Past and Recent Salted Fish and Preserved Food Intakes Are Weakly Associated with Nasopharyngeal Carcinoma Risk in Adults in Southern China

ABSTRACT

Background

Chinese-style salted fish intake in early life is considered an established risk factor for nasopharyngeal carcinoma (NPC). However, results for adult intakes of salted fish and preserved foods are inconsistent.

Objective

The aim of this study was to ascertain the relations of Chinese-style hard and soft salted fish and preserved food intakes with NPC risk.

Methods

We conducted a population-based case-control study in southern China with 2554 NPC cases identified through a rapid case ascertainment system and 2648 healthy controls, frequency-matched on age, sex, and area. Subjects (aged 20–74 y) were interviewed via a food-frequency questionnaire, including information on portion size. Data were also collected on alcohol consumption and potential confounders. Food intake was grouped into 3–5 energy-adjusted intake levels during adulthood (10 y prior) and adolescence (16–18 y). For childhood (at age 10 y), intake frequency of selected food items was collected. Multivariate-adjusted ORs with 95% CIs were estimated via logistic regression.

Results

We found no association between NPC and intake of hard Chinese-style salted fish during adulthood, and an increased risk at the highest level of intake during adolescence (OR: 1.19; 95% CI: 1.03, 1.39). In contrast, we found a decreased risk for the middle intake level of soft salted fish during adulthood (OR: 0.68; 95% CI: 0.57, 0.81) and adolescence (OR: 0.71; 95% CI: 0.59, 0.85). Preserved foods showed contrasting risk profiles, e.g., the highest adult intake level of salted egg (OR: 1.51; 95% CI: 1.22, 1.87) and fermented black beans (OR: 0.67; 95% CI: 0.56, 0.80). Associations with NPC were weaker than previously reported, e.g., for weekly childhood intake of salted fish (OR: 1.56; 95% CI: 1.24, 1.97).

Conclusions

Hard and soft salted fish have different risk profiles. Salted fish and other preserved foods were at most weak risk factors for NPC in all periods and may play a smaller role in NPC occurrence than previously thought.

Introduction

Nasopharyngeal carcinoma (NPC) is endemic in southern China, Southeast Asia, regions of North Africa, and some Arctic populations, with incidence typically 2- to 3-fold higher in males than in females (1). The Epstein–Barr virus (EBV) is considered a necessary but not sufficient cause; but because EBV infection is ubiquitous, it cannot explain the unique geographic disease pattern (2). Chinese-style salted fish, a constituent of the traditional diet in many NPC-endemic areas (3), has been associated with increased risk of NPC (4–8) particularly as a weaning food source (9–12), whereas some studies found no association with adulthood consumption (12–14). Other preserved, dried, salted, or fermented foods have also been associated with NPC risk in several Asian (4, 5, 10–12, 15–18) and northern African countries (6–8). However, some studies have not found an association of preserved food intake with NPC risk (19, 20).

Because epidemiological research on NPC has generally been hampered by methodological weaknesses (21), it remains to be established which of the various preserved food items, if any, are causally associated with NPC. Prior epidemiological studies of the association between Chinese-style salted fish consumption and NPC risk have evaluated frequency of consumption, but not portion size, nor have they distinguished between “soft” and “hard” subtypes of Chinese-style salted fish which may confer different risks due to distinct preparation methods (22). Moreover, most previous studies have not been population-based in design. To address these research gaps, we investigated the associations of intake of Chinese-style salted fish types (the primary exposures of interest) and intake of other preserved foods (the secondary exposures) during adulthood, adolescence, and childhood with risk of NPC (the main outcome) in a large-scale population-based case-control study set in 2 southern Chinese provinces where NPC is endemic.

Methods

Study population

A detailed description of this study has been published previously (23). We carried out a collaborative study entitled “NPC Genes, Environment, and EBV” in the Zhaoqing area of Guangdong Province and the Wuzhou and Guiping/Pingnan areas of Guangxi Autonomous Region, China. Eligible cases were individuals aged 20–74 y living in the study area at the time of diagnosis with no history of malignant disease or congenital or acquired immunodeficiency. A total of 3047 eligible histopathologically confirmed NPC cases were identified between 2010 and 2013 via a rapid case ascertainment network which involved a team of a medical person and study coordinators backed up by 2 local cancer registries; the number of observed cases was similar to the expected number based on previously reported incidence rates. Of these eligible cases, 2554 (84%) agreed to participate. In total, 3202 potential controls frequency-matched to the 5-y age and sex distribution of the cases by residential area were randomly selected between 2010 and 2014. These randomly selected controls were identified from computerized total population registries from our study area. Such registries are present in all large administrative areas in China and contain information on the age, sex, telephone number, and home address of every individual. A total of 2648 eligible controls (83%) consented to participate. After the exclusion of subjects with missing, ineligible, or unreliable data, 2530 cases and 2593 controls remained for evaluation of dietary intake data (Supplemental Figure 1).

