Alcohol, tobacco, diet, mate drinking, and esophageal cancer in Argentina. R Castelletto, X Castellsague, N Muñoz, et al. Cancer Epidemiol Biomarkers Prev 1994;3:557-564. Published online October 1, 1994.
Updated Version
Citing Articles
E-mail alerts Reprints and Subscriptions Permissions
Access the most recent version of this article at: http://cebp.aacrjournals.org/content/3/7/557
This article has been cited by 6 HighWire-hosted articles. Access the articles at: http://cebp.aacrjournals.org/content/3/7/557#related-urls
Sign up to receive free email-alerts related to this article or journal. To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at
[email protected]. To request permission to re-use all or part of this article, contact the AACR Publications Department at
[email protected].
Downloaded from cebp.aacrjournals.org on July 13, 2011 Copyright © 1994 American Association for Cancer Research
Vol.
3 557-564,
October/November
1994
Alcohol,
Cancer
Tobacco, Esophageal
Diet, Mate Drinking, Cancer in Argentina
Roberto Castelletto, Xavier Castellsague, Nubia Mu#{241}oz,’ Jos#{233} Iscovich, Nestor Chopita, and Alejandro Jmelnitsky Departamento de Patologia, Facultad de Ciencias M#{233}dicas de Ia Plata, La Plata, Argentina IR. C.]; Servei d’Epidemiologia i Registre del Cancer, Ciutat Sanit#{224}ria i Universit#{224}ria de Bellvitge, Autovia Castelldefels K. 2,7, E-08907 Hospitalet del Llobregat, Barcelona, Spain IX. C.]; International Agency for Research on Cancer, 1 50 Cours Albert Thomas, F-69372 Lyon cedex Allied 91000, Ciencias
08, France IN. MI; Israel Center for Registration of Cancer Diseases, Ministry of Health, 107 Derech Hebron, Jerusalem Israel
IJ. 1.1; and
Servicio
M#{233}dicasde Ia Plata,
de Gastroenterologia, La Plata,
Argentina
Facultad
and de
IN. C., A. J.]
Abstrad To study the role of hot mate drinking, alcohol, tobacco, and diet in esophageal cancer, a case-control study including 1 31 cases and 262 hospital controls was carried out in La Plata, Argentina. In multivariate analyses, statistically significant increases in risk were detected for alcohol, tobacco, and some dietary factors but not for hot mate drinking. A strong dose-response relationship was observed with the amount of alcohol consumed daily but not with the number of cigarettes smoked. The odds ratio for those drinking more than 200 ml of ethanol/day compared to nondrinkers was 5.7 (95% confidence interval, 2.2-1 5.2). An increased risk was also observed for those eating barbecued meat more than once a week (odds ratio, 2.4; 95% confidence interval, 1 .2-4.8) as compared to those eating it less than once a week, and a redudion in risk was associated with daily consumption of nonbarbecued beef as compared to those eating it less than daily. Concerning mate drinking, the only variable that showed an effect was the temperature at which mate is drunk. Those who reported drinking mate hot or very hot as compared to those drinking it warm had an increase in risk (odds ratio, 1 .7; 95% confidence interval, 1.0-2.9). Our findings strengthen the evidence for an important role of alcohol and tobacco in esophageal carcinogenesis but do not provide strong support for a role of hot mate drinking. Introdudion In Latin America, the highest rates of esophageal cancer are observed in an area that includes Southern Brazil, Uruguay, and Northeastern Argentina (1-3). These high risk populations provide a unique opportunity for assessing the role of thermal injury in the causation of esophageal cancer because a considerable proportion of the population (over 80%) has the habit of drinking mate, a hot infusion made of
Received 5/1 9/94; revised 7/22/94; accepted 7/25/94. 1 To whom requests for reprints should be addressed.
