Abstract
Background:
Incidence of gastric cancer is the highest in Eastern Asia. Multiple modifiable lifestyle factors have been identified as risk factors for gastric cancer. However, their aggregated effect on the risk of gastric cancer has not been examined among populations with high prevalence of Helicobacter pylori.
Methods:
A study was conducted to examine the association between multiple lifestyle factors together and the risk of developing gastric adenocarcinoma in the Singapore Chinese Health Study, a prospective cohort of 63 257 men and women between 45 and 74 years enroled during 1993–1998. Composite score of cigarette smoking, alcohol consumption, obesity, dietary pattern, and sodium intake at baseline was assessed with hazard ratio (HR) and 95% confidence interval (CI) of gastric adenocarcinoma using Cox regression method.
Results:
Higher healthy composite lifestyle scores were significantly associated with reduced risk of gastric adenocarcinoma in a dose-dependent manner. Hazard ratios (95% CIs) for total, cardia, and non-cardia gastric adenocarcinoma for the highest (score 5) vs lowest composite score (score 0/1/2) were 0.42 (0.31–0.57), 0.22 (0.10–0.47), and 0.55 (0.39–0.78), respectively (all Ptrend<0.001). These lifestyles together accounted for 48% of total gastric adenocarcinoma cases in the study population. The inverse association was observed in both genders, and remained after exclusion of first 5 years of follow-up.
Conclusions:
The inverse association between the aggregated healthy lifestyle factors and the risk of gastric adenocarcinoma is in dose-dependent manner in this highly H. pylori-exposed population. These lifestyle factors together may account for up to half of disease burden in this study population.
Keywords: gastric adenocarcinoma, lifestyle factor, composite lifestyle score
Globally, gastric cancer is the fifth most common malignancy. Geographically the highest incidence rate is recorded in Eastern Asia such as China, Japan, and South Korea whereas the lowest in Western Europe and North America (Ferlay et al, 2014). Infection with Helicobacter pylori is an established underlying, but insufficient causal factor for gastric cancer (de Martel et al, 2012). The prevalence of H. pylori infection varies substantially worldwide. The highest infection rate is observed at more than 80% of the entire population in Eastern Asian countries where high incidence rates of gastric cancer are reported. However, <2% of infected people eventually develop gastric cancer over their lifetime (Conteduca et al, 2013), strongly suggesting that other environmental exposures and genetic factors must play an important contributing role to the variation of individual's risk of developing gastric cancer.
Several modifiable lifestyle factors have been identified as risk factors for gastric cancer in various populations. The risk factors include cigarette smoking (Ladeiras-Lopes et al, 2008), heavy alcohol consumption (Tramacere et al, 2012), obesity (Yang et al, 2009), high sodium intake (D'Elia et al, 2012), low physical activity (Abioye et al, 2015), low vegetable and fruit intake (Lunet et al, 2007), and high red meat intake (Zhu et al, 2013). These lifestyle factors usually go hand in hand with each other such as cigarette smoking and alcohol consumption on an individual subject level. We hypothesised that people with multiple healthy lifestyle factors would have lower risk of gastric cancer than those with no or fewer healthy lifestyle factors, and the beneficial effect would be incremental, that is, the healthier lifestyle factors one has, the greater risk reduction one would derive from. Thus it would be more informative for studies that simultaneously examine multiple lifestyle factors together in relation to risk of gastric cancer in populations with different background risk of gastric cancer, especially in population with different prevalence of H. pylori.
Two prospective cohort studies examined the association between combined lifestyle factors and risk of gastric cancer. Both were conducted in the European countries where both the prevalence of H. pylori and the incidence rate of gastric cancer are lower (9.4 per 100 000) than the world average (12.1 per 100 000; Conteduca et al, 2013; Ferlay et al, 2013). In the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, a composite score of three healthy lifestyle factors was significantly associated with reduced risk of gastric adenocarcinoma (Buckland et al, 2015). The other study in France, a part of EPIC, reported a similar association between a composite score of five lifestyle habits and reduced risk of cancers of the oesophagus, stomach, biliary tract, small bowel, and pancreas combined (Dartois et al, 2014).
To our knowledge, there has been no study that prospectively assesses the composite score of multiple lifestyle factors together in relation to risk of gastric cancer in any Asian population that have a higher prevalence of H. pylori infection than western population, where the findings may not be directly applicable to Asian populations. These modifiable lifestyle factors including cigarette smoking, alcohol consumption, obesity, dietary pattern, and sodium intake together may have different impact on the risk of developing gastric adenocarcinoma in a high-risk population. The results of this study can provide important evidence in the development of gastric cancer prevention strategy that promotes healthy lifestyle in populations with high prevalence of H. pylori infection.
Materials and methods
Study population
The details about the Singapore Chinese Health Study (SCHS) had been described previously (Hankin et al, 2001). Briefly, 63 257 Singapore Chinese men and women aged 45–74 years old who were citizens or permanent residents of Singapore and belonged to one of the two major dialect groups (Hokkien or Cantonese) were recruited between 1993 and 1998. After excluding 1936 subjects who had a history of cancer at baseline, the present analysis included 61 321 participants. The SCHS was approved by the Institutional Review Boards at the National University of Singapore and the University of Pittsburgh (Pittsburgh, PA).
Assessment of lifestyle factors
Baseline in-person interviews were conducted using a structured questionnaire for each participant to obtain demographics, body weight and height, tobacco smoking, alcohol consumption, physical activity, diet, medical history, and family history of cancer. Information on dietary consumption was obtained using a structured semi-quantitative food frequency questionnaire (FFQ) with 165 listed dietary items representing majority of food and beverage items commonly consumed by the study population in Singapore. For each dietary item, an individual subject was asked to choose a consumption frequency among eight pre-defined categories from ‘never or hardly ever' to ‘two or more times a day' along with pre-determined portion sizes illustrated in a companion food photo album. This FFQ was validated subsequently in a sub-cohort of our study population (Hankin et al, 2001).
As height and body weight were self-reported, for subjects with missing body weight (N=9971) or height (N=289), a linear regression equation (Weight=α+β*Height) derived from known weight and height of the entire cohort was used to estimate the missing value. Body mass index (BMI) was calculated as weight in kg divided by squared height in m (kg m−2).
For cigarette smoking, a subject who reported smoking at least one cigarette per day for a year or longer (a smoker) was asked the average number of cigarettes smoked per day and the total number of years of smoking on a regular basis. Ex-smokers were asked how many years ago since quitting smoking. Total number of pack-years of smoking was calculated as the number of packs (20 cigarettes per pack) smoked per day multiplied by the number of years of smoking.
For daily ethanol consumption, participants were asked the frequency of drinking type of alcoholic beverages during the past year by the usual serving size. Daily consumption of each type of alcoholic beverage was calculated by frequency multiplied by the serving size, and daily total ethanol was the sum of ethanol over all types of alcoholic beverages consumed. The average ethanol content is 13.5 g in one drink (375 ml) of beer, 10.85 g in one drink (30 ml) of rice wine or hard liquor, and 11.68 g in one glass (118 ml) of grape wine (USDA, 1975).
Daily sodium intake was derived from validated FFQ and the Singapore Food Composition Table that also provides sodium content for each of 165 dietary items (mg per 100 g). The final value of sodium intake was adjusted for total energy intake (mg per 1000 kcal).
