Influence of mate drinking, hot beverages and diet on esophageal cancer risk in south america

Authors


Abstract

To estimate the effects of consuming hot beverages, including mate (an infusion of the herb Ilex paraguayensis), tea, coffee and coffee with milk, and other food items on esophageal cancer risk, we analyzed data from 830 cases and 1,779 controls participating in a series of 5 hospital-based case-control studies of squamous-cell carcinoma of the esophagus conducted in high-risk areas of South America. After adjusting for the strong effects of tobacco and alcohol consumption, both heavy mate drinking (>1 l/day) and self-reported very hot mate drinking were significantly associated with esophageal cancer risk in men and women. The magnitude and strength of the association for mate amount and, to a lesser extent, mate temperature were higher for women than men. The joint effects of mate amount and mate temperature were more than multiplicative, following a statistically significant synergistic interaction (p = 0.02) which was particularly evident among heavy drinkers (>1.50 l/day) of very hot mate (odds ratio = 4.14, 95% confidence interval: 2.24–7.67) compared to light drinkers (<0.50 l/day) of cold/warm/hot mate. Consumption of other very hot beverages, such as tea and coffee with milk but not coffee alone, was also significantly associated with an increased risk, in the 2- to 4-fold range. Statistically significant protective associations were identified for high consumption of vegetables, fruits, cereals and tea. In contrast, frequent consumption of meat, animal fats and salt was associated with a moderately increased risk. This pooled analysis adds evidence for a carcinogenic effect of chronic thermal injury in the esophagus induced by the consumption of very hot drinks, including mate. Our study further confirms the protective effect of a dietary pattern characterized by daily consumption of fruits and vegetables and low consumption of meat and animal fats. Int. J. Cancer 88:658–664, 2000. © 2000 Wiley-Liss, Inc.

The descriptive epidemiology of esophageal cancer is characterized by a wide geographical variability in incidence and mortality rates. Striking differences have been reported not only between world regions and countries but also within smaller areas (Muñoz and Castellsagué, 1994; Muñoz and Day, 1996). There is a geographic cluster of high-incidence areas in South America, which covers northeastern Argentina, southern Brazil, Uruguay and Paraguay (Vassallo et al.,1985; Parkin, 1994; Parkin et al.,1999).

Populations in these high-risk areas share the habit of drinking a local tea known by the folk name of “mate”, an infusion of the herb Ilex paraguayensis that is usually drunk warm, hot or very hot from a gourd or other container through a straw, which places very hot fluid at the oropharynx and esophagus. It has long been hypothesized that hot mate drinking may partially explain this cluster of high-incidence rates in South America.

Driven by these epidemiological observations, The World Health Organization/International Agency for Research on Cancer (IARC, Lyon, France) conducted a series of 5 case-control studies in Argentina, Brazil, Uruguay and Paraguay to assess the role of mate drinking in the etiology of esophageal cancer, taking into account the effects of tobacco and alcohol consumption, the 2 strongest risk factors for esophageal cancer in most Western countries and certain parts of Asia and Africa (IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 1986, IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 1988; Muñoz and Castellsagué, 1994).

However, while these studies identified smoking and drinking as the key risk factors for esophageal cancer, the effects of the mate-related variables yielded less consistent results. This was probably due to the fact that these studies were too small to assess comprehensively the relationship between mate drinking and esophageal cancer risk. In addition, because heavy mate drinking is highly prevalent in each of these populations, either the exposure distribution used in each separate study was skewed toward the high-consumption categories or, in the studies in which tertiles or quartiles were used to categorize dose, the referent groups were “contaminated” by exposed individuals.

Since the 5 studies used the same research protocol and data-collection procedures, we were able to combine the data, providing a unique opportunity to comprehensively quantify the effects of established and putative risk factors for esophageal cancer in a high-risk area. The main results of the pooled analysis have already been reported (Castellsagué et al.,1999, Castellsagué et al.,2000), identifying tobacco and alcohol consumption as the main risk factors for esophageal cancer, even in the absence of each other (Castellsagué and Muñoz, 1999). Here, we report on the specific adjusted effects of mate drinking and consumption of other hot beverages and food items on the risk of developing esophageal cancer in these populations.

MATERIAL AND METHODS

Between 1986 and 1992, 4 hospital-based case-control studies of esophageal cancer were conducted in Argentina, Brazil, Paraguay and Uruguay to identify risk factors that might explain the high incidence rates of the disease in these areas. The main study, in Uruguay, was eventually extended, contributing an additional smaller study. The 5 studies were designed and coordinated by The IARC.

