Diet and prostate cancer : results from the South and East China — case – control study on prostate cancer

Introduction Prostate cancer is the second most diagnosed cancer in men worldwide but the cause remains unclear. Marked differences in incidence between countries point towards exposure to environmental risk factors such as diet. However, studies on the impact of diet on prostate cancer risk remain inconclusive. In China, the prostate cancer incidence is remarkably low and their diet is distinct from that of other countries, hence the Chinese diet may provide clues for identifying protective, or risk-enhancing, effects of diet on prostate cancer risk. The aim of this study is to discuss diet and prostate cancer in South and East China. Materials and methods In a hospital-based case–control study, 285 prostate cancer cases and 370 controls were recruited from four centres in China between August 2005 and April 2009. Cases were men diagnosed with prostate cancer, controls were male patients admitted to the same hospitals. Dietary consumption was assessed using a 31-item food frequency questionnaire. Random effects logistic regression analyses were performed to calculate odds ratios and 95% ­vals.­ Results After adjustment for age, education and intake of other foods within the same dietary groups, we observed a reduced risk of prostate cancer of approximately 50% associated with intake­of­stone­fruit,­flower­and­root­ vegetables,­ garlic,­ dark­ fleshed­ fish­ and soymilk. Conclusion This study strengthens previous evidence of a protective effect of several food types on prostate cancer risk. This­study­confirms­the­importance­ of diet on prostate cancer risk and may spur further research into particular dietary products.


Introduction
Prostate cancer is the most frequently diagnosed cancer in many Western countries and the second most common cancer worldwide 1,2 .Despite extensive research, the cause of prostate cancer remains unclear.One clue may be provided by the observation of considerable dissimilarity in prostate cancer incidence between countries, with >80-100 cases per 100 000 men in Western countries to <10 cases per 100 000 men in Asian countries 1 .Although genetic predisposition is likely to play a role, these marked dissimilarities also clearly suggest that exposure to environmental risk factors may be important in the development of the disease.This assumption is further supported by migration studies that show increases in prostate cancer incidence among those who move from low-incidence countries to high-incidence countries 3,4 .Because dietary patterns are different between countries, diet could be one of these risk factors for developing prostate cancer and may, therefore, be an import factor in the prevention of this disease.
Research into the association between diet and prostate cancer risk seems promising, albeit inconclusive 5 .A number of dietary products have been suspected to increase prostate cancer risk, such as meat 5,6 , milk and other dairy products 7,8 .In contrast, soy 9 , fish 10,11 and vegetable 12 intake were shown to have a protective effect, although not all studies support these findings.An interesting population in which to study the role of diet in prostate cancer risk is China, for the reason that the Chinese diet is different from that of most other countries 13 .Significantly, the Chinese population also exhibits a remarkably low incidence of prostate cancer 1 .By studying differences between the diet of Chinese men who develop prostate cancer and the diet of healthy Chinese controls, we may Competing interests: none declared.Conflict of interests: none declared.
All authors contributed to the conception, design, and preparation of the manuscript, as well as read and approved the final manuscript.
All authors abide by the Association for Medical Ethics (AME) ethical rules of disclosure.
be able to point towards food products that increase or decrease prostate cancer risk in this population.Hence, the aim of this study is to analyse the association between diet and prostate cancer in the Chinese population.

Materials and methods
We used data from the South and A 31-item food frequency questionnaire was used to gather information on intake of fruits, vegetables, protein rich products, fish and seafood, soy products and staple foods.Participants were asked to recall how often they had consumed products in these food groups one year before the interview.In addition, data were collected on socioeconomic status (including level of education, income and occupation), lifestyle factors (including smoking and physical activity), medical history and demographic factors.
For each food product, dietary intake was categorised into four levels: no intake or intake less than once a month; one to three times a month; once a week and twice or more a week.When the number of cases or controls in a category was less than 10, two or more categories were merged.To create consistency in the number of cases and controls within the food groups and to be able to adjust for dietary products within each food group, participants with missing values on one or more questions about intake of products within each group were excluded.