Dietary pattern was determined using the principal component analysis as described previously (Butler et al, 2006). Briefly, we identified two dietary patterns among SCHS participants: vegetable-fruit-soy (VFS) and meat-dim-sum (MDS). Vegetable-fruit-soy was characterised by high intake of fruits, vegetables, and soy foods whereas MDS by high intake of pork, chicken, dim-sum foods, and noodle dishes, respectively (Butler et al, 2006). For each subject, the average of the two VFS (ascending order) and MDS (descending order) ranking scores (range 1–100) was calculated for the present analysis, a high score was indicative of a high intake of VFS and a low intake of MDS (Supplementary Table S1).
Case ascertainment
Gastric cancer cases among the SCHS participants were identified through the linkage analysis with The Singapore Cancer Registry under the National Registry of Diseases Offices (NRDO) of Singapore. This nationwide registry has collected information on individual cancer patients and national cancer trends and patterns since 1968 (Lee, 2015) and has been shown to be complete in recording incident cancer cases (Forman et al., 2014). Gastric cancer was defined using the International Classification of Disease Oncology, 3rd edition (ICD-O-3) as C16.0-C16.9.
Construction of composite protective lifestyle score
Our literature search identified six modifiable risk factors for gastric cancer. They were cigarette smoking, alcohol consumption, obesity, physical inactivity, high intake of sodium, and diet with low fruits/vegetables and high meat. In our study population, we did not find physical inactivity to be a risk factor for gastric adenocarcinoma (P=0.243), hence we excluded it from the composition of final lifestyle score.
Two sets of algorithm were applied for construction of composite protective lifestyle score based on dichotomised cutoff and Z-score of each factor. In the first algorithm, each lifestyle factor was dichotomised and final cut-off value was chosen to reflect the strongest effect size for its univariate association with gastric adenocarcinoma risk. Pack-years of smoking was categorised at median (i.e., 21.9) among ever smokers (low-risk/protective lifestyle score 1=<21.9, 0=⩾21.9). Daily ethanol consumption was categorised at 8.1 g (the third tertile) among ever drinkers (1=<8.1, 0=⩾8.1). Dietary pattern score was categorised at 62 (the fourth quartile) of the entire cohort (1=⩾62, 0=<62). Daily sodium intake was categorised at 782 mg per 1000 kcal (the third tertile; 1=<782, 0=⩾782). BMI was categorised at ⩾27.5 kg m−2 for obesity recommended by the World Health Organisation for Asian populations (WHO, 2004; 1=<27.5, 0=⩾27.5 kg m−2). The percentages of the study population in high or low score of these five individual lifestyle factors are presented in Supplementary Table S2. The final composite protective lifestyle score was the sum of five individual lifestyle factors: 0 represented the lowest and 5 the highest protective lifestyle score.
The second algorithm was to generate sex-specific Z-score for each lifestyle factor to avoid overfitting the model based on dichotomised values of individual lifestyle factors. In the study population, ∼70% of participants were never smokers and 81% never consumed alcoholic beverages. Thus the Z-scores for pack-years of smoking and daily ethanol consumption were derived from ever smokers and ever alcoholic drinkers, respectively. We assigned half of the lowest Z-score for pack-years of smoking and daily ethanol to never smokers and never drinkers, respectively. The composite healthy lifestyle Z-score was the sum of Z-score for all five individual factors for each study subject as follows:
 |
A higher composite Z-score stands for higher score presenting healthier dietary pattern, lower pack-years of smoking, lower ethanol consumption, lower BMI, and lower sodium consumption. The distribution of single lifestyle factor by quartile of composite Z-score was shown in Supplementary Table S3.
H. pylori infection status testing
To further examine the association of composite protective lifestyle score in combination with H. pylori infection status in relation to gastric adenocarcinoma risk, a nested case–control study was conducted involving 522 subjects (133 gastric adenocarcinoma cases and 389 individually matched controls) whose serological H. pylori infection status was determined by the presence or absence of CagA 116 kDa in serum. The details of this case–control study have been previously reported (Ainslie-Waldman et al, 2014).
Statistical analysis
Person-years at risk for each of 61 321 eligible subjects were computed from the date of enrolment to the date of gastric cancer diagnosis, death, migration out of Singapore, or 31 December 2014, whichever occurred first. Cox proportional hazard regression method was employed for calculation of hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs) of gastric adenocarcinoma associated with individual lifestyle factors and both the composite protective lifestyle scores derived from dichotomised cutoff and Z-score of these lifestyle factors together. Test for linear trends was conducted by treating the composite lifestyle score as a continuous variable in the Cox model. The proportional hazards assumption was examined by testing the significance of Pearson's correlation coefficient between Schoenfeld residuals of the composite lifestyle score and ranked survival time (Harrell and Lee, 1986). We found no violation of proportional hazards assumption.
To adjust for potential confounding, age, sex, dialect group, year of recruitment, and level of education were included in all regression models. Further adjustment for other covariates such as intake of individual vegetables and fruit, family history of cancer, and history of gastric/duodenal ulcer did not meaningfully alter the association between the composite lifestyle score and gastric adenocarcinoma risk. Thus the results presented were not adjusted for these variables.
Population attributable risk (PAR) estimate was computed for the proportion of gastric adenocarcinoma cases that would be avoided on a population level attributable to higher composite lifestyle category. On the basis of the HRs from Cox proportional hazard regression model and their variance–covariance matrix and the prevalence of each unique combinations of the covariates in the model, an algorithm developed by Spiegelman et al (2007) was applied to calculate partial PAR along with its 95% CIs while age, sex, dialect group, year of recruitment, and level of education remain unchanged.
To examine the potential modifying effect of subclinical symptoms of gastric cancer on the association between composite protective lifestyle score and gastric adenocarcinoma risk, we conducted sensitivity analyses on subset of data set divided by the length of follow-up, for example, ⩽5 years and >5 years. To investigate the association between composite lifestyle score and gastric adenocarcinoma risk with adjustment of H. pylori infection status, conditional logistic regression model was performed in the nested case–control study for all subjects and for subjects with positive H. pylori infection status defined by positive CagA test results. Stratified analyses were performed by anatomical sites such as cardia and non-cardia of the stomach.
All statistical analysis was performed in SAS 9.4 software package (SAS Institute, Inc., Cary, NC, USA). All P-values reported are two-sided. P-values <0.05 were considered to be statistically significant.
Results
With increasing composite lifestyle score, there was an increase in proportion of women and a decrease in BMI. A higher composite lifestyle score was also characterised by fewer current smokers, lower pack-years of smoking, and lower daily intake of ethanol, sodium, and red meat (Table 1). The distributions of individual lifestyle factors across different composite score of protective lifestyle factors are shown in Supplementary Table S4. These individual lifestyle factors were not or moderately correlated each other (all correlation coefficients <0.23).