Epidemiology,
Biomarkers
& Prevention
and
“Ilex paraguayensis,” and there is a well defined unexposed group who do not drink mate. To assess the role of thermal injury resulting from drinking very hot mate and of other known risk factors for esophageal cancer, such as alcohol and tobacco, a series of case-control studies have been set up in these countries under the coordination of the International Agency for Research on Cancer. Results from the studies in Brazil and Uruguay have identified alcohol drinking and tobacco smoking as the most important risk factors for esophageal cancer in these populations (4-5). A strong association with drinking of hot mate was also detected in Uruguay but not in Brazil. We present here the results of a third case-control study carried out in La Plata, Argentina, to obtain further information on the role of hot mate drinking and other risk factors in esophageal cancer. In this study a protocol and a questionnaire every similar to those utilized in the previous two studies were used.
Subjects
and Methods
Study Population Cases. Cases were
subjects newly diagnosed with esophageal cancer and admitted to any of the 1 0 main hospitals of greater La Plata (population, 620,000 inhabitants) from May 1 986 to June 1 989. Criteria of eligibility were: (a) to have a histological diagnosis of squamous cell carcinoma of the esophagus; (b) to have been diagnosed within the previous 4 months; (c) to have resided in great La Plata area province for at least 5 years; and (d) to be in sufficiently good physical and mental health to give reliable answers to the questionnaire. Cases were ascertained shortly after clinical diagnosis through periodic visits to the participating hospitals. A total of 1 33 cases were identified during the study period; 2 of the patients refused to participate in the study so 1 31 cases were included. They represent about 72% of the cases registered in the area during the study period.
Controls.
For each case, two hospital controls were chosen who fulfilled the following inclusion criteria: (a) to be of the same sex and age (±5 years) as the case; (b) to be admitted to the same hospitals as the corresponding case; (c) not to have a diagnosis oftobaccoand/or alcohol-related disease; and (d) to have resided in great La Plata for at least 5 years. The main diagnostic categories among controls were orthopaedic conditions (38%), infectious diseases (28%), urological conditions (24%), and others (10%). Interviews The study subjects were interviewed by 1 of 4 social workers. The questionnaire elicited information on socioeconomic status (education, occupation, and income), lifetime habits of tobacco smoking (type of tobacco and duration, intensity, and cessation periods), alcohol drinking (daily amount consumed estimated in ml of alcohol and duration
Downloaded from cebp.aacrjournals.org on July 13, 2011 Copyright © 1994 American Association for Cancer Research
557
558
Esophageal
Cancer
in Argentina
and type of alcoholic beverage), and mate drinking (daily amount consumed, duration, temperature of consumption, and mode of preparation). Dietary habits were assessed for two time periods, the period just before the onset of symptoms and 10 years before admission to the hospital. A food frequency questionnaire was used to assess the consumption of 1 0 food groups: fresh meat; processed meat; eggs; cereals; potatoes and sweet potatoes; vegetables; fruits; fats; and dairy products. All questionnaires were checked for consistency and accuracy of coding (J. I.).
Statistical
Analysis
Results
2
The
abbreviations
used
I
Distributio
included 1 31 cases and 262 were invited to participate
are:
OR,
odds
ratio;
CI, confidence
controls. and 262
interval.
n of cases
and
controls
according
to m am
Cases Variable
(19.11
55
55-64
46
(35.1)
83
(31.7)
65-74
39
(29.8)
88
(33.6)
21
(16.01
36
(13.7)
106
(80.9)
212
(80.9)
Male
25
(19.1)
50
(19.1)
San Martin
71
(54.2)
142
(54.2)
de Dios
16
(12.2)
32
(12.2)
DiagnOstico
18
(13.7)
36
(13.7)
26
(19.8)
52
(19.8)
Urban
90
(68.7)
Rural
41
(31.3)
15
(11.5)
103
(78.6)
del
Other
Education
hospitals
None Primary Secondary
Monthly
income
+
0-499 500-999 1000
Smoking
status
Drinking
status
+
Never-smoker Ever-smoker Never-drinker Ever-drinker
196
status
Never-drinker Ever-drinker
(74.8)
66 14 227
(25.2) (5.3) (86.6)
13
(9.9)
21
(8.0)
85
(64.9)
122
(46.6) (23.7)
18
(13.7)
62
28
(21.4)
78
(29.8)
30
123.0)
111
(42.4)
101
(57.6)
(77.1)
151
12
(9.2)
59
(22.6)
1 19
(90.9)
202
(77.3)
Unknown Male
(21.0)
Female
Inst.
residence
262
25
San Juan
Main
(%)
No.
