Smoking and Dietary Risk Factors for Cervical Cancer at Different Age Group in Japan

The importance of the major risk and protective factors for cervical cancer in women by age group was evaluated with the use of data from the Hospital-based Epidemiologic Research Program at Aichi Cancer Center (HERPACC), Japan . This study included 416 cervical cancer cases and 20,985 referents confirmed as free of cancer. Cases and referents were divided into three age groups: the younger, middle and older age groups (30-44, 45-54 and 55-69 years old , respectively). Logistic model was applied separately to the three groups to estimate odds ratios (ORs) of smoking, beverage and dietary habits with adjustment for marital and reproductive factors. The elevated OR of current smokers was observed consistently in all age groups , while alcohol intake did not show any increased ORs. Dietary control for health lowered ORs for all three age groups, and the effect appeared to be more pronounced among the older age group (OR = 0.49, 95% confidence interval: 0.30-0.80). Higher frequency intake of green-yellow vegetables consistently afforded lower ORs among all three age groups . This risk increment of smoking and risk reduction by dietary control were consistent in all age groups . These findings suggested practicable prevention strategy for the cervical cancer by modifying general life style . J Epidemiol, 1998 ; 8 : 6-14.

Cervical cancer (hereafter referred to as CC) is the most common cancer in women in most developing countries and still remains a major public health problem worldwide.The "Cancer Incidence in Five Contin ents, Vol VI" reported that the incidence rate of CC is about twice as common in Japan as in Europe and North America1).The incidence of CC in Japan has been dominant but has recently been decreasing gradually, while the incidence of endometrial cancer is running at lower levels but is gradually increasing.To clarify risk and protective factors of CC in contrast with endometrial cancer, the comparative case-reference study was conducted by applying the data from the Hospital-based Epidemiologic Research Program at Aichi Cancer Center (HERPACC)2) .The results of the comparative study of uterus cancer by subsite gave us the suggestion that there are some modifiable protective factors in dietary habits against CC.
In Japan, the westernization of life style, especially dietary habit, has progressed remarkably since the end of World War II.This remarkable change could be associated with the chronological variation of cancer mortality , but the extent of the influence of change in life style may vary among generations to a considerable degree.The importance of the major risk and/or protective factors for CC in each generation was , therefore, evaluated with the use of data from the HERPACC .Information on the effect of the various risk factors on different generations may well have potentially preventive implications.In order to compare the risk and protective factors between the younger generation and the older generation, a case-reference study of CC was analyzed separately in three age categories (30-44, 45-54 and 55-69 years old group) which could adjust the effect of biological and/or hormonal consequences by age.
Recent evidence obtained from virological studies suggested that certain types of human papillomaviruses (HPV 16, 31 etc.) were one of the major risk factors in the development of CC3).However, HPV is actually spreading within the general population in countries with high incidence of CC, and HPV infection alone could not explain the high incidence of CC, especially, in developing countries4).The measurement of infection with HPV among general women has been fraught with methodological difficulty and epidemiological information on HPV in Japanese women is not available at the moment.Therefore, intervention to prevent exposure to other etiological factors may be practicable for reducing the environmental burden of CC.In particular, we focused on finding the modifiable risk and protective factors of CC, such as smoking, beverage and dietary habits.The present case-reference comparative study was conducted to assess the independent effects of habitual smoking and diet, with adjustment for other factors, such as marital status and reproductive habits in different age groups.

Data Collection
Since 1988, we have conducted the HERPACC study; a self-administered questionnaire survey completed by first-visit outpatients at the Aichi Cancer Center Hospital (ACCH).All questionnaire sheets were collected on the first-visit day after the checking of incomplete responses by a trained interviewer.All the data was loaded into the computer system of the Aichi Cancer Center Research Institute.Details of the questionnaire and data collection procedures are described elsewhere 5,6.7).
Of all the first-visit outpatients, totaling 47,895 between January 1988 and December 1993, 3,393 patients were excluded due to interviewer absence, age exclusion (younger than 18 years old), or visit only for consultation.The questionnaire was administered to 44,502 patients.Among them, 43,759 (98.3%) completed the questionnaire adequately.
The questionnaire included items on occupation, medical history, height and weight (added since 1989), marital status, family history (parents and siblings), smoking and drinking habits, dietary habits, sleeping habits, physical exercise, bowel habits and in females, reproductive history.We did not include questions on socioeconomic status and education level, because Japanese are rather reluctant in general to answer such questions.These conditions prior to the symptoms which outpatients presented with were inquired about and all information was collected before diagnoses were made.

