Adolescent Plant Product Intake in Relation to Later Prostate Cancer Risk and Mortality in the NIH-AARP Diet and Health Study

Abstract

Background:

Although fruit and vegetable intake during adolescence, a potentially sensitive time period for prostate cancer (PCa) development, has been proposed to protect against PCa risk, few studies have investigated the role of adolescent plant product intake in PCa development.

Methods:

Intake of various vegetables, fruit, and grains by males at ages 12–13 y was examined in relation to later PCa risk and mortality in the NIH-AARP Diet and Health Study. Cox proportional hazards regression was used to calculate HRs and 95% CIs of nonadvanced (n = 14,238) and advanced (n = 2,170) PCa incidence and PCa mortality (n = 760) during 1,729,896 person-years of follow-up.

Results:

None of the plant products examined were associated consistently with all PCa outcomes. However, greater adolescent intakes of tomatoes (P-trend = 0.004) and nonstarch vegetables (P-trend = 0.025) were associated with reduced risk of nonadvanced PCa, and greater intakes of broccoli (P-trend = 0.050) and fruit juice (P-trend = 0.019–0.025) were associated with reduced risk of advanced PCa and/or PCa mortality. Positive trends were also observed for greater intakes of fruit juice (P-trend = 0.002), total fruit (P-trend = 0.014), and dark bread (P-trend = 0.035) with nonadvanced PCa risk and for greater intakes of legumes (P-trend < 0.001), fiber (P-trend = 0.001), and vegetable protein (P-trend = 0.013–0.040) with advanced PCa risk or PCa mortality.

Conclusions:

Our findings do not provide strong evidence to suggest that adolescent plant product intake is associated with reduced PCa risk.

Introduction

Prostate cancer (PCa) is a critical public health concern, with almost 1.3 million new cases and 359,000 deaths worldwide in 2018 (1). PCa incidence rates are particularly high in Western countries, such as Australia/New Zealand, Northern and Western Europe, and North America, where a traditional Western diet is consumed, and relatively low in Asian countries, where a plant-based diet is more typical (1, 2). In addition, PCa rates tend to increase when men migrate from low- to high-PCa risk countries [e.g., from Japan to the United States (3)], suggesting that environmental factors, such as a plant-based diet, may potentially protect against PCa risk. This hypothesis is supported by findings from ecologic studies that observed inverse correlations for per-capita pulse (e.g., beans, lentils, and peas) and cereal consumption with PCa mortality (4).

Plant products, such as pulses and cereals, have been proposed to reduce PCa risk by several possible mechanisms. These include 1) protection against antioxidant-mediated DNA damage, such as by lycopene (found primarily in tomatoes), other carotenoids (found in colored fruit and vegetables), sulforaphane and indole-3 carbinol (found in cruciferous vegetables, such as broccoli), and selenium and vitamin E (found in whole grains and nuts) (5, 6). Additional possible mechanisms include 2) increasing sex hormone binding globulin concentrations and decreasing insulin resistance, such as by dietary fiber (7, 8), and 3) increasing the age at onset of puberty and slowing the timing of puberty, such as by vegetable protein and fat (9).

Despite the plausibility of these mechanisms and extensive investigation of diet and PCa over the years, limited evidence exists to suggest that overall adult intake of plant products lowers PCa risk (10). These null findings may potentially be explained by an earlier window of etiologic action for diet—for instance, during adolescence, when the prostate grows and develops rapidly, rather than later in adulthood once it has matured (11). This hypothesis is supported by several lines of evidence. First, mathematical models of human data suggest that initial genomic mutations in PCa occur as early as puberty in some men (12), consistent with a growing body of literature for other cancers (13, 14). Second, PCa precursor lesions and small PCa foci have been observed in men as young as their 20s and 30s (12, 15, 16), and the prevalence of these lesions, as well as rates of PCa incidence and mortality, begins to diverge by race in men as young as their 40s and 50s (17). Third, positive associations have been observed for factors related to childhood and adolescent diet (e.g., greater adult height and younger age at puberty) with PCa risk (9, 18–22), as well as for younger age at migration from low- to high-PCa risk countries (11), further supporting an early life contribution to PCa development. Finally, data from animal models are also supportive, with stronger associations observed for adolescent exposures (e.g., diet) than for adult exposures with prostate lesion development (23–25).