54
75 Sex
factors
Controls
(‘/o)
131
Total Agegroup
risk
Categories No.
Hospital
Data were analyzed using unconditional logistic regression (6) with maximum likelihood estimation of parameters values allowing for the potential effects of the three study design variables: age; sex; and hospital. All logistic regression models systematically included sex, age (categorized into four levels), and hospital (one category for each of the three main hospitals from which most of the cases came from and a fourth category grouping the remaining seven hospitals). An unconditional analysis was chosen because it is generally accepted that appropriate adjustment for the design variables in unconditional logistic regression yields similar results to those obtained through conditional logistic regression. We nevertheless checked this assumption by fitting the main crude and adjusted models using conditional logistic regression (6). The results were similar to those using unconditional regression. The measure of association used to compare the risk of developing esophageal cancer between exposed and unexposed subjects was the OR2 as an approximation of the relative risk. Potential risk factors for esophageal cancer were divided into five groups of variables: socioeconomic; tobacco smoking; alcohol consumption; mate drinking; and frequency of grouped food items consumption. Current smokers were defined as those subjects who smoked at the time of the interview and those who stopped smoking less than 1 year before the date of interview. Accordingly, ex-smokers were those who stopped smoking for at least 1 year before the date of interview. The same definition was applied for alcohol and mate-drinking status. For final risk estimates, a model was selected which simultaneously included those variables that showed a statistically significant effect in each group of risk factors. Tests for dose-response relationships were performed by coding variable categories in successive integers and testing the hypothesis that the parameter estimate for that variable was zero. Tests for linear trends were performed only in the final model once we adjusted for all the potential confounders. An a-value of 0.05 was used as the indicator of statistical significance, and accordingly 95% CIs are reported. The fitting of all logistic regression models was done using the software package GLIM (Numerical Algorithms Group, Oxford, UK) and for conditional logistic regression the software package EGRET (Cytel Software Corp.).
The study population A total of 292 controls
Table
1 15 1 16
(1 1 .4) (88.5)
24
(9.2)
238
(90.8)
(89.7%) agreed to reply to the questionnaire. Nonresponders were replaced to complete the ratio of two controls/case. No information was collected for nonrespondents. Table 1 summarizes the distribution of the main demographic and socioeconomic variables, as well as some risk factors by case-control status. Overall, the distribution of risk factors in this population is similar to that of other case-control studies of esophageal cancer. Thus, as cornpared to controls, cases of esophageal cancer tended to have received less education, have a lower self-reported monthly income, and come from rural areas. Cases were also more likely than controls to have drunk alcohol and to have smoked.
Effeds of Tobacco Smoking. Crude and adjusted ORs for tobacco smoking variables are shown in Table 2. Overall, tobacco smoking was moderately associated with esophageal cancer risk. After adjustment for the effects of education and alcohol consumption, ex-smokers and current smokers were about three times more likely than controls to develop esophageal cancer. No differences in risk were observed between ex-smokers and current smokers and between light smokers and smokers of 1 5 cigarettes or more/day. Similar results were observed when ex-smokers were defined as those who stopped smoking at least 6 years before the interview instead of 1 year. Neither duration nor amount of cigarettes smoked were statistically significant after adjusting for each other. In the crude analysis, smokers of black tobacco presented a statistically significant 2-fold increased risk of esophageal cancer as compared to smokers of blond
Downloaded from cebp.aacrjournals.org on July 13, 2011 Copyright © 1994 American Association for Cancer Research
Cancer
Table
2
Crude
and
adjus ted effec ts of tobacco
Cases Categories
Biomarkers
& Prevention
variables
Controls
N
(%)
No.