The design, methods of investigation and principal results for each of the 4 main studies have been reported (Victora et al.,1987; De Stefani et al.,1990; Castelletto et al.,1994; Rolón et al.,1995). Similar research protocols and questionnaires, adapted to the local situation, were used for the 5 studies, allowing for a valid pooled analysis of the data. The methods for the pooled analysis and the results on smoking and drinking for the combined data have also been reported (Castellsagué and Muñoz, 1999; Castellsagué et al.,1999, Castellsagué et al.,2000). The recruitment periods, the number of cases and controls and the key exposure information on mate drinking are summarized in Table I.

Table I. STUDY AND SUBJECTS' CHARACTERISTICS AND EFFECT OF SELECTED MATE DRINKING VARIABLES ON ESOPHAGEAL CANCER RISK BY STUDY
Study and subjects' characteristicsStudyTotal
ArgentinaBrazilParaguayUruguay 1Uruguay 2
  • 1

    Adjusted for age group, hospital, residency, years of education, gender, average number of cigarettes smoked/day and average amount of pure ethanol consumed/day.

  • 2

    OR is for hot vs. cold/warm mate drinking as there were too few or no subjects exposed to very hot.

Recruitment period4/86 to 12/884/85 to 2/861/88 to 3/917/85 to 9/881/89 to 2/924/85 to 2/92
Number of subjects (cases/controls)131/262171/342131/381261/522136/272830/1,779
Gender (% of male subjects)80.978.984.076.277.278.9
Mean age (years)63.559.562.266.766.664.0
Mate prevalence (% of current drinking among controls)86.367.081.973.874.376.1
Mate amount consumed (daily mean liters among controls)0.630.930.751.041.050.90
Mate temperature (% hot/% very hot among controls)48.5/0.469.8/26.765.3/088.8/6.181.9/6.773.3/7.5
Effect of mate temperature OR (95% CI)1 (very hot vs. cold/warm/hot)1.942 (1.15–3.25)0.72 (0.42–1.26)1.642 (0.89–3.00)3.05 (1.79–5.21)3.10 (1.44–6.68)1.76 (1.27–2.44)
Effect of mate amount OR (95% CI)1 (>1 l vs. 0.1–0.5 l)1.73 (0.80–3.74)0.93 (0.51–1.71)0.88 (0.43–1.80)2.76 (1.72–4.43)2.38 (1.24–4.58)1.57 (1.22–2.03)

Cases

Case eligibility criteria for the 5 studies included histologically confirmed squamous-cell carcinoma of the esophagus (in Paraguay a cytological or radiological diagnosis of esophageal cancer was also acceptable), diagnosis within the previous 3 to 6 months, residence in the area for at least 5 years and sufficiently good physical and mental health to give reasonably reliable answers to the questionnaires.

In Argentina, cases were identified in any of the 10 main hospitals of greater La Plata (population 620,000). In Brazil, cases were identified in any of the 8 main hospitals and in any of the 3 radiotherapy units of Port Alegre and Pelotas, the 2 largest cities in the state of Rio Grande do Sul. In Uruguay, cases were identified in any of the 4 main hospitals in Montevideo, which cover about 45% of the population of Montevideo and about 55% of the population of the rest of the country. In Paraguay, subjects were identified in any of the 4 hospitals, private clinics, pathology laboratories and radiology clinics in Asuncion. Participation rates among eligible cases were 90.0% in Brazil, 97.4% in Uruguay, 98.5% in Argentina and 99.2% in Paraguay.

Controls

Controls were identified and recruited from the same hospitals where the cases were identified. In Argentina, Brazil and Uruguay, 2 controls were selected for each eligible case. In Paraguay, the control:case ratio was 3. Eligibility criteria for controls included matching by gender and age (±5 years), admission to the same hospital and during the same period as the corresponding case, residence in the area for at least 5 years and sufficiently good physical and mental health to give reasonably reliable answers to the questionnaires. Patients with diseases related to alcohol or tobacco were not accepted as controls. The distribution of the main diagnostic categories among controls has been published for each study (Victora et al.,1987; De Stefani et al.,1990; Castelletto et al.,1994; Rolón et al.,1995). In Brazil, Uruguay, Argentina and Paraguay, if a selected control refused to participate, a replacement with the same matching criteria was used. In Paraguay, no replacements were accepted, but the participation rate among eligible controls was very high (97.0%).