Table 1. Distribution of selected baseline characteristics of all participants by composite lifestyle scores (five factors), the Singapore Chinese Health Study, 1993–2014.
|
Composite lifestyle scores |
||||
|---|---|---|---|---|
| Characteristics | 0/1/2 | 3 | 4 | 5 |
| N | 6223 | 20 017 | 24 799 | 10 282 |
| Age in years, mean (s.d.) | 56.4 (7.7) | 55.9 (7.9) | 56.6 (8.1) | 56.9 (7.9) |
| BMI (kg m−2), mean (s.d.) | 24.8 (4.5) | 23.3 (3.5) | 22.8 (2.8) | 22.5 (2.4) |
| Female, % | 21.8 | 50.5 | 61.5 | 71.2 |
| Education level, % | ||||
| No formal education | 20.2 | 24.7 | 30.0 | 30.0 |
| Primary school | 52.8 | 44.8 | 43.0 | 41.5 |
| ⩾Secondary level | 27.0 | 30.5 | 27.0 | 28.5 |
| Smoking status, % | ||||
| Never smoker | 25.4 | 63.5 | 78.2 | 87.0 |
| Former smoker | 21.3 | 12.8 | 8.6 | 6.3 |
| Current smoker | 53.3 | 23.7 | 13.2 | 6.7 |
| Smoking amount among ever smokers, mean (s.d.) | ||||
| Number of cigarettes per day | 24.1 (11.2) | 18.1 (11.1) | 12.3 (9.0) | 9.2 (6.2) |
| Number of years of smoking | 36.4 (8.8) | 33.6 (11.4) | 30.5 (12.7) | 28.2 (13.6) |
| Pack-years of smoking | 43.8 (21.9) | 31.2 (22.4) | 18.5 (16.4) | 11.8 (7.4) |
| Alcohol consumption status, % | ||||
| Non-drinkers | 58.9 | 77.9 | 86.4 | 91.4 |
| Drinkers | 41.1 | 22.1 | 13.6 | 8.6 |
| Ethanol intake (g per day) among drinkers, mean (s.d.) | 20.6 (22.3) | 8.4 (14.2) | 3.3 (5.9) | 2.0 (1.8) |
| Daily sodium intake (mg), mean (s.d.) | 1607.9 (784.4) | 1361.0 (663.3) | 915.1 (414.8) | 816.9 (319.4) |
| Food consumption in grams, mean (s.d.) | ||||
| Total vegetables | 113.5 (67.9) | 110.2 (62.3) | 99.2 (58.3) | 137.3 (67.0) |
| Total fruits | 183.9 (174.5) | 194.7 (161.9) | 191.1 (160.3) | 256.8 (189.9) |
| Total red meat | 46.9 (32.0) | 37.7 (26.0) | 25.6 (19.2) | 18.7 (15.2) |
Abbreviation: BMI=body mass index.
After more than 1 million cumulative person-years of follow-up (mean 16.9 years per subject), as of 31 December 2014, a total of 801 incident cases of gastric cancer were identified among all participants of SCHS. Among them, 32 were sarcomas, 29 lymphomas, and 49 malignancies with unspecified histology; all of them were excluded from the present analysis. Thus the present study included 691 cases of gastric adenocarcinoma; among them, 118 were cardia, 491 were non-cardia, and 82 were unspecified subsite of the stomach. The mean duration between baseline interview and the diagnosis of all gastric adenocarcinoma cases was 6.9 years (s.d.=3.9).
Individual scores of all five protective lifestyle factors separately were significantly associated with a 18–34% reduction in HR of gastric adenocarcinoma (Table 2). The association was stronger for BMI with risk of cardia than non-cardia cancer (Table 2). HRs (95% CIs) of gastric adenocarcinoma for individual risk factor (before dichotomised) are presented in Supplementary Table S5. The cut-off value of each lifestyle factors was chosen based on their risk association with gastric adenocarcinoma for the creation of composite score. High composite score of protective lifestyle factors was significantly associated with reduced HR of gastric adenocarcinoma in a dose-dependent manner (Table 2). Compared with the lowest composite scores using dichotomous cutoff (0–2), HRs (95% CIs) of gastric adenocarcinoma for composite scores of 3, 4, and 5 protective lifestyle factors were 0.68 (0.52–0.88), 0.51 (0.40–0.66), and 0.42 (0.31–0.57), respectively (Ptrend<0.001). This association was stronger for cardia than non-cardia cancer. (Table 2). On the basis of the distribution of the composite score, we estimated that 48% of total gastric adenocarcinoma, including 72% in cardia and 43% in non-cardia, could be attributable to these five risk factors combined.
Table 2. Lifestyle factors and gastric adenocarcinoma risk among all participants (n=61 321), the Singapore Chinese Health Study, 1993–2014.
|
All cases |
Cardia |
Non-cardia |
|||||
|---|---|---|---|---|---|---|---|
| Characteristics | Person-years | Cases | HR (95% CI)a | Cases | HR (95% CI)a | Cases | HR (95% CI)a |
|
Smoking | |||||||
| Pack-years of smoking >21.9 | 132 949 | 187 | 1.00 (reference) | 37 | 1.00 (reference) | 128 | 1.00 (reference) |
| Pack-years of smoking ⩽21.9 | 906 112 | 504 | 0.66 (0.55–0.83) | 81 | 0.54 (0.35–0.84) | 363 | 0.66 (0.53–0.83) |
|
Alcohol | |||||||
| Daily ethanol intake >8.1 g | 62 238 | 70 | 1.00 (reference) | 11 | 1.00 (reference) | 51 | 1.00 (reference) |
| Daily ethanol intake ⩽8.1 g | 976 823 | 621 | 0.69 (0.53–0.89) | 107 | 0.86 (0.45–1.63) | 440 | 0.66 (0.49–0.89) |
|
BMI | |||||||
| ⩾27.5 kg m−2 | 89 879 | 66 | 1.00 (reference) | 16 | 1.00 (reference) | 47 | 1.00 (reference) |
| <27.5 kg m−2 | 949 182 | 635 | 0.80 (0.62–1.03) | 102 | 0.54 (0.32–0.93) | 444 | 0.79 (0.58–1.07) |
|
Dietary pattern score | |||||||
| 1st–3rd Quartile (<62) | 774 142 | 549 | 1.00 (reference) | 101 | 1.00 (reference) | 383 | 1.00 (reference) |
| 4th Quartile (⩾62) | 264 919 | 142 | 0.82 (0.68–0.99) | 17 | 0.60 (0.35–1.01) | 108 | 0.89 (0.71–1.11) |
|
Sodium intake | |||||||
| ⩾782 mg per 1000 kcal energy | 340 708 | 242 | 1.00 (reference) | 48 | 1.00 (reference) | 161 | 1.00 (reference) |
| <782 mg per 1000 kcal energy | 698 353 | 449 | 0.80 (0.68–0.94) | 70 | 0.64 (0.44–0.93) | 330 | 0.87 (0.72–1.05) |
|
Composite scores (five factors)b | |||||||
| 0/1/2 | 97 712 | 123 | 1.00 (reference) | 30 | 1.00 (reference) | 77 | 1.00 (reference) |
| 3 | 334 135 | 246 | 0.68 (0.52–0.88) | 41 | 0.49 (0.30–0.79) | 179 | 0.80 (0.61–1.04) |
| 4 | 427 141 | 237 | 0.51 (0.40–0.66) | 38 | 0.36 (0.22–0.60) | 169 | 0.57 (0.43–0.76) |
| 5 | 180 074 | 85 | 0.42 (0.31–0.57) | 9 | 0.22 (0.10–0.47) | 66 | 0.55 (0.39–0.78) |
| Ptrend | <0.001 | <0.001 | <0.001 | ||||
| PAR (95% CI) | 0.48 (0.36–0.59) | 0.72 (0.51–0.84) | 0.43 (0.27–0.57) | ||||
|
Composite
Z-score (five factors) | |||||||
| 1st Quartile | 251 028 | 212 | 1.00 (reference) | 44 | 1.00 (reference) | 140 | 1.00 (reference) |
| 2nd Quartile | 258 747 | 167 | 0.76 (0.62–0.93) | 29 | 0.64 (0.40–1.02) | 129 | 0.88 (0.70–1.12) |
| 3rd Quartile | 263 404 | 162 | 0.72 (0.58–0.88) | 25 | 0.54 (0.33–0.88) | 109 | 0.72 (0.56–0.92) |
| 4th Quartile | 265 881 | 150 | 0.64 (0.51–0.78) | 20 | 0.42 (0.25–0.71) | 113 | 0.71 (0.55–0.91) |
| Ptrend | <0.001 | 0.001 | 0.002 | ||||
Abbreviations: BMI=body mass index; CI=confidence interval; HR=hazard ratio; PAR=population attributable risk.