Data collection

An electronic structured questionnaire was administered to study participants by trained interviewers in person or by telephone. Each interviewer was assigned a similar number of cases and controls. The interviewers were unaware of the study hypotheses but could not be blinded to case-control status. Collected information included demographics; body size; residential and occupational history; history of chronic ear, nose, and respiratory tract conditions; family history of NPC and other cancers; cigarette smoking; alcohol consumption; tea consumption; dietary habits; and use of Chinese herbal medicine. Additional efforts to ensure data quality, including checking for the consistency and plausibility of all FFQ items, were carried out across all available time periods. Upon completion of all data cleaning, 2502 cases and 2559 controls remained with dietary data during adulthood, and 2244 cases and 2309 controls contributed dietary data during adolescence. For the childhood data set, we used the set of subjects with reliable dietary data during adulthood.

Assessment of dietary intake

The FFQ included 77 items selected from the most commonly consumed foods in the local study areas, based on the results of a pilot study and a previously validated FFQ (24). The same FFQs were used for both adult and adolescent dietary assessment periods. Subjects reported frequency (5 categories) of consumption and overall actual portion size [in standard units of jin (∼500 g) or liang (∼50 g)] for 9 food categories: staple foods, pickled foods, eggs, fresh meat, fresh fish, dairy, cakes and nuts, beans and soy, and fresh fruits and vegetables. Chinese-style salted fish consisted of both “hard” and “soft” subtypes. The “hard” subtype is prepared by directly salting and then drying; the “soft” subtype, also referred to as “molded” salted fish, is allowed to undergo a decomposition phase and softening before salting and drying. Intake of other preserved foods assessed in the FFQ included dried fish, pickled vegetables, fermented bean curds, preserved egg, salted egg, fermented black beans or bean paste, fish sauces, salty shrimp paste, and preserved meats. An illustrated booklet was used to aid study participants in estimating portion size. Dietary intake of all food items was assessed 10 y before the interview (adulthood) and at 16–18 y old (adolescence). In addition, we collected data on frequency only of dietary intake at 10 y of age (childhood) for 5 food items: salted fish, pickled vegetables, fermented pastes, fresh vegetables, and fresh fruits. For seasonal food items, interviewees were instructed to estimate intake for the times of year when each food item was available.

We also gathered detailed information on frequency and overall amount of alcoholic beverage intake (beers, wines, and spirits), as well as information on the amount and type of cooking oil consumed in the subject's household during adulthood, and cooking methods used during adulthood and adolescence.

To derive nutrients and energy from food intake, we used detailed China food composition tables (25). Where multiple food composition table entries were present for foods assessed as a single item in our FFQ, e.g., rice, we took the average of the available estimates. For dietary items consisting of several foods in our FFQ but listed as a single entry in the food composition tables, e.g., animal meat (including pork, beef, mutton, etc.), we first estimated the nutrient content of each food item individually before creating a summary value for the overall food item.

Daily total energy intake

We calculated daily total energy intake (TEI) based on energy contributions from FFQ items and alcohol intake during adulthood and adolescence. A sensitivity analysis in which TEI also included the energy contribution from cooking oil, which was available at the household level for ∼80% of study participants, produced similar results (data not presented) to those from our main analysis.

Ethics

The current study was approved by the Institutional Review Board or Ethical Review Board at all participating centers. All study participants provided written or oral informed consent.

Statistical methods

For crude comparisons between cases and controls, we used chi-square tests for categorical variables and Student's t tests for continuous variables. We used multivariable unconditional logistic regression models to estimate ORs and 95% CIs for the risk of NPC associated with dietary intake during adulthood, adolescence, and childhood. The minimally adjusted multivariable model included the frequency matching variables age (in 5-y groups), sex, and residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan). The main model was also adjusted for known or suspected risk factors of NPC that are plausibly related to patterns of food intake: socioeconomic status (SES, operationalized as education, housing and occupation) (26), smoking (27), and family history of NPC (2) . It was also adjusted for intake of fruits and vegetables as potential protective factors (24) to account for possible confounding via correlated food patterns, all coded as shown in Table 1 (see also the footnotes for Tables 2–4). Adult and adolescent models were both adjusted for childhood intake of preserved food as listed in Table 3, and the adult model was also adjusted for adolescent intake of preserved foods to account for persistent patterns of food intake along the life trajectory. Our main model was also adjusted for the residual-adjusted intake (26) of nonsalted preserved foods listed in Table 4.

TABLE 1.

