Associations between Preserved Foods and Breast Cancer Risk in Hong Kong Chinese Women

Female breast cancer is a major public health burden in the world. During 2012 through 2016, the incidence rate of overall breast cancer increased by 0.3% per year worldwide (1), and it becomes the most frequently diagnosed cancer among women in 2020 (2). Breast cancer is driven by several well-established risk factors, including reproductive factors and family breast cancer history. Some dietary and lifestyle factors were also found to be associated with increased risk of breast cancer (3–5), whereas many of them are the modifiable risk factors such as alcohol consumption, smoking, lack of physical activity levels, and meat intake rich in fat, which collectively contributed to around 25%–30% of breast cancer etiology (4).

Food preservation refers to the process to extend the longevity of food use by salting, pickling, sugaring, and canning to keep from spoilage for the prolonged storage. Canned food is the popular preserved foods in Hong Kong and many other countries due to its easy storage and low price to satisfy the busy schedule of modern life. Nitrate and nitrite are commonly used as additives to improve quality of preserved foods, including cured foods, and protect against microbial contamination, but they are precursors of exogenous sources of N-nitroso compounds that are probable carcinogen defined by the International Agency for Research on Cancer (5–7). In addition, canned foods may be contaminated by free bisphenol A (BPA) released from the lining of cans (8). BPA has a potential to mimic the endogenous estrogen by activating estrogen receptors leading to proliferation of breast cells but suppression of apoptosis, and thereby contributing breast cancer development particularly for the hormone receptor–positive breast cancer (9).

Recent epidemiological studies tend to support that consumption of preserved foods products may increase the risk of hormone-regulated cancer (prostate, ovarian and breast) and colorectal cancer (10–16), although these evidences were not completely consistent. A case–control study in 2,376 Spain women supported that processed/cured meat increased the risk of breast cancer (17), but results from a case–control study in 1,448 women in China did not support an association between breast cancer and preserved meat or vegetable (18). Furthermore, a small case–control study in Turkey among 243 women found that canned food was not associated with breast cancer risk, and this lack of association may be related to low statistical power due to a low prevalence of canned food consumption in Turkish society (19). Nevertheless, a retrospective cohort study among 140 women in Iran provided interesting result that avoiding consumption of canned foods could improve the survival of patients with breast cancer (20). It is also important to note the differences in preserved food habits between Asian and Western countries. As the types of cured meat and vegetables as well as their consumption habits are different between Western and Asian countries (Lap Cheong in Hong Kong vs. Salami in Western countries), the associations of consumption of these preserved food types with breast cancer risk in Western countries may differ from those of the Asian countries, and hence, the findings obtained from the Western population may not be directly applicable to the Hong Kong population. This case–control study aimed to examine the association of preserved foods consumption (i.e., cured meat, pickled vegetable, canned meat, and canned fruit/vegetable) with the risk of breast cancer and its biology subtypes among Hong Kong Chinese women.

Breast cancer cases of this study were consecutively recruited from the Department of Surgery or Department of Clinical Oncology of three public hospitals in Hong Kong from November 2011 to January 2018. Eligible cases were Chinese women ages 24–84 years old who were newly diagnosed with primary invasive breast cancer (International Classification of Disease, Tenth Revision, code 50) within three months before the interview, with a response rate of 87.5%. Only patients with newly diagnosed invasive breast cancer were treated as the eligible cases, and 134 in situ cases of breast cancer were excluded from this study. Controls were frequency matched with the cases by 5-year age group who were free from history of any cancer. All controls were recruited from the same hospital as the cases from the Department of Medicine or Surgery with medical conditions unrelated to breast cancer diseases (i.e., diseases of digestive system, diseases of circulatory system, diseases of the genitourinary system, endocrine, nutritional and metabolic diseases, and diseases of the nervous system). Overall, the response rate of controls was 88.4%. We also excluded the controls with prior history of physician-diagnosed cancer in any site or recently diagnosed breast-related diseases. We interviewed controls within 6 months after the cases were recruited. All cases and controls were interviewed face-to-face by our trained interviewers. We used the standardized questionnaire to obtain socio-demographic information, lifestyles habits, dietary habits, reproductive-related factors, anthropometric factors, and family history of breast cancer (21).

