Regional risk factors for stomach cancer in the FRG.

A multicentric, hospital-based, case-control study was conducted in high- and low-risk areas for stomach cancer in the Federal Republic of Germany, by which means a low intake of dietary vitamin C (relative risk [RR] = 2.32, 95% confidence interval [CI] 1.22-4.43 for lowest against highest quintile), noncentralized water supply (RR = 2.17, CI 1.14-4.13 against central water supply), refrigerator use for less than 25 years (RR = 1.33, CI 0.82-2.15 against more than 30 years), and the use of spruce for smoking meat at home (RR = 3.33, CI 1.56-7.12 against not smoking meat at home), were identified as factors potentially causally related to stomach cancer occurrence. The attributable risk for gastric carcinoma among the population (AR) was 37.5% for low vitamin C intake, 37.2% for noncentralized water supply, 10.6% for late refrigerator use, and 4.15% for use of spruce for smoking meat at home in this analysis. The overall ARp amounted to 68.3%. These personally linked factors also showed a strong regional distribution, in that the low-risk area had more favorable categories of exposure. Traditional nutritional habits around 1910 were recorded during a survey by ethnologists in 1965. This material was used to contrast those in high and low stomach cancer risk areas with the habits in the south of Germany in general. Vegetable use was most common in the low-risk area, whereas mashed potatoes, cabbage, and farinaceous dishes dominated in the high-risk area. Tomatoes were introduced several years later into the high-risk area, both in terms of consumption and cultivation.(ABSTRACT TRUNCATED AT 250 WORDS)


Introduction
Dietary facors are considered as potentially important in the carcinogenesis of some neoplastic malignancy sites (1). Special interest has been concentrated on the organs that are lined along the epithelial tube of the digestive system. Here the relationship between the ingested food items and pathophysiological mechanisms in surrounding tissues is more obvious and is consequently thought to be of particular importance in most types of gastrointestinal cancer (2).
Notwithstanding the ongoing decline of the mortality and incidence of stomach cancer in Europe, there are important differences of incidence between different countries. In order to identify specific risk factors, several epidemiological casecontrol studies for stomach cancer have either already been terminated or are underway in countries with differing stomach cancer mortality (3)(4)(5)(6)(7)(8).
Compared with the rest of Germany, high rates of mortality from stomach cancer prevail in southeast Bavaria, whereas in some regions of the centrally located state of Hesse the lowest rates were found (9) (Fig. 1). This pattern has been stable since the 1950s (10). It is also noteworthy that age-adjusted mortality for stomach cancer is declining similar to other industrialized countries and that the declining trend has been linked to birth cohorts, especially in Bavaria, where younger cohorts showed halfthe risk within 20 years seen for both males and females (11 ).
A case-control study was performed simultaneously in both Bavaria and Hesse. Besides investigating risk factors on an individual level, the study aimed at explaining the regional differences in stomach cancer mortality. This was accomplished by calculating the preventable proportion based on the differences in prevalences of exposure between highand low-risk area.
In addition to the material collected within the case-control study, we had the opportunity to analyze data that had been collected in 1965 about traditional habits by a detailed questionnaire. Going back to the situation in 1910, some of these data referred to issues being of potential importance in the case of stomach cancer. These data were also useful to contrast highand low-risk areas for stomach cancer.

Materials and Methods
The multicentric, hospital-based, case-control study was performed in fourareas ofGermany: three ina region typical for high stomach cancer mortality in the southeast ofthe Federal Republic (Deggendorf, Straubing, Ingolstadt) andone in an area in the midst ofHesse (Giessen) showing anotoriously low mortality from this malignancy (10). During the study periods from 1985 to 1987,143 stomach cancer cases and 579 controls contributed a complete interview. Thecontrol groupconsisted ofotherhospital patients and visitors coming to these hospitals. A detailed description of the study design and the distribution ofcases and controls by region and age group is given elsewhere (12; Boeing and Frentzel-Beyme, in preparation). <  The information about exposure was obtained by an interview employing a standardized questionnaire with preformulated questions and lasted 55 min on average, covering sociodemographic characteristies, residential, occupational, medical, and smoking histories, water supply, food conservation methods, and the intake of food before onset of the disease. Dietary intake was assessed as usual intake in the last 5 years before signs of onset of a severe disease. Constant taining ofthe interviewers ensured a uniform performance of the interviews.