Questionnaire

All eligible cases and controls were interviewed at the hospitals using a face-to-face, pre-tested, standardized questionnaire administered by specially trained interviewers. Proxy interviews were not accepted.

The questionnaire elicited detailed information on demographic and socio-economic characteristics and on lifetime habits of tobacco smoking, alcohol drinking and consumption of hot beverages including mate, tea, coffee and coffee with milk (daily amount consumed, duration and cessation periods and self-reported drinking temperature). Consumption temperature of hot drinks was self-reported. Subjects were simply asked the temperature at which the corresponding hot drink was usually drunk: cold, warm, hot or very hot. Dietary habits were assessed for 2 time periods, the period just before the onset of symptoms and 10 years before admission to the hospital. A food-frequency questionnaire with 50 food items was used to assess the consumption of 10 primary food groups: potatoes, vegetables, fruits, fats, dairy products, fresh meat, processed meat, eggs, cereals, and barbecue. The results for recent and past consumption were virtually the same. Thus, only those on recent consumption are reported.

Statistical analyses

A single data file including all common comparable items of information for each study was compiled, and the following variables were computed: average number of cigarettes smoked/day and average consumption amounts and duration of pure ethanol, hot mate, tea, coffee and coffee with milk consumption. For ex-drinkers of alcohol and for ex-drinkers of mate, years since stopping the habit was also computed. The consumption frequencies of the 50 food items included in the questionnaire were appropriately grouped to yield the consumption frequency of the above-mentioned 9 food groups. Data were analyzed using unconditional multiple logistic-regression models with maximum likelihood estimation of parameter values to obtain odds ratios (ORs) as an approximation of relative risk (Breslow and Day, 1980). Unconditional analysis was performed since the results obtained using conditional regression in each separate study were practically identical to those obtained with the unconditional approach, except for Brazil, which reported the results from the conditional analysis. It is generally accepted that in a matched case-control study unconditional analysis can be performed if appropriate adjustment is made for the matching variables. Thus, all logistic-regression models included as co-variates the 3 matching variables: age (4 categories: ≤54 years, 55–64 years, 65–74 years, ≥75 years), gender and hospital group (the 26 participating hospitals were classified into 12 groups). All risk estimates were further adjusted by residency (rural/urban), years of schooling completed (0–3 years, 4–6 years and ≥7 years), average number of cigarettes smoked/day and average amount of pure ethanol consumed/day.

Effect modification and subgroup analyses within and between exposure variables were assessed by including in the fully adjusted model the main terms for each variable and the corresponding interaction term.

The Generalized Linear Interactive Modelling (GLIM) package was used to obtain maximum likelihood estimates from main-term and interaction-term models. For multiple levels of exposure, tests for dose-response relationships were performed by coding the corresponding variable categories in ordinal integers and using the likelihood ratio test statistic with 1 degree of freedom to test the hypothesis that the parameter estimate for that variable was 0.

Statistical significance was set at 0.05, and 95% confidence intervals (CIs) around the OR are presented. All p values were derived from 2-sided statistical tests.

RESULTS

Subjects' characteristics by study

Table I shows, for each study, the recruitment periods and the distribution of participants (830 cases and 1,779 controls) according to selected exposure characteristics. The effects of mate temperature and amount are also shown by study and overall. For mate temperature, statistically significant associations were found in Argentina and the 2 studies in Uruguay. A positive association was also found in Paraguay, but it was not statistically significant. The OR for mate temperature in Brazil was slightly protective but far from statistically significant.

For mate amount, the same pattern was observed: strong associations in Uruguay, a moderate association in Argentina and no association in Brazil and Paraguay. The pooled OR estimate for self-reported drinking temperature (very hot vs. lower temperatures) was 1.76 (95% CI 1.27–2.44). The pooled OR estimate for amount (>1 l vs. <0.5 l) was 1.57 (95% CI 1.22–2.03).

Effects of mate-related variables

Table II shows the adjusted effects of selected mate-related variables for men and women.

Table II. ADJUSTED EFFECTS OF MATE-RELATED VARIABLES ON ESOPHAGEAL CANCER RISK BY GENDER AND OVERALL
Variable and categoriesMalesFemalesAll
Number of cases/controlsOR1 (95% CI)Number of cases/controlsOR1 (95% CI)OR2 (95% CI)
  • Bold figures indicate statistical significance at the 0.05 level.

  • 1

    Adjusted for age group, hospital, residency, years of education, average number of cigarettes/day, average amount of pure ethanol/day.