This study was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committees (CREC no. 2013.207) and the Kowloon West Cluster [CREC no. KW/EX-12–081 (54–13)]. Written informed consent has been obtained before the interview. The studies were conducted in accordance with Declaration of Helsinki.

Information on habitual dietary habits was obtained by a validated Block food frequency questionnaire developed by the National Cancer Institute with some modification (22), with the frequencies reported according to categories of: no consumption/less than once per year, less than once per month, 1–3 times per month, 1–3 times per week, 4–6 times per week, and more or equal to once per week. Block food frequency questionnaire has been used in two previous studies in Hong Kong population (23, 24), showing significant results with biological plausibility. We invited our participants to report the frequency of preserved food consumption five years before our recruitment time frame. We asked each participant to report their frequency of consumption of cured meat (e.g., the Chinese-style dry-cured sausage, preserved meat and duck leg, etc.), pickled vegetables (e.g., the Chinese-style preserved vegetables, pickled cabbage, salt mustard greens, etc.), canned meat (e.g., luncheon pork, fried dace with salted black beans), and canned fruit/vegetable (e.g., canned pineapple, canned corn, canned mixed fruit, etc.). We further stratified cured meat, pickled vegetable, canned meat and canned fruit/vegetables into no consumption/< 1 per year, ≥1 per year ≤1 per week, and ≥1 per week.

We extracted tumor characteristics from breast cancer pathological report according to hospital medical records, including estrogen receptor (ER), progesterone receptor (PR), HER2, Ki67 and histological grade. Positive status of ER or PR was defined as: Allred score of 3/8 or above; or H score of 51/300 or above; or at least 1% positivity of nuclear staining (25, 26). Positive HER2 status was defined as IHC score of 3+ or FISH or chromogenic in situ hybridization (CISH) showed HER2 gene amplification, which was tested when IHC score was equivocal (i.e., score of 2). IHC score of 0 and 1+ were defined as HER2-negative (HER2⁻) Given that about 83% of HER2 IHC result of 2+ patients with breast cancer were found to be HER2⁻ after subsequent FISH/CISH test we defined IHC 2+ result with missing FISH or CISH information as HER2⁻.

On the basis of the clinicopathologic surrogate definition as recommended by 2013 St. Gallen International Expert Consensus (27), we further classified our cases into four biology subtypes: Luminal A-like (ER⁺ and PR⁺ and HER2⁻ and either Ki-67 <20% or histological grade I/II); luminal B-like [(ER⁺ and HER2⁻ and either PR⁻ or Ki-67 ≥20% or histological grade III) OR (ER⁺ and HER2⁺)]; HER2-positive (nonluminal; ER⁻ and PR⁻ and HER2⁺); and triple-negative (ER⁻ and PR⁻ and HER2⁻; Supplementary Table S1).

We used independent t tests and the χ² test to compare the differences between cases and controls for the continuous and categorically selected characteristics, respectively. Selected characteristics included sociodemographic information {age at interview (continuous), education attainment (primary school or below, secondary educational level, tertiary education level), tobacco smoking (smoker, ex-smoker, and non-smoker), alcohol drinking (drinker and non-drinker), other dietary habits [green vegetable (<1 per day and ≥1 per day), meat consumption (<1 per day and ≥1 per day), citrus consumption (<1 per week and ≥1 per week), dairy product consumption (no consumption/<1 per year and ≥1 per year ≤1 per week, and ≥1 per week), soybean product consumption (no consumption/<1 per year and ≥1 per year ≤1 per week, and ≥1 per week), deep fried food consumption (no consumption/<1 per year, ≥1 per year ≤1 per week, and ≥1 per week)], menopause (no or yes), weight and height [converted to body mass index (BMI): weight (kilograms)/height (meter²; underweight <18.5, normal 18–24.9, overweight 25–29.9, and obese≥30)], chronic diseases condition [hypertension status (no or yes), cardiovascular diseases status (CVD; no or yes)], family history of breast cancer (no or yes), and physical activity (inactive and regular exercise)}.