Statistical analysis
Random effect logistic regression analyses were performed to assess the relationship between diet and prostate cancer.City was considered to be a random effect because dietary intake within cities tended to have greater similarity than dietary intake in the other cities.For each of the six food groups, two regression models were fitted.The first model derived odds ratios (OR 1 ) adjusted for age and level of education as potential confounders.The second model (OR 2 ) was additionally adjusted for other food products within the same group, as food intake within the food groups was significantly correlated.The lowest intake of a product formed the reference category, except for meat and egg intake.Here, the highest intake was used as reference due to insufficient numbers in the lowest category of intake.
To derive p-values for linear trend (P trend ), the log-likelihood of a model with the food product fitted as an ordinal variable was compared to the log-likelihood of a nested model without the food products.Potential departures from linearity were assessed by comparing the loglikelihood of a model with dietary intake as an ordinal variable with the log-likelihood of a model with dietary intake as categorical variable (P non-linearity ).A statistically significant P trend and a non-significant P non-linearity suggested a dose-response relationship between food intake and prostatecancerodds.Abothsignificant P trend and P non-linearity suggested an association, but no dose-response.A non-significant P trend and a non-significantP non-linearity meant the effect was independent of the amount of intake.To compare sample means between cases and controls we performed independent samples t-tests.Statistical analyses were performed using STATA version 10.0.

Characteristics of the study population
A total of 285 prostate cancer cases and 370 controls were included in our study.The response rate of eligible cases was approximately 95% while over 98% of eligible controls participated in this study.Information on non-responders is scarce so it remains unknown if non-responders differed from responders on various characteristics.Our cases had a mean age of 72 years (SD = .70)and mean age of the controls was 65 years (SD = .68).Almost one-third of the controls suffered from digestive diseases.Other descriptive statistics of the study population are described in Table 1.

Main results
Tables2-7presentthemainfindings per food group.In Table 2, the effect of fruit on prostate cancer in our population is presented.Men who ate stone fruit more than once a month had approximately half the odds (OR 2 ) of prostate cancer compared to those who consumed these fruits less often, with an inverse dose-response relationship.The apparent protective effect of stone fruit consumption was highest at an intake of twice or more a week, with a corresponding OR of .1-3/month or ≥2/week seemed to significantly decrease the prostate cancer odds, compared with intake of less than once a month, with an OR 1 of, respectively, .52 and .40.However, after adjustment for other fruit intake the OR for intake ≥2/week was no longersignificantandtheP trend became non-significant.Intake of soft, fleshy and vine fruit was not associated with prostate cancer risk.
Among vegetable groups (Table 3), an inverse effect on prostate cancer odds of approximately 50% and higher for flower and root vegetable consumption was found.With regard to leafy and bulb vegetables we noticed a significant dose-response after adjustment for other vegetable intake.Furthermore, a mushroom intake of two or more times a week showed a significant reduction in prostate cancer odds (OR 1 = .45),althoughthisdidnotholdsignificance after adjustment for other vegetable intake.For garlic, we observed that consumption twice or more a week halved the prostate cancer odds when adjusted for other vegetable intake.
We found no significant association between legume intake (Table 3) and protein food intake (Table 4) and the risk of prostate cancer.
In the category fish and seafood consumption (Table 5), we found that Chinese men who consumed white fleshed fish once a month or more had over 50% less odds of prostate cancer than men who never or rarelyatethiskindoffish.Conversely, the p values showed no relation and the significant effect disappeared after adjustment for dark fleshed fish and other seafood intake (OR 2 ).An intake of dark fleshed fish twice or more times per week showed a significantly reduced prostate cancer odds (OR 1 = .45)and an increase in protection with increased intake.For the category seafood, other than white or darkfleshedfish,wedidnot findan association with prostate cancer.
Men who drank soybean milk (Table 6) once a week had half the odds of prostate cancer compared to men with a more or less frequent soy milk intake.The related ORs were .46 and .39 after adjustment for other soy intake groups.For both soybean milk and 'other soy products' we found a significant P trend adjusted for other soy intake.There were lower odds with higher intake of soybean milk and higher odds with higher intake of 'other soy products'.Intake of soybean curd (tofu) was not associated with prostate cancer odds in our study.
Finally, for staple foods (Table 7) we observed that frequent consumption(≥2/week)ofpotatoesandbread significantly reduced the odds of prostate cancer with respectively an OR 1 of .49(95% CI: .26,.95) and .55(95% CI: .32,.93).The effect for frequent potato intake disappeared but the effect for bread remained after adjustment for other staple food intake.However, none of the p values were significant.Corn and pasta intake appeared not to be associated with prostate cancer odds.