  • 2

    Further adjusted for gender.

Mate status     
   Never48/1741.0012/591.001.00
   Ever607/1,2341.34  (0.92–1.96)163/3122.20  (1.08–4.47)1.52  (1.10–2.12)
   Ex93/1391.91  (1.20–3.06)22/531.74  (0.73–4.11)1.87  (1.25–2.80)
   Current514/1,0951.27  (0.87–1.87)141/2592.30  (1.13–4.71)1.47  (1.06–2.05)
   Trend test p = 1.00 p = 0.01p = 0.20
Mate amount     
(l/day)
   048/1741.0012/591.001.00
   0.01–0.50173/4161.23  (0.81–1.86)59/1421.93  (0.92–4.08)1.39  (0.98–1.98)
   0.51–1.00232/5471.20  (0.80–1.79)51/1221.86  (0.85–4.05)1.34  (0.95–1.90)
   1.01–1.5072/1141.69  (1.03–2.77)16/204.03  (1.47–11.04)1.96  (1.27–3.03)
   1.51–2.0073/1051.69  (1.03–2.77)23/223.95  (1.53–10.20)2.03  (1.32–3.13)
   >2.0055/492.30  (1.32–4.03)13/511.65  (2.96–45.79)3.04  (1.84–5.02)
   Unknown2/3 1/1  
   Trend test among ever-drinkers p = 0.004 p = 0.001p = 0.0001
Years of drinking     
   048/1741.0012/591.001.00
   1–2974/1901.23  (0.76–2.00)19/381.83  (0.69–4.88)1.40  (0.91–2.13)
   30–39105/2241.32  (0.85–2.07)16/471.28  (0.49–3.37)1.39  (0.93–2.07)
   40–49174/3471.38  (0.91–2.10)33/702.10  (0.91–4.87)1.53  (1.06–2.21)
   50–59146/3091.27  (0.81–1.98)46/802.39  (1.06–5.37)1.47  (1.00–2.17)
   ≥60107/1621.61  (0.96–2.72)48/762.71  (1.18–6.22)1.92  (1.25–2.96)
   Unknown1/2 1/1  
   Trend test among ever-drinkers p = 0.30 p = 0.10p = 0.11
Years since quitting     
   0514/1,0951.00141/2591.001.00
   1–950/671.78  (1.16–2.74)16/281.11  (0.55–2.25)1.53  (1.07–2.19)
   ≥1035/671.07  (0.67–1.69)6/220.39  (0.14–1.09)0.90  (0.59–1.36)
   Unknown8/5 0/3  
   Trend test p = 0.19 p = 0.16p = 0.59
Mate temperature     
   Cold/warm103/2291.0024/671.001.00
   Hot426/9091.06  (0.77–1.45)110/2181.33  (0.71–2.49)1.11  (0.84–1.47)
   Very hot73/911.77  (1.09–2.86)26/242.47  (1.03–5.91)1.89  (1.24–2.86)
   Unknown5/5 3/3  
   Trend test p = 0.046 p = 0.05p = 0.008

Some, but not all, measures of mate drinking were significantly related to risk. Ever mate drinking increased the risk 1.5-fold compared with never-drinkers. Mate amount was strongly related to risk, with a statistically significant dose-response relationship that was particularly evident for drinkers of more than 1 l/day. This pattern was stronger in women than in men. Thus, women drinking more than 2 l of mate/day had an 11.6-fold increased risk compared to never-drinkers. Among men, the corresponding estimate was 2.30 (95% CI 1.32–4.03). Subjects self-reporting drinking mate at very hot temperatures had an almost 2-fold increased risk compared with drinkers of cold or warm mate. This effect was also higher in women than in men. Moderate effects were detected with duration, and a borderline protective effect with years since quitting mate drinking was found only among women. Because mate amount and temperature had independent effects on risk, even adjusting for each other, we further explored the interaction between these 2 variables to estimate the effect of amount in each stratum of temperature and the effect of temperature in each stratum of amount (Table III). The joint effects of mate amount and temperature were more than multiplicative, following a statistically significant synergistic interaction (p = 0.02). The increased risk was particularly evident among heavy drinkers (>1.50 l/day) of very hot mate, in whom the risk was increased more than 4-fold (OR = 4.14, 95% CI 2.24–7.67) compared with light drinkers (<0.50 l/day) of cold/warm/hot mate. The trend test of mate amount was statistically significant in each stratum of mate temperature (p = 0.01 among drinkers of cold/warm/hot mate, p = 0.002 among drinkers of very hot mate). The overall effect of mate amount after adjusting for the effects of mate temperature and tobacco and alcohol consumption was moderate, but the trend test was highly significant (p = 0.0006). The overall effect of mate temperature was also moderate (OR = 1.62), but it was particularly increased in the highest consumption category (OR = 4.14), suggesting that temperature acts synergistically when the volume consumed is also high (>1.50 l/day).

