Intake of Alcohol and Tea and Risk of Nasopharyngeal Carcinoma: A Population-Based Case–Control Study in Southern China

The incidence of nasopharyngeal carcinoma (NPC) varies substantially by geographic location. Recent data from the International Agency for Research on Cancer showed that 82.9% of incident NPC cases in 2018 occurred in East Asia, South-East Asia, and South-Central Asia (1). Studies especially among high-risk areas that can clarify the associations of modifiable risk factors with the risk of NPC are of potential importance for primary cancer prevention (2–6). However, so far there are still few well-established modifiable risk factors for NPC, except for smoking, poor oral health, and perhaps some dietary components (7–9).

Previously reported associations between alcohol or tea intake and the risk of NPC remain inconsistent. Most case–control studies reported no significant association with alcohol intake (10–15), whereas two systematic reviews and some other studies showed that alcohol drinking had a J-shaped dose–response trend: light or moderate alcohol drinkers exhibited a significant or nonsignificant decreased NPC risk (16), while heavy drinkers had an increased risk (17–22). Several studies investigated the potential association between tea drinking and the risk of NPC: studies in Guangdong and Taiwan or in Los Angeles (14, 17, 23, 24) reported that tea drinkers had a decreased NPC risk, whereas one study in Turkey found an increased risk among tea drinkers who consumed <10 glasses per day (10).

However, most of these studies lacked detailed evaluation of intake of alcohol and tea, or they did not collect precise information on different types of alcoholic beverages or tea (2, 10, 19). This information might be important for understanding the dose–response pattern of a potential causal association between consumption of ethanol or tea and NPC risk. To this end, we took advantage of our NPC Genes, Environment, and Epstein-Barr virus (NPCGEE) study, a large population-based case–control study conducted in southern China, to investigate associations between detailed alcohol and tea drinking over time and NPC risk.

NPCGEE is a population-based case–control study conducted in three areas of southern China. The rationale and design of the NPCGEE study have been described in detail elsewhere (8, 9, 25, 26). Briefly, between 2010 and 2014, we enrolled a total of 2,554 incident NPC cases and 2,648 population-based controls. All histopathologically diagnosed cases of NPC were ascertained by a rapid reporting system. Controls were randomly selected every 6 to 12 months from total population registries covering three geographic areas, with frequency matching to the 5-year age and sex distribution of the cases by geographic region. The overall participation rates for cases and controls were 83.8% and 82.7%, respectively. Inclusion criteria for both cases and controls were age 20 to 74 years, current residence in the study area during the recruitment period, and no previous malignant disease or congenital or acquired immunodeficiency. All eligible participants were interviewed in-person by trained interviewers using an electronic questionnaire to collect data on individual lifestyle factors.

Institutional review boards or ethics vetting boards from all study centers approved this study. Eligible subjects granted written or oral informed consent.

All eligible cases and controls with available alcohol drinking and tea drinking information were included in this analyses. We excluded subjects with data lost during the upload process (n = 18), missing information on alcohol drinking or tea drinking (n = 57), and those with interview of poor quality or missing on confounding factors listed in Table 1 (n = 140). The final analytic population included 2,441 NPC cases and 2,546 controls (Supplementary Fig. S1).

All participants were first asked whether they had ever consumed alcohol at least once a week for 6 months. Those who responded “yes” further reported their drinking patterns. Information was collected for seven types of alcoholic beverages: beer, yellow rice wine, fruit wine, moderate distillate spirits, strong distillate spirits, and imported liquors. The corresponding ages at starting and stopping (if applicable), cups per day (for beer, yellow rice wine, fruit wine, and other wines) or liang per day (for distillate spirits and imported liquor), and weekly and monthly frequency of intake were ascertained. One cup was defined as 200 mL and one liang (Chinese unit) as 50 mL.

We assumed an ethanol content of 4.3 g/100 mL for beer, 10.35 g/100 mL for yellow rice wine, 10.2 g/100 mL for fruit wine or wines, 30.8 g/100 mL for moderate distillate spirits, 37.2 g/100 mL for strong distillate spirits, and 42.6 g/100 mL for imported liquors, based on China Food Composition Table (27). For each specific beverage and for combined alcoholic beverages, mean daily ethanol intake (g/day) was calculated by multiplying ethanol content (g/mL), and volume of alcohol intake (mL), weighted by lifetime days of intake.