For single lifestyle factor, model included all factors simultaneously and adjusted age at baseline interview (in years), baseline interview year (1993–1995, 1996–1998), father's dialect (Cantonese, Hokkien), gender, and education (no formal education, primary education, ⩾secondary education).
Model adjusted for age at baseline interview (in years), baseline interview year (1993–1995, 1996–1998), father's dialect (Cantonese, Hokkien), gender, and education (no formal/primary, ⩾secondary education).
We also examined the association between Z-score of single lifestyle factor and gastric adenocarcinoma risk (Supplementary Table S6). High Z-score for all individual lifestyle factors except for dietary pattern was associated with significantly increased risk of gastric cancer
When these Z-scores were summed up after reversing the Z-scores of risk lifestyle factors (see Materials and Methods section), high composite Z-score representing healthier lifestyle factors was associated with statistically significant, reduced risk of gastric adenocarcinoma (Table 2). Although weaker than the composite score of the dichotomised lifestyle protective factors, the inverse association was strong and in dose-dependent manner, and present for both cardia and non-cardia cancers. (Table 2). This inverse association between composite score of protective lifestyle factors, either derived from dichotomised categories or Z-scores, and gastric adenocarcinoma risk was present in both men and women (Table 3), and for both short (⩽5 years) and long (>5 years) duration of follow-up of the entire cohort (Table 4).
Table 3. Composite lifestyle score and gastric cancer risk by gender, the Singapore Chinese Health Study, 1993–2014.
| Characteristics | N | Cases | Person-years | HR (95% CI)a |
|---|---|---|---|---|
|
Male (N=27 293) | ||||
| Composite scores (five factors) | ||||
| 0/1/2 | 4866 | 103 | 74 708 | 1.00 (reference) |
| 3 | 9929 | 159 | 158 964 | 0.72 (0.56–0.92) |
| 4 | 9543 | 117 | 157 918 | 0.53 (0.40–0.69) |
| 5 | 2955 | 33 | 49 597 | 0.46 (0.31–0.69) |
| Ptrend | <0.001 | |||
| PAR (95% CI) | 0.50 (0.34–0.63) | |||
| Composite Z-scores (five factors) | ||||
| 1st Quartile | 6823 | 127 | 105 697 | 1.00 (reference) |
| 2nd Quartile | 6824 | 101 | 109 148 | 0.76 (0.58–0.99) |
| 3rd Quartile | 6823 | 97 | 112 193 | 0.74 (0.56–0.96) |
| 4th Quartile | 6823 | 87 | 114 148 | 0.62 (0.47–0.82) |
| Ptrend | 0.001 | |||
|
Female (N=34028) | ||||
| Composite scores (five factors) | ||||
| 0/1/2 | 1357 | 19 | 23 004 | 1.00 (reference) |
| 3 | 10 088 | 88 | 175 171 | 0.67 (0.41–1.09) |
| 4 | 15 256 | 120 | 269 223 | 0.52 (0.32–0.84) |
| 5 | 7327 | 52 | 130 477 | 0.47 (0.28–0.79) |
| Ptrend | 0.003 | |||
| PAR (95% CI) | 0.41 (0.18–0.60) | |||
| Composite Z-scores (five factors) | ||||
| 1st Quartile | 8507 | 85 | 145 331 | 1.00 (reference) |
| 2nd Quartile | 8507 | 66 | 149 599 | 0.76 (0.55–1.05) |
| 3rd Quartile | 8507 | 65 | 151 211 | 0.70 (0.50–0.96) |
| 4th Quartile | 8507 | 63 | 151 733 | 0.66 (0.48–0.92) |
| Ptrend | 0.011 | |||
Abbreviations: CI=confidence interval; HR=hazard ratio; PAR=population attributable risk.
Model adjusted for age at baseline interview (in years), baseline interview year (1993–1995, 1996–1998), father's dialect (Cantonese, Hokkien), and education (no formal education, primary education, ⩾secondary education).
Table 4. Sensitivity analysis for composite lifestyle score and gastric adenocarcinoma risk by length of follow-up, the Singapore Chinese Health Study, 1993–2014.
| Follow-up time | Cases | Person-years | HR (95% CI)a |
|---|---|---|---|
|
Follow-up ⩽5 years | |||
| Composite score (five factors) | |||
| 0/1/2 | 36 | 30 149 | 1.00 (reference) |
| 3 | 68 | 97 487 | 0.71 (0.47–1.07) |
| 4 | 51 | 121 335 | 0.44 (0.28–0.68) |
| 5 | 15 | 50 499 | 0.33 (0.18–0.61) |
| Ptrend | <0.001 | ||
| Composite Z-score (five factors) | |||
| 1st Quartile | 66 | 74 552 | 1.00 (reference) |
| 2nd Quartile | 40 | 74 837 | 0.60 (0.41–0.89) |
| 3rd Quartile | 29 | 75 022 | 0.43 (0.28–0.66) |
| 4th Quartile | 35 | 75 058 | 0.50 (0.33–0.76) |
| Ptrend | <0.001 | ||
|
Follow-up >5 years | |||
| Composite score (five factors) | |||
| 0/1/2 | 86 | 67 563 | 1.00 (reference) |
| 3 | 178 | 236 648 | 0.71 (0.55–0.92) |
| 4 | 187 | 305 805 | 0.57 (0.43–0.74) |
| 5 | 70 | 129 575 | 0.52 (0.38–0.73) |
| Ptrend | <0.001 | ||
| Composite Z-score (five factors) | |||
| 1st Quartile | 146 | 176 476 | 1.00 (reference) |
| 2nd Quartile | 127 | 183 910 | 0.83 (0.66–1.06) |
| 3rd Quartile | 133 | 188 382 | 0.84 (0.66–1.07) |
| 4th Quartile | 115 | 190 823 | 0.70 (0.54–0.89) |
| Ptrend | 0.007 | ||
Abbreviations: CI=confidence interval; HR=hazard ratio.
Model adjusted for age at baseline interview (in years), baseline interview year (1993–1995, 1996–1998), father's dialect (Cantonese, Hokkien), gender, and education (no formal education, primary education, ⩾secondary education).
To take into account the impact of H. pylori infection on the observed risk association, we conducted similar analysis in a nested case–control study within the SCHS whose serological status of H. pylori infection was determined by the presence or absence of CagA in serum. There was a statistically significant inverse association between composite score of dichotomised lifestyle factors and gastric adenocarcinoma risk among all subjects of the case–control study after adjustment for H. pylori infection status as well as among subjects with positive CagA status only. A similar inverse association was observed for both cardia and non-cardia cancer. For composite Z-score, the association slightly attenuated after adjustment for H. pylori infection status and among subjects with positive CagA status, especially for non-cardia cases, given the small sample size (Table 5).