The purpose of this study was to advance our understanding of the role of adolescent diet, particularly adolescent intake of plant products, in PCa development and progression to fatal disease. To our knowledge, this exposure has only been investigated in 4 small studies to date, with generally mixed findings for PCa (26–29). Therefore, we took advantage of data from the large NIH-AARP Diet and Health (AARP) Study to investigate the relation between adolescent intake of a number of plant products and nutrients in relation to later PCa risk and mortality.

Methods

Study population and design

In 1995, the US National Cancer Institute invited 3.5 million AARP members 50–70 y of age residing in 6 states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and 2 metropolitan areas (Atlanta and Detroit) to participate in the AARP Study. Consenting participants completed a mailed baseline questionnaire regarding their demographics, lifestyle characteristics, and current diet. Six months later, participants completed a supplementary risk factor questionnaire regarding their diet during adolescence (ages 12–13 y), and in 2004–2005, they completed a follow-up questionnaire to update and expand their medical and lifestyle information.

For the present investigation, we limited the analysis to men who completed the risk factor questionnaire (n = 213,900) and then further excluded men who 1) completed questionnaires by proxy (n = 19,184); 2) reported any cancers, including PCa, before the date of completion of the risk factor questionnaire (n = 15,703); 3) reported extreme energy intake during adolescence (<500 or >4500 kcal/d, n = 10,003); 4) had 4 or more missing adolescent food items (n = 4121); or 5) had any missing adolescent plant product items (n = 13,408). The final analytic cohort included 150,671 men. This study was approved by the National Cancer Institute’s Special Studies Institutional Review Board.

Assessment and definitions of adolescent plant product and nutrient intake

The supplementary risk factor questionnaire included a 37-item FFQ that assessed participants’ diet at ages 12–13 y. This FFQ was designed to assess food items that are major sources of nutrients hypothesized to be important for carcinogenesis at the time of FFQ development: calcium, vitamin A, fiber, vitamin C, and fat (30). For each food item, participants were asked to select their frequency of intake from 8 possible responses: “never,” “1–11 times/year,” “1–3 times/month,” “1–2 times/week,” “3–4 times/week,” “5–6 times/week,” “1 time/day,” and “2 or more times/day.” Portion size was not assessed on the FFQ; therefore, we used data from the USDA 1965–1966 Household Food Consumption Survey to estimate portion size for each item as the median size for boys 12–13 y of age. We also used these data to estimate the maximum frequency of intake as 2 times/d.

Plant products were evaluated as individual items and as groups of items. Groups included tomatoes (tomato soup and fresh tomatoes, as well as tomato soup, fresh tomatoes, and pizza), nonstarch vegetables (tomato soup, fresh tomatoes, lettuce salads, broccoli, carrots, and green beans), total vegetables (nonstarch vegetables and legumes), and fruit (oranges, orange juice, apples, and canned fruits). Intakes of nutrients, such as vegetable fat, vegetable protein, carbohydrates, and fiber, were also estimated and adjusted for total energy intake using the nutrient density method (31). These nutrient intakes were estimated as percentage of total adolescent energy intake and then divided into quintiles based on the overall study population.

Prostate cancer ascertainment and definitions

PCa diagnosis, grade (histologic), and stage [Surveillance, Epidemiology, and End Results Program (SEER) disease extent and summary stage] were obtained from 11 state cancer registries, including those in the 8 original states and regions, as well as those in states where participants commonly moved during follow-up [Arizona, Nevada, and Texas; 97.5% of participants captured by these registries (32)]. Vital status was ascertained through linkage with the US Social Security Administration Death Master File, National Death Index, and questionnaire responses from next of kin. This method of cancer ascertainment was found to be 89.2% sensitive and 99.5% specific for cancer in a previous validation study in AARP (32). PCa cases were defined based on clinical stage, lymph node involvement, and metastasis. Specifically, 1) total PCa was defined as a first primary diagnosis of PCa (International Classification of Diseases, Ninth Revision code 185 or International Classification of Diseases, Tenth Revision code C61); 2) advanced PCa as clinical stages T3–T4, lymph node involvement (N1), metastasis (M1), or PCa death; and 3) PCa mortality as PCa death. Participants were followed through 31 December, 2011.

