic factors made by inflammatory cells performed in 40 local laboratories on automated counters.
in homing during metastasis.2,3 Therefore, inflamma- Multiple measurements in a central laboratory would have
tion seems to play a key role in all stages of cancer from reduced measurement error and increased the precision
initiation to distant spread. of our results; therefore, our present results are likely to
Several limitations of this analysis must be consid- be underestimates of the true associations due to non-
ered. Differential cell counts were not performed for the differential misclassification. The participants in the WHI
WHI, so we do not have information about what types were generally healthy, well-educated, postmenopausal
of leukocytes were more prone to be elevated in women women; therefore, our results may not apply to the gen-
who subsequently developed cancer. Only 1 measure- eral population or to men, despite the broad geographic
ment of WBC count was performed, and the analyses were representation of the 40 clinical centers.
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1843
(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 17), SEP 24, 2007 WWW.ARCHINTERNMED.COM
1844