Cases and Referents
The data collected were linked with the hospital-based cancer registry files.Among 9,172 cancer patients, 416 women aged 30-69 years who were first diagnosed within 12 months after first-visit as having CC by histological examinations were recruited as the case group.Referents were 20,985 female first-visit outpatients aged 30-69 years who had had neither a previous diagnosis of cancer nor hysterectomy.Table 1 shows the 20,985 referents, 416 CCs by age group.To compare the differences and similarities of the risk and protective factors between the younger generation and the older generation, cases and referents were divided into three age groups: the younger age group (30-44 years old), the middle age group (45-54 years old) and the older age group (55-69 years old).The three groups were divided by average age at menopause of reference group, 49.9 years old, to focus on the independent risk and protective factors of CC in pre-, peri-and post-menopausal status.The biological effect of menopausal status was eliminated by using for these age categories.

Score of Food Intake Frequency
Score of food intake frequency employs a five-digit code, which was organized in the following way: 5; Every day, 4; 3-4 times per week, 3; 1-2 times per week, 2; 1-3 times per month, and 1;never.The score means of food consumption for cases and referents of each of the food items were calculated.The differences between means for cases and referents scores were tested by Wilcoxon-Mann-Whitney test.

Statistical Analysis
Odds ratio (OR) and its 95% confidence interval (95%CI) for each exposure variable were estimated by using an unconditional logistic regression model.The P value for trend corresponded to the estimate of the slope derived from the logistic model.Utilizing the LOGISTIC procedure provided by SAS (SAS Institute), multivariate unconditional logistic regression analysis was carried out applying the statistically significant variables from the prior univariate analysis.

General Background
Compared with referents, women with CC were younger at first pregnancy, had higher parity and a lesser proportion of Table 1.Age distribution of cases and referents.
single status (Table 2), although there were few differences in the middle age group.Cases were also more likely to report as current smoker among all three age groups.Therefore, we calculated ORs of beverage and dietary habits in further analysis adjusting by these 4 variables for each age group and 5 variables including age for all subjects.
Smoking and Drinking Habits Current smokers were consistently displayed a higher OR among all three groups (Table 3).The OR of current smoking, 10 or more cigarettes per day, was approximately 2.6 among all age groups.Although smoking did not elevate the risk in a dose-response pattern, a risk increment was observed to be related to passive exposure to environmental tobacco smoke from husband's smoking among the younger and middle age groups.Our analyses provided no evidence that alcohol intake significantly altered risk, whereas a slightly higher OR (OR=1.79,95%CI: 0.71-4.48)was noted in those aged 45-54 years who consumed more than one "go" (180 ml) of Japanese "sake" (15 -16% degree of alcohol) per day (Table 3) .