All interviews were centrally coded by one person, the data entered into a data bank, and data checks were performed to identify coding and punching errors.
Risk estimates were obtained by unconditional logistic regression methods (13). All estimates were adjusted for age (continuous), sex, and hospital. The relative risk estimates were calculated for the pooled group of controls and separately by comparison with both hospital and visitor controls. Results of both subgroup analyses were, in most instances, in agreement and, therefore, only the results for the total group are presented.
The calculation of the attributable risk among the population was derived from the cases or the controls by considering each individual as separate strata. The ARp was, derived from the controls as ARp = E pi (RRi-1)/E pi (RRi-1) + 1 (14) and from the cases as ARp = E pi (RR,-1)/RR, (15).
First, the AR.p was calculated for each risk factor separately. In addition, an overall ARp was also derived based on the suggestions ofBruzzi et al. (16). In practice this was accomplished by generating an overall relative risk by multiplying for each participant the single risks obtained from a logistic model with his or her exposure categories (RRi, = RR, x RR2 X ... RRk). The RRsi obtained in this way was treated as presented above.
The proportion of stomach cancer cases possibly preventable (PP) by changing exposure were calculated using the formula of Wahrendorf (17). Here also each individual was considered as separate strata.
A second source of data for a regional comparison between highand low-risk areas for stomach cancer was created by a national survey that was conducted in 1965 by a group of ethnologists who contacted several hundred individuals who participated in earlier surveys on traditions. These individuals (usually ministers or teachers) filled in a detailed questionnaire about traditional habits prevailing or having prevailed in their village, town, or quarter. Most of the questions referred to nondietary aspects of daily life and festivities. However, three complexes referred to issues being of interest in the case of stomach cancer etiology. One question concerned traditional dishes during weekdays and weekends in 1910. Other questions referred to the vegetable tomato, when it was eaten the first time in the respective area, and when the tomato was first planted locally. Also, in a series of questions referring to meat conservation, information on the kind of wood used for smoking meat was obtained. The original, completed questionnaires are currently stored at the Seminar of Ethnology in Bonn (H. L. Koks). Up to now only selected aspects of this enquiry had been processed and presented. Therefore, the written responses to the questions mentioned above had to be abstracted from the original questionnaires. Old German writing was still in use at the time of the survey, hence, deciphering the information required a specialist trained in reading this old style of writing. The abstraction of material, therefore, was restricted to particular areas in the south of Germany including Hesse as the low-risk area and Bavaria as the high-risk area for stomach cancer (Fig. 2).
Altogether, 644 questionnaires were found reporting on traditions in this defined area. Sometimes the particular questions of interest had been skipped by the respondent. Thus, different sample sizes occur for different questions. From the question regarding usual dishes, only the information of weekday dishes was used. The dishes were first categorized into 31 different food items. Later these items were combined to 15 food groups.
The area from which questionnaires had been abstracted was divided into four regions. The first region is located in the lowrisk area and contains also Giessen (one site ofthe case-control study). The second area constitutes the areas of high stomach cancer mortality in lower Bavaria (Niederbayem) and upper Palatinate (Oberpfilz) including Deggendorf, Straubing, and Ingolstadt (further sites ofthe case-control study). A third region around the Alpes was contrasted with this Bavarian region. The remaining areas formed the fourth region. The prevalences of particular habits were calculated for each region.

General Risk Model and Attributable Risk
In this paper a risk model is presented considering simultaneously those factors identified by the case-control study in low-and high-risk areas for stomach cancer in the FRG from a series ofvariables being associated with and most likely causally related to stomach cancer. Out of the array of variables, vitamin C, the type ofwater supply, years of refrigerator use, and type of wood used for smoking meat were selected for this presentation.
Other factors found to be associated with stomach cancer in this study such as particular food items or food groups, tobacco smoking, or the consumption of particular alcoholic beverages were not considered in this analysis because of their uncertain mode of action or the possibility of spurious associations. A description of the complete results and a discussion in relation to the outcome ofother studies is beyond the scope ofthis paper and can be found elsewhere (12).