Similarly, habitual tea drinkers were defined as those who ever drank tea at least once a week for 6 months. Tea drinkers further reported their drinking habits for five types of tea: black tea, green tea, jasmine tea, oolong tea, and pu'er tea. For each type of tea, the collected information included serving size [small cup (10 mL), medium cup (50 mL), or large cup (300 mL)], ages at starting and stopping (if applicable), cups per day, temperature, and concentration. Temperature of tea was defined according to the waiting time before drinking, for example, “hot” referred to drinking boiled tea immediately, whereas “cold” referred to drinking tea after more than 30 minutes or as a chilled drink. The concentration for tea intake (strong or not) was recorded according to the subjective taste of individual subject.

Lifetime volume (mL) of intake of each specific type of tea was calculated by multiplying cups per day by serving size and by days of drinking each specific type of tea, and mean daily tea intake for all kinds of tea combined was calculated by summing the cumulative lifetime volume of intake of each specific type of tea, divided by days of lifetime exposure.

Two of the most challenging aspects of conducting a valid population-based case–control study were to ensure complete case ascertainment and unbiased control selection. Therefore, we first chose a study setting with a relatively high incidence of NPC and a relatively low rate of residential mobility. Then we developed a rapid case ascertainment system involving a network of physicians who diagnosed and/or treated NPC. Frequency-matched controls were randomly selected from a computerized, continuously updated total population registry to ensure the representativeness of controls, and we used multiple ways to contact the selected controls to improve the participation rate. In addition, we used several approaches to reduce the risk of recall bias due to the use of self-reported, retrospectively collected data. First, we developed an electronic questionnaire with built-in logic checking and several rounds of pilot testing. Second, we used an interviewer manual that described standard survey techniques to be implemented for all participants, common errors to avoid, and pictures of serving sizes for food and beverage items, including alcohol and tea drinking. Third, interviews were audiotaped for quality control, and subjects were recontacted, if necessary, to clarify responses. The detailed quality-control approaches of the NPCGEE study have been described previously (26).

Characteristics among NPC cases and population controls were compared using the χ² test for categorical variables. To accommodate different patterns of alcohol and tea drinking between men and women, we evaluated associations stratified by sex. We used adjusted unconditional logistic regression models to estimate ORs and their corresponding 95% confidence intervals (CI). Those who never habitually drank alcohol (i.e., habitual nondrinkers) were used as the reference category for analyses of alcohol, and those who never habitually drank tea were the reference category for analyses of tea.

For both analyses, former drinkers were defined as ever drinkers with >2 years since cessation. Intake of total ethanol was categorized as 0.1–9.9, 10.0–19.9, 20.0–49.9, 50.0–89.9, and ≥90.0 g/day, according to cut-off points commonly used for ethanol intake. Intake of total tea was categorized as ≤150.0, 150.1–500.0, 500.1–900.0, 900.1–1800.0, or >1800.0 mL/day according to quintiles among control subjects. Duration of alcohol drinking or tea drinking was categorized into quartiles among male controls and all controls, respectively. For the analysis of specific tea types, black and pu'er teas (fermented teas) were combined and green and jasmine teas (nonfermented teas) were combined.

Potential confounders were selected on the basis of known or suspected risk factors for NPC that are plausibly related to patterns of tea or alcohol intake. Minimally adjusted models included age at diagnosis/interview, sex, and residential area. Fully adjusted models were additionally adjusted for education level, current housing type, current occupation, cigarette smoking, body mass index (BMI) at age 20 years, salt-preserved fish consumption during 2000–2002, NPC among first-degree relatives, vegetable consumption in 2000–2002, frequency of tooth brushing, and frequency of herbal soup consumption.

Tests of linear trend with NPC risk were performed using the median value of each category in the logistic regression models. We also used restricted cubic splines with three knots as a nonparametric approach to examine the potential nonlinear relation between ethanol consumption or tea intake and risk of NPC (28). Overall and nonlinear associations in the nonparametric regression analysis were tested using the Wald χ² test. We also evaluated the presence of effect modification by age, sex, socioeconomic status indicators, and current smoking status by evaluating stratified results.

