Fruit and vegetable intake and head and neck cancer risk in a large United States prospective cohort study

Authors


Abstract

Squamous head and neck cancers include cancers of the oral cavity, pharynx and larynx are the sixth leading cause of cancer mortality worldwide, resulting in more than 350,000 deaths annually. Intake of fruit and vegetables may protect against head and neck cancer incidence, although few prospective studies have examined this association. We investigated this relation in 490,802 United States participants of the NIH-AARP Diet and Health cohort using Cox proportional hazard models adjusted for potential confounders. During 2,193,751 person years of follow-up from 1995/1996–2000, 787 participants were diagnosed with head and neck cancer. We found an inverse association between total fruit and vegetable intake and head and neck cancer risk (per serving/day/1,000 calories, Hazard Ratio, 95% Confidence interval: 0.94, 0.89–0.99). In models mutually adjusted for fruit and vegetable intake, the association was stronger for vegetables (fifth vs. first quintile: 0.65, 0.50–0.85) than for fruits (fifth vs. first quintile: 0.87, 0.68–1.11). When further subclassified into botanical groups, those in the highest tertile of leguminosae (dried beans, string beans and peas, 0.80, 0.67–0.96), rosaceae (apples, peach, nectarines, plums, pears and strawberries, 0.60, 0.49–0.73), solanaceae (peppers and tomatoes, 0.82, 0.69–0.98) and umbelliferae (carrots, 0.73, 0.60–0.89) had decreased risk of head and neck cancer, but no significant associations were seen for 9 other botanical groups. Results from this large prospective observational study are consistent with previous case-control studies and support the hypothesis that total fruit and vegetable intake is associated with reduced risk of head and neck cancer. © 2007 Wiley-Liss, Inc.

Squamous cancers of the head and neck are the sixth leading cause of cancer related mortality worldwide, resulting in more than 350,000 deaths annually.1 Head and neck cancer includes tumors at multiple anatomical sub-sites, including the oral cavity, pharynx and larynx, with substantially different rates of 5 year survival, ranging from 30% for cancers of the hypopharynx to 63% for cancers of the larynx.2

The main causes of head and neck cancer are thought to be alcohol and tobacco use.3 Diet may also play an etiological role.3–6 Fruits and vegetables, rich in potentially anticarcinogenic compounds, including carotenoids, fiber, folate, flavonoids, plant sterols, phenolic acids and vitamin C, have been hypothesized to protect against head and neck cancer incidence.4, 6 Most previous studies have supported this hypothesis but have had a case-control study design. Recall and selection bias can affect the results of case-control studies. The few existing prospective investigations7–11 have had small case numbers,8–10 grouped head and neck cancer sites together with the esophagus,7–10 not investigated total fruit or vegetable intake,8–10 or examined precursor lesions.11

We investigated the association between fruit and vegetable intake and individual head and neck cancer sub-sites in a large United States prospective cohort, the NIH-AARP Diet and Health study.

Material and methods

Study population

The establishment and recruitment procedures of the National Institutes of Health (NIH)-AARP Diet and Health Study have been described.12 Between 1995 and 1996, a questionnaire eliciting information on dietary intake and health-related behaviors was mailed to 3.5 million members of AARP, formerly known as the American Association of Retired Persons-a US organization whose membership is open to those greater than 50 years of age, who resided in one of six U.S. states (California, Florida, Louisiana, New Jersey, North Carolina and Pennsylvania) or two metropolitan areas (Atlanta, Georgia and Detroit, Michigan) and were between 50 and 71 years of age. 566,407 respondents (339,671 men and 226,736 women) filled out the survey in satisfactory detail and consented to be in the study. After exclusion of subjects with cancer at baseline (n = 51,219), proxy respondents (n = 15,760), those with box-cox log transformed total energy, total fruit, vegetable, or fruit and vegetable intake more than two interquartile ranges from the median (8,616), and those who died or were diagnosed with cancer on the first day of follow-up (n = 10), the resulting analytical cohort comprised 490,802 participants: 292,898 men and 197,904 women. The NIH-AARP Diet and Health Study was approved by the Special Studies Institutional Review Board of the U.S. National Cancer Institute (NCI).

Cohort follow-up

As described previously,13 addresses for members of the NIH-AARP cohort were updated annually by matching the cohort database to that of the National Change of Address database maintained by the U.S. Postal Service. Follow-up time extended from study baseline (between 1995 and 1996) to diagnosis of head and neck, esophageal or stomach cancer, as a diagnosis of one of these cancers would be associated with increased surveillance of the other sites, date of death, end of study (December 31, 2000), or the date moved out of the registry ascertainment area.

