American Joumal of Epidemiology Vol. 139, No.

3
Copyright © 1994 by The Johns Hopkrns Unrversity School of Hygiene and Public Health Printed m U.S.A.
AII rights reserved

Chili Pepper Consumption and Gastric Cancer in Mexico:

A Case-Control Study

Lizbeth López-Carrillo,1·2 Mauricio Hernández Avila,2 and Robert Dubrow1

Laboratory studies indicate that capsaicin, the hot-tasting component of chili

peppers, may be carcinogenic. A population-based case-control study was conducted
in Mexico City during 1989-1990 to evaluate the relation between chili pepper
consumption and gastric cancer risk. The study included 220 incident cases and 752
controls randomly selected from the general population. lnformation was collected by
interview. Chili pep- per consumers were at high risk for gastric cancer compared with
nonconsumers (age- and sex-adjusted odds ratio= 5.49, 95% confidence interval (CI)
2.72-11.06). Among consumers, there was a highly significant trend of increasing risk
with increasing self-
rated level of consumption (low, medium, and high) (p = 2 x 10-7). The odds ratio for
high-level consumers compared with nonconsumers was 17.11 (95% CI 7.78-37.59).
However, when consumption was measured as frequency per day, a significant trend
among consumers was not observed. Multivariable adjustment increased the
magnitude of the chili pepper-gastric cancer association, but a significant trend among
consumers (measured as frequency per day) was still not observed. Chili pepper
consumption may be a strong risk factor for gastric cancer, but further studies are
needed to test this hypothesis. Am J Epidemio/ 1994;139:263-71.