We analyzed the associations between different preserved foods (i.e., cured meat, pickled vegetable, canned meat, and canned fruit or vegetable) and breast cancer risk with an unconditional logistic regression models [in odds ratios (OR) and 95% confidence intervals (95% CI)], using no consumption/<1 per year as reference category. We further performed multivariable models for the associations between preserved foods and breast cancer, using a backward stepwise approach (determined by Akaike information criterion) using potential confounders. Potential confounders included age at interview, green vegetable intake, dietary product consumption, soybean product consumption, BMI, family history of breast cancer, and physical activity. We used the multivariable models to test the association of the frequency of preserved foods consumption with breast cancer risk, and used trend test to evaluate a dose–response association between the frequencies of preserved foods intake with an ordered relationship across the categories and the risk of breast cancer development.

We performed a stratified analysis according to the breast cancer subtypes. We awared that some previous studies only use hormone receptor status, HER2 status, and the Ki67 index to classify breast cancer biology subtypes (28), and thus we further performed a sensitive analysis according to the classification of luminal A-like (ER⁺ and PR⁺ and HER2⁻ and Ki-67 <20%) and luminal B-like [(ER⁺ and HER2⁻ and PR⁻ or Ki-67 ≥20%) OR (ER⁺ and HER2⁺)] and compared with all controls. We also conducted a stratified analysis according to specific ER and PR, compared with all controls.

In addition, a previous study suggested that preserved food may increase the risk of overweight and obesity (29), and overweight or obesity was associated with increased breast cancer risk regardless of subtypes in Hong Kong Chinese women (30). Thus, we further conducted a stratified analysis by BMI from the multivariable models in both cases and controls. Furthermore, preserved food may increase the risk of hypertension and CVD (31–33), and a study reviewed that hypertension and CVD was associated with increased breast cancer risk, especially among postmenopausal women (34, 35). Therefore, we also performed a sensitive analysis by excluding women with physician diagnosed hypertension or CVD, and conducted a stratified analysis according to the menopausal status in both cases and controls.

Furthermore, about 5.0% of participants had a missing value for educational attainment, 5.5% had missing value of menopausal status information, 11.0% of them had missing values for height or weight, 3.0% of them had missing values of hypertension or CVD, and 18.4% missing their breast cancer biology subtypes. We used multiple imputations with five imputed datasets to address the problem of missing data by using the mice package in R (36). All statistical analyses were conducted with RStudio version 1.4.1103 (RStudio, PBC), and two-sided P value less than 0.05 was considered statistically significant.

The results and materials described previously in the article, and the relevant data related to the consumption of preserved food that used in this article, are available from the corresponding author on reasonable request.

In this case–control study, we recruited 1,307 incident breast cancer cases and 1,050 hospital controls. Table 1 shows the distribution of selected characteristics among breast cancer cases and hospital controls. Women with breast cancer were more likely to be older (57.6±11.7 vs. 54.7±11.7), overweight (27.2% vs. 19.7%)/obese (6.8% vs. 4.6%), have family history of breast cancer (12.2% vs. 4.0%), and more likely to consume dairy and soybean product. Other selected variables, including educational attainment, menopausal status, tobacco smoking, alcohol consumption, meat and green vegetable intake, deep dried food and citrus consumption, and hypertension and cardiovascular diseases were similar between cases and controls.