Identification of cancer cases

Incident cases of cancer were identified by probabilistic linkage between the NIH-AARP cohort membership and 8 state cancer registry databases, which have been certified by the North American Association of Central Cancer Registries (NAACCR) for meeting the highest standards of data quality. Approximately 90% of cancer cases are detected in the cohort by this approach.13 Cancer sites were identified by anatomic site and histologic code of the International Classification of Disease for Oncology (ICD-O, third edition)14 as previously described.15 The total head and neck cancer category included those diagnosed with a cancer of the oral cavity, oro-hypopharynx, larynx and those with squamous cell carcinomas of other anatomical head and neck sites, including those overlapping these regions.

Dietary assessment

The baseline questionnaire has been described.12 Participants were asked to report their usual frequency of intake and portion size over the last 12 months, using 10 frequency categories ranging from “never” to “6+ times/day” for beverages and from “never” to “2+ times/day” for solid foods and 3 categories of portion size. The food items, portion sizes and nutrient database (Pyramid Servings Database) were constructed16 using data from the US Department of Agriculture's 1994–1996 Continuing Survey of Food Intake by Individuals Survey.17 The Pyramid Servings Database utilized a recipe file to disaggregate food mixtures into their component ingredients and assign them to the appropriate food group. One pyramid serving corresponded to one serving in the US Department of Agriculture's food guide Pyramid–one medium sized fresh fruit, half a cup cut fruit, 6 oz juice, 1 cup leafy vegetables, or half a cup other vegetables.18 The FFQ was validated using two nonconsecutive 24-hr recalls in 1,953 participants. For fruit and vegetable intake combined, the energy-adjusted Pearson correlations between the FFQ and the 24-hr recalls were 0.72 and 0.61 for men and women, respectively.19

We excluded white potatoes from the vegetable group and created quintiles of fruit and vegetable variables. Similar results were observed for variables that included potatoes. We also examined groups based on botanical taxonomy to help identify possible associations according to chemopreventive phytochemicals.20 We used tertiles for the botanical groups because of the large number of participants who did not regularly consume food items from some of the botanical groups.

Statistical analysis

Analyses were performed with SAS version 8.2. An alpha level of less than 0.05 was considered statistically significant and all tests were two-sided. Hazard ratios and 95% confidence intervals were calculated using Cox proportional hazards regression.21 We tested for and found no deviations from the proportional hazards assumption. Follow-up time was used as the underlying time metric. Using age as the underlying time metric did not alter results. Linear trend tests across increasing categories of intake were conducted by assigning participants the median intake for their categories and entering that term as a continuous variable in the regression model.

All models included adjustment for continuous age, and categorical variables of sex, education (<high school education, completion of high school, some post-high school training, completion of college and completion of graduate school), body mass index (BMI; <18.5, 18.5 to <25, 25 to <30, 30 to <35 and ≥35 kg/m2), alcohol intake (0, >0–1, >1–3 and > 3 drinks/day), smoking (never cigarette smoker, quit ≤1 pack/day, quit >1 pack/day, currently smoking ≤1 pack/day, and currently smoking >1 pack/day), vigorous physical activity (never, rarely, 1–3 times/month, 1–2 times/week, 3–4 times/week, 5 or more times/week), usual activity throughout the day (sit all day, sit much of the day/walk some, stand/walk often/no lifting, lift/carry light loads and carry heavy loads). Adding variables for race/ethnicity, pipe/cigar smoking or intake of total meat, red meat, total fat, saturated fat, total grain or whole grains did not affect risk estimates (data not shown).

To adjust for energy intake, we used the nutrient density method.22 We divided fruit and vegetable intake by total energy intake (per 1,000 calories) and also included total energy from all dietary sources in the models. As a sensitivity analysis, we also analyzed data using a standard multivariate model that included fruit and vegetable intake (servings/day), total energy and other covariates and the results were similar to those obtained with the nutrient density method (data not shown). Fruits and vegetables were mutually adjusted for each other. Missing values for adjusting covariates were included as dummy variables in the models. Excluding those with missing values for covariates reduced the number of cases slightly but did not change point estimates appreciably, as less than 4% of the cohort lacked data for each covariate.

To examine effect modification, we multiplied the continuous intakes of fruit and vegetables by the appropriate stratifying variables and calculated p-values for statistical interactions using a likelihood ratio test with the appropriate degrees of freedom. Multivariate models stratified by smoking status (current, former and never) included adjustment for smoking dose and years since smoking cessation where appropriate as well as the covariates used for the overall models including alcohol use; likewise, models stratified by alcohol use were adjusted by smoking and other covariates.