Characteristics of NPC cases and controls from south China, used for adulthood dietary analysis¹

	Cases (n = 2502)	Controls (n = 2559)	P value2
Residential area			0.31
Zhaoqing	1268 (50.6)	1304 (50.9)
Wuzhou	680 (27.1)	654 (25.5)
Guiping/Pingnan	554 (22.1)	601 (23.4)
Sex			0.99
Male	1836 (73.3)	1878 (73.3)
Female	666 (26.6)	681 (26.6)
Age, y	48.5 ± 10.6	49.7 ± 10.8	<0.0001
Age distribution			0.01
20–24 y	27 (1.0)	33 (1.2)
25–29 y	55 (2.1)	47 (1.8)
30–34 y	143 (5.7)	123 (4.8)
35–39 y	276 (11.0)	242 (9.4)
40–44 y	415 (16.5)	395 (15.4)
45–49 y	489 (19.5)	486 (19.0)
50–54 y	329 (13.1)	342 (13.3)
55–59 y	348 (13.9)	378 (14.7)
60–64 y	235 (9.3)	274 (10.7)
65–69 y	122 (4.8)	150 (5.8)
70–74 y	63 (2.5)	89 (3.4)
Education level, y			0.005
≤6	990 (39.5)	919 (35.9)
7–9	1003 (40.0)	1025 (40.0)
10–12	405 (16.1)	478 (18.6)
>12	104 (4.1)	137 (5.3)
Current housing type			<0.0001
Building (concrete)	1800 (71.9)	1988 (77.6)
Cottage (clay brick)	692 (27.6)	568 (22.1)
Boat	10 (0.3)	2 (0.07)
Missing	—	1 (0.03)
Current occupation			0.0002
Unemployed	77 (3.0)	96 (3.7)
Farmer	848 (33.8)	970 (37.9)
Blue-collar	774 (30.9)	669 (26.1)
White-collar	238 (9.5)	217 (8.4)
Other/unknown	565 (22.5)	607 (23.7)
First-degree NPC family history			<0.0001
Yes	271 (10.8)	69 (2.6)
No	2181 (87.1)	2447 (95.6)
Unknown	46 (1.8)	42 (1.6)
Cigarette smoking			0.08
Current	1113 (44.4)	1196 (46.7)
Former	181 (7.2)	152 (5.9)
Never	1208 (48.2)	1211 (47.3)
Fruit and vegetable intake, g/d	675 ± 343	676 ± 371	0.06
Energy intake, kcal/d			0.04
Adult3	1445 ± 519	1453 ± 541
Adult II4	1620 ± 508	1610 ± 514
Adolescent5	1246 ± 562	1248 ± 588

¹Values are mean ± SD or frequency (%). NPC, nasopharyngeal carcinoma.

²Categorical variables were analyzed using a chi-squared test, other variables using a 2-sided t test.

³Daily energy intake calculated from food items and alcohol during adulthood.

⁴Estimated hypothetical daily energy intake. For ∼80% of individuals we had dietary information from cooking oil for adulthood from which we could calculate its contribution to dietary energy intake. Based on this average energy contribution from cooking oil from these individuals, we could estimate a hypothetical dietary energy intake which utilized all dietary energy sources in adulthood.

⁵Adolescent daily energy intake calculated from food items and alcohol intake.

TABLE 2.

ORs and 95% CIs of NPC associated with Chinese-style salted fish intake during adulthood and adolescence, adjusted for total energy intake using the residual method¹

	Intake range per level (g/d)2
Level3	Male	Female	n cases/controls	OR (95% CI)
Adult dietary intake4
Total5 Chinese-style salted fish
1	0	0	846/797	1.00 (ref)
2	>0.00 to ≤0.41	>0.00 to ≤0.34	489/585	0.81 (0.68, 0.96)
3	>0.41 to ≤1.64	>0.34 to <1.36	489/582	0.73 (0.62, 0.87)
4	>1.64	≥1.36	678/596	0.93 (0.78, 1.10)
			P-trend	0.16
Hard salted fish
1	0	0	1049/1121	1.00 (ref)
2	>0.00 to ≤0.68	>0.00 to ≤0.45	667/723	1.02 (0.88, 1.19)
3	>0.68	>0.45	786/715	1.05 (0.91, 1.22)
			P-trend	0.90
Soft salted fish
1	0	0	1705/1658	1.00 (ref)
2	>0.00 to <0.54	>0.00 to ≤0.34	336/452	0.68 (0.57, 0.81)
3	≥0.54	>0.34	461/449	0.88 (0.74, 1.04)
			P-trend	0.45
Adolescent dietary intake6
Total5 Chinese-style salted fish
1	0	0	834/876	1.00 (ref)
2	>0.00 to ≤0.54	>0.00 to ≤0.41	437/471	0.99 (0.83, 1.17)
3	>0.54 to ≤2.19	>0.41 to ≤1.72	426/473	0.89 (0.75, 1.06)
4	>2.19	>1.72	547/488	1.10 (0.93, 1.31)
			P-trend	0.55
Hard salted fish
1	0	0	1017/1133	1.00 (ref)
2	>0.00 to ≤0.82	>0.00 to <0.68	561/588	1.07 (0.92, 1.22)
3	>0.82	≥0.68	666/588	1.19 (1.03, 1.39)
			P-trend	0.01
Soft salted fish
1	0	0	1522/1531	1.00 (ref)
2	>0.00 to ≤0.82	>0.00 to ≤0.68	289/390	0.71 (0.59, 0.85)
3	>0.82	>0.68	403/388	0.97 (0.82, 1.15)
			P-trend	0.27

¹NPC, nasopharyngeal carcinoma.