Assessment of covariates

All covariates were assessed by self-administered questionnaire. Information on participants’ age (years), race/ethnicity (white, black, and other), education (12 y or less, post–high school or some college, college graduate, and postgraduate), marital status (married, separated, widowed, and never married), cigarette smoking history (never, former, and current), adolescent and adult physical activity levels (never, rarely, 1–3 times per month, 1–2 times per week, 3–4 times per week, and 5 or more times per week), personal history of diabetes (yes, no), family history of cancer (yes, no), current waist circumference (inches) and height (m), and weight (kg) at ages 18 y, 35 y, 50 y, and currently was assessed at baseline. Information on participants’ prostate-specific antigen (PSA) testing (never, once, and more than once) and digital rectal examination (DRE) screening history (never, once, and more than once) was assessed on the supplementary risk factor questionnaire, and information on participants’ father’s occupation [professional or technical (e.g., doctor, lawyer), managerial (e.g., plant manager, CEO), other nonmanual (e.g., bank teller), manual, in a trade (e.g., carpenter), and other manual (e.g., farm laborer)] was assessed on the follow-up questionnaire.

Statistical analysis

Cox proportional hazards regression was used to estimate HRs and 95% CIs for the association between each plant product and PCa risk and mortality. Person-time was calculated from the date of completion of the risk factor questionnaire until the earliest of the following dates: PCa diagnosis, death, relocation outside of the cancer registry catchment area, or 31 December, 2011. Frequencies of plant product intake were modeled as categorical variables, and P values for linear trends were calculated by assigning the median to each frequency category, unless categories were combined, in which case weighted medians were used (e.g., upper category for legumes). Energy-adjusted intakes of nutrients were categorized into quintiles, and P-trends were calculated using median values. The shape of all associations (i.e., linear compared with nonlinear) was explored by spline regression (Supplemental Figure 1).

We performed our statistical analysis in a sequential manner, similar to previous analyses of adolescent diet and PCa risk in the AARP Study (33, 34). In model 1, we adjusted for age. In model 2, we adjusted for covariates that could serve only as potential confounders, as well as those that could serve as either potential confounders or mediators. Covariates included race/ethnicity, education, marital status, smoking history, waist circumference, attained height, father’s occupation, family history of PCa, screening PSA and DRE history, diabetes history, adolescent energy intake, adult alcohol intake, adolescent and adult BMI (in kg/m²), adolescent and adult physical activity, and adult intake of the exposures of interest (e.g., adult tomato intake for adolescent tomato intake) to control for the correlation between adolescent and adult dietary exposures. In model 3, we explored the effect of plant products independent of other food groups by controlling for adolescent intake of other food groups (e.g., red meat and dairy). As no meaningful differences were observed in the estimates from models 1–3, only those from model 3 are presented.

As a sensitivity analysis, we restricted the analyses to heavily screened men to investigate the potential for detection bias. Heavily screened men were defined as those with a history of at least 2 PSA tests and 2 DREs in the 3 y before baseline. We also performed analyses restricted to men aged <55 y at the time of completion of the adolescent FFQ, as younger individuals (<55 y of age) have been found to have better accuracy of recalled adolescent diet than older individuals (35–40). Additionally, we performed analyses stratified by calendar time that passed through puberty 1) completely before the Great Depression or World War II (WWII), 2) partially during WWII, and 3) completely following WWII. This stratified analysis is to capture food availability during adolescence, and to check the likelihood that diet recalled for early puberty (ages 12–13 y) also reflected diet later in puberty when the prostate grows and develops rapidly (41–43).

Results

During a total of 1,729,896 person-years of follow-up, 14,238 men were diagnosed with nonadvanced PCa and 2170 with advanced disease; 760 men died of their disease (Table 1). The baseline characteristics of these men are summarized in Table 1 by adolescent vegetable, fruit, and fiber intake. Overall, participants tended to be in their early 60s at baseline, and most were non-Hispanic white, married, and former or current smokers and had been screened for PCa more than once. Participants with a higher frequency of adolescent vegetable, fruit, and fiber intake were less likely to be non-Hispanic white and married and more likely to engage in physical activity (in both adolescence and adulthood) and to consume lesser frequencies of red meat as adolescents and higher amounts of vegetables, fruit, and fiber as adults. Considering vegetable and fruit intake separately from fiber, men who consumed vegetables and fruit with greater frequency as adolescents were more likely to have received a postcollege education, to have fathers with professional or technical occupations, to have a lower BMI as adults, and to be never smokers, whereas the opposite pattern was observed for fiber. Participants with higher frequencies of vegetable and fiber intake as adolescents tended to be older, were more likely to have passed through puberty before or during World War II, and consumed sweets and dairy products with lesser frequency in adolescence.