capsaicin; capsicum; case-control studies; stomach neoplasms

Animal and in vitro studies indicate that
capsaicin, the pungent principie of hot chili
peppers (members of the genus Capsicum),
may be carcinogenic. In mice, capsaicin has
been shown to induce duodenal adenocar-
cinomas (1), to actas a promoter of stomach
and liver tumors (2), and to induce micro-
nuclei formation (3). In vitro, capsaicin was
mutagenic to Chinese hamster lung cells (4)
and exhibited mutagenic activity in the
Ames test (1, 3).
Received for publication March 18, 1993, and in
final form October 22, 1993.
Abbreviations: CI, confidence interval; OR, odds
ratio.
1
Department of Epidemiology and Public Health,
Vale University School of Medicine, New Haven, CT.
2 National lnstitute of Public Health (Mexico), Cuer-
navaca, Morelos, Mexico.
Reprint requests to Dr. lizbeth López-Carrillo, Na-
tional lnstitute of Public Health, Center for Public
Health Research, Av. Universidad 655, Col. Sta.
Maria Ahua-
catitlan, Cuernavaca, Morelos, C.P. 62508, Mexico.
Mexico produces about 650,000 tons of
chili peppers annually, 95 percent of which
is consumed locally. The per capita con-
sumption is about 20 g (one chili pepper)
per day (5).
In 1990, gastric cancer accounted for 10
percent of cancer deaths in Mexico (6).
The high leve! of chili pepper consump-
tion in Mexico, along with reports of the
carcinogenicity and mutagenicity of capsa-
icin, prompted the authors to conduct a
case- control study of adenocarcinoma
of the stomach in the Mexico City
metropolitan area to determine whether
chili pepper con- sumption is a risk factor
far gastric cancer in
humans.
MATERIALS ANO METHODS
Cases
Between September 17, 1989 and June
30, 1990, 267 newly diagnosed cases of gas-
263
264 López-Carrillo et al.
trie cancer, 20 years of age and older, were
identified among residents of the Mexico
City metropolitan area. These cases were
identified from 15 Mexico City metropoli-
tan area hospitals and represented about 80
percent of the stomach cancer cases reported
to the Mexican National Cancer Registry by
hospitals in the Mexico City metropolitan
are a.
Twenty-two of the 267 cases were un-
available far interview. Five cases were ex-
cluded because the tumors turned out not to
be adenocarcinomas of the stomach; 20
cases were excluded because the pathology
slides could not be obtained far review. The
slides from the remaining 220 cases were
reviewed by Dr. José Jessurun (Mexico City
General Hospital) and were classified ac-
cording to the system of Laureo (7). Ninety-
eight cases were classified as intestinal-type
adenocarcinoma, 95 cases as diffuse-type,
and 27 cases as indeterminate.
Controls
Controls were an age-stratified random
sample of residents of the Mexico City met-
ropolitan area selected from the household
sampling frame of the 1986-1987 Mexican
National Health Survey (8). The sampling
units were 1,375 blocks of households ran-
domly selected from the Mexico City met-
ropolitan area, each block containing 40-óO
households. One household was randomly
selected from each block, and was randomly
assigned a 5-year age group (from age
20-24 to 60-64 years) according to the age
distribution of gastric cancer cases reported
to the National Cancer Registry in 1987 (9).
The 1,375 households were visited by
study interviewers, who obtained listings of
the ages of every adult who lived in the
households. Persons in the assigned age
group for a given household were consid-
ered to be eligible controls. In addition, due
to the old age distribution of gastric cancer
cases combined with the young age distri-
bution of the Mexico City population, all
persons in the selected households who were
65 years of age and older were considered to
be eligible controls. If more than one eli-
gible control lived in a household, only one
was chosen to be interviewed, generally an
eligible control who was at home at the time
of the interviewer visit.
No one was found to be home after three
visits to 75 households. In 510 households,
there were no eligible controls. In 38 house-
holds, the eligible controls refused to be in-
terviewed. Eligible controls from the re-
maining 752 households were interviewed.
Data collection
Interviewers, using a structured question-
naire, conducted direct interviews with 78
percent of the cases; far the remainder of
cases, who were severely ill, surrogate re-
spondents (usually the spouse, although
sometimes an adult son or daughter) were
interviewed. The median number of days
between date of diagnosis and interview was
30.5 days (range: 0-281 days). A random
sample of 7 percent of the controls was se-
lected for surrogate interviews. Control in-
terviews were conducted during the same
time period as the case interviews.
For logistic reasons, cases and controls
were interviewed by different interviewers.
Although interviewers were aware of the
subjects' case-control status, they were
blind to the study hypothesis. The duration
of interview differed between cases (30.2 :±::
13.0 minutes) and controls (24.1 :±:: 8.8
minutes).
The questionnaire included a semiquan-
titative food frequency questionnaire based
on the methodology developed by Willett et
al. (10), and adapted to the Mexican diet, as
well as specific questions about chili pepper
consumption. Cases were queried about
their dietary habits during the 12-month pe-
riod prior to the onset of symptoms. Con-
trols were queried about their dietary habits
during the 12-month period preceding the
interview. lt is our assumption that dietary
habits during these periods reflected longer-
term adult dietary habits. Subjects were also
asked about sociodemographic variables,
smoking, and medica) history.
Chili pepper consumption was first as-
sessed with the question, "Do you eat chili
 Chili Pepper Consumption and Gastric Cancer 265
peppers or chili sauces with your meals?" tervals were estimated by unconditional lo-
(choices were yes, sometimes, or no). Con- gistic regression. Age (5-year age groups)
sumers of chili (those who answered yes or
sometimes) were then asked about the fre-
quency of consumption of the three types of
chili pepper that they most often consumed.
Chili pepper consumers were also asked to
rate themselves as consumers (low, me-
dium, or high) and whether they frequently
enjoyed the burning sensation caused by
chili pepper consumption (yes or no).