²Intake range per food level (grams per day) based on controls’ intake for males and females separately. Although intake cutoffs are presented in general to 2 decimal places, intake ranges do not overlap for different food levels.

³Level 1 was the reference group, which consisted of individuals who did not consume the food item. Levels 2 and above consisted of individuals with increasing intakes of the food item (residually adjusted for total energy intake).

⁴We used unconditional logistic regression to estimate ORs with 95% CIs. ORs were adjusted for sex, age, residential area, education level, current housing type, current occupation, first-degree family history of NPC, cigarette smoking, adult daily energy intake (log transformed), energy-adjusted intake of other foods (adult fruit and vegetable intake, adult intake of non–salted fish preserved foods, adolescent total preserved food intake), and childhood frequency of intake of total preserved foods.

⁵Hard and soft salted fish combined.

⁶We used unconditional logistic regression to estimate ORs with 95% CIs. ORs were adjusted for sex, age, residential area, education level, current housing type, current occupation, first-degree family history of NPC, cigarette smoking, adolescent daily energy intake (log transformed), energy-adjusted intake of other foods (adolescent fruit and vegetable intake and adolescent intake of non–salted fish preserved foods), and childhood frequency of intake of total preserved foods.

TABLE 4.

ORs and 95% CIs of NPC associated with intake of preserved foods other than Chinese-style salted fish during adulthood and adolescence, adjusted for total energy intake using the residual method¹

	Intake range per level (g/d)2
Level3	Male	Female	n cases/controls	OR (95% CI)
Adult dietary intake4
Pickled vegetables5
1	0.00 to ≤0.82	0.00 to ≤1.36	501/613	1.00 (ref)
2	>0.82 to ≤3.33	>1.36 to <3.57	620/649	1.19 (0.99, 1.42)
3	>3.33 to ≤10.0	≥3.57 to ≤14.2	632/649	1.15 (0.96, 1.39)
4	>10.0	>14.2	749/648	1.24 (1.02, 1.52)
			P-trend	0.055
Dried fish
1	0	0	993/1026	1.00 (ref)
2	>0.00 to ≤0.82	>0.00 to ≤0.68	647/766	1.00 (0.84, 1.18)
3	>0.82	>0.68	862/767	1.22 (1.02, 1.46)
			P-trend	0.04
Salted egg
1	0	0	1530/1701	1.00 (ref)
2	>0.00 to ≤0.89	>0.00 to <0.53	431/429	1.18 (0.97, 1.42)
3	>0.89	≥0.53	541/429	1.51 (1.22, 1.87)
			P-trend	0.0002
Fermented bean curds
1	0	0	856/902	1.00 (ref)
2	>0.00 to ≤0.13	>0.00 to <0.10	516/546	0.99 (0.83, 1.18)
3	>0.13 to <0.66	≥0.10 to ≤0.33	638/548	1.17 (0.99, 1.40)
4	≥0.66	>0.33	492/563	0.85 (0.71, 1.02)
			P-trend	0.81
Fermented black beans or bean paste
1	0	0	831/691	1.00 (ref)
2	>0.00 to <0.61	>0.00 to ≤0.54	529/617	0.69 (0.58, 0.82)
3	≥0.61 to <1.66	>0.54 to <1.66	559/617	0.72 (0.60, 0.85)
4	≥1.66	≥1.66	583/634	0.67 (0.56, 0.80)
			P-trend	<0.0001
Sausage/ham/related meats
1	0	0	578/521	1.00 (ref)
2	>0.00 to <0.68	>0.00 to ≤0.54	389/510	0.71 (0.58, 0.86)
3	≥0.68 to ≤1.36	>0.54 to ≤1.36	512/510	0.88 (0.73, 1.07)
4	>1.36 to ≤3.33	>1.36 to ≤3.33	536/508	0.93 (0.76, 1.13)
5	>3.33	>3.33	487/510	0.77 (0.62, 0.94)
			P-trend	0.25
Preserved eggs
1	0	0	1446/1526	1.00 (ref)
2	>0.00 to ≤0.89	>0.00 to <0.65	505/516	0.92 (0.77, 1.11)
3	>0.89	≥0.65	551/517	0.84 (0.68, 1.04)
			P-trend	0.11
Adolescent dietary intake6
Pickled vegetables5
1	≥0.00 to ≤1.36	≥0.00 to <1.66	547/605	1.00 (ref)
2	>1.36 to ≤5.00	≥1.66 to ≤5.00	486/567	0.88 (0.73, 1.05)
3	>5.00 to ≤14.28	>5.00 to ≤14.28	524/570	0.95 (0.79, 1.13)
4	>14.28	>14.28	687/567	1.30 (1.09, 1.55)
			P-trend	0.001
Dried fish
1	0	0	1017/1133	1.00 (ref)
2	>0.00 to <0.82	>0.00 to ≤0.68	561/588	1.07 (0.92, 1.22)
3	≥0.82	>0.68	666/588	1.19 (1.03, 1.39)
			P-trend	0.01
Salted egg
1	0	0	1682/1863	1.00 (ref)
2	>0.00 to <0.71	>0.00 to <0.62	248/223	1.12 (0.89, 1.41)
3	≥0.71	≥0.62	314/223	1.32 (1.03, 1.69)
			P-trend	0.02
Fermented bean curds
1	0	0	969/1106	1.00 (ref)
2	>0.00 to <0.16	>0.00 to ≤0.13	453/398	1.30 (1.09, 1.54)
3	≥0.16 to ≤0.66	>0.13 to ≤0.54	449/397	1.21 (1.01, 1.45)
4	>0.66	>0.54	373/408	0.95 (0.79, 1.15)
			P-trend	0.96
Fermented black beans or bean paste
1	0	0	957/926	1.00 (ref)
2	>0.00 to <0.66	>0.00 to <0.66	457/441	0.83 (0.70, 0.99)
3	≥0.66 to ≤2.50	≥0.66 to ≤2.50	456/412	0.75 (0.63, 0.90)
4	>2.50	>2.50	434/470	0.78 (0.65, 0.94)
			P-trend	0.002
Sausage/ham/related meats
1	0	0	1159/1199	1.00 (ref)
2	>0.00 to ≤0.41	>0.00 to ≤0.41	215/278	0.77 (0.62, 0.95)
3	>0.41 to ≤0.82	>0.41 to ≤0.82	265/277	0.89 (0.73, 1.10)
4	>0.82 to <2.77	>0.82 to <2.77	284/277	0.98 (0.79, 1.20)
5	≥2.77	≥2.77	321/278	1.06 (0.86, 1.30)
			P-trend	0.66
Preserved eggs
1	0	0	1710/1880	1.00 (ref)
2	>0.00 to ≤0.71	>0.00 to <0.53	242/214	1.11 (0.88, 1.40)
3	>0.71	≥0.53	292/215	1.23 (0.96, 1.57)
			P-trend	0.10