In Table 1, the relative risk estimates are presented for the selected variables. These relative risks can deviate from previous estimates because here all factors are considered simultaneously. It is also important to note that the low-risk category was selected as baseline to facilitate the calculation of the attributable risk. Table 2 presents the attributable risk among the population (ARp) for the four risk factors. Attributable risks were calculated from the prevalences of exposure among the controls as described in "Materials and Methods." In addition to these calculations based on the prevalences among the controls, the approach, based on the prevalences of exposure among the cases, is also presented showing slightly different estimates. These differences may be caused by the effect ofconfounder variables such as age, sex, and hospital.  The overall ARp combining the effects of vitamin C, type of watersupply, years ofuseofa refrigerator, and kindofwood used for smoking meat at home amounted to 68.3 % for these four factors.
Regional Differences in Risk Factors Based on Data from the Case-Control Study The regional distributionofthe exposure categories for the four risk factors is shown in Table 3. All risk factors show a gradient between low-andhigh-risk area in respect to lower prevalenceof categories associated with high risk in the low-risk area.
Thedifferences in exposure prevalencesbetweenhighandlowrisk areas were used to calculate the preventable proportion (Table  4). For this calculation, the relative riskestimates obtained forthe total group were used because ofthe small numbers appearing in each region. When the exposureprevalences ofthehigh-riskarea for stomach cancer shifted to the exposure prevalences ofthe low-riskareasthepreventableproportionofstomachcancerinBavaria was highest for the type ofwater supply (15.4%), followed by use of spruce for smoking meat at home (10.8%), vitamin C intake (10.6%), and periodofrefrigeratoruse (4.1%). By usingtheoverall relative risk estimate approach, the preventable proportion amounted to 38.8 %.

Regional Comparison of Dietary Habits in 1910
The eating profiles around 1910 of four regions with distinct stomach cancer risk were investigated for differences in these areas (Fig. 3). Region 1 (Hesse) represents the habits in the lowrisk area, and region 2 (southeast Bavaria) represents the habits in the high-risk area for stomach cancer. The eating pattern of other regions (3 = south Bavaria with intermediate stomach cancer risk, and region 4 = remaining areas [ Fig. 2]) are not shown but referred to in the text. The most impressive differences between high-and low-risk area for stomach cancer (region 1 against region 2) were seen in regard to  The mean year in wh cultivated locally was a planting of tomatoes b 5 years later than in tht 1928 for region 1, 1921 tomato consumption a tomato cultivation). Toi showed no particular regions.
Information in the su ing meat is presented possibilities and also c region in Hesse was linked with the use ofbeechwood, whereas in the other regions, different kinds of wood were in use, including spruce. However, it has to be considered that individual households can deviate from this pattern. The inquiry in the casecontrol study revealed that in the low-risk area spruce was not being used.  (18). Epidemiological research on risk factors for stomach cancer is supposed to generate evidence for mechanisms most likely to explain the high occurrence of this type of cancer in particular regions in the world. Simultaneous studies on risk factors in highand low-risk areas m culturally similar environments may, therefore, not only -----identify risk factors for individuals but also generate hypotheses for reasons and even causes of regional differences (6,8,12).

OX s
The results from the case-control study in high-and low-risk \** areas for stomach cancer in the FRG clearly revealed that factors associated with stomach cancer in individuals also showed a risk gradient between highand low-risk area. When differences in exposure between highand low-risk areas were quantified and found that these differences can explain a major part of the regional differences in cancer occurrence. In the situation FREQOUENCY PERCENT described here, personally linked factors and regional factors region ---+ -+ 2 supplemented each other, whereas by theory, personal and Hesse S.-E. Bavaria regional risk factors may deviate. "Regional" risk factors being commonly adopted by the enumption of dishes around 1910 by study areas with tire local population may not appear as risk factors in regionalmortality (Hess, low risk; southeast Bavaria, high ly conducted studies and are therefore difficult to identify. By contrasting highand low-risk areas alone, causally relevant factors cannot be differentiated from cultural differences. vegetable soup, vegetables, legumes, and This difficulty of interpretation of contrasting regions aparea stood out. The use of farinaceous peared when highand low-risk areas were contrasted according ired to the other regions. The high-risk to prevalences of "habits in 1910." The authors were not able to lowest in the consumption of vegetables, conclude whether discernible differences were ofetiological imuit, and bacon. portance. Nevertheless, to make use of such additional informa-Ifthe four regions in 1910 can be described tion has its worth in that certain consistencies with current sted were mentioned in more than 65 % of knowledge and concepts may be observed. kregion): in the low-risk area (Hesse): No coincidence was seen between the results of the case-.getable soup, potatoes, meat; in the highcontrol study that revealed that high consumption of cheese, Bavaria): potatoes, mashed potatoes, fruit, especially citrus fruit, and whole-meal bread was negativethe region near the Alpes, not being parly related and high consumption ofprocessed meat was positively or stomach cancer (south Bavaria): milk related to stomach cancer risk (12) and the differences in habits dishes, and meat; and in the remaining around 1910 in highand low-risk regions. These differences seen hes, potatoes, and meat.