Table III. JOINT EFFECTS OF AMOUNT AND TEMPERATURE OF MATE DRINKING ON ESOPHAGEAL CANCER RISK1
Mate drinking amount (daily liters)Mate drinking temperatureSummary OR for amount3
Cold/warm/hotVery hotp for trend (2-sided)
  • 1

    First number indicates OR (95% CI) for esophageal cancer adjusted for age group, gender, hospital, years of schooling, residency, average amount of pure ethanol consumed/day and average number of cigarettes smoked/day. Bold figures denote statistical significance at the 0.05 level. Bottom number indicates number of cases/controls in that combined exposure category. p value for the interaction between mate amount and mate temperature = 0.016.

  • 2

    Further adjusted for mate amount.

  • 3

    Further adjusted for mate temperature.

≤0.501.0  (reference)0.99  (0.48–2.02)0.971.0  (reference)
212/51814/33
0.51–1.000.91  (0.71–1.16)1.59  (0.96–2.63)0.030.95  (0.75–1.20)
243/62139/48
1.01–1.501.50  (1.05–2.14)0.73  (0.24–2.26)0.21.40  (0.99–1.98)
83/1195/14
> 1.501.38  (1.00–1.90)4.14  (2.24–7.67)0.00071.58  (1.17–2.14)
124/16240/19
p for trend (2-sided)0.010.002 0.0006
Summary OR for temperature21.0  (reference)1.62  (1.16–2.25)0.006 

We further explored the combined effects of hot mate drinking and tobacco and alcohol consumption by fitting appropriate logistic regression models with interaction terms for these variables. The effect of drinking large amounts of hot mate was somewhat higher among persons unexposed or lightly exposed to tobacco and alcohol. However, the interactions between mate temperature and amount with tobacco and alcohol consumption were not statistically significant.

Effects of consumption of other hot beverages

Table IV shows the adjusted effects of consumption of other hot beverages on esophageal cancer risk by gender and overall. While tea consumption was associated with a protective effect, drinking tea and coffee with milk (but not coffee) at a very hot temperature was significantly associated with an increased risk. Thus, the adjusted OR for drinking very hot tea was 3.73 (95% CI 1.41–9.89) compared with drinkers of cold/warm/hot tea. The corresponding OR for drinking very hot coffee with milk was 2.20 (95% CI 1.37–3.81).

Table IV. ADJUSTED EFFECTS OF OTHER HOT BEVERAGES ON ESOPHAGEAL CANCER RISK BY GENDER AND OVERALL1
Variable and categoriesMalesFemalesAll
Number of cases/controlsOR1 (95% CI)Number of cases/controlsOR2 (95% CI)OR3 (95% CI)
  • 1

    Bold figures indicate statistical significance at the 0.05 level. Subjects in Uruguay-2 study were excluded from these analyses as data on other hot beverages were not collected.

  • 2

    Adjusted for age group, hospital, residency, years of education, average number of cigarettes/day and average amount of pure ethanol/day.

  • 3

    Further adjusted for gender.