Dietary Habit
Table 4 presents the mean scores of categorical food intake frequencies for cases and referents of food items and z-values.Cases combining three age groups reported lower frequencies for 14 of the 18 foods compared with those of all referents.This deficiency was particularly large for green vegetables, carrot and boiled, broiled or raw fish.The consumptions of green vegetables and carrot that contain substantial amounts of vitamin A and carotenoids, were consistently lower in cases than referents among all age groups in statistically significant manner.When data of the three age groups were combined, the referents reported consuming 11 foods significantly more frequently than did cases.On the other hand, we observed an excess in cases in consumption of such salty foods as pickles and salted or dried fish, and of instant noodle.
Table 5 showed the summarized adjusted ORs of dietary habits for CC by age group.Daily intake of raw vegetables did not significantly alter the association with risk of CC.Daily intake of fruit, milk and dietary control for health (on a diet controlling salty food, fatty food and/or total energy) afforded lower ORs among all three age groups.The risk reduction through the daily intake of milk seemed to be clearer in the middle age group (OR=0.48,95%CI:0.30-0.75).The protective effect of dietary control for health appeared to be somewhat more pronounced among the older age group (OR=0.47,95%CI:0.29-0.75).Women under dietary control due to any disease/condition were excluded from this analysis.The levels of risk reduction through dietary control on salty foods, fat and fatty foods or total calorie did not differ from each other, because most people control all three factors as a set of dietary modification.
The adjusted ORs of selected food items for CC are compiled in Table 6.The consumption of green-yellow vegetables (green vegetables, carrot and pumpkin) consistently showed lowered ORs for CC among all age groups, while that of lightgreen vegetables (cabbage and lettuce) did not appear to influence the risk.When data of the three age groups were combined, the ORs for CC by higher frequence intake ( three or more times per week) of bean curd and boiled, broiled or raw  (Those who have no information of husband's smoking status were excluded.)c) The odds ratios for alcohol were adjusted for marital status, age at first pregnancy, number of pregnancies and smoking.d) The unit of Japanese sake (1 go is epuivalent to 180m1 and contains 28.8m1 of net alcohol).e) Adjusted for age, marital status, age at first pregnancy and number of pregnancies.f) Adjusted for marital status, age at first pregnancy and number of pregnancies.*P<005, **pct<0.01fish were 0.65 (95%CI: 0.49-0.88),0.56 (95%CI: 0.42-0.74),respectively.These risk reductions were more prominent among the older age group.
The adjusted ORs by multiple unconditional logistic regression analysis of 10 main items were shown in Table 7.When analyzed for all subjects, items (dietary control and frequent intake of fish) remained as lowered ORs and smoking habit as elevated OR after consideration of other factors.Especially in the older age group, the ORs for smoking, dietary control and higher frequency of fish intake were marked.

DISCUSSION
A methodological issue of the present study was the possible bias caused by using hospital-based non-cancer patients as referents.It is sometimes pointed out that there are discrepant characteristics between the general population and hospitalbased referents.
In Japan, outpatients, in general, visit hospitals directly when they have symptoms.This situation is much different from that in Western countries, where people visit local general clinics first, and then they are referred to hospitals which function as secondary and/or special facilities for further medical treatment.Patients with incident cancer comprised only 11 percent of all new outpatients in women visited ACCH.Sixty percent of the non-cancer group visited the hospital three times or less.Among the randomly sampled non-cancer outpatients (n=1,000) in 1988-89, only 34% were found to have some specific diseases.The most common diseases in that group were benign tumor and/or non-neoplastic polyp (13.1%), mastitis (7.5%), digestive disease (4.1%), benign gynecological disease (4.1 %) and so on 8).We were, therefore,   sexual behavior and HPV infection 3,4).However, a questionnaire study on sexual behavior among general women is not very easy in our HERPACC study, and furthermore, a massive virological examination on HPV infection among outpatients is now impossible methodologically.
In future, we believe , epidemiological information on HPV infection among general women might become available, and then we might be able to evaluate an actual relationship between HPV infection through sexual behavior and risk impact of CC among Japanese women epidemiologically.Therefore, intervention to prevent exposure to other etiological factors may be practicable for reducing the burden of CC under present circumstance.Numerous epidemiologic studies have shown a positive association between smoking and CC 1).Simons et al .reported a relation between smoking and the proportions of DNA adducts in cervical epithelium, and provided direct biochemical evidence of potentially carcinogenic agents 13).Previous studies thus, have shown that current smoking is related strongly to the risk of CC 12) and the present result was entirely consistent with this among all age groups (Table 3).The present study failed to find the increased OR of passive smoking among the older age group, which was contradictory to our hypothesis that the effect of passive smoking may be more pronounced among the older age group because of the longterm exposure to environmental tobacco smoke.The exposure conditions of the younger and middle age groups , such as the environmental conditions of their housing , might be different from those of the older age group.Effects other than a tobacco effect may influence the older generation more .In this study, however, alcohol intake did not show any increased OR of CC .
At any rate, the alcohol intake could account for a small etio-logic fraction of CC because the proportion of female drinkers is very small and also alcohol consumption by females in Japan is much lower than that in Western countries 7.14).
After adjusting for other factors, dietary control for health lowered the risk of CC in all age groups.Significant association with dietary control for health was the strongest in the older age group.Perceived changes in dietary habits, such as salty food, fat intake and/or total calorie control, might modulate the risk of CC.Further studies are required to confirm the relation between dietary control for health and CC in terms of the possible biological effect of salty food, fat intake and/or calorie control, per se, and other possible confounding factors.
Recently the possibility that diet may be important in the causality and prevention of cancer in human beings has received major attention.A causal role for beta-carotene intake under natural condition in cancer prevention seems acceptable from a viewpoint of biological plausibility.Six case-control studies 15-20) of CC also appeared to indicate a decreased risk at higher intake of beta-carotene, although only two studies showed a statistically significant decrease.Dietary studies on beta-carotene and CC have yielded inconsistent results, however, the serologic studies were slightly more consistent").Palan et al. reported that cervicovaginal cells and plasma beta-carotene levels were found to be significantly decreased in women with CC as compared with controls 22).A recent experimental study by using newly established cervical dysplasia-derived cell lines demonstrated the protective effect against cervical carcinogenesis by beta-carotene 23).Our analysis consistently indicated risk reductions in CC on frequent intake of green-yellow vegetables among all age groups.This finding suggested that vitamins and trace minerals, both of which are found in good quantity in vegetables, might play a role in the protective effect against CC.
Little published information has specifically examined the relationship between fish intake and risk for CC.We observed risk decrements in CC on frequent intake of fish after adjusting for other factors among the older age group.This fording must be interpreted with considerable caution, not only because all dietary survey methods present problems with respect to reliability and validity, but also because of the limited number of dietary variables collected in this study.Further studies are required to confirm the relation.Daily intake of milk appeared to be a protective factor for the middle age group.It is well known that milk is a effective supplier of calcium which might have a healthy effect on humans via a mechanism of estrogens, especially for the older generation.It should also be recognized that daily intake of milk was a index of health consciousness.This effect might be explained by a further study on biochemical properties of milk.
The present study focused on fording common risk and protective factors of CC over generations.We found consistent factors over three age groups, and identified generation-specif-ic protective factors in dietary habits.In the present study we can not clearly explain that the prominent protective effects by dietary habits among the older age group may be caused by cumulative effect over age and/or cohort effect.The question of the role of diet in the etiology of CC seems as yet to be unresolved, however, the results of this study suggested clues to the strategy for CC prevention, especially in the older age group.It is important to quit smoking for CC prevention regardless of generation.Moreover, the finding from this study suggested dietary recommendations, i.e. adequate intake of fruits and green-yellow vegetables which are good sources of vitamins and minerals.The observed risk reduction in CC by dietary control for health is noteworthy from the public health standpoint.