Analyses were performed with SAS version 9.4 (SAS Institute). The two-sided significance level for all statistical tests was 0.05.

Table 1 presents the distribution of characteristics among 2,441 NPC cases and 2,546 controls, overall and stratified by sex. Both among overall participants and stratified by sex, in comparison with controls, cases tended to be younger, to have a lower education level, to live in a cottage (clay brick structure) or boat, to be blue-collar workers, to have higher BMI at age 20 years, to be more likely to have a family history of NPC, to brush teeth less frequently, and to have lower consumption of herbal soup. Male cases were more likely to be ever smokers.

Table 2 shows the ORs and 95% CIs for associations between alcohol drinking status and risk of NPC among subjects combined and stratified by sex. Similar estimates were found in minimally and fully adjusted models. Compared with nondrinkers of alcohol, former drinkers had a nonsignificantly increased risk of NPC (OR, 1.31; 95% CI, 0.99–1.74), and no excess risk was found among current drinkers (OR, 1.08; 95% CI, 0.93–1.25). We conducted further analyses for heavy ethanol intake and found no excess NPC risk for alcohol drinkers of 90.0–119.9 g/day (OR, 1.09; 95% CI, 0.68–1.73), but a significantly 1.57-fold (95% CI, 1.01–2.46) increased risk for alcohol drinkers of ≥120.0 g/day. No significant associations were found for former or current alcohol drinkers compared with nondrinkers after stratification by sex. Among women, only 5.5% of NPC cases and 5.1% of control subjects reported ever habitual consumption of alcohol (Table 2). Alcohol intake among women was not analyzed in further detail due to the low frequency of ever consumption.

Table 3 shows the ORs and 95% CIs for associations between current alcohol drinking and risk of NPC among men. No clear exposure-response trends were seen for duration of alcohol drinking or intake of specific alcoholic beverages (P_trend > 0.05). Nonparametric regression curves indicated a nonsignificant association between current daily ethanol intake and NPC risk in men (P for nonlinear association = 0.200; Fig. 1A).

In contrast, among former drinkers, there was no trend either with ethanol intake or with duration of drinking in fully adjusted models (Supplementary Table S1).

Table 2 also shows the ORs and 95% CIs for associations between tea drinking status and risk of NPC among overall subjects and stratified by sex. Compared with the nondrinkers, current tea drinking was significantly associated with a 27% decrease in risk of NPC (95% CI, 16%–36%) in the fully adjusted model. Similar results were found after stratification by sex, although men were more likely than women to have ever had a tea drinking habit (Table 2).

Table 4 lists the ORs and 95% CIs for NPC risk in relation to current tea drinking among overall participants. Mean daily current intake of tea (mL/day) showed no clear pattern of association with NPC risk; the ORs were significantly lower than 1 for first three categories, compared with never drinkers, but not significant for the next two categories. Nonparametric regression curves by restricted cubic spline analysis with three knots suggested a significant nonlinear dose–response relation between volume of current daily tea intake and NPC risk (P for overall association = 0.038; P for nonlinear association = 0.033; Fig. 1B).

Analyses by tea temperature, concentration of tea, and types of tea revealed that those who consumed warm or hot tea, those who did not drink strong tea, and those who drank black and green tea tended to have a decreased NPC risk (Table 4). We found no significant interaction between tea drinking and age, sex, educational level, current housing type, and current occupation (Supplementary Table S2).

To evaluate whether tea cup size played an effect-modification role, we conducted further analyses stratified by tea cup size. We found a linear decreasing trend between daily tea intake and risk of NPC among tea drinkers using small tea cups (Supplementary Fig. 2A), whereas tea drinkers using medium or large tea cups had attenuated or no decreased NPC risk (Supplementary Fig. S2B and S2C).