Data analysis

Frequency of chili pepper consumption

(times/day) was calculated far each subject
by summing the frequencies of
consumption of the three types of chili
pepper most often consumed by each
subject. Consumers of chili peppers were
classified into three lev- els according to
the approximate tertile dis- tribution among
the controls.
Individual food items were grouped into
eight food groups: dairy products, fruits,
vegetables, meat, processed meat, fish, ce-
reals and breads, and soda and candy. The
frequency of consumption (times/day) of
each food group was calculated by
summing the frequencies of consumption of
the indi- vidual food items in the group.
The daily intake of total calories was
com-
puted by multiplying the daily frequency of
consumption of each food item by the calo-
rie content of the specified portian and sum-
ming over the food items.
Alcohol consumption (g/day) was calcu-
lated by summing the frequencies of con-
sumption ofbeer, wine, rum and brandy,
and
tequila after converting the frequencies to
grams of alcohol per day according to the
average alcohol content of the beverages. A
scorc far cigarette consumption was ob-
tained as the product of years of smoking
and average number of cigarettes smoked
per day (cigarette-years),
Far each subject, a socioeconomic index
(faur levels) was calculated, based on the
types of water supply and sewage system.
Odds ratios and 95 percent confidence in-
was included in ali models because controls
were frequency matched to cases by age.
Sex was included in ali models a priori.
To test far the significance of linear
trends, frequency of chili pepper consump-
tion (times/day) was included as a continu-
ous variable in a logistic model; chili pep-
per consumption (no, sometimes, yes), self-
rated level of chili pepper consumption
(none, low, medium, high), and frequent
enjoyment of "getting burned" (noncon-
sumer, no, yes) were included in models as
ordinal variables. To test far the signifi-
canee of linear trends among consumers, a
dichotomous variable far chili pepper con-
sumption (yes or sometimes vs. no) was
included in each model along with the chili
pepper consumption variable of interest. In
each case, the p value far trend was derived
from the likelihood ratio statistic for the
variable of interest.
To determine whether the relation be-
tween chili pepper consumption and gastric
cancer differed by histologic type, intesti-
nal-type and diffuse-type cases were com-
pared separately with the control group
with respect to chili pepper consumption.
The statistical significance of differences
between intestinal-type and diffuse-type
cases was determined by the case-case
comparison.

RESULTS
Toe case group consisted of 122 males
(55 percent) and 98 females (45 percent).
The control group consisted of 296 males
(39 percent) and 456 females (61 percent).
The cases and controls had similar age
distribu- tions by study design. The mean
age of the cases was 57.2 ::.t 14.3 years, and
the mean age of the controls was 59.2 ::.t
14.7 years.
After adjusting far age and sex, chili pep-
per consumption was strongly associated
with gastric cancer (odds ratio (OR) = 5.49,
95 percent confidence interval (CI) 2.72-
11.06) (table 1). However, when chili pep-
per consumption was measured as fre-
quency of consumption per day, the odds
ratio far tertile 3 (5.01) was actually lower
than that far tertile 2 (6.14). Subjects who
266 López-Carrillo et al. Chili Pepper Consumption and Gastric Cancer 267

o
"'
N
N 1
e,
e o
o ll)
v
E
a,
o X
~ r--
~ ío "'1

ai
(')
o
l.
o
oo
e X ó X X
> (')
a,
l.>
C\IN v
l.> ......
.: v,...
¡¡; ll) CX) CX)
a,
e,
....:N ,...: ,...: (')
,... C\I .-C\j(\j

e
o cb .J. .J. d, el,
C) CX) ex> v ll)
eo óM C\Í si (')
aE o o co ,- OCONa:>
,.ºº._ ._
l.()