Table 5. Composite lifestyle score and gastric adenocarcinoma risk among subjects with measurement of Helicobacter pylori infection (CagA) status, the Singapore Chinese Health Study, 1993–2014.
|
All subjects |
H. pylori-positive onlya |
||||
|---|---|---|---|---|---|
| Characteristics | Ca/Co | OR (95% CI)b | OR (95% CI)c | Ca/Co | OR (95% CI)b |
| All subjects | 133/389 | 128/329 | |||
| Composite score | |||||
| 0/1/2 | 26/46 | 1.00 (reference) | 1.00 (reference) | 24/39 | 1.00 (reference) |
| 3 | 47/112 | 0.77 (0.43–1.40) | 0.78 (0.43–1.42) | 46/92 | 0.78 (0.42–1.46) |
| 4 | 44/161 | 0.44 (0.24–0.81) | 0.43 (0.23–0.81) | 43/139 | 0.42 (0.21–0.82) |
| 5 | 16/70 | 0.37 (0.16–0.82) | 0.36 (0.16–0.82) | 15/59 | 0.34 (0.15–0.80) |
| Ptrend | 0.017 | 0.017 | 0.019 | ||
| PAR (95% CI) | 0.62 (0.26, 0.82) | 0.62 (0.27, 0.83) | 0.64 (0.28, 0.84) | ||
| Composite Z-score | |||||
| 1st Quartile | 48/110 | 1.00 (reference) | 1.00 (reference) | 46/94 | 1.00 (reference) |
| 2nd Quartile | 28/89 | 0.71 (0.41–1.22) | 0.68 (0.40–1.19) | 28/76 | 0.70 (0.39–1.23) |
| 3rd Quartile | 34/90 | 0.79 (0.46–1.38) | 0.80 (0.46–1.41) | 32/77 | 0.75 (0.42–1.35) |
| 4th Quartile | 23/100 | 0.53 (0.28–0.98) | 0.56 (0.29–1.06) | 22/82 | 0.54 (0.27–1.05) |
| Ptrend | 0.068 | 0.117 | 0.094 | ||
| Cardia cases | 24/69 | 21/60 | |||
| Composite score | |||||
| 0/1/2 | 6/8 | 1.00 (reference) | 1.00 (reference) | 5/8 | 1.00 (reference) |
| 3 | 7/22 | 0.46 (0.09–2.42) | 0.45 (0.08–2.41) | 7/17 | 0.58 (0.10–3.29) |
| 4 | 9/27 | 0.28 (0.06–1.30) | 0.28 (0.06–1.33) | 8/25 | 0.33 (0.06–1.68) |
| 5 | 2/12 | 0.20 (0.02–1.63) | 0.20 (0.02–1.61) | 1/10 | 0.11 (0.01–1.57) |
| Ptrend | 0.074 | 0.075 | 0.065 | ||
| Composite Z-score | |||||
| 1st Quartile | 11/23 | 1.00 (reference) | 1.00 (reference) | 10/20 | 1.00 (reference) |
| 2nd Quartile | 4/15 | 0.47 (0.12–1.84) | 0.47 (0.12–1.87) | 4/13 | 0.53 (0.13–2.17) |
| 3rd Quartile | 5/17 | 0.29 (0.06–1.43) | 0.29 (0.06–1.44) | 4/14 | 0.33 (0.07–1.65) |
| 4th Quartile | 4/14 | 0.62 (0.15–2.59) | 0.62 (0.15–2.59) | 3/13 | 0.42 (0.08–2.24) |
| Ptrend | 0.296 | 0.297 | 0.195 | ||
| Non-cardia cases | 88/253 | 87/212 | |||
| Composite score | |||||
| 0/1/2 | 17/27 | 1.00 (reference) | 1.00 (reference) | 16/23 | 1.00 (reference) |
| 3 | 34/74 | 0.69 (0.33–1.44) | 0.73 (0.34–1.57) | 34/60 | 0.76 (0.35–1.67) |
| 4 | 27/106 | 0.31 (0.14–0.70) | 0.34 (0.14–0.79) | 27/90 | 0.35 (0.15–0.83) |
| 5 | 10/46 | 0.24 (0.08–0.70) | 0.28 (0.10–0.81) | 10/39 | 0.29 (0.10–0.85) |
| Ptrend | 0.009 | 0.023 | 0.026 | ||
| Composite Z-score | |||||
| 1st Quartile | 31/68 | 1.00 (reference) | 1.00 (reference) | 30/59 | 1.00 (reference) |
| 2nd Quartile | 20/58 | 0.72 (0.37–1.39) | 0.72 (0.36–1.43) | 20/49 | 0.72 (0.36–1.44) |
| 3rd Quartile | 24/59 | 0.76 (0.38–1.51) | 0.86 (0.42–1.75) | 24/50 | 0.87 (0.42–1.80) |
| 4th Quartile | 13/68 | 0.36 (0.15–0.83) | 0.44 (0.19–1.05) | 13/54 | 0.45 (0.19–1.08) |
| Ptrend | 0.030 | 0.121 | 0.139 | ||
Abbreviations: Ca=Cases; CI=confidence interval; Co=Controls; OR=odds ratio; PAR=population attributable risk.
H. pylori positive defined by positive serum CagA status.
Conditional logistic regression model adjusted for age at baseline interview (in years), baseline interview year (1993–1995, 1996–1998), father's dialect (Cantonese, Hokkien), gender, and education (no formal education, primary education, ⩾secondary education).
Further adjusted for serum H. pylori CagA status (positive, negative).
Discussion
The present study demonstrates that a high composite score of five healthier lifestyle factors including smoking, alcohol consumption, BMI, a diet high in vegetables/fruit and low in red meat, and low intake of dietary sodium is significantly associated with reduced risk of developing gastric adenocarcinoma in an Asian population with high prevalence of H. pylroi. The highest composite score was associated with a statistically significant 58% decreased risk of gastric adenocarcinoma compared with the lowest composite score. These lifestyle factors together can account for up to half of the disease burden in this study population, of which ∼85% had a history of infection with H. pylori.
To our knowledge, the present study is the first prospective study to examine the association between combined lifestyle factors and gastric adenocarcinoma risk in an Asian population with high H. pylori prevalence. There are only two previous reports, both in low-risk European populations, on the composite lifestyle factors and gastric cancer risk. In the EPIC study with 11.4 years of follow-up, highest score of three lifestyle factors (cigarette smoking, alcohol consumption, and adherence to a Mediterranean diet) was associated with a significant 50% decrease in risk of gastric adenocarcinoma and a PAR of 19% if all subjects were in the healthiest lifestyle score category (Buckland et al, 2015). The E3N (Etude Epidémiologique auprès de femmes de la Mutuelle Générale de l'Education Nationale) study in French women with 15 years of follow-up, a part of EPIC study, showed similar results; the highest score of protective lifestyle factors, including abstinence from smoking, low to moderate alcohol drinking (<2 drinks per day), normal range of BMI (18.5–25 kg m−2), high recreational physical activity, and high vegetable and fruit intake, was associated with a statistically significant 40% decrease in risk of cancer in the digestive system excluding large bowel and a PAR of 12% (Dartois et al, 2014). The relative risk findings of the present study are consistent with those in low-risk populations. Compared with previous studies, we found a higher PAR of 48%. Possible explanations include (1) more lifestyle factors included (five for present study; three factors for EPIC study); (2) stronger association for certain lifestyle factors (e.g., HR for low alcohol intake=0.69 in present study vs 0.83 in EPIC study); and (3) higher prevalence of risk factors in our study population.