Results

Fifty percent of participants consumed more than 5 servings of fruits and vegetables/day, with a median intake of 2.8 servings/day in the lowest quintile and 9 servings/day in the highest quintile. For analysis, we divided fruit and vegetable intake by energy intake and present results per serving per 1,000 calories/day. Table I shows baseline characteristics of participants by quintile of total fruit and vegetable intake per 1,000 calories/day. Participants in the highest quintile of fruit and vegetable intake combined were more likely to be women, college graduates, participants in vigorous physical activities and less likely to be smokers and consumers of more than three alcoholic beverages per day than those in the lowest quintile. BMI and age were similar across quintiles.

Table I. Study Characteristics by Quintiles of Fruit and Vegetable Intake Per 1,000 Calories
 Quintiles of fruit and vegetable intake1
Q1Q2Q3Q4Q5
  • 1

    Categories may not add up to 490,802 persons due to missing data.

  • 2

    Servings/day per 1,000 calories.

Cohort (Number, %)98,16020.098,16120.098,16020.098,16120.098,16020.0
Total fruit and vegetable intake2 (Median, interquartile range)1.51.2–1.82.42.2–2.63.23.0–3.44.13.8–4.45.85.2–6.8
Sex (Number, %)          
 Male72,56424.866,68122.859,79420.452,36017.941,49914.2
 Female25,59612.931,48015.938,36619.445,80123.156,66128.6
Age (Median, interquartile range)61.656.8–66.062.457.6–66.562.857.9–66.76358.1–66.863.158.2–66.9
BMI, m/kg2 (Median, interquartile range)26.624.2–29.826.624.3–29.726.524.0–29.526.323.8–29.325.823.4–28.9
Total daily calories (Median, interquartile range)19551447.0–2600.51,8021377.5–2333.91,6941306.6–2176.61,5921231.8–2046.71,4451105.0–1870.4
Education (Number, %)          
 Less than high school7,94127.35,92120.35,20717.95,01817.25,05217.3
 12 years (completed high school)22,82723.819,25920.118,22719.017,95118.717,60518.4
 Some post-high school training34,19521.132,56520.132,00819.731,50619.431,93419.7
 Completed college16,88918.319,48721.119,39721.018,86720.417,80319.3
 Completed graduate school13,52914.018,42119.020,65921.322,04722.822,23522.9
Alcohol intake (Number, %)          
 Zero drinks/day22,19018.921,39018.222,04518.823,72520.228,10023.9
 0< to 1 drinks/day42,65216.449,92619.253,57120.656,32021.757,50922.1
 1< to 3 drinks/day14,96020.117,44323.416,74922.514,55519.610,70014.4
 >3 drinks/day17,95948.69,07124.55,49614.93,1468.51,3063.5
Cigarette smoking Status (Number, %)          
 Never smoked24,72914.232,45618.636,07620.739,01022.441,76424.0
 Former46,46519.449,38620.649,16420.548,32020.146,65219.4
 Current25,09036.814,67021.511,34616.79,14813.47,84411.5
Usual activity throughout the day (Number, %)          
 Sit during the day/little walking9,39224.28,22421.27,57919.57,16018.56,43516.6
 Sit during the day/walk a fair amount31,79320.132,73320.732,40120.531,55620.029,54318.7
 Stand/walk a lot-no lifting34,21318.535,84519.436,93020.038,05320.639,70721.5
 Lift/carry light loads, stairs, hills16,59319.716,46919.616,69319.916,89220.117,37220.7
 Do heavy work/carry loads4,05528.92,96121.12,50817.92,27016.22,22015.8
Vigorous physical activity (Number, %)          
 Never6,88531.64,45820.53,86017.73,30615.23,24814.9
 Rarely18,84428.314,58921.912,69019.110,95416.59,48714.3
 1–3 times/month16,50924.814,86922.313,12719.711,99818.010,16115.2
 1–2 times/week21,62420.522,75021.522,03620.920,89719.818,37917.4
 3–4 times/week19,90715.224,85918.927,37920.929,20922.329,89522.8
 5 or more times/week13,40514.315,65216.718,16419.420,76122.225,62527.4

During 2,193,751 person-years of follow-up (mean = 4.5 years), 787 participants were diagnosed with squamous cancers of the head and neck (608 men and 179 women). Increasing consumption of fruit and vegetables was associated with reduced head and neck cancer risk, (Hazard Ratio (HR) per serving per 1,000 calories/day: 0.94, 95% CI: 0.89–0.99). Consumers in the highest quintile of fruit and vegetable intake had reduced risk of head and neck cancer relative to those in the lowest quintile of intake (0.71, 0.55–0.92) and the test for trend was also significant (p = 0.018) (Table II). We observed similar associations between fruit and vegetable intake and head and neck cancer in men and women. Men in the fifth quintile of fruit and vegetable intake had 0.75 times (0.55–1.02) the risk of head and neck cancer of men in the first quintile. Similarly, women in the fifth quintile had 0.67 times the risk of women in first quintile (0.40–1.11) (p for statistical interaction = 0.1).