• <O ,...: o,...: co

ll) CX)
::, a: . .: . . . : ,. . : o;
111
N
O) ....
Q) u
e! (') CX) b
e :, e: e: Q)
E oo C)
o v o
"'~
-
>
Q)
o:,
"' ó ó X ó
o..,E E"'e
l.>
ai +- : C\J
8
a. "'
. ; o ....
a.
CI>
a. u e ío N
oo X
b
X o X ......
(') LO
<D -e- ex>
.-o ll)
o ......
co.C\10. ll) . ,-o
~,...:
ll) ,... C),... ,...
iq (')
C) ...
J, el,
<D ......
J. cb cb
N e, (')
¿, J.
(')o
ÓC\Í C\Í C\Í C\Í C\Í C')
a: oco oov-
o O<X>
....: ....: lO
C)
v O CO .-
....:..;:tc.ólt')
O
OO>C\J..-
O C1> C"')
~(')si,-.:
,-
º"'"'
o
....: ~ (D
(O C\I

o"'~
N e, O> N ex> eo N
zu-.
C) CX) C) (')
o"' coCX) ll) <D a:, ll) C\I CX)

ll) ll) C\I

v ll) ll)
ll)

~ e:
~g
c. c.
., E
c.:::,

Bs=(1)
rated themselves as having a high leve! of
chili pepper consumption were at dramati-
cally increased risk (OR = 17.11, 95 percent
CI 7.78-37.59). Chili pepper consumers
who reported frequent enjoyment of "get-
ting burned" were at higher risk (OR =
6.27) than consumers who did not enjoy
"getting burned" frequently (OR = 4.67).
There was a significant linear trend when
chili pepper consumption was measured as
no/sometimes/yes (p = 3 X 10-10), as fre-
quency of consumption per day (p = 0.002),
as none/low/medium/high (p = 5 X 10-14),
or as frequent enjoyment of "getting
burned" (nonconsumers/no/yes) (p = 7 X
10-1) (table 1). Among consumers, the trend
was significant for sometimes/yes (p =
0.003) and low/medium/high (p = 2 X
10-1), but not for frequency per day (p =
0.48) or frequent enjoyment of "getting
burned" (no/yes) (p = 0.09).
In separate multivariable models, each of
the chili pepper consumption variables was
adjusted simultaneously for age, sex, total
calories, variables known or suspected a
priori to alter gastric cancer risk (socioeco-
nomic status; history of peptic ulcer; and
consumption of fruits, vegetables, processed
meat, salt, alcohol, and cigarettes), and bean
consumption (which is associated with chili
pepper consumption in the Mexican diet)
(table 1). The odds ratios for chili pepper
consumption from these multivariable rnod-
els were appreciably higher than the age-
and sex-adjusted odds ratios (e.g., odds ratio
for yes vs. no = 9.22; odds ratio for high-
level of consumption vs. nonconsumption =
27.88). Further modeling indicated that this
negative confounding was due to bean con-
sumption and history of peptic ulcer.
As in the age- and sex-adjusted models,
there was a significant linear trend for each
of the four chili pepper consumption vari-
ables in the multivariable models. However,
the trend among consumers was still not sig-
nificant when chili pepper consumption was
measured as frequency per day (p = 0.45)
or as frequent enjoyment of "getting
burned" (p = 0.45).
Of concern were potential biases resulting
from case-control differences in duration of
interview and type of interview (direct vs.
surrogate ). In addition, the control selection
scheme was biased toward selecting con-
trols who tended to stay at home. These po-
tential biases were assessed by simulta-
neously adding the variables duration of
interview, type of interview, and occupation
(occupations that tend to stay at home, i.e.,
housewife, retired, and unemployed vs. ali
others) to the multivariable models. Adding
these variables resulted in no meaningful
change in the odds ratios for the chili pepper
consumption variables.
Table 2 presents age- and sex-adjusted
odds ratios for the effects of chili pepper
consumption on gastric cancer risk by his-
tologic type. The odds ratios for intestinal-
type gastric cancer were substantially higher
than the odds ratios for diffuse-type gastric
cancer. However, differences between the
intestinal and diffuse types were not statis-
tically significant.
DISCUSSION
To our knowledge, this is the first epide-
miologic study to examine the relation be-
tween chili pepper consumption and gastric
cancer in humans. Chili pepper consump-
tion was found to be a strong risk factor for
gastric cancer in Mexico. Chili pepper con-
sumers were at 5.5-fold greater risk for gas-
trie cancer than nonconsumers. Persons who
rated themselves as heavy consumers of
chili peppers were at 17-fold greater risk.
Multivariable adjustment for variables
known or suspected a priori to alter gastric
cancer risk actually increased these risk
estimates.
One troublesome point, however, was the
flatness of the dose-response curve among
chili pepper consumers, as measured by fre-
quency of consumption per day. The trend
for this variable among consumers was not
significant in either the age- and sex-ad-
justed model or in the multivariable model.
On the other hand, a trend among consumers
was observed when chili pepper consump-
tion was measured as sometimes/yes or as
low/medium/high. If the chili pepper con-
sumption-gastric cancer association is real
268 López-Carrillo et al. Chili Pepper Consumption and Gastric Cancer 269