When we examined the association between preserved food and risk of breast cancer (Table 2), we found that cured meat intake was positively associated with breast cancer risk [adjusted OR (AOR), 1.49; 95% CI, 1.23–1.80], and showed an exposure–response relationship with increasing frequency of consumption (P_trend <0.01). Compared with no cured meat consumption/<1 per year of cured meat intake, cured meat intake ≥1 per week was associated with an AOR of 2.17 (95% CI, 1.20–4.07; Table 2). The results remained consistent when we conducted a stratified analysis by BMI or sensitivity analysis by excluding women with physician diagnosed hypertension or CVD, despite the numbers of underweight and obese women are too small to find significant results (Supplementary Tables S2 and S3). Post-menopausal women with cured meat consumption ≥1 per week had a higher risk of breast cancer (AOR, 3.44; 95% CI = 1.46–9.14; Table 3). The estimated association between cured meat consumption and the risk of breast cancer was similar after performing the multiple imputation analyses (Supplementary Table S4). In addition, compared with women with no canned fruit/vegetable intake habit, women with canned fruit/vegetable intake ≥1 per week had a higher risk of breast cancer (OR, 1.28; 95% CI, 1.00–1.65), but it was no longer significant after the adjustment (Table 2). Similar associations were retained between canned fruit/vegetable consumption and breast cancer after performing the multiple imputation analyses (Supplementary Table S4).

Of the 1,307 cases, 1,067 (81.6%) breast cancer cases had known status of ER, PR, HER2, Ki67 or histological grade (Supplementary Table S1). Among cases with known pathological data, 334 (31.3%) of them were diagnosed with luminal A-like, 467 (43.8%) were with luminal B-like, 115 (10.8%) were with HER2-positive, and 151 (14.2%) were diagnosed with triple-negative breast cancer (TNBC) subtypes.

We used multiple imputations to perform further analysis stratified by biology subtypes. Consistent with the main findings, cured meat was associated with the risk of all breast cancer subtypes, with an AOR ranged from 1.15 to 1.95 (Table 4; Supplementary Tables S4 and S5). These association remained positive regardless the specific ER and PR (AOR ranged from 1.29 to 1.38). Pickled vegetable intake was positively associated with TNBC subtype exclusively (AOR, 1.98; 95% CI, 1.14–3.67). Habitually canned fruit/vegetable intake was associated with an increased risk of HER2-positive breast cancer subtype (AOR, 1.80; 95% CI, 1.18–2.75).

This case–control study demonstrated that cured meat consumption was associated with increased risk of breast cancer; there was an exposure–response relationship with increasing frequency of consumption. We did not observe any association between the consumption of canned meat and risk of breast cancer. High frequency of canned fruit/vegetable consumption was associated with risk of breast cancer, and this association was particularly strong for HER2-positive breast cancer subtype; however, this association was not statistically significant after further adjustment.

Limited epidemiological evidence has evaluated the association between preserved food and breast cancer risk. We found that high frequency of cured meat intake was associated with 2.66 odds of breast cancer, and this association presented a positive gradient of breast cancer risk with increasing frequency of consumption. The results remained significant or borderline significant after narrowed down the CI (i.e., 99% CI) when performing logistic regression. Previous studies reported that cured meat consumption increased the risk of breast cancer (17), especially for hormone receptor–negative breast cancer (17, 37); and several other types of cancer, including prostate and colorectal cancer (13–16). Previous studies adopted different definitions of preserved meat and the types of cured meat, showing inconsistent findings for the association between preserved meat and the risk of breast cancer (18, 38). In Hong Kong, cured meat commonly preserved with salt. Increased salt intake may increase the risk of hypertension and CVD (32, 39) that were found to be associated with an increased risk of breast cancer (34, 35). Our findings remained consistent after excluding women with physician diagnosed hypertension or CVD. Several biological mechanisms are proposed for the link between cured meat and breast cancer risk. One putative mechanism is related to endogenously formed N-nitroso compounds during the curing process. Cured meat contains high levels of N-nitroso compounds, including N-nitrosodimethylamine (NDMA), N-nitrosarcosine (NSAR), and N-nitrosoproline (NPRO; refs. 40, 41), in which N-nitroso compounds are probably carcinogenic to human (7, 42, 43). The acceptable daily intake level of nitrate and nitrite is set to below 3.7 and 0.07 mg/kg body weight, respectively (44). According to a study in Hong Kong in 2010, the average exposure to nitrate from dietary intake of vegetables for adult consumer was estimated to be 3.5 mg kg⁻¹ body weight per day (45). In spite of the fact that there is no clear evidence showing the linkage between N-nitroso compounds and breast cancer development, a study demonstrated that ranitidine, which may yield NDMA, was positively associated with risk of breast cancer (46). In addition, most of the Chinese-style cured meat is prepared with a mixture of chunks of fat and lean pork meat or liver stuffed with fatty pork. Previous studies showed that red meat consumption increased the risk of breast cancer (38, 47). The positive association between cured meat intake and breast cancer risk in this study might contributed to the red meat content of cured meats. In addition, fat intake is directly associated with increased risk of developing breast cancer, which is shown to be independent from BMI (48). Taken together, consumption of cured meat products containing high level of N-nitroso compounds, which are also rich in fat contents may result in increased risk of breast cancer.