Table II. Multivariate Hazard Ratios and 95% Confidence Intervals for Fruit and Vegetable Intake and Head and Neck Cancer
Category12Per serving per 1,000 caloriesQ1 (reference)Q2Q3Q4Q5p for trend
HR3, 95% CINo.HRNo.HR3, 95% CINo.HR3, 95% CINo.HR3, 95% CINo.HR3, 95% CI
  • 1

    Fruit and Vegetable constituents—Total Fruit and vegetables (no potatoes): total fruits + vegetables; Total Vegetables (no potatoes): spinach, turnip, collard greens, mustard, kale, cole slaw, cabbage, sauerkraut, carrots, dried beans, string beans, peas, corn, broccoli, cauliflower, brussel sprouts, mixed vegetables, tomatoes, sweet papers, lettuce salad, sweet potatoes, yams, tomato juice, tomato sauce, chili and salsa; Total Fruits: Whole fruits + Fruit Juice; Whole fruits: apples, apple sauce, pears, bananas, dried fruit excluding apricots, peaches, nectarines, plums, cantaloupe, other melons, strawberries, oranges, tangerines, tangelos, grapefruit and grapes; Fruit juice: orange and grapefruit juice, and other fruit juices and drinks.

  • 2

    Median intake per quintile: Total fruit and vegetable intake–1.5, 2.4, 3.2, 4.1, 5.8; Vegetables–0.7, 1.2, 1.6, 2.1, 3.2; Total fruits–0.4, 1.0, 1.5, 2.1, 3.2; Whole fruits–0.2, 0.6, 0.9, 1.3, 2.2; Fruit juice–0, 0.2, 0.4, 0.7, 1.4.

  • 3

    Adjusted for age at entry into cohort, alcohol intake, BMI, cigarette-smoke-dose, education, sex, total energy intake, usual activity throughout the day and vigorous physical activity.

  • 4

    Additionally adjusted for continuous fruit intake.

  • 5

    Additionally adjusted for continuous vegetable intake.

  • 6

    Additionally adjusted for continuous vegetable intake and continuous fruit juice intake.

  • 7

    Additionally adjusted for continuous vegetable intake and continuous whole fruit intake.

Total fruit and vegetables0.94, 0.89–0.992671.001600.83, 0.68–1.021480.90, 0.73–1.111220.84, 0.67–1.06900.71, 0.55–0.920.018
Vegetables40.89, 0.82–0.972441.001650.85, 0.70–1.041490.86, 0.70–1.061410.91, 0.73–1.13880.65, 0.50–0.850.005
Total fruits50.98, 0.91–1.062551.001600.87, 0.71–1.071460.95, 0.76–1.171210.89, 0.71–1.121050.87, 0.68–1.110.324
Whole fruits60.94, 0.85–1.042571.001690.90, 0.74–1.101470.93, 0.75–1.151220.88, 0.70–1.12920.77, 0.59–1.000.063
Fruit juice71.03, 0.93–1.152281.001440.76, 0.62–0.941510.81, 0.66–1.001180.70, 0.56–0.881460.94, 0.76–1.170.930

In models mutually adjusted for fruit and vegetable intake, the association was stronger with vegetable (Q5 vs. Q1: 0.65, 0.50–0.85), than fruit intake (Q5 vs. Q1: 0.87, 0.68–1.11). Whole fruit intake (Q5 vs. Q1: 0.77, 0.59–1.00; p for trend = 0.063), but not fruit juice (Q5 vs. Q1: 0.94, 0.76–1.17; p for trend = 0.93), showed a borderline inverse association with head and neck cancer. When examined by head and neck cancer sub-site, associations between fruit and vegetables and cancers of the oral cavity, oro-hypopharynx and larynx were similar to the overall estimates (Table III).

Table III. Multivariate Hazard Ratios and 95% Confidence Intervals for Fruit and Vegetable Intake and Head and Neck Cancer Sub-Sites
Category1SitePer serving per 1000 caloriesQ1 (reference)Q2Q3Q4Q5p for trend
HR2, 95% CINo.HRNo.HR2, 95% CINo.HR2, 95% CINo.HR2, 95% CINo.HR3, 95% CI
  • 1

    Fruit and vegetable constituents—Total Fruit and vegetables (no potatoes): total fruits + vegetables; Total vegetables (no potatoes): spinach, turnip, collard greens, mustard, kale, cole slaw, cabbage, sauerkraut, carrots, dried beans, string beans, peas, corn, broccoli, cauliflower, brussel sprouts, mixed vegetables, tomatoes, sweet papers, lettuce salad, sweet potatoes, yams, tomato juice, tomato sauce, chili and salsa; Total Fruits: Whole fruits + Fruit Juice; Whole fruits: apples, apple sauce, pears, bananas, dried fruit excluding apricots, peaches, nectarines, plums, cantaloupe, other melons, strawberries, oranges, tangerines, tangelos, grapefruit and grapes; Fruit Juice: orange and grapefruit juice and other fruit juices and drinks.