To our knowledge, this is among the first endeavors to assess the association of detailed alcohol and tea consumption with the risk of NPC in a large population-based study. We found no increased risk of NPC among current alcohol drinkers, and analyses by daily consumption, duration, and type of alcoholic beverage provided no compelling evidence for a causal association. In contrast, tea intake was associated with a decreased risk, although the lack of a monotonic dose–response trend in relation to amount or duration of tea drinking complicates causal inference and suggests that residual confounding or bias may play a role.

Consistent with our findings, most studies reported no association between overall alcohol drinking and NPC risk (10–16, 18, 21, 29). A J-shaped dose–response relationship was found in a meta-analysis based on six case–control studies (19), and results from a subsequent case–control study indicated that light or moderate alcohol intake was associated with a reduced risk, while heavy consumption increased risk (17). Although we found a moderately increased risk for heavy ethanol intake of ≥120.0 g/day, caution is needed when interpreting this result, as the number of heavy alcohol drinkers was small. Another recent meta-analysis of two cohort studies and population-based case–control studies showed no significant association between alcohol consumption and NPC (22). Studies conducted in other high-risk areas for NPC, such as Hong Kong, Taiwan, Malaysia, Thailand, and Singapore, as well as some low-risk areas, showed similar null results (2–4, 6, 24, 30, 31). Most of these studies, however, had a small sample size and could not calculate specific ethanol intake.

Several previous studies examined four types of alcoholic beverages in association with NPC risk, but reported inconsistent associations for beer or spirits (13, 17, 24, 32), and consistently no increased risk for wine or Chinese rice wine (17, 24, 33). Most of these studies were unable to evaluate dose–response trends for each type of alcoholic beverage. We found an increased risk of NPC among current alcohol drinkers with wines or low intake of strong distillate spirits/liquors (<10 g/day), but these associations were attenuated and not significant after additional adjustment for dietary pattern (Supplementary Table S3), suggesting that the positive association could be partly explained by dietary pattern or other unknown residual confounding.

We found a significant inverse association between current intake of tea, especially black/pu'er tea, and risk of NPC; this decreased risk was confined to drinkers of hot/warm/mild tea or weak tea. Consistent with our results, four case–control studies also reported a significantly decreased risk of NPC associated with overall tea drinking (10, 14, 17, 23, 24); this association differed by specific tea type. One study reported a reduced risk among green tea drinkers (29); another found an inverse association among drinkers of any of six tea types (17). The inconsistencies may arise due to different patterns of tea drinking or tea manufacturing across areas, or they may be attributable to chance or residual confounding. To our knowledge, no prospective cohort studies have explored the relationship between tea drinking and NPC risk.

Our results showed that there seemed a nonlinear relationship between daily tea intake and NPC risk after adjusting for most confounders. Misclassification or differential recall bias may partly contribute to this nonlinear relationship, although exposure misclassification often leads to linearization of apparent exposure-response trends, rather than nonlinearization (34). However, we cannot eliminate the influence of residual confounding, even after adjusting for numerous confounders. Another possible reason for our findings may be that different tea drinking patterns in the study area are correlated with the use of characteristic tea cups of distinct sizes. Those with low daily tea intake were more likely to use small tea cups and drink Kungfu tea, which was typically consumed at a higher concentration with a short brewing time. Those with high daily tea intake more often used medium or large tea cups, in which tea was brewed for a relatively long time in a large teapot and, potentially allowing for oxidation of protective substances such as tea polyphenols. Although we did not find a significant association between highly concentrated tea drinking and NPC risk, this characteristic was subjectively reported in our study. We found a significant effect modification by tea cup size, which support this interpretation; however, these results were based on small numbers and require independent confirmation.

A protective effect of tea drinking is, nevertheless, biologically plausible. In animal studies, the epigallocatechin gallate, which can be extracted from tea, may reactivate methylation-silenced genes in cancer cell lines and inhibit cancer growth (35, 36). In addition, the tea polyphenol (-)-epigallocatechin-3-gallate has potent antimicrobial activity against viruses and block EBV infection–induced cytokine expression and EBV-induced B-lymphocyte transformation (37). These catechins and others are found in multiple tea types at various concentrations; for example, epigallocatechin-3-gallate levels in oolong tea were found to be lower than in other tea types, although findings may be method dependent (38).