Nonsmoker
30
(23.0)
1 11
(42.4)
1 .0
Ex-smoker
41
(31 .3)
62
(23.7)
3.4
(1 .7-6.5)
2.8
(1 .4-5.6)
60
(45.8)
89
(34.0)
3.5
(1 .9-6.6)
2.9
(1 .5-5.6)
0
30
(23.4)
1 11
(43.0)
1 .0
1-14
28
(21.9)
50
(19.4)
2.8
(1.4-5.7)
3.0
(1.4-6.3)
15
70
(54.7)
97
(37.6)
3.8
(2.1-7.0)
2.9
(1.5-5.5)
(0.3-0.8)
0.6
Smoking
Average
smoker
ORc’
(95%
Cl)
ORa”
(95%,
CI)
1 .0
no. of cigarettes/day
Unknown Type
(‘/o)
status
Current
3
1 .0
4
of tobacco
Mainly
black
57
(58.2)
59
(40.1)
1.0
Mainly
blond
41
(41 .8)
88
(59.9)
0.5
Unknown Years
3
1.0 (0.3-1
.0)
4
of smoking
0
30
(22.9)
11 1
(42.4)
1 .0
1-29
26
(19.9)
28
(10.7)
5.0
(2.3-10.9)
5.1
12.2-11.8)
30-39
19
(14.5)
41
(15.7)
2.4
(1.1-5.4)
2.0
(0.9-4.6)
40
56
(42.8)
82
(31.3)
3.3
(1.8-6.3)
2.5
(1.3-4.9)
0-1
60
(45.8)
89
(34.0)
1 .0
2-7
16
(12.2)
16
8-19
12
(9.2)
29
20
13
(9.91
17
(6.5)
Nonsmoker
30
11 1
(42.4)
Years
a ORc, b
smo king
Epidemiology,
ORa,
since
quitting
smoking
crude OR (logistic regression model adjusted OR (adjusted for education
(22.9)
includes only the study and mi/day of alcohol,
(6.1) (11.1)
1 .0
1.6
(0.7-3.4)
1.5
(0.6-3.3)
0.6
(0.3-1.3)
0.5
(0.2-1.2)
1 .0
(0.4-2.3)
1 .3
(0.5-3.3)
0.3
(0.2-0.5)
0.4
(0.2-0.7)
design variables age, sex, and hospital). in addition to the design variables).
tobacco. After adjustment for education and alcohol consumption the increase persisted but was marginally significant. Although subjects who smoked for 1 -29 years presented a 5-fold increase in risk as compared to nonsmokers, no trend was observed with duration of tobacco smoking. The time elapsed since quitting smoking had no effect on esophageal cancer risk. Of the tobacco smoking-related variables in Table 2, the average number of cigarettes smoked/day was considered in the final model because it showed the largest reduction in deviance.
Effeds
1 .0
of Alcohol Consumption. Table 3 presents the resuits from the crude and adjusted analyses for alcoholrelated variables. Alcohol consumption was found to be positively associated with esophageal cancer risk. Thus, as compared to nondrinkers, current drinkers were about three times more likely to develop esophageal cancer. Ex-drinkers showed an intermediate OR of 2.2. The strongest positive association observed was, however, with the amount of pure ethanol consumed. Drinkers of 200 ml of ethanol/day were 8 times more likely to develop esophageal cancer than nondrinkers. As expected, lower alcohol consumption categories presented lower magnitudes or risk estimates. The effect of amount of alcohol consumed remained highly significant after adjusting for duration (P= 0.0001). Although duration of alcohol consumption was positively associated with esophageal cancer risk, the increase in risk was mainly due to the effects of ever-versus neverdrinking, and the effect of duration was no longer significant after adjusting for amount (using the likelihood ratio test, P = 0.3). Time since quitting alcohol drinking did not
significantly alter the subsequent risk of developing esophageal cancer. With regard to type of beverage, the only type of drink that showed a statistically significant effect on cancer risk was wine. Drinkers of wine had a 2-fold risk increase as compared to nondrinkers of wine. Wine is the most common alcoholic drink in this population. While 80% of subjects were wine drinkers, only 3-6% were drinkers of other types of alcoholic beverages (beer and liquor). Consumption of other alcoholic beverages such as spirits and beer was associated with an increased risk but because prevalences were very low (between 2 and 9%), OR estimates were not statistically significant (data not shown). Since amount of pure ethanol consumed was the alcohoi-reIated variable that showed the largest reduction in deviance and the strongest association with cancer risk it was included in the final model.