Epidemiological evidence and biological plausibility lend support for the observed association between each of the lifestyle factors studied and gastric cancer risk. For tobacco smoking, a meta-analysis of 27 cohort studies showed a 62% increased relative risk (RR) of gastric cancer for current vs never smokers among males (RR=1.62, 95% CI: 1.50–1.75) and 20% among females (RR=1.20, 95% CI: 1.01–1.43; Ladeiras-Lopes et al, 2008), which was consistent with our findings (HR=1.56, 95% CI: 1.27–1.93). Tobacco smoking has been listed as group I carcinogen by IARC (IARC, 2004). Besides more than 70 known carcinogens in cigarette smoke including tobacco-specific nitrosamines and polycyclic aromatic hydrocarbons (FDA, 2012), tobacco smoke also contains high level of nicotine, which could promote carcinogenesis. In vitro studies showed that nicotine could activate nicotinic acetylcholine receptors and induce cellular proliferation in gastric cancer cell lines by upregulating cyclooxygenase 2 (COX-2; Shin et al, 2004). It could also increase phosphorylating extracellular signal-regulated kinase-1/2 (ERK1/2) to further activate the downstream signalling pathways involving COX-2 and ERK (Shin et al, 2007).
Heavy alcohol use is associated with elevated gastric cancer risk. In humans, ethanol is metabolised to acetaldehyde, a group I carcinogen classified by IARC (IARC, 2010) and further oxidised into nontoxic acetate (Klyosov, 1996). A review of 15 cohort and 44 case–control studies found that heavy drinkers (⩾4 drinks per day) had a 20% higher gastric cancer risk than non-drinkers (RR=1.20, 95% CI: 1.01–1.44; Tramacere et al, 2012). However the association was not found in Asian populations (Tramacere et al, 2012). In the present study, we found that even relatively moderate alcohol consumption (approximately one drink per day) was associated with a significant 30% increase in the risk of gastric adenocarcinoma. In the present study, one drink of alcoholic beverage per day already rendered the damage that may only be detectable in a population, of which majority do not drink (81% never drinker). The underlying mechanism is that alcohol can damage gastric mucosa, resulting in enhanced adherence of H. pylori bacteria. A previous study reported a high H. pylori infection rate in people with high consumption of alcohol in a Chinese population (Zhang et al, 2010). Alternatively regular drinkers may choose an unhealthy lifestyle that is associated with risk of gastric adenocarcinoma. For example, daily drinkers were more likely to be current smokers (58% vs 18%), consumed higher level of red meat (43.3 vs 30.0 g), and consumed low vegetables/fruit and high red meat, all of them are risk factors for gastric cancer. A composite score of these lifestyle factors can summarise their overall effect on gastric cancer risk without worrying about potential confounding effect.
Mounting evidence supports that high intake of red meat or processed meat is associated with increased gastric cancer risk. On the basis of a review of 12 cohort and 30 case–control studies, both red meat and processed meat was associated with a significant 45% increase in risk of gastric cancer (Zhu et al, 2013). Haem iron intake from red meat can endogenously be converted to carcinogenic compounds such as N-nitroso compounds (NOCs). An experimental study demonstrated a high correlation between haem iron from red meat and endogenous NOCs formation (Jakszyn et al, 2006). In the EPIC cohort, highest level of haem iron intake from red meat was found to be associated with a significant 67% increase in gastric adenocarcinoma risk (HR=1.67, 95% CI: 1.20–2.34) compared with the lowest level (Jakszyn et al, 2012).
Antioxidants such as vitamin C in fruit and vegetables can inhibit the endogenous NOC formation (Mirvish, 1996; Iijima et al, 2003). A meta-analysis of 24 cohort studies found a significant 10% reduction in risk of gastric cancer associated with highest levels of fruit (RR=0.90, 95% CI: 0.83–0.98) and fresh vegetables (RR=0.90, 95% CI: 0.79–1.01; Wang et al, 2014). In the present study, the combination of the two dietary patterns, VFS and MDS, was associated with significantly decreased risk of gastric adenocarcinoma.
The association between sodium intake and gastric cancer risk has been extensively studied in different populations. Two in vivo studies in rats showed that administration of high concentration of sodium chloride (1.3–4.5 M) caused immediate damage to gastric mucosa, increased cellular proliferation, and altered the viscosity of gastric mucosa (Sorbye et al, 1988; Furihata et al, 1996). A review of 11 cohort studies found a statistically significant positive association between salt intake and gastric cancer risk, especially in Japan (Wang et al, 2009) where both salt intake and gastric cancer incidence rate are highest in the world. Given the ubiquitous presence and lack of information on salt content in some prepared foods, it is very challenging, especially using FFQ or 24-h dietary recalls, for accurately estimating daily salt intake. Urinary excretion of sodium over 24 h is recommended as a better measure for daily salt intake (Wiseman, 2008). A cross-sectional study reported a positive association between urinary sodium excretion and prevalence of gastric cancer in a Korean population (P=0.006), but no association was found when survey-assessed dietary sodium intake was used (Park et al, 2014). Studies using urinary biomarkers of sodium with a prospective design are warranted to clarify the role of dietary sodium on gastric adenocarcinoma in humans.
Overweight and obesity (BMI⩾25) are significantly associated with increased risk of gastric cancer in a comprehensive review of 10 cohort studies (Yang et al, 2009). The association was weaker in Asian than western populations (Yang et al, 2009). Compared with Caucasians, Asians have relatively lower BMI (Deurenberg et al, 1998; Pan et al, 2004). Using a WHO-recommended cutoff for obesity for Asian populations (⩾27.5 kg m−2; WHO, 2004), we found obesity was associated with borderline significant increase in risk of overall gastric adenocarcinoma and significant increase in risk of adenocarcinoma in the cardia. These results are consistent with previous finding (Yang et al, 2009). The biological link between obesity and gastric cancer risk may be through the pro-inflammatory cytokines produced from excess body fat that can lead to chronic inflammation (Roberts et al, 2010). Overexpression of interlukin-1β resulted in inflammation in gastric cells and eventual formation of carcinoma in experimental studies (Tu et al, 2008). Excess body fat may also upregulate insulin-like growth factor-1 that stimulates cellular proliferation and inhibits apoptosis (Bianchini et al, 2002).
Z-score-based lifestyle composite score is significantly associated with risk of gastric adenocarcinoma in a dose-response manner although its association was slightly weaker than the categorical-based composite score. As a complementary approach to the categorical-based composite score, Z-score method corroborates the inverse association between adaption of healthy lifestyle factors and reduced risk of gastric adenocarcinoma. Z-score-based composite score was based on the standardised values for each factor with equal weight, which addressed the issue of data overfitting. However, this complete data-driven approach may not optimise the stratification of study subjects at risk, thus could lead to a weaker association between the summed Z-score and gastric cancer risk.
The strengths of our study include the prospective study design, unique study population (Southeast Asians), a relatively large sample size (63 000 participants), long-term follow-up (17 years), and serological status of H. pylori infection. A summary lifestyle factor score can classify individuals into more homogenous groups by their risk profile that minimises potential misclassification and residual confounding effect. The main limitation is that all the information on lifestyle factors was self-reported at baseline with inherent non-differential misclassification that could bias the risk estimates towards null. Therefore, the observed risk estimates may be lower than the true effect of these lifestyle factors on the risk of gastric adenocarcinoma. It should also be noted that our estimate of daily sodium intake may have missed some dietary sources, which may cause some biased results. Given a prospective study design, both cancer cases and non-cancer individuals answered to the same dietary questionnaire. Thus, if there is any misclassification, it would be non-differential and lead to an underestimated risk association. Future studies using urinary excretion of sodium over 24 h are warranted to confirm our findings.