Tea status and amount (ml/day)
   Never418/8891.0093/1861.001.00
   Ever132/3090.80  (0.58–1.10)51/1230.74  (0.45–1.22)0.81  (0.62–1.06)
      Amount     
         1–50073/1670.95  (0.67–1.35)30/810.69  (0.40–1.19)0.90  (0.67–1.20)
         500+59/1420.53  (0.32–0.88)21/420.96  (0.39–2.34)0.62  (0.40–0.96)
         Trend test p = 0.045 p = 0.41p = 0.04
Coffee status and amount (ml/day)
   Never365/8471.00103/2271.001.00
   Ever185/3511.09  (0.84–1.41)41/820.89  (0.53–1.49)1.04  (0.83–1.30)
      Amount     
         1–500122/2421.05  (0.79–1.40)27/620.70  (0.38–1.26)0.96  (0.74–1.24)
         500+63/1091.19  (0.80–1.78)14/201.68  (0.72–3.93)1.26  (0.88–1.81)
         Trend test p = 0.40 p = 0.78p = 0.39
Coffee with milk status and amount (ml/day)
   Never287/6231.0065/1491.001.00
   Ever263/5751.09  (0.86–1.38)79/1601.44  (0.90–2.29)1.15  (0.94–1.42)
      Amount     
         1–500223/4951.08  (0.84–1.39)65/1421.32  (0.81–2.16)1.12  (0.90–1.40)
         500+40/801.14  (0.72–1.81)14/182.12  (0.91–4.93)1.31  (0.89–1.95)
         Trend test p = 0.45 p = 0.07p = 0.12
Tea temperature     
   Cold/warm19/471.008/181.001.00
   Hot32/1050.85  (0.37–1.95)19/580.58  (0.16–2.07)0.66  (0.35–1.25)
   Very hot13/58.73  (1.95–39.10)7/72.20  (0.42–11.56)3.73  (1.41–9.89)
   Unknown68/152 17/0  
   Trend test p = 0.045 p = 0.47p = 0.11
Coffee temperature     
   Cold/warm45/591.003/141.001.00
   Hot117/2540.44  (0.26–0.76)29/581.50  (0.27–8.51)0.54  (0.33–0.87)
   Very hot23/380.76  (0.35–1.64)9/103.46  (0.45–26.57)1.01  (0.52–1.98)
   Trend test p = 0.15 p = 0.18p = 0.60
Coffee with milk temperature
   Cold/warm60/1171.0012/281.001.00
   Hot158/4080.87  (0.57–1.32)48/1141.05  (0.43–2.58)0.89  (0.62–1.29)
   Very hot45/502.22  (1.20–4.10)19/172.82  (0.94–8.45)2.29  (1.37–3.81)
   Unknown  0/1  
   Trend test p = 0.058 p = 0.05p = 0.009
Any very hot beverage (other than mate)
   Never very hot319/7561.0085/2121.001.00
   Ever very hot62/682.28  (1.48–3.50)28/243.21  (1.66–6.23)2.45  (1.72–3.49)
   Unknown34/72 2/21  
Any very hot beverage (including mate)
   Never very hot442/1,0491.00112/2641.001.00
   Ever very hot103/1182.10  (1.49–2.96)32/352.18  (1.20–3.94)2.07  (1.55–2.76)
   Unknown5/31 0/10  

When combining the effect of all hot beverages, associations were found for drinking any very hot beverage both including mate (OR = 2.07, 95% CI 1.55–2.76) and excluding mate (OR = 2.45, 95% CI 1.37–3.81).

Effects of dietary factors

Table V shows the adjusted effects of current consumption of selected food groups on esophageal cancer risk. Protective associations were found with frequent consumption of tea (Table IV), fruits, vegetables and, to a lesser extent, cereals. Daily/almost daily consumption of fruits was associated with a 63% risk reduction compared with subjects rarely consuming fruits, a protective association that followed a strong dose-response relationship (p < 0.00001). In contrast, frequent consumption of meat, fat and salt was associated with a moderately increased risk.

Table V. ADJUSTED EFFECTS OF CONSUMPTION OF SELECTED FOOD GROUPS ON ESOPHAGEAL CANCER RISK BY GENDER AND OVERALL1
Variable and categoriesMalesFemalesAll
Number of cases/controlsOR2 (95% CI)Number of cases/controlsOR2 (95% CI)OR3 (95% CI)
  • 1

    Bold figures indicate statistical significance at the 0.05 level. No association was found for consumption of milk, sausage, eggs or potatoes.

  • 2

    Adjusted for age group, hospital, residency, years of education, average number of cigarettes/day and average amount of pure ethanol/day.

  • 3

    Further adjusted for gender.

  • 4

    Subjects in the Uruguay-2 study were excluded from these analyses.

  • 5

    Includes subjects from Brazil and Uruguay-1.