Joint Effects of Alcohol
and Tobacco. Table 4 shows the joint effects of alcohol and tobacco. Although the interaction between these two factors is not statistically significant (P= 0.45), higher risks are associated with higher exposures to both factors. Those drinking more than 200 ml of alcohoi/ day and smoking over 1 5 cigarettes/day have a risk 1 9 times higher than nonexposed subjects, and a higher risk than those exposed to alcohol alone (OR, 8.0) or tobacco alone (OR, 2.7), suggesting no departure from the multiplicative model. Effects of Mate Consumption. As shown in Table 5, virtually none ofthe mate-related variables were associated with oesophageal cancer risk. Drinkers of more than 1 .5 liters/ day had a nonsignificant increase in risk (OR, 1 .6; 95% Cl,
Downloaded from cebp.aacrjournals.org on July 13, 2011 Copyright © 1994 American Association for Cancer Research
559
560
Esophageal
Cancer
in Argentina
Table
3
Crude
and
adjusted
effects
of alcohol
Cases
consumption
variables
Controls
Categories No. Drinking
(%)
No.
(‘/o)
ORca
(95%
CI)
ORa”
(95%
Cl)
status
Nondrinker
12
(9.2)
59
(22.6)
1.0
Ex-drinker
23
(17.6)
45
(17.2)
2.9
(1.2-6.7)
2.2
(0.9-5.3)
96
(73.3)
157
(60.1)
3.5
(1.7-7.2)
2.9
(1.4-6.1)
Current
drinker
Alcohol
amount
1.0
(mI/day)
0
12
(9.4)
59
122.9)
1.0
1-99
56
(43.7)
144
(55.8)
2.4
(1.1-5.0)
2.0
100-199
33
(25.8)
31
(12.0)
7.6
(3.2-18.4)
5.5
12.2-13.7)
200
27
(21.1)
24
(9.3)
8.4
13.3-21.3)
8.0
13.0-21.2)
1.6-13.1
4.0
Unknown Alcohol
3 duration
1.0 (0.9-4.2)
4
(years)
0
12
19.2)
59
(22.7)
1.0
1-29
16
112.3)
23
(8.8)
4.6
30-39
24
118.5)
43
(16.5)
3.5
(1.4-8.4)
2.8
(1.1-6.9)
40
78
(60.0)
135
(51.9)
3.0
(1.4-6.3)
2.4
(1.1-5.1)
Unknown Years
1
since
quiting
1.0 (1.4-11.6)
2
drinking
Nondrinker
12
(9.2)
59
(22.6)
1.0
8
11
(8.4)
19
(7.3)
3.2
(1.2-8.6)
2.5
(0.9-7.2)
2-7
12
(9.2)
26
(10.0)
2.6
(1.0-6.9)
2.0
(0.7-5.5)
Current
96
(73.3)
157
(60.2)
3.5
(1.7-7.2)
2.9
(1.4-6.0)
(1.4-5.2)
2.3
Unknown
1.0
1
Wine Nondrinker Ever-drinker a
h
ORc, ORa,
crude OR (logistic regression model adjusted OR (adjusted for education
no. of
cigarettes/day
0
(11.5)
63
(24.1)
1.0
188.5)
199
(75.9)
2.7
includes only the design variables age, and average number of cigarettes/day,
Table
Average
15 116
4
Combined
0 OR3
195%
Cl)
OR
Cases/controls
1.0(reference)
1.6 (0.6-4.2)
0.7(0.1-6.5) 1/15
AIV
1 .0 (reference) 12/59
of cigarette
smoking
Average
consumption
CI)
OR
ethanol
consumption”
of ethanol
(mI/day)
OR
13.3
OR (95% Cl) Cases/controls 1.0(reference)
4/7
(2.8-62.3)
29/1 00
5.0(0.6-39.1)
2.8 (1.3-5.8)
2/3
28/50
11.8(3.7-37.7)
19.0(5.4-66.9)
2.7(1.4-5.2)
23/19
21/11
68/96
23/51
5.5 (2.2-1
55/144
3.7)
8.0 (3.1-21
31/31
=
All” Cl)
5.7 (1.1-28.6)
7/5
2.0 (0.9-4.2)
(95%
Cases/controls
1/7
1.3-11.0)
0.5-4.5) which disappeared after adjusting for smoking. Although there was a marginally significant positive effect of mate temperature consumption, the association was not strong (OR, 1 .7; 95% Cl, 1 .0-2.9). Subjects who always drank the first mate, as compared to those who threw it away or drank it only sometimes, had a significantly decreased risk of developing esophageal cancer (OR, 0.5; 95% Cl, 0.3-0.9). No variables from this group were considered in the final model.