  • 2

    Adjusted for age at entry into cohort, alcohol intake, BMI, cigarette-smoke-dose, education, sex, total energy intake, usual activity throughout the day, and vigorous physical activity.

  • 3

    Additionally adjusted for continuous fruit intake.

  • 4

    Additionally adjusted for continuous vegetable intake.

  • 5

    Additionally adjusted for continuous vegetable intake and continuous fruit juice intake.

  • 6

    Additionally adjusted for continuous vegetable intake and continuous whole fruit intake.

Total fruit and vegetablesOral cavity0.93, 0.86–1.001041.00640.81, 0.59–1.11570.81, 0.58–1.14580.90, 0.64–1.28360.61, 0.41–0.930.052
Oro-hypopharynx0.95, 0.85–1.07541.00270.72, 0.45–1.16260.84, 0.51–1.38150.57, 0.31–1.04200.90, 0.51–1.580.500
Larynx0.95, 0.87–1.03971.00580.89, 0.64–1.24571.07, 0.76–1.51410.91, 0.62–1.34260.69, 0.44–1.110.190
 Vegetables3Oral cavity0.84, 0.73–0.951001.00600.72, 0.52–1.00640.84, 0.61–1.15590.83, 0.59–1.16360.56, 0.37–0.840.017
Oro-hypopharynx0.90, 0.74–1.09551.00210.49, 0.30–0.82300.80, 0.50–1.26200.60, 0.35–1.01160.56, 0.31–1.010.082
Larynx0.91, 0.79–1.05781.00761.29, 0.94–1.78460.91, 0.62–1.31511.15, 0.80–1.67280.77, 0.49–1.220.242
 Total fruits4Oral cavity1.00, 0.90–1.12991.00640.86, 0.63–1.19530.83, 0.58–1.17591.02, 0.72–1.45440.84, 0.57–1.250.632
Oro-hypopharynx1.00, 0.84–1.18451.00341.10, 0.70–1.74311.23, 0.76–2.00100.47, 0.23–0.96221.19, 0.68–2.080.875
Larynx0.98, 0.86–1.11961.00540.82, 0.58–1.15531.00, 0.71–1.43451.00, 0.68–1.46310.80, 0.51–1.230.527
 Whole fruits5Oral cavity0.96, 0.82–1.12981.00720.98, 0.71–1.33580.90, 0.64–1.27500.87, 0.60–1.27410.81, 0.54–1.220.268
Oro-hypopharynx0.89, 0.69–1.15461.00361.11, 0.71–1.75291.09, 0.66–1.77170.74, 0.41–1.35140.72, 0.37–1.390.181
Larynx0.96, 0.80–1.14981.00580.85, 0.61–1.18490.89, 0.62–1.27440.95, 0.65–1.40300.77, 0.49–1.220.393
 Fruit juice6Oral cavity1.06, 0.90–1.25811.00620.90, 0.64–1.25620.91, 0.65–1.27490.78, 0.54–1.12651.10, 0.78–1.530.556
Oro-hypopharynx1.12, 0.88–1.43401.00240.74, 0.45–1.23341.09, 0.69–1.74180.66, 0.37–1.16261.05, 0.63–1.740.809
Larynx1.00, 0.82–1.21881.00490.68, 0.48–0.97500.72, 0.50–1.02450.74, 0.51–1.07470.85, 0.59–1.220.742

We further examined the association using subgroups of fruit and vegetables based on botanical classifications. Participants in the highest tertile of Leguminosae (beans and peas, HR: 0.80, 0.67–0.96), Rosaceae (apples peaches, nectarines and strawberries, HR: 0.60, 0.49–0.73), Solanaceae (peppers and tomatoes, HR: 0.82, 0.69–0.98) and Umbelliferae (carrots, HR: 0.73, 0.60–0.89) consumption had lower risk of head and neck cancer than those in the lowest tertile (Table IV). We observed similar results in the oral cavity, oro-hypopharynx and larynx (data not shown).