Main results
In this study, the association between diet and the risk of prostate cancer in Licensee OA Publishing London 2013.Creative Commons Attribution License (CC-BY)  the Chinese population was studied.The results are in line with previous studies in other populations that found a protective effect of stone fruit, flower vegetables, root vegetables, garlic, dark fleshed fish and soybean milk on prostate cancer risk.On the other hand, soy foods, excluding soybean curd and soybean milk, appeared to be possible threads to prostate cancer risk in this population.
The inverse association of citrus fruit intake with prostate cancer risk that we found is consistent with some previous studies 14 , but the potential protective effect of stone fruits has not been studied much.The effects of these fruits have typically been ascribed to the vitamin C and flavonoids they contain.Vitamin C is an antioxidant that may protect cell membranes and DNA from oxidative damage 14 .Flavonoids are associated with antioxidant,anti-inflammatoryandantitumour activities 15 .Interestingly, the observed significant protective effect of citrus fruit in our study disappeared after adjustment for other fruit intake, which suggests that the established effect is confounded by consumption of other fruits.
In contrast to our results, the majority of studies have failed to demonstrate a protective effect of vegetables on prostate cancer risk 16,17 .Some studies, however, suggest that protection of vegetables may be due to antioxidative effects 5 , and inhibition of carcinogen activation 18,19 .The WCRF/AICR expert panel concluded that there was insufficient evidence for a possible association between consumption of vegetables and prostate cancer 5 .Our findings regarding the increased risk from leafy and bulb vegetable intake are contrary to other epidemiological research that observed no effects 20 , and remains unexplained.Our finding of a protective effect of high garlic intake is consistent with and explained by other studies 21,22 .
Fatty fish are rich in long chain omega-3 fatty acids, EPA and DHA, and have a stimulating effect on  OR 1 = Odds ratio adjusted for age and education.Hospital is included as random effect.OR 2 = Odds ratio adjusted for age and education and the other types of meat in this table.Hospital is included as random effect. 1 Categories <1/month and 1-3/month are merged because no. of cases and controls in the first category was <10. 2 Reference category is highest meat intake (≥2/week) because by far the most cases are in this category (can be found in file 'additional analyses'). 3Categories <1/month and 1-3/month are merged because no. of cases in the first category was <10. 4 Reference category is highest egg intake (≥2/week) because by far the most cases are in this category (can be found in file 'additional analyses').* = P for trend, P for non-linearity.
prostate cancer cell growth 23 .This study is the first to investigate an associationwithdifferentfishtypes, and the data suggests that the type of fish does matter in the protective effect against prostate cancer.The difference in effect between typesoffishmaybeduetofattyacid composition: dark fish are richer in EPA and DHA than the mostly lean whitefishspecies 24 .
While there is evidence of a protective effect of soy food intake in general on prostate cancer risk 5,9 , the current data only demonstrated the protective effect of soy milk.Soy food intake may lead to inhibition of tumour growth 25 , suppression of DNA synthesis and induction of apoptosis in human prostate cell lines 26 .In our study, prostate cancer risk showed an increase, instead of the expected decrease as with 'other soy food'.In some studies of fermented soy foods, such as miso, the same effect was found 25,27 .Fermentation might reduce the important anti-carcinogenic effect ofisoflavones 25 , which is suspected to be one of the protective ingredients in soy food 9 , and it has been shown to inhibit endogenous formation of N-nitroso compounds 28 .Our category of 'other soy products' may have included fermented soy products.
Studies in various populations generally suggest a greater risk of prostate cancer with an higher intake of meat 12 .However, this association was not found in our study, but also not in  other studies among Asian men 29,30 .Why there appears to be a difference in effect of meat intake on prostate cancer risk in Asian males compared to other populations remains unclear.Dairy intake was also found not to be significantlyassociatedwithprostate cancer risk in this study.Although it is biologically plausible that higher dairy intake increases prostate cancerrisk,otherstudiesalsofailtofind this association 31 .A little research has been done on starch in association with prostate cancer.Starch is an important component of staple foods.One study in Italy has, however, found that high intakes of starch is directly associated with prostate cancer risk.This is contradictory to our study and needs to be investigated further 32 .