In conclusion, we observed a strong, statistically significant association between high composite score of protective lifestyle factors and reduced risk of gastric adenocarcinoma. Altogether these factors can account for up to almost half of disease burden in this Asian population with a very high prevalence of H. pylori. These findings are very encouraging for a comprehensive strategy for promoting healthy living that could be effective for primary prevention of gastric cancer even in populations with a relatively high background risk of gastric cancer and high prevalence of H. pylori.
Acknowledgments
We thank Siew-Hong Low of the National University of Singapore for supervising the field work of the Singapore Chinese Health Study and the Singapore Cancer Registry for assistance with the identification of cancer outcomes. The study was supported by grants from the National Institute of Health (NIH)/National Cancer Institute (NCI; grant numbers: R01 CA144034 and UM1 CA182876 (to J-M Yuan)).
This work is published under the standard license to publish agreement. After 12 months the work will become freely available and the license terms will switch to a Creative Commons Attribution-NonCommercial-Share Alike 4.0 Unported License.
Footnotes
Supplementary Information accompanies this paper on British Journal of Cancer website (http://www.nature.com/bjc)
The authors declare no conflict of interest.
Supplementary Material
References
- Abioye AI, Odesanya MO, Abioye AI, Ibrahim NA (2015) Physical activity and risk of gastric cancer: a meta-analysis of observational studies. Br J Sports Med 49(4): 224–229. [DOI] [PubMed] [Google Scholar]
- Ainslie-Waldman CE, Koh WP, Jin A, Yeoh KG, Zhu F, Wang R, Yuan JM, Butler LM (2014) Coffee intake and gastric cancer risk: the Singapore Chinese health study. Cancer Epidemiol Biomarkers Prev 23(4): 638–647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bianchini F, Kaaks R, Vainio H (2002) Overweight, obesity, and cancer risk. Lancet Oncol 3(9): 565–574. [DOI] [PubMed] [Google Scholar]
- Buckland G, Travier N, Huerta JM, Bueno-de-Mesquita HB, Siersema PD, Skeie G, Weiderpass E, Engeset D, Ericson U, Ohlsson B, Agudo A, Romieu I, Ferrari P, Freisling H, Colorado-Yohar S, Li K, Kaaks R, Pala V, Cross AJ, Riboli E, Trichopoulou A, Lagiou P, Bamia C, Boutron-Ruault MC, Fagherazzi G, Dartois L, May AM, Peeters PH, Panico S, Johansson M, Wallner B, Palli D, Key TJ, Khaw KT, Ardanaz E, Overvad K, Tjonneland A, Dorronsoro M, Sanchez MJ, Quiros JR, Naccarati A, Tumino R, Boeing H, Gonzalez CA (2015) Healthy lifestyle index and risk of gastric adenocarcinoma in the EPIC cohort study. Int J Cancer 137(3): 598–606. [DOI] [PubMed] [Google Scholar]
- Butler LM, Koh WP, Lee HP, Tseng M, Yu MC, London SJ Singapore Chinese Health Study (2006) Prospective study of dietary patterns and persistent cough with phlegm among Chinese Singaporeans. Am J Respir Crit Care Med 173(3): 264–270. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Conteduca V, Sansonno D, Lauletta G, Russi S, Ingravallo G, Dammacco F (2013) H. pylori infection and gastric cancer: state of the art (review). Int J Oncol 42(1): 5–18. [DOI] [PubMed] [Google Scholar]
- D'Elia L, Rossi G, Ippolito R, Cappuccio FP, Strazzullo P (2012) Habitual salt intake and risk of gastric cancer: a meta-analysis of prospective studies. Clin Nutr 31(4): 489–498. [DOI] [PubMed] [Google Scholar]
- Dartois L, Fagherazzi G, Boutron-Ruault MC, Mesrine S, Clavel-Chapelon F (2014) Association between five lifestyle habits and cancer risk: results from the E3N cohort. Cancer Prev Res (Phila) 7(5): 516–525. [DOI] [PubMed] [Google Scholar]
- de Martel C, Ferlay J, Franceschi S, Vignat J, Bray F, Forman D, Plummer M (2012) Global burden of cancers attributable to infections in 2008: a review and synthetic analysis. Lancet Oncol 13(6): 607–615. [DOI] [PubMed] [Google Scholar]
- Deurenberg P, Yap M, van Staveren WA (1998) Body mass index and percent body fat: a meta analysis among different ethnic groups. Int J Obes Relat Metab Disord 22(12): 1164–1171. [DOI] [PubMed] [Google Scholar]
- FDA US (2012) Harmful and Potentially Harmful Constitutents in Tobacco Products and Tobacco Smoke: Established List 03/23 2016 Available at www.fda.gov/TobaccoProducts/GuidanceComplianceRegulatoryInformation/ucm297786.htm.
- Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D, Bray F (2013) GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 11. International Agency for Research on Cancer: Lyon, France, 2016(02/14). [Google Scholar]
- Ferlay J, Soerjomataram II, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman DD, Bray F (2014) Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer 136(5): E359–E386. [DOI] [PubMed] [Google Scholar]
- Forman D, Bray F, Brewster DH, Gombe Mbalawa C, Kohler B, Piñeros M, Steliarova-Foucher E, Swaminathan R, Ferlay J (2014) Cancer Incidence in Five Continents Vol. X. International Agency for Research on Cancer: Lyon. [DOI] [PubMed] [Google Scholar]
- Furihata C, Ohta H, Katsuyama T (1996) Cause and effect between concentration-dependent tissue damage and temporary cell proliferation in rat stomach mucosa by NaCl, a stomach tumor promoter. Carcinogenesis 17(3): 401–406. [DOI] [PubMed] [Google Scholar]
- Hankin JH, Stram DO, Arakawa K, Park S, Low SH, Lee HP, Yu MC (2001) Singapore Chinese Health Study: development, validation, and calibration of the quantitative food frequency questionnaire. Nutr Cancer 39(2): 187–195. [DOI] [PubMed] [Google Scholar]
- Harrell FE, Lee KL (1986) Verifying Assumptions of the Cox Proportional Hazards Model. Proceedings of the Eleventh Annual. SAS User's Group International Conference. SAS Institute, Inc: Cary, NC, USA. [Google Scholar]
- IARC (2004) IARC Monographs on the Evaluation of Carcinogenic Risks to Humans,Tobacco Smoke and Involuntary Smoking. Vol. 83 p1187. Lyon, France. Available at: https://monographs.iarc.fr/ENG/Monographs/vol83/mono83.pdf. [PMC free article] [PubMed]
- IARC (2010) IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Alcohol consumption and ethyl carbamate. Vol. 96 p1278-1279. Lyon, France. Available at: http://monographs.iarc.fr/ENG/Monographs/vol96/mono96.pdf. [PMC free article] [PubMed]
- Iijima K, Fyfe V, McColl KE (2003) Studies of nitric oxide generation from salivary nitrite in human gastric juice. Scand J Gastroenterol 38(3): 246–252. [DOI] [PubMed] [Google Scholar]