e
ºocaº~, <Q)
O)
C\I
<Q)
o
co
ó ó
a dE co,
t
::, (7>
en ,.... 'í' co
e::,:; o O) o o
O;t: o
o
uo X ó X ó
~o C\I co
Qg. ·; LO C\I C\I LO l'-
o (1)
:E
LO O . co
~~ :::,
(\j O)
,....
I~
(1)

i
<J) . • LO

:E.!€ ;!: e J, ..- .},

uo e ..,. ~~C\I

o~ o,....: ,.....:,.....:M a
o
-uuo a: o (\j .....(\j
E:::,
CII "O o OMl'-
0 00 M OC0.._,.00 e
<J)

:t: CII ,-: ci M ..:e\iC\iai 8

CII en
CII l'CI o <J)

:;... -9e +r ~
o co co C\I ..... COl'-COLO
zº 00 (\j (') (\j

.2~
-ol'C:I i 1 "í'
o'í'
CI) (')

-
en C l.
o
o oo o
-¡¡; Cl. X
ó
X
~ .5:e
N
"O o CX) (\j
(')

CII
u C
l'CI
o
co
U),.....:
co
¿;¡¡
Ol
e en
:C121 C'E:IIu O> LO
C\ÍMC\i
00
o~
u u
'#. e O C\llllM
~B OLO'<tLO
,.....: ,.....: <O ,.....:
.!:! o C\I C\I C\J
"¡ O ¡ a:
..,.(')
.
-l'CI
CC CI
Oc-,
-a,
o+- <J)
O)

,.... •
o co
(1)
<J)