Meat     
   Never or 1–3/week105/3231.0044/1011.001.00
   Almost daily, daily550/1,0851.46  (1.11–1.92)131/2700.98  (0.62–1.53)1.28  (1.02–1.61)
Cereals4     
   Never, rarely55/681.0014/181.001.00
   1–3/week or month90/2450.61  (0.37–1.02)23/590.50  (0.18–1.40)0.63  (0.40–0.98)
   Almost daily, daily405/8850.61  (0.33–1.12)107/2320.49  (0.14–1.66)0.64  (0.38–1.10)
   p for trend 0.17 0.360.19
Vegetables     
   Never, rarely71/1021.0015/201.001.00
   1–3/week or month352/7810.56  (0.39–0.82)94/1890.62  (0.28–1.39)0.60  (0.43–0.84)
   Almost daily, daily232/5250.64  (0.44–0.95)66/1620.48  (0.21–1.08)0.62  (0.44–0.88)
   p for trend 0.32 0.070.08
Fruits     
   Never, rarely114/891.0015/191.001.00
   1–3/week or month282/5800.39  (0.27–0.55)60/1290.64  (0.28–1.45)0.43  (0.31–0.58)
   Almost daily, daily259/7390.31  (0.22–0.45)100/2230.68  (0.30–1.52)0.37  (0.27–0.51)
   p for trend <0.00001 0.67<0.00001
Fat     
   Never, rarely89/2211.0030/611.001.00
   1–3/week or month188/3831.18  (0.84–1.65)49/1140.63  (0.33–1.20)1.03  (0.77–1.38)
   Almost daily, daily378/8041.57  (1.09–2.26)96/1961.09  (0.56–2.10)1.42  (1.03–1.94)
   p for trend 0.015 0.820.03
Barbecue     
   Never, rarely126/3071.0058/931.001.00
   1–3/week or month432/9900.95  (0.72–1.26)106/2590.57  (0.36–0.89)0.84  (0.66–1.06)
   Almost daily, daily96/1101.44  (0.95–2.18)11/190.87  (0.35–2.15)1.22  (0.85–1.76)
   Unknown1/1    
   p for trend 0.17 0.080.73
Salt5     
   1–3/week or month240/5691.0089/1801.001.00
   Almost daily, daily93/992.11  (1.48–3.01)9/141.17  (0.45–3.04)1.94  (1.40–2.69)

As shown in Table V, this dietary pattern (frequent consumption of fruits and vegetables decreasing risk and frequent consumption of meat and fat increasing risk) was more statistically demonstrable among men than women. No associations were found for consumption of milk, sausage, eggs or potatoes. Analyses of past consumption of the same food groups yielded virtually the same results (data not shown).

DISCUSSION

The 5 studies included in this analysis were designed and co-ordinated by a single research team at IARC. Similar research protocols and data-collection procedures adapted to the local situation were used, thus allowing for a pooled analysis of the data. This strategy makes this the largest and most comprehensive assessment to date of the effect of hot mate drinking and other dietary factors on esophageal cancer risk in South America. Furthermore, pooling the data allowed us to assess with sufficient statistical power the effect of mate drinking in women, in whom esophageal cancer is far less common and epidemiological information on risk factors is scarce.

Drinking of mate and other hot beverages

Although hot mate drinking has been classified as “probably carcinogenic to humans”, it is not clear whether its potential carcinogenic effect is due to the components of the herb, to the temperature at which it is consumed or to both (IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 1991).

After taking into account the strong effects of tobacco and alcohol consumption, both heavy mate drinking and very hot mate drinking were significantly associated with esophageal cancer risk. Although not statistically significant, the overall pattern of associations (status, amount, duration and temperature) was stronger and more coherent in women than in men. The associations with amount and temperature remained statistically significant even after adjusting for the effects of each other. In a further attempt to disentangle their effects, we assessed the interactive effects of amount and temperature by fitting a logistic-regression model with the corresponding main and multiplicative terms. The joint effects of mate amount and temperature were compatible with a synergistic model. As shown in Table III, the effect of mate temperature was significantly increased among heavy drinkers (>1.5 l/day) and the effect of mate amount was much stronger among very hot mate drinkers than among drinkers of mate at lower temperatures.

Consistent with the thermal injury hypothesis, consumption of other hot beverages, such as coffee with milk and tea at a high temperature, was also associated with a 2- to 4-fold increased risk (Table IV).

We did not find an increased risk for drinking very hot coffee. This is consistent with the hypothesis that the volume at which hot beverages are usually consumed plays a role in the thermal injury theory of esophageal carcinogenesis. De Jong et al. (1972) already suggested this hypothesis. These investigators determined, in humans, the temperatures at the lower end of the esophagus during ingestion of hot coffee at varying temperatures and varying volumes, using a modified duodenal tube. Intra-esophageal temperatures were higher than expected (43% of the measurements reached 44°C or higher, up to 53°C), and a linear relationship between initial coffee temperatures and subsequent intra-esophageal temperatures was also shown. They demonstrated that the increase in the intra-esophageal temperature depended more on the volume swallowed than on the temperature of the coffee. These findings are consistent with our epidemiological observation that the temperature effect is particularly evident when the hot beverage is drunk in large amounts. The daily average consumption amounts and the volume provided in each serving are usually larger for mate, tea and coffee with milk than for coffee. A population-based survey conducted in Brazil measured the temperature at which mate was consumed in 1,400 adults, showing that the mean temperature just before consumption was 69.5°C. (Victora et al.,1990). These 2 studies thus add objective evidence to our findings as they show that hot beverages may be swallowed extremely hot and that the esophageal mucosa may be exposed to temperatures that may cause chronic thermal injury, a risk factor for esophageal cancer.