- Jakszyn P, Agudo A, Lujan-Barroso L, Bueno-de-Mesquita HB, Jenab M, Navarro C, Palli D, Boeing H, Manjer J, Numans ME, Igali L, Boutron-Ruault MC, Clavel-Chapelon F, Morois S, Grioni S, Panico C, Tumino R, Sacerdote C, Quiros JR, Molina-Montes E, Huerta Castano JM, Barricarte A, Amiano P, Khaw KT, Wareham N, Allen NE, Key TJ, Jeurnink SM, Peeters PH, Bamia C, Valanou E, Trichopoulou A, Kaaks R, Lukanova A, Bergmann MM, Lindkvist B, Stenling R, Johansson I, Dahm CC, Overvad K, Olsen A, Tjonneland A, Skeie G, Broderstad AR, Lund E, Michaud DS, Mouw T, Riboli E, Gonzalez CA (2012) Dietary intake of heme iron and risk of gastric cancer in the European prospective investigation into cancer and nutrition study. Int J Cancer 130(11): 2654–2663. [DOI] [PubMed] [Google Scholar]
- Jakszyn P, Bingham S, Pera G, Agudo A, Luben R, Welch A, Boeing H, Del Giudice G, Palli D, Saieva C, Krogh V, Sacerdote C, Tumino R, Panico S, Berglund G, Siman H, Hallmans G, Sanchez MJ, Larranaga N, Barricarte A, Chirlaque MD, Quiros JR, Key TJ, Allen N, Lund E, Carneiro F, Linseisen J, Nagel G, Overvad K, Tjonneland A, Olsen A, Bueno-de-Mesquita HB, Ocke MO, Peeters PH, Numans ME, Clavel-Chapelon F, Trichopoulou A, Fenger C, Stenling R, Ferrari P, Jenab M, Norat T, Riboli E, Gonzalez CA (2006) Endogenous versus exogenous exposure to N-nitroso compounds and gastric cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC-EURGAST) study. Carcinogenesis 27(7): 1497–1501. [DOI] [PubMed] [Google Scholar]
- Klyosov AA (1996) Kinetics and specificity of human liver aldehyde dehydrogenases toward aliphatic, aromatic, and fused polycyclic aldehydes. Biochemistry 35(14): 4457–4467. [DOI] [PubMed] [Google Scholar]
- Ladeiras-Lopes R, Pereira AK, Nogueira A, Pinheiro-Torres T, Pinto I, Santos-Pereira R, Lunet N (2008) Smoking and gastric cancer: systematic review and meta-analysis of cohort studies. Cancer Causes Control 19(7): 689–701. [DOI] [PubMed] [Google Scholar]
- Lee HP (2015) Singapore Cancer Registry Interim Report. Trends in Cancer incidence in Singapore 2010-2014. Available at: https://www.nrdo.gov.sg/docs/librariesprovider3/default-document-library/cancer-trends-report-2010---2014_web.pdf?sfvrsn=0.
- Lunet N, Valbuena C, Vieira AL, Lopes C, Lopes C, David L, Carneiro F, Barros H (2007) Fruit and vegetable consumption and gastric cancer by location and histological type: case-control and meta-analysis. Eur J Cancer Prev 16(4): 312–327. [DOI] [PubMed] [Google Scholar]
- Mirvish SS (1996) Inhibition by vitamins C and E of in vivo nitrosation and vitamin C occurrence in the stomach. Eur J Cancer Prev 5(Suppl 1): 131–136. [PubMed] [Google Scholar]
- Pan WH, Flegal KM, Chang HY, Yeh WT, Yeh CJ, Lee WC (2004) Body mass index and obesity-related metabolic disorders in Taiwanese and US whites and blacks: implications for definitions of overweight and obesity for Asians. Am J Clin Nutr 79(1): 31–39. [DOI] [PubMed] [Google Scholar]
- Park JH, Kim YC, Koo HS, Oh SW, Kim S, Chin HJ (2014) Estimated amount of 24-hour urine sodium excretion is positively correlated with stomach and breast cancer prevalence in Korea. J Korean Med Sci 29(Suppl 2): S131–S138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Roberts DL, Dive C, Renehan AG (2010) Biological mechanisms linking obesity and cancer risk: new perspectives. Annu Rev Med 61: 301–316. [DOI] [PubMed] [Google Scholar]
- Shin VY, Wu WK, Chu KM, Koo MW, Wong HP, Lam EK, Tai EK, Cho CH (2007) Functional role of beta-adrenergic receptors in the mitogenic action of nicotine on gastric cancer cells. Toxicol Sci 96(1): 21–29. [DOI] [PubMed] [Google Scholar]
- Shin VY, Wu WK, Ye YN, So WH, Koo MW, Liu ES, Luo JC, Cho CH (2004) Nicotine promotes gastric tumor growth and neovascularization by activating extracellular signal-regulated kinase and cyclooxygenase-2. Carcinogenesis 25(12): 2487–2495. [DOI] [PubMed] [Google Scholar]
- Sorbye H, Svanes C, Stangeland L, Kvinnsland S, Svanes K (1988) Epithelial restitution and cellular proliferation after gastric mucosal damage caused by hypertonic NaCl in rats. Virchows Arch A Pathol Anat Histopathol 413(5): 445–455. [DOI] [PubMed] [Google Scholar]
- Spiegelman D, Hertzmark E, Wand HC (2007) Point and interval estimates of partial population attributable risks in cohort studies: examples and software. Cancer Causes Control 18(5): 571–579. [DOI] [PubMed] [Google Scholar]
- Tramacere I, Negri E, Pelucchi C, Bagnardi V, Rota M, Scotti L, Islami F, Corrao G, La Vecchia C, Boffetta P (2012) A meta-analysis on alcohol drinking and gastric cancer risk. Ann Oncol 23(1): 28–36. [DOI] [PubMed] [Google Scholar]
- Tu S, Bhagat G, Cui G, Takaishi S, Kurt-Jones EA, Rickman B, Betz KS, Penz-Oesterreicher M, Bjorkdahl O, Fox JG, Wang TC (2008) Overexpression of interleukin-1beta induces gastric inflammation and cancer and mobilizes myeloid-derived suppressor cells in mice. Cancer Cell 14(5): 408–419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- US Department of Agriculture (USDA) (1975) Agriculture Handbook. Vol. 456, p 31.
- Wang Q, Chen Y, Wang X, Gong G, Li G, Li C (2014) Consumption of fruit, but not vegetables, may reduce risk of gastric cancer: results from a meta-analysis of cohort studies. Eur J Cancer 50(8): 1498–1509. [DOI] [PubMed] [Google Scholar]
- Wang XQ, Terry PD, Yan H (2009) Review of salt consumption and stomach cancer risk: epidemiological and biological evidence. World J Gastroenterol 15(18): 2204–2213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- WHO (2004) Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 363(9403): 157–163. [DOI] [PubMed] [Google Scholar]
- Wiseman M (2008) The second World Cancer Research Fund/American Institute for Cancer Research expert report. Food, nutrition, physical activity, and the prevention of cancer: a global perspective. Proc Nutr Soc 67(3): 253–256. [DOI] [PubMed] [Google Scholar]
- Yang P, Zhou Y, Chen B, Wan HW, Jia GQ, Bai HL, Wu XT (2009) Overweight, obesity and gastric cancer risk: results from a meta-analysis of cohort studies. Eur J Cancer 45(16): 2867–2873. [DOI] [PubMed] [Google Scholar]
- Zhang L, Eslick GD, Xia HH, Wu C, Phung N, Talley NJ (2010) Relationship between alcohol consumption and active Helicobacter pylori infection. Alcohol Alcohol 45(1): 89–94. [DOI] [PubMed] [Google Scholar]
- Zhu H, Yang X, Zhang C, Zhu C, Tao G, Zhao L, Tang S, Shu Z, Cai J, Dai S, Qin Q, Xu L, Cheng H, Sun X (2013) Red and processed meat intake is associated with higher gastric cancer risk: a meta-analysis of epidemiological observational studies. PLoS One 8(8): e70955. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.