¹NPC, nasopharyngeal carcinoma.

²Intake range per food level (grams per day) based on controls’ intake for males and females separately. Although intake cutoffs are presented in general to 2 decimal places, intake ranges do not overlap for different food levels.

³Level 1 was the reference group, with higher levels representing increased intake residually adjusted for total energy intake.

⁴We used unconditional logistic regression to estimate ORs with 95% CIs. ORs were adjusted for sex, age, residential area, education level, current housing type, current occupation, first-degree family history of NPC, cigarette smoking, adult daily energy intake (log transformed), energy-adjusted intake of other foods (adult fruit and vegetable intake, adult intake of all other preserved foods, adolescent total preserved food intake), and childhood frequency of intake of total preserved foods.

⁵Because <10% of individuals had 0 g/d intake for this food item, level 1 included individuals who did not consume the food item and individuals with the lowest residually adjusted intake.

⁶We used unconditional logistic regression to estimate ORs with 95% CIs. ORs were adjusted for sex, age, residential area, education level, current housing type, current occupation, first-degree family history of NPC, cigarette smoking, adolescent daily energy intake (log transformed), energy-adjusted intake of other foods (adolescent fruit and vegetable intake, adolescent intake of all other preserved foods), and childhood frequency of intake of total preserved foods.

TABLE 3.

ORs and 95% CIs of NPC associated with intake frequency of Chinese-style salted fish, pickled vegetables, and fermented pastes during childhood¹

	n cases/controls	OR (95% CI)
Chinese-style salted fish
Never eaten	623/771	1.00 (ref)
Yearly	1217/1103	1.37 (1.19, 1.58)
Monthly	389/461	1.04 (0.87, 1.26)
Weekly	251/202	1.56 (1.24, 1.97)
Daily	21/20	1.19 (0.61, 2.30)
	P-trend	0.01
Pickled vegetables
Never eaten	159/190	1.00 (ref)
Yearly	736/736	1.14 (0.89, 1.45)
Monthly	555/727	0.89 (0.66, 1.09)
Weekly	686/569	1.32 (1.02, 1.70)
Daily	365/335	1.20 (0.91, 1.57)
	P-trend	0.03
Fermented pastes
Never eaten	1129/1267	1.00 (ref)
Yearly	1020/767	1.46 (1.28, 1.66)
Monthly	177/300	0.60 (0.49, 0.75)
Weekly	157/203	0.79 (0.62, 1.00)
Daily	18/20	0.88 (0.45, 1.72)
	P-trend	0.01

¹We used unconditional logistic regression to estimate ORs with 95% CIs. ORs were adjusted for sex, age, residential area, education level (6, 7–9, 10–12, or >12 y), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), first-degree family history of NPC (yes, no, or unknown), cigarette smoking (current, former, or never), childhood frequency of intake of fresh fruits and fresh vegetables, and childhood frequency of intake of other preserved items (Chinese-style salted fish, pickled vegetables, and fermented pastes). NPC, nasopharyngeal carcinoma.