Table 2 .
Comparison of age, anthropometric, reproductive and other nondietary factors among cases and referents.

Table 3 .
Adjusted odds ratios and 95% confidence intervals of habitual smoking and drinking for cervical cancer by age group.
OR : Odds Ratio, 95% Cl : 95% Confidence Interval a) Number of cigarattes per day.b) Nonsmoker without or with smoker husband.

Table 4 .
Distribution of the mean scores of categorical food intake frequenciesa) for cervical cancer cases and referents.

Table 5 .
Adjusted odds ratios and 95% confidence intervals of dietary habits for cervical cancer by age group .
OR : Odds Ratio, 95% CI : 95% Confidence Intervals a) On a diet controlling salty food, fatty food or total calorie .b) Adjusted for age, marital status, age at first pregnancy , number of pregnancy and smoking, c) Adjusted for marital status, age at first pregnancy, number of pregnancy and smoking .* P<0 .05,**P<0.01

Table 6 .
Adjusted odds ratios and 95% confidence intervals of selected food items for cervical cancer by age group.
This study should be free of case response bias to the questionnaire because all data were collected prior to diagnoses; questionnaire, the average life-style of first-visit outpatients who visited the ACCH located in Nagoya City was not very

Table 7 .
Adjusted odds ratios and 95% confidence intervals for selected life-style factors included in all adjusted logistic regression models.OR: Odds Ratio, 95%CI:95% Confidence Intervals a) Odds ratio for each variable is adjusted for age and all of the other variables shown .b) Odds ratio for each variable is adjusted for all of the other variables shown .
Other factors except dietary habits may carry more weight with CC among the younger generation, e.g .