in 1910 referred in particular to the use of vegetables, which did Lich tomatoes had first been consumed and not show a negative association with stomach cancer risk in the inalyzed for the four regions. The use and case-control study except (in tendency) for raw vegetables.
egan in the high-risk area on the average However, in view of the results of many other individual-based e other regions (1933 for region 2 against studies, the finding ofhigh vegetable consumption in the low-risk 6 for region 3, and 1925 for region 4 for area fits well in the picture in that these studies showed a negative nd 1934 against 1929, 1930, and 1927 for association of high vegetable intake with stomach cancer risk mato use and growing in the low-risk area, (3)(4)(5)(6)(7)(19)(20)(21)(22). It has to be taken into account that cabbage, which differences compared with the other was consumed in higher amounts in the high-risk area, may usually be stored by processing to sauerkraut. This procedure irvey regarding the use of wood for smokmay destroy most of the vitamins found in fresh cabbage. In the in Figure 4, which lists all the different high-risk area, also, the habit of processing freshly cooked onsiders the combined use of wood. The potatoes to mashed potatoes may diminish the provision of _dh Att w oui I*S.1.  vitamin C by destroying this vitamin by oxygen and heat.
One can only speculate about the particular role of legumes because this item did not appear particularly protective in individually based studies. In the case-control study, legumes had also been found to be consumed in higher amounts in the low-risk area (data not shown).
The regionally confined typical distribution ofkinds of wood used for smoking meat is in line with the observation in the casecontrol study linking spruce use with stomach cancer in the highrisk area for stomach cancer; however, n6particular observation can be reported for the region in the southeast of Bavaria based on the situation in 1910.
Not enough is known about dietary deficiencies of cohorts born around 1900 compared with those born around 1920 and thereafter clearly experiencing a decreased mortality risk (9,11). It is well known that during World War I and in the subsequent years (1916)(1917)(1918)(1919), nutrition was very poor in wide areas ofGermany. Furthermore, one ofthe rare sources ofanimal protein in the first decades oftis century that was readily available in every region and low-priced was salt herring, preserved in brine. The availability ofthis well-preserved merchandise around the year even in places remote from the seacoast was one typical feature ofthe past, when meat, milk, and eggs were less affordable for low-income groups. This seems to be in contrast to the data from the survey about eating habits around 1910, which did not reveal that herring played an important role in the eating pattern at the beginning of the 20th century. However, the survey was completed usually by middle-class persons who may not have considered the eating pattern of the part of the population with low income. Salted fish is considered as one source of nitrosamide precursors which form with nitrite direct-acting mutagens (23). Nitrate contents in private water sources, a precursor ofnitrite, was described to be high in earlier periods (24). Private water sources lasted until recently in circumscript areas, included to some extent in the case-control study ( Table 3).
The distribution ofvegetable growing and the intake ofcertain types of fresh or processed greens have not been surveyed with the same coverage in the early years of this century as today. To our knowledge, roots such as turnips and cheap cabbage were, next to potatoes, often the only available food during war and post-war times. The contents of nitrate in these, as well as in the water supply of those years, can only be guessed.
The regional analysis ofpotential risk factors indicate that the higher risk for stomach cancer in southeast Bavaria can be partly explained by a low provision of dietary vitamin C through food such as vegetables and fruits that are able to block the formation ofdirect-acting mutagens in the stomach ifconsumed in higher amounts (25). In addition, particular regional habits such as smoking meat with spruce and possibly an increased intake of nitrate in connection with lack ofvitamin C as it is when nitraterich water is consumed may also have been of importance.