en zº
CI

- LO O '<t O
..,. C\10 .....
~g ,- (\J C\J ,-

<J)
og
:8 e
o
Zu
~ E
en l'CI
:g ~
o~
"O
C11 U
,: ; ;:, · m
' 5 l'Cl-'o
>C .!! - eo
( 1)
CII~ a..::
en >, c. c.
"O .e ., E
c.:::,
e~
.
ca en
CII ._
CI C11
;:: :E ~
Uu
<C eu
c..;r 1
wu
..J;::
~!
1- CI
and a dose-response relation does, in fact,
exist, these results indicate that there was
substantial misclassification in the measure-
ment of frequency of consumption per day
among consumers.
Alternatively, the possibility should be
considered that the association was due to a
strong bias in favor of the cases answering
yes to having consumed chili peppers. Such
a bias could be due to differential recall of
chili pepper consumption between the cases
and controls. Although cases were queried
about chili pepper consumption during the
12 months prior to the onset of symptoms,
it is nevertheless possible that symptoms did
influence their responses. Perhaps chili pep-
per consumption among persons with pre-
clinical stomach cancer during the year prior
to symptoms caused unusual gastric dis-
tress, making cases particularly aware of
their chili pepper consumption. In addition,
sorne people may relate chili pepper con-
sumption to stomach ailments. Thus, cases
may have reported chili pepper consumption
to a greater degree than controls because
they believed that eating chili peppers
caused their stomach cancer. This type of
recall bias, which is difficult to rule out, is
always of concern in case-control studies.
However, chili peppers have been a tradi-
tional part of the Mexican diet, and the au-
thors know of no evidence that the public
specifically associates stomach cancer with
chili pepper consumption.
There were other potential sources of bias,
many of which are inherent in the case-con-
trol study design. Selection bias would oc-
cur if stomach cancer cases who consumed
chili peppers were more likely to be selected
as study subjects than potential controls who ·
consumed chili peppers. About 80 percent
of stomach cancer cases reported to occur in
the Mexico City metropolitan area were
identified from the 15 participating hospi-
tals, and more than 80 percent of these cases
were study subjects. lt is possible that not ali
cases of gastric cancer in the Mexico City
metropolitan area were histologically iden-
tified and reported. While the authors cannot
prove that the cases who participated in this
study were representative of all cases with
respect to chili pepper consumption, a bias
would have to be very strong to explain the
association observed.
Toe method of control selection was bi-
ased toward selecting controls who tended
to stay at home. Selection bias would have
occurred if staying at home were associated
with chili pepper consumption. However,
including occupation (housewife, retired,
and unemployed vs. ali other occupations)
in the multivariable models did not change
the results, indicating that this potential bias
was not a factor in this study.
For logistic reasons, different interview-
ers were used for the cases and controls. If
the case interviewers solicited information
about chili pepper consumption differently
than the control interviewers, a bias could
result. Such a bias is unlikely, however. The
case and control interviewers were trained
together in the same manner, and the inter-
view was structured. Although the ínter-
viewers knew the case/control status of the
study subjects, they were not aware of the
specific hypothesis of this study. Thus, it is
unlikely that they influenced the answers in
a particular way.
A matter of concern was that the case in-
terviews were significantly longer than the
control interviews. However, including du-
ration of interview in the multivariable
model resulted in no meaningful change in
the odds ratios for the chili pepper consurnp-
tion variables.
lnformation obtained from surrogates
may be less accurate than information ob-
tained directly from study subjects. More
surrogate interviews were obtained among
cases than among controls. However, in-
cluding type of interview in the multivari-
able models did not change the results.
Assuming that the chili pepper-gastric
cancer association is real, with about 80 per-
cent of the general population consuming
chili peppers and an overall relative risk of