Table IV. Multivariate Hazard Ratios and 95% Confidence Intervals for Fruit and Vegetable Botanical Groups and Head and Neck Cancer
Botanical groupTertile 3 versus 1p for trend
No.Median (servings/1,000 calories) in Tertile 1/Tertile 3HR,1 95% CI
  • 1

    Adjusted for age at entry into cohort, alcohol intake, BMI, cigarette-smoke-dose, education, sex, total energy intake, usual activity throughout the day and vigorous physical activity.

Chenopodiaceae: raw spinach and cooked spinach218/2900.0/1.00.96, 0.80–1.140.403
Compositae: lettuce212/3340.0/0.50.92, 0.77–1.100.487
Convolvulaceae: sweet potatoes and yams215/3260.0/0.10.88, 0.74–1.060.236
Cruciferae: broccoli, cauliflower, brussell sprouts, turnip, cabbage, coleslaw, collard, mustard and kale197/3360.1/0.50.90, 0.75–1.080.252
Cucurbitaceae: cantaloupe, watermelon and honeydew melon217/3290.0/0.20.96, 0.80–1.140.859
Gramineae: corn231/3080.0/0.10.92, 0.77–1.100.421
Leguminosae: dried beans, string beans and peas205/3210.1/0.60.80, 0.67–0.960.015
Musaceae: bananas219/3450.0/0.50.92, 0.77–1.090.479
Rosaceae: apples, peach, nectarines, plums, pears and strawberries149/4050.1/0.60.60, 0.49–0.73<0.001
Rutaceae (citrus): oranges, tangerines, tangelos and grapefruits214/3360.1/1.10.90, 0.75–1.080.293
Solanaceae: tomatoes, peppers201/3100.1/0.50.82, 0.69–0.980.023
Umbelliferae: carrots164/3470.0/0.20.73, 0.60–0.890.001
Vitaceae: grapes215/3610.0/0.10.87, 0.73–1.040.436

We examined fruit and vegetable intakes in relation to head and neck cancer by strata of alcohol consumption and smoking status, because alcohol and smoking are strong risk factors for head and neck cancer. Among never smokers who did not drink alcohol (n = 39 cases), the multivariate HR for total fruit and vegetable intake per serving per 1,000 calories was 0.93 (0.77–1.13), whereas among ever smokers who also drank alcohol (n = 478 cases), the HR was 0.94 (0.88–1.01). These stratified estimates were similar to the overall estimates: 0.94, 0.89–0.99. We further examined associations with fruits and vegetables within strata of alcohol and cigarette use (Table V). Risk estimates appeared similar among alcohol drinkers and nondrinkers (p for statistical interaction = 0.68); the multivariate HR for total fruit and vegetable intake per serving per 1,000 calories was 0.95 (0.89–1.01) among alcohol drinkers and 0.92 (0.85–1.01) among nondrinkers. No significant differences by smoking status (never, former and current) were observed for total fruit and vegetable intake, total vegetable intake, total fruit intake or whole fruits (Table V). In contrast, we observed significant effect modification for fruit juice intake by smoking status (p = 0.014). Greater intake of fruit juice had no association with head and neck cancer risk in never smokers, a nonsignificant inverse association in former smokers and a significant positive association in current smokers (Table V).

Table V. Multivariate Hazard Ratios and 95% Confidence Intervals for Fruit and Vegetable Intake and Head and Neck Cancer Stratified by Smoking Status
CategoryOverallSmoking statusAlcohol use
NeverFormerCurrentP for interaction1NoYesp for interaction
  • 1

    p for statistical interaction from −2 log likelihood test with 2 degrees of freedom for smoking and 1 degree of freedom for alcohol (see methods section).

  • 2

    HR, 95% CI per serving/1,000 calories: Adjusted for age at entry into cohort, alcohol intake, BMI, cigarette-smoke-dose, education, sex, total energy intake, usual activity throughout the day and vigorous physical activity.

No.787109334317 219568 
Total fruit and vegetables20.94, 0.89–0.990.97, 0.86–1.100.90, 0.84–0.971.01, 0.94–1.100.3250.92, 0.85–1.010.95, 0.89–1.010.682
Vegetables20.89, 0.82–0.971.11, 0.94–1.320.83, 0.73–0.940.88, 0.77–1.020.1420.84, 0.72–0.980.92, 0.83–1.010.364
Total fruits20.98, 0.91–1.060.85, 0.70–1.030.97, 0.87–1.081.13, 1.01–1.270.0880.99, 0.88–1.120.97, 0.89–1.060.907
Whole fruits20.94, 0.85–1.040.77, 0.59–1.011.02, 0.89–1.180.99, 0.82–1.200.4300.92, 0.77–1.100.94, 0.83–1.070.245
Fruit juice21.03, 0.93–1.150.96, 0.72–1.280.88, 0.73–1.071.26, 1.09–1.460.0141.08, 0.90–1.291.01, 0.88–1.160.523

After excluding the first 2 years of follow-up, the association of fruit and vegetable intake and head and neck cancer remained similar (437 cases, HR per serving per 1,000 calories: 0.92, 0.86–0.99).