Cl)
1.2 (0.1-12.01
1 7/36
a Analysis restricted to subjects with known information for both exposures. In P= 0.45. ‘, Summary effect of cigarette smoking (OR adjusted for ethanol consumption). C Summary effect of ethanol consumption (ORs adjusted for cigarette smoking). d Body table ORs are adjusted for design variables and education.
200
(95%
Cases/controls
4.3 (1.4-13.2) 3.7
and
(1.2-4.4)
variables).
100-199
1 5/57
2/6 15
195%
Cases/controls
2.5 (0.4-16.5)
sex, and hospital). in addition to the design
1-99
9/38 1-14
effects
1.0
380).
.2)
n = 380
27/21
Test for interaction
between
125/255
ethanol
consumption
and
cigarette
smoking,
Among controls mate consumption was not associated with level of education or income. However, mate consumption was associated with diet; most heavy mate drinkers (drinking over 1 liter/day) ate beef daily and fruits frequently but ate barbecued meat and raw and cooked vegetables infrequently (data not shown).
Effects those
of Frequency food
items
that
of Food Items. were
significantly
Downloaded from cebp.aacrjournals.org on July 13, 2011 Copyright © 1994 American Association for Cancer Research
Table 6 summarizes or marginally
asso-
Cancer
Table
S
Crude
and
adjusted
Cases
effects
of mate
consumption
Epidemiology,
Biomarkers
& Prevention
variables
Controls
Categories No. Mate
drinking
ORc”
(95%
CI)
ORa”
(95’!,,
15
(11.4)
24
(9.2)
1.0
(11.4)
14
(5.3)
1.7
(0.6-4.6)
1.7
(0.6-5.0)
101
(77.1)
224
(85.5)
0.7
(0.4-1.41
0.5
(0.2-1.1)
0
15
(11.4)
24
(9.2)
1.0
0.01-0.4
42
(32.1)
66
(25.2)
1.1
(0.5-2.3)
1.0
(0.4-2.3)
0.5-0.9
40
130.5)
125
(47.7)
0.5
(0.2-1.1)
0.4
10.2-0.8)
1.0-1.4
23
(17.6)
35
(13.4)
1.1
(0.5-2.5)
0.8
(0.3-2.1)
1
11
(8.4)
12
(4.6)
1 .6
(0.5-4.5)
1 .0
10.3-3.3)
15
(1 1 .4)
24
(9.2)
1 .0
5
(3.8)
7
(2.7)
1.3
(0.3-4.9)
1.1
(0.2-5.1)
30-39
16
(12.2)
33
(12.6)
0.9
(0.3-2.5)
0.8
(0.3-2.3)
40-49
27
(20.6)
52
(19.8)
0.9
(0.4-2.1)
0.6
(0.2-1.7)
50
68
(51.9)
146
(55.7)
0.7
(0.4-1.5)
0.5
(0.2-1.2)
42
(36.8)
121
(51.0)
1.0
72
163.2)
1 16
148.91
1 .8
amount
1.0
Iliters/day)
.5 duration
1.0
(years)
0 1-29
Mate
Cl)
status 15
Mate
1 .0
temperature consumption
Warm Hot/very
hot
Unknown ORc, ORa,
1%)
Ex-drinker
Mate
b
No.