about five, 76 percent of stomach cancer
cases in the Mexico City metropolitan area
can be attributed to chili pepper consump-
tion. Of course, the etiology of gastric can-
cer is multifactorial (11-13), and gastric
cancer cases attributable to chili pepper con-
270 López-Carrillo et al.
sumption most likely arose in conjunction
with other risk factors. Such factors could
include low intake of fruits, vegetables, and
fiber-rich breads; consumption of processed
and smoked meats, dried, salted, and
smoked fish, and salty and pickled faods;
and possibly alcohol consumption and ciga-
rette smoking (11-13). In addition, Helico-
bacter pylori infection recently has been im-
plicated as a gastric cancer risk factor (14-
17).
Nevertheless, in spite of the heavy chili
pepper consumption, the rate of gastric can-
cer in Mexico City appears to be relatively
Iow (18). This may be due to Iow consump-
tion of processed meats and salted, smoked,
and pickled foods along with high consump-
tion of fruits, vegetables, and beans (beans
were highly protective against stomach can-
cer in this case-control study (unpublished
data)). Thus, chili pepper consumption may
confer a high relative risk on a background
of very low absolute risk due to a Iow preva-
Ience of other important stomach cancer risk
factors and a high prevalence of protective
factors.
Hot chili peppers are the only dietary
source of capsaicin. It is capsaicin, in par-
ticular, that makes chili peppers distinctive
from other vegetables. Given that capsaicin
has been found to be carcinogenic and mu-
tagenic in animal and in vitro studies, it is a
reasonable conclusion that chili pepper con-
sumption could be a stomach cancer risk
factor due to the capsaicin content of chili
peppers.
To our knowledge, only two other epide-
miologic studies have reported on chili pep-
per consumption and cancer risk. In an Ital-
ian case-control study, chili was briefly
mentioned as being protective against stom-
ach cancer (19). No information was pre-
sented on the leve! of consumption that
showed a protective effect. Chili peppers are
not heavily consumed in northem Italy (20),
where this study was conducted. It is pos-
sible that chili consurnption was correlated
with other spices, such as onion and garlic,
that are more heavily consumed in Italy, and
that may be protective (19, 21-23). In a
case-control study in India, red chili powder
was found to be a risk factor for cancers of
the oral cavity, pharynx, esophagus, and Iar-
ynx using population controls, but not using
hospital controls (24).
Chili pepper consumption was a stronger
risk factor far intestinal-type gastric cancer
than far diffuse-type gastric cancer. This
was consistent with the hypothesis that
intestinal-type gastric cancer is more closely
related to dietary and environmental factors
than is the diffuse-type (12). However, due
to small numbers, the difference in risk was
not statistically significant, and more work
is needed to clarify this point.
Although the authors can identify no ob-
vious bias that can explain the strong asso-
ciation between chili pepper consumption
and gastric cancer, the lack of a dose-
response relation when chili pepper con-
sumption was measured as frequency per
day was of particular concem. Further stud-
ies of chili pepper consumption and gastric
cancerare clearly needed. If the association
is confirmed, it would have great public
health importance in Mexico and in other
parts of the world where chili peppers are
heavily consumed.
ACKNOWLEDGMENTS
This work was supported by the Mexican Min-
istry of Health, the National Council of Science
and Technology (Mexico), the Mexican Health
Foundation, and the Advisory Board in Epide-
miology (Mexico). Dr. Dubrow received support
from a National Cancer lnstitute Preventive On-
cology Academic Award (K07-CA01463).
The authors thank Dr. Malaquias López-
Cervantes for technical comments and sugges-
tions; Ing. Joaquín Papaqui for computing as-
sistance; Dr. José Jessurun for reading the pa-
thology slides; Drs. Jaime Sepulveda and
Roberto Tapia-Conyer for providing access to
the National Health Survey sampling frame; Dr.
María del Pilar López-Acuña for providing ac-
cess to and support from the National Cancer