Strengths of our study include its large size and strictly population-based design in an NPC-endemic area, with high and equal participation rate among eligible cases and controls. Potential but nonquantifiable limitations include recall bias and reverse causality. We undertook, however, extensive efforts to reduce or eliminate such bias through numerous approaches described in detail in earlier publications (7, 8, 25, 26). Furthermore, alcohol and tea are not well-established risk or protective factor for NPC. Thus, the influence of recall bias might not be a big concern. We did not collect information about the specific reasons for quitting alcohol or tea use among former drinkers, which hampers assessment of reverse causality. However, we defined former ethanol drinkers as those who had stopped for >2 years, thus symptoms of NPC might not be the major reasons for them to stop drinking ethanol. We were unable to collect detailed information on the ethanol content of individual alcoholic beverages; instead, we used the mean or most commonly used concentrations (39, 40) to calculate cumulative ethanol consumption. Finally, we were unable to distinguish between patterns of regular drinking and episodic drinking of either alcohol or tea.

In conclusion, a statistical association between alcohol drinking and the risk of NPC is more likely nonexisting. Tea drinking may be associated with moderately reduced NPC risk. However, these conclusions warrant further investigation and should be interpreted with caution for some residual confounding.

E.T. Chang reports grants from NCI during the conduct of the study and other from Exponent outside the submitted work. I. Ernberg reports grants from Swedish Cancer Society during the conduct of the study. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.

R. Feng: Conceptualization, data curation, formal analysis, visualization, writing–original draft. E.T. Chang: data curation, writing–review and editing. Q. Liu: Conceptualization, data curation. Y. Cai: Resources, data curation. Z. Zhang: Conceptualization, data curation, funding acquisition. G. Chen: Conceptualization, data curation. Q.-H. Huang: Conceptualization, resources, data curation. S.-H. Xie: Resources, data curation, project administration. S.-M. Cao: Conceptualization, resources, data curation. Y. Zhang: Conceptualization, resources, data curation. J.-P. Yun: Resources, data curation. W.-H. Jia: Conceptualization, resources, data curation, methodology. Y. Zheng: Conceptualization, resources, data curation. J. Liao: Conceptualization, resources, data curation, investigation. Y. Chen: Resources, data curation, investigation. T. Huang: Resources, data curation, formal analysis, investigation. L. Lin: Conceptualization, resources, data curation. I. Ernberg: Conceptualization, resources, data curation, methodology. G. Huang: Conceptualization, resources, data curation, supervision, methodology. Y.-X. Zeng: Conceptualization, resources, data curation, supervision, methodology. H.-O. Adami: Conceptualization, resources, data curation, supervision, funding acquisition, methodology, writing–review and editing. W. Ye: Conceptualization, resources, data curation, supervision, methodology, project administration, writing–review and editing.

The NCI at the U.S. NIH [grant number R01 CA115873; principal investigator (PI): H.-O. Adami; co-PIs: Y. Zeng, Y.-X. Zeng]; the Swedish Research Council (2015-02625, 2015-06268, 2017-05814, to W. Ye); and the Karolinska Institutet Distinguished Professor Award (Dnr: 2368/10-221, to H.-O. Adami). The work in the Guiping/Pingnan area was supported by grants from the New Century Excellent Talents in University (no. NCET-12-0654, to Z. Zhang), National Basic Research Program of China (no. 2011CB504300, to G. Huang), and Guangxi Natural Science Foundation (2013GXNSFGA 019002, to Z. Zhang). R. Feng was partly supported by the Postdoctoral Research Foundation of China (2018M640874).

The authors sincerely thank their deceased colleague, Yi Zeng (Institute for Viral Disease Control and Prevention, Chinese Center for Disease Control and Prevention), who played a pivotal role during the planning and conduct of this study. They also thank the members of the External Advisory Board of the NPC Genes, Environment, and EBV Study, including Curtis Harris (U.S. NCI), Xihong Lin (Harvard T.H. Chan School of Public Health), Allan Hildesheim (U.S. NCI), Wei-Cheng You (Peking University), and You-Lin Qiao (Chinese Academy of Medical Sciences) for their guidance of the overall case–control study.

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.