Discussion

In this large prospective study, we observed significant inverse associations between fruit and vegetable intake and head and neck cancer incidence. Risk in the fifth quintile of fruit and vegetable consumption for head and neck cancer was 29% lower than risk in the first quintile. The association appeared stronger with vegetables than with fruits. We observed similar associations when we examined cancers of the oral cavity, oro-hypopharynx and larynx separately.

Our results are generally consistent with the results of previously published, mostly case-control, studies of fruit and vegetable intake and cancers of the oral cavity, pharynx and larynx. Two meta-analyses have summarized the majority of these studies.4, 5 A meta-analysis including cancers of both the oral cavity and pharynx4 observed significant associations with both fruit (combined odds ratio per portion: 0.51, 0.40–0.65) and vegetable intake (combined OR per portion: 0.50, 0.38–0.65), but did not examine laryngeal cancer. A second meta-analysis5 included fewer studies but examined cancers of both the oral cavity/pharynx and the larynx and observed significant inverse associations between fruits and oral cavity/pharynx (per 100 g/day, combined odds ratio of 0.53, 0.37–0.76) and laryngeal cancer (OR of 0.73, 0.64–0.84), and nonsignificant inverse associations with vegetables (per 100 g/day: oral cavity/pharynx: 0.84, 0.67–1.07; laryngeal cancer: 0.92, 0.83–1.02).

To our knowledge, three previous prospective cohort studies examined associations of total fruit or vegetable intake and incidence of some or all head and neck cancer sub-sites. One study of Japanese men in Hawaii observed a suggestive inverse association between fruit intake (≥5/week vs. ≤1/week, HR: 0.65, 0.39–1.07) and upper aerodigestive tract cancers, including squamous cancers of the head and neck and the esophagus, but did not investigate total vegetable intake.8 The EPIC cohort study observed a borderline significant association between intake of total fruits and vegetables and cancers of the upper aerodigestive tract–squamous cancers of the head and neck and esophagus (Quintile (Q) 5 vs. Q1: 0.60, 0.37–0.99). The association was stronger for fruits (Q5 vs. Q1: 0.60, 0.38–0.97) than for vegetables (Q5 vs. Q1: 0.80. 0.49–1.31).7 Finally, a study of oral premalignant lesions observed no association with total fruit and vegetable intake (Q5 vs. Q1: 0.99, 0.61–1.61), a suggestive association with fruit (Q5 vs. Q1: 0.77, 0.47–1.27), and no association with vegetables (Q5 vs. Q1: 1.05, 0.65–1.71).11

It is not clear why hazard ratios in our study appear stronger for vegetable than fruit intake, in contrast to some previous studies that have shown the opposite. Most previous studies have had a case-control design, potentially affected by recall and selection bias. Alternative explanations include chance and differences in the types or preparations of fruits or vegetables consumed in different populations. For example, a large proportion of the fruits consumed by participants in our study were in the form of juice. Whole fruits and fruit juice have different constituents and had distinct associations with head and neck cancer in our study.

Fruit and vegetables contain numerous components, including carotenoids, fiber, folate, flavonoids, plant sterols, phenolic acids and vitamin C,6, 23 that might reduce the risk of developing head and neck cancer. We observed significant inverse associations with 4 botanical groups of fruits and vegetables: Leguminosae (beans and peas), Rosaceae (apples, peaches, nectarines and strawberries), Solanaceae (peppers and tomatoes) and Umbelliferae (carrots). Each of these foods has numerous potentially beneficial compounds that could play a protective role and so it is not possible to determine which phytochemicals are important in the current analysis. We did not observe an association with fruit juice or citrus fruit intake (Rutaceae) in our study, which argues against a role for vitamin C. Future studies of specific phytochemicals and other fruit and vegetable constituents, in this and other cohorts, may clarify the mechanisms by which fruit and vegetables could reduce head and neck cancer risk.

Unmeasured or poorly measured confounders are generally a concern in observational studies. This is especially true for head and neck cancer given that both alcohol and cigarette smoking are strong risk factors3 and are inversely associated with fruit and vegetable intake (Table I). Even though we adjusted the fruit and vegetable estimates for alcohol and cigarette use, residual confounding by alcohol and tobacco could explain our results. To assess this possibility, we conducted additional analyses that were stratified by alcohol use and cigarette smoking status. One way of assessing residual confounding by smoking and alcohol use is to examine the association only in never smokers who also did not drink alcohol. Though limited to 39 cases not exposed to the main risk factors for this disease, we observed a non-significant inverse association of fruit and vegetable intake and head and neck cancer among never-smokers who also did not drink alcohol that was similar in magnitude to the estimate from all subjects. But, future prospective studies with larger numbers of head and neck cancer cases that did not smoke or drink alcohol are needed to determine whether these results reflect a true association or chance.