All statistical tests were 2-sided, with a P value < 0.05 considered to be statistically significant. Exposure–response associations were evaluated using tests for linear trend, and also for nonlinear association by means of an orthogonal quadratic contrast for the exposures where nonlinear patterns were strongest. All statistical analysis was performed using SAS version 9.4 (SAS Institute).

To achieve an approximately unskewed distribution, a natural logarithmic transformation was applied to TEI (kilocalories per day) and food intake (grams per day). Because the majority of food items had a substantial proportion of individuals with zero values for intake, the residual adjustment was not applied to food intake values of 0. Instead, subjects reporting zero intake were assigned to the lowest intake category (level 1) for this food item. All nonzero intake values were transformed and adjusted for TEI (based on energy intake calculated from FFQ items and alcohol consumption) using the residual method (26). Subjects with nonzero intakes were then classified into categories of equal size based on the distribution of these adjusted intake values among controls (as representative for the intake in the general population), separately for males and females. The number of categories varied between 3 and 5, based on the number of nonzero intakes observed. The same number of levels was applied to food items for both adult and adolescent intakes to facilitate comparison. Sensitivity analyses using actual intake values instead of residual-adjusted values to classify individuals into groups, an alternative strategy for residual adjustment, were also explored (results not presented). None of the sensitivity analyses yielded substantively different findings from the primary analysis.

Results

Study population characteristics

Table 1 shows the distribution of basic demographic characteristics and potential risk factors for NPC among cases and controls with data on dietary intake during adulthood. The distribution of these factors was similar among subjects with available data on dietary intake during adolescence (data not shown). Cases were less educated, more likely to live in cottages, more likely to have blue-collar jobs, and more often reported a first-degree family history of NPC than controls. We found little difference in smoking history and consumption of fresh fruits and vegetables, and only a minor, although statistically significant, difference in TEI (<10 kcal/d).

Intake of Chinese-style salted fish

In adulthood

Although the test for a linear trend in the association between total intake of Chinese-style salted fish in adulthood and NPC risk was nonsignificant (P = 0.16), our results indicated a nonlinear, U-shaped pattern of association: compared with the lowest intake level, risk was significantly reduced for levels 2 and 3 but not for level 4 (Table 2, Supplemental Figure 2); see “Assessment of nonlinear trends” below. We found no association with adult consumption of the hard Chinese-style salted fish subtype (P = 0.90). In contrast, we found a significant 32% reduced risk of NPC in association with the middle of 3 categories of adult consumption of soft salted fish, but no association with the highest intake level and no significant linear trend (P = 0.45).

In adolescence

Adolescent intake of total salted fish was not significantly associated with risk of NPC (Table 2). We found an ∼20% excess risk in association with the highest intake level compared with nonintake (level 1) of hard salted fish during adolescence, with a weak but statistically significant linear exposure–response trend (P = 0.01). In contrast, we found an almost 30% reduction in NPC risk in association with the middle level of adolescent intake of soft salted fish, but no significant association with the highest intake level.

In childhood

Compared with never eaters, we found a significantly increased risk of NPC associated with weekly and yearly intake of Chinese-style salted fish during childhood (Table 3). A statistically significant (P-trend = 0.01) but nonmonotonic exposure–response trend was found, with increased ORs for yearly and weekly but not monthly or daily consumption compared with none.

Intake of other preserved foods

In adulthood and adolescence

Table 4 shows the associations of intake of various preserved food items during adolescence and adulthood with NPC risk. Because we observed a considerable consistency of results across time periods for the majority of these food items, results are described in terms of whether associations with NPC risk were generally positive, inverse, or null.

All but 1 of the food items that demonstrated a significantly increased risk of NPC were associated only at the highest level of intake in both adolescence and adulthood. For pickled vegetables, the highest level of intake conferred ORs of 1.24 (95% CI: 1.02, 1.52) for adult intake and 1.30 (95% CI: 1.09, 1.55) for adolescent intake compared with the lowest level. The highest level of intake of dried fish (prepared differently from Chinese-style salted fish) conferred ORs of 1.22 (95% CI: 1.02, 1.46) during adulthood and 1.19 (95% CI: 1.03, 1.39) during adolescence. Likewise, the highest level of salted egg intake was associated with ORs of 1.51 (95% CI: 1.22, 1.87) and 1.32 (95% CI: 1.03, 1.69) for adult and adolescent intake, respectively. Fermented bean curds intake in adolescence increased NPC risk but not in adulthood, without any consistent trend (Table 4).