In addition, we found that pickled vegetable was associated with increased risk of TNBC subtype. Despite the underline mechanism is not clear, two previous studies from Korea and Iran found that pickled vegetables intake were positively associated with risk of breast cancer (49, 50). Although Chinese pickles included pickled vegetables that can also generate potentially carcinogenic N-nitroso compounds, the concentration is much lower than cured meat (40). In addition, the results may be unstable due to only a small proportion of women (n = 13) with TNBC subtype in the reference group who did not consume pickled vegetables. Taken together, the association between pickled vegetables and the risk of breast cancer may be weaker than such association with cured meat.

Canned food may contain free BPA (8) that can mimic the actions of endogenous estrogen by activating estrogen receptors. BPA was found to cause proliferation of breast cells and led to an increased risk of breast cancer in rat study (9). In our study, consuming canned fruit or vegetable was associated with increased risk of breast cancer, which however was not statistically significant after adjustment. On the other hand, most fruit or vegetable canned in syrup contains high levels of sugar, the contribution of added sugar in canned fruits or vegetable may increase their body weight. Obesity is a strong risk factor of breast cancer (30, 51); women with a high frequency of canned fruit or vegetable intake may increase their body weight thus increasing the risk of breast cancer. Furthermore, the consumption of canned food was low and the statistical power may not be strong enough to test the hypothesis that BPA related to canned food consumption links to the increased risk of breast cancer, especially for receptor-positive breast cancer.