When we examined alcohol intake independently from smoking, we observed similar inverse estimates for fruit and vegetable intake in alcohol drinkers and nondrinkers, reducing the likelihood that our findings are due to residual confounding by alcohol use. The results stratified by smoking status, however, were more complex. In never smokers, we observed a nonsignificant protective effect for total fruit and vegetable intake, a nonsignificant adverse effect for vegetables, and a borderline significant protective effect for fruit. One interpretation of this result is that fruit is associated with reduced risk of head and neck cancer and that the association with vegetables in the overall analysis is due to residual confounding by smoking. Two alternative explanations, though, merit consideration. First, the estimates in never smokers were not statistically different from the estimates in smokers, and the numbers of cases in never smokers were far fewer than the number in smokers; differences between the overall estimates and those in never smokers could be the result of chance. Second, compared to never smokers, current smokers eat fewer servings of both fruits vegetables. Therefore, we would expect that residual confounding, if it occurred, would result in an inverse association with both food groups among current smokers. In our study, however, vegetables, but not fruits, were protective in current smokers, suggesting that the inverse association with vegetables in current smokers is not due to residual confounding by smoking.

We note that limiting the analysis to never smokers or those who do not drink alcohol may circumvent residual confounding by smoking or alcohol drinking, but such restriction also precludes assessing whether fruits and vegetables reduce the risk of smoking and alcohol related head and neck cancer, which constitute the large majority of cases. Head and neck cancer caused only by other occupational and environmental exposures may have a different relation with fruit and vegetable intake than cancers caused by the carcinogens in cigarette smoke or alcohol intake. In our study, fruit and vegetable intake had an inverse association in ever smokers that also drank alcohol.

We found a significant statistical interaction between fruit juice intake and smoking status (Table V). There was no association seen between fruit juice intake and head and neck cancer risk in never smokers, a non-significant inverse association in former smokers, and a significant positive association in current smokers. Whether this unexpected finding is the result of chance or reflects an adverse interaction between carcinogens in cigarette smoke and components of fruit juice may be determined in future studies.

This study has a number of strengths. It is one of the first prospective examinations of fruit and vegetable intake and the risk for squamous cancers of the head and neck. The large number of cases afforded considerable statistical power for the investigation of this association and allowed the examination of oral cavity, oro-hypopharyngeal, and laryngeal cancers separately. Questionnaire data were collected before cancer diagnoses, minimizing the possibility of recall bias. To limit confounding, we utilized detailed covariate information and adjusted for important head and neck cancer risk factors, including cigarette smoking, alcohol use, education and physical activity. We also examined botanical groups. This study also has several limitations. Fruit and vegetable intake was measured at baseline via a food frequency questionnaire which is subject to measurement error.24 However, random non-differential misclassification of fruit and vegetable intake would most likely result in underestimating as opposed to overestimating the observed association. Also, separating fruit and vegetable intake into botanical groups resulted in multiple comparisons increasing the likelihood of a Type I error. We also lacked information on smoking initiation, marijuana use, and past alcohol consumption.

In conclusion, results from this large prospective observational study are consistent with previous studies and support the hypothesis that total fruit and vegetable intake is associated with reduced risk of head and neck cancer.

Acknowledgements

Cancer incidence data from the Atlanta metropolitan area were collected by the Georgia Center for Cancer Statistics, Department of Epidemiology, Rollins School of Public Health, Emory University. Cancer incidence data from California were collected by the California Department of Health Services, Cancer Surveillance Section. Cancer incidence data from the Detroit metropolitan area were collected by the Michigan Cancer Surveillance Program, Community Health Administration, State of Michigan. The Florida cancer incidence data used in this report were collected by the Florida Cancer Data System under contract to the Department of Health (DOH). The views expressed herein are solely those of the authors and do not necessarily reflect those of the contractor or DOH. Cancer incidence data from Louisiana were collected by the Louisiana Tumor Registry, Louisiana State University Medical Center in New Orleans. Cancer incidence data from New Jersey were collected by the New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey State Department of Health and Senior Services. Cancer incidence data from North Carolina were collected by the North Carolina Central Cancer Registry. Cancer incidence data from Pennsylvania were supplied by the Division of Health Statistics and Research, Pennsylvania Department of Health, Harrisburg, Pennsylvania. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations or conclusions. The authors are indebted to the participants in the NIH-AARP Diet and Health Study for their outstanding cooperation.

Ancillary