TABLE 1.

Baseline demographic, lifestyle, and medical characteristics of men by adolescent total vegetable, fruit, and fiber intake, NIH-AARP Diet and Health Study, 1995–2011¹

		Frequency of total vegetable intake			Frequency of total fruit intake			Quintile of energy-adjusted fiber intake, g/1000 kcal
Characteristic	Total	≤4/wk	1/d	≥3/d	≤2/wk	5–6/wk	≥2/d	Q1 (<5.2)	Q3 (6.3–7.4)	Q5 (>8.9)
N	150,671	19,748	44,391	20,533	39,600	22,452	27,947	30,134	30,135	30,134
Person-years	1,729,896	226,738	508,461	235,305	453,047	259,088	319,900	351,045	346,426	340,915
Nonadvanced PCa cases, n	14,238	1772	4368	1884	3611	2089	2672	2775	2882	2767
Advanced PCa cases, n	2170	268	640	325	567	306	415	412	444	480
Person-years for the PCa mortality analysis	2,045,929	266,524	604,145	277,627	532,991	306,244	379,111	413,990	409,601	402,788
PCa mortality, n	760	91	230	122	224	104	138	125	149	189
Demographic characteristics
Age, y	61.8 ± 5.3	61.4 ±5.4	61.8 ± 5.3	62.0 ± 5.3	62.1 ± 5.2	61.6 ± 5.3	61.6 ± 5.4	60.8 ± 5.4	61.8 ± 5.3	62.8 ± 5.1
Non-Hispanic white	94.8	95.2	95.4	91.8	94.2	95.3	93.8	97.1	96.0	88.8
Education2
12 or less y	17.7	23.9	17.4	15.4	24.6	15.6	14.0	16.0	16.9	21.8
Post-high school or some college	31.1	32.9	31.3	29.4	35.1	31.5	27.1	29.7	30.4	33.4
College graduate	23.4	21.7	24.0	22.4	19.8	24.3	24.5	25.3	24.1	20.1
Postgraduate	25.8	19.6	25.4	30.5	18.2	26.6	32.5	27.7	26.6	22.0
Married	85.7	86.3	86.5	82.9	86.7	86.0	83.4	86.1	85.8	84.9
Adolescent characteristics
Father's occupation
Professional or technical (e.g., doctor, lawyer)	6.0	4.0	5.9	7.5	3.3	6.2	8.7	6.4	6.4	4.9
Managerial (e.g., plant manager, CEO)	9.7	8.3	9.9	9.9	6.2	9.8	12.7	11.1	10.1	7.5
Other nonmanual (e.g., bank teller)	9.7	9.9	9.8	8.5	7.8	10.3	10.5	11.5	9.7	7.5
Manual (e.g., carpenter, electrician)	18.4	18.4	18.7	18.1	19.1	19.2	16.8	17.4	18.6	18.5