Intakes of 2 food items, fermented black beans/bean paste and sausage/ham/related meats, were associated with reduced risk of NPC in both adulthood and adolescence. Intake of fermented black beans/bean paste in adulthood was associated with an ∼30% reduction of NPC risk for all categories of adult intake (P-trend < 0.0001) when compared with those who did not consume the item (level 1). A similar but weaker association was seen for adolescent consumption, with ORs of 0.83–0.75 (P-trend = 0.002). Intake of sausage and ham during both adulthood and adolescence was inversely associated with NPC risk, with no consistent linear exposure–response trend. Preserved egg consumption in either adulthood or adolescence was not significantly associated with NPC risk. We did not estimate associations of the other 2 preserved foods (fish sauce and salted shrimp paste) with NPC risk, because 99% of individuals did not consume these food items.

In childhood

Weekly childhood consumption of pickled vegetables was associated with significantly increased risk of NPC by 32% compared with never eaters, whereas other intake frequencies were not related to NPC risk (Table 3). For the childhood consumption of fermented pastes, an inconsistent relation with NPC was seen. Compared with no intake, yearly consumption of fermented paste was associated with significantly increased risk of NPC (OR: 1.46; 95% CI: 1.28, 1.66), whereas monthly consumption was associated with reduced risk (OR: 0.60; 95% CI: 0.49, 0.75).

Assessment of nonlinear trends

We tested for evidence of a nonlinear association for the exposures where nonlinear patterns were strongest, and found evidence of a quadratic trend in all such instances. These included adult total salted fish intake (quadratic P-trend = 0.0005); soft salted fish intake among adults (quadratic P-trend = 0.004) and adolescents (quadratic P-trend = 0.0003); and fermented bean curds intake among adults (quadratic P-trend = 0.01) and adolescents (quadratic P-trend = 0.001).

Supplementary results

Results for the minimally adjusted model can be found in the Supplementary data (Supplemental Tables 1–3). In short, the results of the minimally adjusted model did not differ substantially from our main analysis results. The supplementary data also contain figures illustrating the risk patterns from Tables 2 and 4 (Supplemental Figures 2 and 3).

Discussion

In this large population-based case-control study in southern China, we investigated the associations of intakes of Chinese-style salted fish and various other preserved foods during different age periods with risk of NPC. We found evidence that the 2 salted fish subtypes have different risk profiles and the intakes of some preserved foods appear to increase the risk of NPC. In contrast, other preserved foods may reduce or not affect NPC risk. In general, OR estimates were lower than those reported from previous studies (9–12, 15, 16, 18, 28). The excess risks were indeed so modest and the exposure–response trends so indistinct that causality remains uncertain for all periods of intake and food items examined. Because prior studies did not evaluate portion size, whether the weaker associations are due to lower overall quantity of intake in this study population cannot be assessed.

Although Chinese-style salted fish consumption during early life has long been considered an accepted risk factor for NPC (11, 12), whether consumption in adulthood affects risk remains unclear. Some studies have indicated no association (12, 14, 20, 29, 30), whereas others have found OR estimates between 1.2 and 3.1 (15, 31, 32). Our study revealed a complex nonlinear relation. To better understand these results, we further investigated the 2 subtypes of Chinese-style salted fish that are commonly consumed in southern China. We found that the overall nonlinear association with consumption in adulthood and adolescence was driven by the soft subtype of Chinese-style salted fish. In adulthood, consumption of the hard subtype was not associated with NPC, whereas intake in adolescence was significantly associated with an increased risk. To our knowledge, no previous study has shown Chinese-style salted fish to have protective qualities in relation to NPC, nor has any evaluated the association between the hard compared with soft salted fish subtypes and risk of NPC.

The 2 subtypes of salted fish are prepared in different manners. Hard salted fish undergoes a direct salting and drying process over a number of days in the outside heat after gutting (22). In contrast, the soft variety is left to decompose naturally first, followed by salting later when soft. Overall, Chinese-style salted fish has been shown to induce nasopharyngeal tumors in animal models (22, 33, 34). The primary proposed mechanism involves the presence of large amounts of N-nitrosamines. These are known to be genotoxic and carcinogenic (35–37) and present in Chinese-style salted fish (38, 39) and other preserved foods in NPC-endemic regions (40, 41). Substances found in preserved foods have also been shown to reactivate EBV in cell lines (42). However, no research has systematically classified the constituents of the salted fish subtypes individually to see if their carcinogenic properties differ.

Our findings, especially related to overall and soft Chinese-style salted fish, may arise because of residual confounding by SES after adjustment for income, housing type, and education. The rapid cultural and economic changes in China may have reduced our ability to correctly classify SES. Traditionally, soft Chinese-style salted fish has not been considered affordable by all. Approximately one-third of our study population consumed soft salted fish. Those in the middle level of intake in adulthood and adolescence, both of which were inversely associated with NPC risk, tended to be of higher SES, which is associated with lower risk of NPC (26).