Studies in other populations have also found a significant effect of hot drinking on esophageal cancer and its pre-cancerous lesions. In China, preference for drinking “burning hot” beverages was identified as the strongest risk factor for esophagitis (Wahrendorf et al.,1989), and in Hong Kong, preference for consuming drinks or soups at high temperatures was associated with esophageal cancer risk (Cheng et al.,1992). In northeast China, scalding temperature of meals and drinks was identified as the strongest dietary risk factor for esophageal cancer (Hu et al.,1994). A study in France showed that hot Calvädos was the type of alcoholic beverage carrying the highest risk for esophageal cancer (Launoy et al.,1997). Ingestion of drinks at high temperatures has also been shown to increase the risk of this cancer in earlier studies in Puerto Rico (Martinez, 1969), Singapore (De Jong et al.,1974) and Iran (Cook-Mozaffari et al.,1979).

In our study, measurement of the temperature at which drinks were consumed was subject to misclassification since it was self-reported and therefore subjective. However, as the association between hot temperature and esophageal cancer risk was practically unknown in the study population, there is no reason to think that misclassification was differential with regard to case-control status, which is the type of misclassification that could spuriously over-estimate the association. Indeed, if misclassification existed, it is likely that it was non-differential (i.e., cases and controls would be equally likely to misclassify temperature status and in the same direction), and the reported risk estimates might be an under-estimation of the real underlying effect of temperature. A substantial under-estimation of the temperature effect was suggested in a validation study conducted in Brazil and Uruguay (data not shown). Furthermore, as shown in a study conducted in Iran, when tea temperature is measured, a convincing association between temperature and risk is also found (Ghadirian, 1987).

Dietary factors

Consistent with many epidemiological studies on diet and cancer, we found the well-established dietary pattern characterized by a protective effect of diets rich in fruits and vegetables and by an increased risk associated with diets rich in meat and animal fat (De Jong et al.,1974; Winn et al.,1984; Tuyns et al.,1987; De Carli et al.,1987; Hebert et al.,1993; Hu et al.,1994).

Also consistent with previous studies indicating that green tea can inhibit esophageal carcinogenesis (Gao et al.,1994; Yang and Wang, 1993), we found a protective effect of consuming large amounts of tea. However, we also found that drinking tea at very hot temperatures was an important risk factor that increased esophageal cancer risk almost 4-fold.

In summary, the present pooled analysis of 5 case-control studies involving over 2,500 subjects confirms the importance of hot drinking and adds further evidence to the chronic thermal injury hypothesis of esophageal carcinogenesis. Concerning mate, our data suggest that risk is largely increased only when it is drunk very hot and in large amounts. Assuming that a causal relationship exists, consumption of very hot drinks, including mate, would account for 10% and 12% of esophageal cancer incidence in men and women, respectively. These attributable risk estimates may, however, be low as consumption of other hot foods, such as soups, was not considered in our study. Despite this, tobacco smoking and alcohol drinking remain by far the most important risk factors for esophageal cancer in these populations, accounting for about 90% of its incidence (Castellsagué et al.,1999). Also, in these populations, it has been shown that cessation of these habits, particularly smoking cessation, significantly reduces the subsequent risk of developing esophageal cancer (Castellsagué et al.,2000). To effectively reduce the burden of esophageal cancer in these high-risk countries, priority should be placed on implementing health-promotion strategies to prevent and reduce these widespread exposures in the population.

Acknowledgements

Field work was supported by funds provided by the IARC. The pooled project was funded by the IARC and the Fondo de Investigaciones Sanitarias (FIS) of the Spanish government (FIS 97/0662). We are indebted to the local collaborators who made the 5 studies possible: Drs. L.B. Barcelos, D.A. Peccin and N.M. Braga, in Brazil; Dr. A. Vasallo, in Uruguay; and Drs. J. Iscovich, N. Chopita and A. Jmelnitsky, in Argentina.

Ancillary