To the best of our knowledge, this is the first study examined the association between different preserved foods (i.e., cured meat, pickled vegetables, canned meat and canned fruit/vegetable) and breast cancer risk in Hong Kong Chinese women. We recruited our participants from three large public hospitals with a high response rate (88.4%). Test–retest reliability conducted at least one month by telephone interview after the initial recruitment presented a good agreement for tea drinking (Kappa = 83%) among 158 cases and 153 controls (52) with the initial interview. However, we acknowledge some limitations to this study. Limited sample size should be aware especially in statistical analysis stratified by subtypes of breast cancer and BMI categories. We did not collect data on the proportion of preserved food intakes, in which the amount of food intake may influence the results. In addition, we did not collect the information of type of preserved methods and thus we cannot estimate the levels of nitrate exposure in the cured meat. However, methods of producing cured meat in Hong Kong, involving sodium chloride or dried cured (e.g., dry-cured sausage), may also contain sodium nitrate (53, 54), which could be converted into toxic nitrite by bacteria during processing. We collected the total meat intake; however, we did not specify the consumption of red meat and processed meat from the questionnaire as this part of information was not collected. We understood that the Block food frequency questionnaire was not validated in Chinese, but it demonstrated sensitive to detect a significant association of intake of tea, deep fried food or pickled vegetable with breast cancer and prostate cancer (23, 24). Recall bias is a major concern of most case–control studies. We tried to minimize this possible bias by introducing this study as a “general women health study.” We also asked them to recall their regular dietary habit to reduce the recall bias. However, women tend to over-report their height and under-report their weight, thus the BMI may be underestimated (55). Selection bias is possible but it was not a major issue of this study, as we have a high response rate of cases (87.5%) and controls (88.4%), and also the structure of our cases was similar to that was reported by the Hong Kong Cancer Registry (56). Because of various health conditions across different diseases, the dietary habits in hospital controls may not represent that of the general population. Nevertheless, a previous case–control study found a similar magnitude of risk between using the general population controls and hospital controls with a variety of disease types (57); thus, we recruited controls with a broad range of disease diagnoses that were not related to breast diseases from the same hospitals. Misclassification bias may be a concern as we classified HER2 IHC 2⁺ patients with missing FISH/CISH data into HER2 negative. However, we noted that about 83% of HER2 IHC 2+ cases were found to be HER2⁻ on subsequent FISH/CISH test among our HER2 IHC 2⁺ cases of this study as well as cases in the database of one of our collaborating hospitals, suggesting that our major results are less likely to be affected by the misclassification. In addition, the tumor markers will be missing if the patients were not to undergo surgery at our recruitment hospital. It may be because the patients with breast cancer chose other hospitals to complete their operation (e.g., private hospital) as the waiting time in the public hospital was too long (i.e., missing at random); or the tumor stage is too advanced and the physician did not recommend patient to do the surgery (i.e., missing at not random), and therefore we may miss the data for the patients with advanced stages when we performed the stratified analysis according to breast cancer subtypes. However, the results from multiple imputations of tumor data yielded consistent results, suggesting that the bias due to missing tumor data is unlikely to have impact on our results. For the problem of multiple testing, we narrowed down the CI when performing logistic regression (i.e., 99%CI) in the main model to compensate the number of inferences that made and reduced the probability of getting a false positive result. The association between cured meat and risk of breast cancer remained significant or borderline significant, but no longer significant for the association between canned fruit/vegetable intake and breast cancer risk. Despite we used backward stepwise approach to use potential confounders, we cannot eliminate plausible alternative explanations by unmeasured confounders for the associations between preserved foods and the risk of breast cancer. Lack of data on cancer stages might also be a concern as the association with varies stages may be suggesting different effect (e.g., the association with a higher-stage disease might support an effect in tumor progression whereas a consistent association across stages might suggest an early carcinogenic effect). We used multiple imputations to account for missing data and found robust associations between consumption of preserved food and the risk of breast cancer.

In conclusion, this study demonstrated a positive association between cured meat intake and breast cancer risk that had never been reported in Chinese women, thus have added further evidence to the current literature that cured meat intake might be a potential novel risk factor for breast cancer, nevertheless, the findings from this case–control study would have to be confirmed by future large prospective cohort studies.

L.A. Tse reports grants from Research Grants Council of Hong Kong during the conduct of the study. No disclosures were reported by the other authors.

The funders had no role in study design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval.

P.M.Y. Lee: Conceptualization, formal analysis, investigation, writing–original draft, project administration. C.-H. Kwok: Supervision, writing–review and editing. Y.-K. Tsoi: Writing–review and editing. C. Wu: Writing–review and editing. S.-H. Law: Supervision, writing–review and editing. K.-H. Tsang: Writing–review and editing. Y.-C. Yeung: Writing–review and editing. W.C. Chan: Writing–review and editing. G.M. Tse: Writing–review and editing. K.K.-W. Yuen: Writing–review and editing. R.K.W. Hung: Writing–review and editing. F. Wang: Methodology, writing–review and editing. L.A. Tse: Resources, data curation, funding acquisition, methodology, writing–review and editing.

This study was supported by the funding from Research Grants Council of Hong Kong (grant number 474811). The authors acknowledge Magdalene Yin Shan Leung, Tess Hiu Man Tsoi, Ivy Hung Kuen Hsu, and Apple Kit Ping Kwok for their assistance on data collection. We thank all physicians, pathologists, nurses, and participants from participated clinics in Princess Margaret Hospital, Yan Chai Hospital and North District Hospital. L.A. Tse received the grant which supported by the Research Grants Council of Hong Kong (grant number 474811).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.