Limitations and recommendations for future research
A limitation of this study was the probable heterogeneity within the food categories that we studied.Food types were allocated to heterogeneous food categories which may have contained both protective products and products that do not affect prostate cancer risk.The products that had no effect on prostate cancer may have alteredtheresultsandbiasedfinding a protective effect.A recommendation for future research to avoid this problem would be to assess the intake of individual products instead of pooling products into groups.Furthermore, only frequency of intake was measured while the quantity of food intake was not taken into account.However, the absolute amount of food intake in the different categories may have been fairly similar, particularly in the more frequent intake categories (e.g., high intake 1x/week compared to low intake 2x/week).This may have given a distorted view on the odds.In this study, we focused on whether a more frequent intake of dietary products was associated with the odds on prostate cancer.Another interesting question could have been whether an intake that differs from the most common intake would have been associated.With this, it is also possible to study whether unbalanced diets or unusual intake are associated with prostate cancer odds.It is recommended that future studies take this into account.No data on total energy intake or body mass index (BMI) was available in this sample.Studies on these variables are inconclusive 12 , but point to an increased risk of prostate cancer with an increased total energy intake and BMI.Adjustment for these variables could have confounded our results as dietary intake is evidently highly correlated to total energy intake and BMI.
As another disadvantage of this study, selection bias may have occurred.This is a common problem to hospital-based studies.However, as the response rate in this study is considerably high, the effect of possible selection bias will be limited.As the majority of controls suffered from digestive diseases, we performed a sensitivity analysis by excluding these patients from all analyses to check if the disease of the control may have been a source of bias.This was not the case (data not shown).Finally, recall bias is a common problem with food frequency intake assessments.This is further complicated by the fact that patients may have changed their dietary intake after diagnosis of a disease.We tried to minimise this bias by asking for food intake one year ago.To minimise interviewer bias, the interviewers were repeatedly trained in their interview techniques.

Table 1 Descriptive statistics of cases and controls No. of cases % No. of controls %
32 after adjustment for other fruit intake.Citrus fruit consumption at a rate of Licensee OA Publishing London 2013.Creative Commons Attribution License (CC-BY) For citation purposes: Verboom CE, Zhong Z, Xie LP, Hu Z, Zhong WD, Wong YC, et al.Diet and prostate cancer: results from the South and East China-case-control study on prostate cancer.OA Epidemiology 2013 Aug 01;1(2):11.Competing interests: none declared.Conflict of interests: none declared.All authors contributed to the conception, design, and preparation of the manuscript, as well as read and approved the final manuscript.All authors abide by the Association for Medical Ethics (AME) ethical rules of disclosure.

Table 2 Odds ratios and 95% CIs for fruit intake on prostate cancer risk Frequency No. of cases No. of controls OR 1 95% CI 1 OR 2 95% CI 2
1 = Odds ratio adjusted for age and education.Hospital is included as random effect.OR 2 = Odds ratio adjusted for age and education and the other types of vegetables in this table.Hospital is included as random effect.
OR * = P for trend, P for non-linearity.

Table 3 Odds ratios and 95% CIs for vegetable intake on prostate cancer risk Frequency No. of cases No. of controls OR 1 95% CI 1 OR 2 95% CI 2
1 = Odds ratio adjusted for age and education.Hospital is included as random effect.OR 2 = Odds ratio adjusted for age and education and the other types of vegetables in this table.Hospital is included as random effect.
OR * = P for trend, P for non-linearity.

Table 5 Odds ratios and 95% CIs for fish and seafood intake on prostate cancer risk Frequency No. of cases No. of controls OR 1 95% CI 1 OR 2 95% CI 2
6 = Odds ratio adjusted for age and education.Hospital is included as random effect.OR 2 = Odds ratio adjusted for age and education and the other types of fish in this table.Hospital is included as random effect.5Categories1/weekand≥2/week are merged because no. of cases and controls in the third category was <10.6Categories 1/week and ≥2/week are merged because no. of cases in the last category was <10.* = P for trend, P for non-linearity. OR

Table 6 Odds ratios and 95% CIs for soy product intake on prostate cancer risk
OR 1 = Odds ratio adjusted for age and education.Hospital is included as random effect.OR 2 = Odds ratio adjusted for age and education and the other types of soy products in this table.Hospital is included as random effect.* = P for trend, P for non-linearity.