Registry; Dr. Susan Vandale and Ms. Blanca
Pelcastre for technical assistance with the manu-
script; and to our research staff and the partici-
pating hospitals: Hosp. Adolfo López Mateas,
 Chili Pepper Consumption
Inst. Nac. de Cancerología, Hosp. Español,
Hosp. Especialidades C.M. "La Raza", Hosp.
GEA González, Hosp. General de Mexico,
Hosp. Ignacio Zaragoza, Hosp. Gineco-
Obstetricia C.M. "La Raza", Clínica Londres,
Hosp. Los Angeles, Hosp. Metropolitano, Inst.
Nac. de la Nutrición, Hosp. de Oncología
C.M.N., Hosp. Pemex Central Sur, and Hosp.
20 de Noviembre.
REFERENCES
l. Toth B, Rogan E, Walker B. Tumorigenicity and
mutagenicity studies with capsaicin of hot pep-
pers. Anticancer Res 1984;4:117-20.
2. Agrawal RC, Wiessler M, Hecker E, et al. Tu-
mour-promoting effect of chili extract in Balb/c
mice. lnt J Cancer 1986;38:689-95.
3. Nagabhushan M, Bhide SV. Mutagenicity of
chili extrae! and capsaicin in short-term tests.
Enviran Mutagen 1985;7:881-8.
4. Lawson T, Gannett P. The mutagenicity of cap-
saicin and dihydrocapsaicin in V79 cells. Cancer
Lett 1989;48:109-13.
5. Coordinación General de Abasto y Distribución
del DDF. Servicio Nacional de Información de
Mercados. Banco Nacional del Pequeño Comer-
cio. Sistema producto chile serrano para el Dis-
trito Federal. (In Spanish). Mexico City: Coor-
dinación General de Abasto y Distribución del
DDF. Servicio Nacional de Información de Mer-
cados. Banco Nacional del Pequeño Comercio,
1990.
6. Dirección General de Estadística, Informática y
Evaluación. Dirección de Estadistica. Principales
causas de defuncion por entidad federativa y gru-
pos de edad. (In Spanish). Mexico City: Direc-
ción General de Estadística, Informática y Evalu-
ación, 1990.
7. Lauren P. The two histological main types of
gastric carcinoma: diffuse and so-called intesti-
nal-type carcinoma. Acta Pathol Microbio!
Scand 1965;64:31-49.
8. Dirección General de Epidemiología. Encuesta
nacional de salud. (In Spanish). Mexico City:
Dirección General de Epidemiología, 1988.
9. Dirección General de Epidemiología. Registro
Nacional de Cancer-Resultados 1987, Ciudad
and Gastric Cancer 271
de Mexico. (In Spanish). Mexico City: Dirección
General de Epidemiología, 1989.
10. Willett WC, Sampson L, Stampfer MJ, et al.
Reproducibility and validity of a semiquantita-
tive food frequency questionnaire. Am J Epide-
mial 1985;122:51-65.
11. Nomura A. Stomach. In: Schottenfeld D, Frau-
meni JF, eds. Cancer epidemiology and preven-
tion. Philadelphia: WB Saunders, 1982:624-37.
12. Howson CP, Hiyama T, Wynder EL. The decline
in gastric cancer: epidemiology of an unplanned
triumph. Epidemial Rev 1986;8:l-27.
13. Boeing H. Epidemiological research in stornach
cancer: progress over the last ten years. J Cancer
Res Clin Oncol 1991;117:133-43.
14. Nomura A, Stemrnermann GN, Chyou PH, et al.
Helicobacter pylori infection and gastric carci-
noma among Japanese Americans in Hawaii. N
Engl J Med 1991;325:1132-6.
15. Parsonnet J, Friedman GD, Vandersteen DP, et
al. Helicobacter pylori infection and the risk of
gastric carcinoma. N Engl J Med 1991;325:
1127-31.
16. Forman D, Newell DG, Fullerton F, et al. As-
sociation between infection with Helicobacter
pylori and risk of gastric cancer: evidence
from a prospective investigation. BMJ 1991;302:
1302-5.
17. Talley NJ, Zinsmeister AR, Weaver A, et al.
Gastric adenocarcinoma and Helicobacter pylori
infection. J Natl Cancer lnst 1991;83: 1734-9.
18. Muñoz N, Correa P, Cuello C, et al. Histologic
types of gastric carcinoma in high- and low-risk
areas. lnt J Cancer 1968;3:809-18.
19. Buiatti E, Palli O, Decarli A, et al. A case-control
study of gastric cancer and diet in ltaly. lnt J
Cancer 1989;44:611-16.
20. DeWitt D, Gerlach N. The whole chili pepper
book. Boston: Little, Brown and Co, 1990:172.
21. Trichopoulos D, Ouranos G, Day NE, et al. Diet
and cancer of the stomach: a case-control study
in Greece. lnt J Cancer 1985:36:291-7.
22. You WC, Blot WJ, Chang YS, et al. Allium
vegetables and reduced risk of stomach cancer. J
Natl Cancer lnst 1989;81 :162-4.
23. Graham S, Haughey B, Marshall J, et al. Diet in
the epidemiology of gastric cancer. Nutr Cancer
1990;13:19-34.
24. Notani PN, Jayant K. Role of diet in upper aero-
digestive tract cancers. Nutr Cancer 1987; 10:
103-13.