Prior research into the adult consumption of other preserved foods in association with NPC risk has yielded inconsistent results. A meta-analysis demonstrated an OR of ∼2.0 for consumption of all preserved foods; however, the authors questioned the quality and comparability of the underlying studies (21). Our results indicate that only some preserved foods may increase NPC risk. One large study in Shanghai (17) found a 43% increased risk of NPC associated with the adult consumption of preserved vegetables, whereas in Guangdong research showed that NPC cases more frequently consumed plum vegetable, known locally as “Chan Pai Mui” (16). In the same Shanghai study, they found preserved meat to increase the risk of NPC by 77%. In contrast, we found preserved meats to be inversely associated with NPC risk. We also found an inverse association with the consumption of fermented black beans/bean paste in both adulthood and adolescence, a result that contrasts with those of other studies (11, 17). Results could potentially differ across regions owing to differences in preparation methods and constituents for similar food items, as well as differences in background NPC risk.

We collected frequency of dietary intake in childhood for only a few selected items and we could not contact subjects’ mothers to obtain childhood dietary information. Although we found that early childhood intake of Chinese-style salted fish and other preserved food items was associated with increased NPC risk, our OR estimates for salted fish were lower than previously published estimates, which ranged from 2- up to a 37-fold increase in risk (4, 10, 12, 14, 15, 18). Our lower estimates could be driven in part by a reduction in absolute intake in recent years. Research in Guangzhou (12) in the 1950s–1970s reported that ≤85% of 10-y-old children consumed salted fish on at least a monthly basis, and a Hong Kong study in the 1960s and 1970s reported that this figure was 75% (10). More recent figures in Guangzhou showed that only 36% of children consumed this item monthly (18) and our data revealed that only 27% of 10-y-olds up until the 1990s ate Chinese-style salted fish monthly. Childhood may represent a unique period of susceptibility when these foods contribute to NPC risk. Previous research using birth order as a proxy for timing of EBV infection in childhood revealed associations with NPC risk (43), pointing to the etiologic importance of risk factors in early childhood.

Our study has some notable strengths, including its population-based design, large sample size, high participation rates, collection of data on portion size, and estimation of daily caloric intake over multiple life time points. Ours also is the first study, as far as we know, to investigate both adult and adolescent diet and the salted fish subtypes separately. In addition, to minimize information bias, we developed a structured electronic questionnaire, audio-recorded all interviews for quality checking, and trained and monitored all interviewers in conducting face-to-face interviews. Some weaknesses are inherent owing to the retrospective nature of case-control studies. Self-reported retrospective questionnaire data can be inaccurate, especially for distant past exposures. There is also the possibility of recall bias, which we attempted to minimize by training interviewers to interact with cases and controls in the same manner. We attempted to mitigate potential reverse causality by asking subjects to report dietary intake from 10 y before the interviews. Although our FFQ was developed based on a previously validated FFQ (24, 44), it was not itself validated. It was also not possible in this study to validate dietary intake with traditional methods such as 24-h recall. However, studies have shown that actual past diet is related to recalled past diet with low to moderate correlations for 15–25 y previously, and moderate to strong correlations for dietary recall from 6–14 y previously (45). Finally, there is always the possibility that some of our results may be influenced by residual confounding, with an unpredictable direction and magnitude of bias.

In conclusion, our results indicate that salted fish subtypes and other preserved foods may have different risk profiles, and that overall these foods appear at most to be weak risk factors for NPC. Our findings therefore question the current causal model of NPC and imply a less substantial role for these food items in the endemic occurrence of NPC than previously thought. As a corollary, the identification of novel environmental causes, which presumably interact with EBV and genetic factors, is imperative to explain the enigmatic descriptive epidemiology of NPC. Because various salted and preserved foods have also been associated with NPC in other regions, e.g., other Asian and African countries (4–8, 10–12, 15–18), our results may have wider implications, with differences in food constituents and preparation methods warranting further investigation.

Notes

Supported by National Cancer Institute (NIH) grant R01 CA115873, Swedish Research Council grants 2015-02625, 2015-06268, and 2017-05814, and Karolinska Institutet Distinguished Professor Award Dnr: 2368/10-221 (to H-OA). The work from the Guiping/Pingnan area was supported by New Century Excellent Talents in University grant NCET-12-0654, National Basic Research Program of China grant 2011CB504300, and Guangxi Natural Science Foundation grant 2013GXNSFGA 019002 (to ZZ). DB was supported by a KID scholarship from Karolinska Institutet.

Author disclosures: DB, AP, ETC, ZL, C-XZ, QL, YC, ZZ, GC, Q-HH, S-HX, S-MC, J-YS, W-HJ, Y Zheng, JL, YC, LL, IE, H-OA, GH, Y Zeng, Y-XZ, and WY, no conflicts of interest.

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

Supplemental Tables 1–3 and Supplemental Figures 1–3 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

H-OA, GH, Y Zeng, Y-XZ, and WY contributed equally to this work.

Abbreviations used: EBV, Epstein–Barr virus; NPC, nasopharyngeal carcinoma; SES, socioeconomic status; TEI, total energy intake.