Other manual (e.g., farm laborer)	17.3	18.5	17.6	16.3	22.4	17.1	12.5	16.3	16.9	19.4
Unknown	38.9	40.2	38.1	40.3	41.2	37.5	38.8	37.4	38.3	42.2
Participants' age in 1945, y
<12	47.5	50.5	47.6	45.9	44.5	48.6	49.2	55.4	47.2	39.4
12 to 18	45.5	43.0	45.4	46.3	47.9	44.6	44.1	39.3	46.2	51.1
>18	7.1	6.5	7.0	7.9	7.6	6.8	6.7	5.3	6.7	9.5
BMI at age 18, kg/m2	21.7 ± 3.3	21.7 ±3.4	21.7 ± 3.3	21.8 ± 3.4	21.5 ± 3.3	21.8 ± 3.2	21.9 ± 3.3	21.9 ± 4.2	21.7 ± 3.2	21.5 ± 3.2
Physical activity ≥5 times/wk	50.2	43.9	48.7	57.9	46.5	50.1	55.8	47.2	50.2	52.9
Food type consumed,3 times/d
Tomato	0.46 ± 0.36	0.16 ± 0.12	0.34 ± 0.19	0.98 ± 0.45	0.37 ± 0.30	0.46 ± 0.31	0.61 ± 0.45	0.29 ± 0.25	0.46 ± 0.33	0.61 ± 0.45
Broccoli	0.05 ± 0.10	0.01 ± 0.02	0.04 ± 0.06	0.13 ± 0.19	0.03 ± 0.06	0.05 ± 0.08	0.09 ± 0.15	0.03 ± 0.05	0.05 ± 0.08	0.08 ± 0.15
Nonstarch vegetables	1.33 ± 0.83	0.38 ± 0.15	1.04 ± 0.22	2.80 ± 0.76	1.08 ± 0.65	1.31 ± 0.70	1.71 ± 1.00	0.88 ± 0.57	1.35 ± 0.74	1.71 ± 1.02
Legumes	0.24 ± 0.25	0.07 ± 0.08	0.19 ± 0.17	0.46 ± 0.36	0.27 ± 0.27	0.23 ± 0.23	0.23 ± 0.26	0.08 ± 0.08	0.22 ± 0.18	0.45 ± 0.34
Potatoes	0.68 ± 0.38	0.61 ± 0.35	0.67 ± 0.34	0.77 ± 0.46	0.74 ± 0.39	0.67 ± 0.36	0.62 ± 0.40	0.61 ± 0.36	0.70 ± 0.37	0.70 ± 0.42
Total vegetables	1.57 ± 0.93	0.45 ± 0.16	1.23 ± 0.17	3.26 ± 0.80	1.34 ± 0.77	1.55 ± 0.80	1.94 ± 1.11	0.96 ± 0.58	1.57 ± 0.79	2.16 ± 1.11
Total fruit	0.92 ± 0.77	0.71 ± 0.57	0.86 ± 0.64	1.28 ± 1.01	0.20 ± 0.10	0.78 ± 0.08	2.15 ± 0.68	0.60 ± 0.50	0.93 ± 0.69	1.20 ± 1.01
Dark bread	0.18 ± 0.34	0.14 ± 0.27	0.16 ± 0.30	0.30 ± 0.48	0.14 ± 0.32	0.17 ± 0.31	0.28 ± 0.42	0.10 ± 0.21	0.18 ± 0.33	0.27 ± 0.45
White bread	0.92 ± 0.62	0.99 ± 0.59	0.93 ± 0.60	0.81 ± 0.67	1.11 ± 0.63	0.88 ± 0.60	0.70 ± 0.62	0.93 ± 0.61	0.95 ± 0.62	0.79 ± 0.61
Sweets	0.82 ± 0.71	0.96 ± 0.69	0.86 ± 0.66	0.67 ± 0.84	0.80 ± 0.62	0.84 ± 0.63	0.83 ± 0.84	0.98 ± 0.84	0.88 ± 0.69	0.55 ± 0.53