Nondrinker Currentdrinker
a
1%)
2
crude OR (logistic regression model adjusted OR ladjusted for education,
1.0 (1 .1-2.9)
1 .7
(1.0-2.9)
1 includes average
only the design variables number of cigarettes/day,
age, sex, and hospital). and alcohol consumption
(mI/day),
in addition
to the design
variablesl.
ciated with esophageal cancer risk either in the crude or in the adjusted analyses. Consumption of beef was found to be inversely associated with esophageal cancer risk but the protective effect lost significance after adjusting for other dietary variables. A positive effect was found for subjects who ate barbecued meat 1 or more times/week as compared to those who ate it less than weekly (OR, 2.1 ; 95% CI, 1 .0-4.4). Frequent consumption of raw vegetables was associated with a slight decrease in esophageal cancer risk but the effect was not significant. As shown in Table 6, consumption of oil, poultry, dairy products, citrus and noncitrus fruits, and sweets was not associated with esophageal cancer after adjustment for confounding variables; neither was consumption of sausage, eggs, and cereals (data not shown). Of all the food-related variables that showed a significant effect on risk, only beef and barbecued meat remained significantly associated with esophageal cancer after simultaneously adjusting for the effect of other relevant food items, and, consequently, both were considered in the final model.
Three variables remained significantly associated with esophageal cancer risk in the final model: tobacco; ethanol; and barbecued meat. As noted in Table 7, there was a clear, highly significant trend of average mi/day of ethanol consumed with esophageal cancer risk (P = 0.00003). Although the test for trend of average number of cigarettes smoked/day was significant (P = 0.001), no clear dose-response relationship was observed among the smokers. These findings indicate that ethanol consumption and to a lesser extent cigarette smoking were the two factors most strongly associated with the risk of developing esophageal cancer. The potential modifying effects (i.e., statistical interaction) of tobacco smoking, alcohol consumption, and mate drinking were formally tested against each other but were found to be not significant. Thus, the risk estimates for subjects presenting different combinations of these risk factors are not presented.
Final Model.
This study identified alcohol, tobacco, and barbecued meat as the main risk factors for esophageal cancer in La Plata, Argentina, in agreement with the findings of similar studies carried out earlier in Brazil and Uruguay (4-5) and in a study on precancerous lesions of the esophagus carried out in the same study population in Argentina (7). However, contrary to the findings in Uruguay, no association was detected with drinking of hot mate. The lack of effect of mate drinking even in those drinking over 1 .5 liters/day is surprising. The fact that there was a moderate increase in risk in this category (OR, 1 .6; 95% Cl, 0.5-4.5), which disappeared after adjusting for tobacco (OR, 1 .0; 95% CI, 0.3-3.3) indicates that the association is partially confounded by this variable. Those who reported drinking the mate hot or very hot as compared to those
A final
model was chosen that included, in addition to the design variables, education, the average number of cigarettes smoked/day, the average amount of ethanol consumed/day, and the frequency of consumption of barbecued meat and beef. Table 7 presents the final ORs and 95% CIs for each variable. To check for negative confounding, the effects of each mate variable were tested again using this final model. None of the mat&related variables showed any statistically significant effect after adjusting for all the variables associated with esophageal cancer (data not shown). The marginally protective effect of blond tobacco found in the adjusted analysis was still present when included in this final model. However, the OR was not statistically significant (data not shown).
Discussion
Downloaded from cebp.aacrjournals.org on July 13, 2011 Copyright © 1994 American Association for Cancer Research
561
562
Esophageal
Cancer
in Argentina
Table
6
Crude
a nd adjust
ed effects
of select ed food
frequen
cy variables
Controls
Cases Variables No.
No.
(%)
1%)
ORc”
(95%
CI)
ORa!’
(95’/o
CI)
Beef )
81
(31 .0)
1 .0
169.0)
0.5