Total grains	1.12 ± 0.65	1.16 ± 0.62	1.11 ± 0.62	1.13 ± 0.71	1.27 ± 0.66	1.08 ± 0.62	1.00 ± 0.70	1.07 ± 0.61	1.16 ± 0.66	1.08 ± 0.68
Dairy products	1.64 ± 0.84	1.80 ± 0.78	1.67 ± 0.81	1.47 ± 0.89	1.62 ± 0.78	1.62 ± 0.79	1.66 ± 0.83	2.08 ± 0.78	1.65 ± 0.82	1.14 ± 0.69
Total red meat	1.22 ± 0.68	1.24 ± 0.06	1.23 ± 0.60	1.19 ± 0.85	1.25 ± 0.62	1.25 ± 0.62	1.13 ± 0.77	1.35 ± 0.72	1.28 ± 0.66	0.93 ± 0.59
Fiber, g/1000 kcal	7.22 ± 2.53	4.58 ± 1.83	6.80 ± 2.12	10.02 ± 3.05	6.21 ± 2.84	7.16 ± 2.18	8.99 ± 2.58	4.40 ± 0.66	6.84 ± 0.31	11.02 ± 2.42
Adult (baseline) lifestyle and medical characteristics
BMI, kg/m2	27.1 ± 4.2	27.4 ± 4.2	27.1 ± 4.1	27.2 ± 4.4	27.4 ± 4.3	27.1 ± 4.1	27.0 ± 4.3	27.1 ± 4.2	27.1 ± 4.2	27.1 ± 4.2
Height, m	1.78 ± 0.07	1.78 ± 0.07	1.79 ± 0.07	1.78 ± 0.08	1.78 ± 0.07	1.79 ± 0.07	1.79 ± 0.08	1.79 ± 0.07	1.79 ± 0.07	1.78 ± 0.07
Physical activity ≥5 times/wk	22.4	20.6	21.7	26.2	21.2	22.0	25.5	22.0	22.2	23.4
Intake of
Total vegetables, servings/d	2.0 ± 1.3	1.6 ± 1.0	1.9 ± 1.1	2.7 ± 1.7	1.8 ± 1.1	2.0 ± 1.2	2.4 ± 1.6	1.8 ± 1.1	2.0 ± 1.1	2.3 ± 1.1
Total fruit, servings/d	2.1 ± 1.7	1.8 ± 1.5	2.0 ± 1.5	2.6 ± 2.1	1.8 ± 1.6	2.0 ± 1.5	2.7 ± 2.1	1.9 ± 1.5	2.0 ± 1.5	2.4 ± 1.5
Fiber, g/1000 kcal	10.6 ± 3.7	9.7 ± 3.6	10.4 ± 3.5	11.7 ± 4.0	10.1 ± 3.6	10.6 ± 3.6	11.3 ± 3.9	9.9 ± 3.6	10.5 ± 3.6	11.4 ± 3.6
Family history of PCa
Yes	9.6	8.9	9.5	10.2	9.4	9.6	9.4	9.4	9.8	9.5
Unknown	18.9	19.8	18.7	19.4	19.3	18.4	19.6	18.5	18.6	20.4
Never smoker	30.1	28.0	29.9	32.2	26.0	30.9	33.6	30.0	30.6	29.8
Diabetes mellitus status	9.4	10.1	9.0	9.8	10.2	9.1	9.9	9.1	8.9	10.1
History of PSA screening
Never	20.6	21.6	20.2	20.7	21.4	20.4	19.9	20.1	20.5	21.5
Once	24.1	24.5	24.0	24.3	24.6	24.5	23.7	24.1	24.1	24.6
More than once	47.8	46.4	48.3	47.3	46.2	47.8	48.9	48.1	48.0	46.1
Unknown	7.5	7.6	7.5	7.7	7.8	7.3	7.5	7.7	7.4	7.7

¹Values are means ± SDs for continuous variables and percentages for categorical variables unless otherwise indicated. PCa, prostate cancer; PSA, prostate-specific antigen.

²Some percentages may not sum to 100% because of missing values.

³Means for vegetable and fruit intake are adjusted for adolescent energy intake. Standard deviations are not adjusted for adolescent energy intake.

Overall, participants reported consuming vegetables often during adolescence, with ~29% of participants consuming vegetables once/d, 31% twice/d, and smaller proportions consuming vegetables ≤4 times/wk (13%), 5–6 times/wk (13%), and ≥3 times/d (14%). For fruit, these values were more equally distributed, with 27% consuming fruit ≤2 times/wk, 18% 3–4 times/wk, 15% 5–6 times/wk, 22% once/d, and 19% ≥twice/d. Overall, fiber intake was low in the cohort, ranging from a median of 4.89 g/1000 kcal in the lowest quintile to 16.04 g/1000 kcal in the highest quintile. The largest contributor to fiber intake was legumes (19.8%), followed by white bread (10.2%), citrus fruit (9.3%), peanut butter (8.5%), apples (7.8%), and potatoes (7.0%).

Table 2 presents fully adjusted HRs for risks of nonadvanced and advanced PCa, as well as PCa mortality by increasing frequency of vegetable intake during adolescence. Overall, none of the vegetables or vegetable groups considered were associated consistently with all PCa outcomes. However, a few were associated with individual outcomes. Greater adolescent intakes of tomatoes (P-trend = 0.004) and nonstarch vegetables (P-trend = 0.025) were associated with significantly reduced risk of nonadvanced PCa, whereas no trends were observed for risk of advanced PCa or PCa mortality. In contrast, greater adolescent intake of broccoli was associated with reduced PCa mortality (P-trend = 0.050) and possibly reduced risk of advanced PCa (P-trend = 0.158) but not with reduced risk of nonadvanced PCa. Other plant products, on the other hand, were associated with increased risk of PCa outcomes. For legumes, fiber, and vegetable protein, no clear linear trends were observed with risk of nonadvanced PCa, whereas suggestive or significant positive trends were observed with risk of advanced PCa and/or PCa mortality (legumes: P-trend < 0.001 for advanced PCa and P-trend = 0.149 for PCa mortality; fiber: P-trend = 0.001 for advanced PCa and P-trend = 0.072 for PCa mortality; and vegetable protein: P-trend = 0.013 for advanced PCa and P-trend = 0.040 for PCa mortality). For all other plant products [total vegetables, carrots, lettuce, green beans, peanut butter, pizza (a contributor to tomato intake), and vegetable fat], no clear trends were observed with any PCa outcome (Table 2, Supplemental Tables 1–2, and Supplemental Figure 1).

TABLE 2.

Fully adjusted HRs and 95% CIs of prostate cancer by vegetable intake at 12–13 y of age in the NIH-AARP Diet and Health Study, 1996–2011¹

Food item	HR (95% CI) by intake frequency					P-trend
Tomato soup and fresh tomato	≤1/wk	2/wk	3–4/wk	5–6/wk	≥1/d
Nonadvanced PCa	1	1.02 (0.98, 1.07)	1.01 (0.96, 1.06)	0.96 (0.91, 1.02)	0.91 (0.85, 0.97)	0.004
Advanced PCa	1	1.02 (0.91, 1.14)	0.96 (0.84, 1.08)	0.94 (0.81, 1.10)	1.05 (0.89, 1.24)	0.857
PCa mortality	1	0.87 (0.71, 1.07)	0.92 (0.75, 1.14)	0.80 (0.62, 1.04)	1.00 (0.77, 1.30)	0.585
Pizza2	≤11/y	≥1/mo
Nonadvanced PCa	1	0.97 (0.93, 1.03)				0.312
Advanced PCa	1	0.91 (0.80, 1.04)				0.151
PCa mortality	1	0.88 (0.69, 1.12)				0.300
Broccoli	≤11/y	2/mo	≥1/wk
Nonadvanced PCa	1	1.01 (0.97, 1.06)	1.03 (0.98, 1.09)			0.284
Advanced PCa	1	1.01 (0.91, 1.12)	0.90 (0.78, 1.04)			0.158
PCa mortality	1	0.79 (0.65, 0.95)	0.77 (0.60, 1.00)			0.050
Nonstarch vegetables3	≤4/wk	5–6/wk	1/d	2/d	≥3/d
Nonadvanced PCa	1	1.05 (0.99, 1.11)	1.01 (0.96, 1.07)	0.98 (0.93, 1.04)	0.94 (0.87, 1.02)	0.025
Advanced PCa	1	0.98 (0.84, 1.13)	1.01 (0.89, 1.15)	1.04 (0.90, 1.20)	1.14 (0.94, 1.37)	0.140
PCa mortality	1	0.89 (0.70, 1.15)	1.06 (0.86, 1.31)	0.89 (0.70, 1.14)	1.09 (0.80, 1.48)	0.906
Legumes	≤11/y	2/mo	1–2/wk	≥3/wk
Nonadvanced PCa	1	1.06 (1.00, 1.12)	1.04 (0.98, 1.10)	0.98 (0.92, 1.05)		0.016
Advanced PCa	1	1.09 (0.94, 1.26)	1.12 (0.97, 1.30)	1.25 (1.06, 1.46)		<0.001
PCa mortality	1	0.95 (0.73, 1.22)	1.12 (0.88, 1.44)	1.16 (0.88, 1.52)		0.149
Total vegetables	≤4/wk	5–6/wk	1/d	2/d	≥3/d
Nonadvanced PCa	1	1.05 (0.98, 1.12)	1.06 (1.00, 1.12)	1.00 (0.94, 1.06)	0.96 (0.89, 1.03)	0.013
Advanced PCa	1	0.94 (0.79, 1.12)	1.02 (0.88, 1.18)	1.02 (0.87, 1.19)	1.12 (0.93, 1.34)	0.133
PCa mortality	1	1.15 (0.86, 1.52)	1.10 (0.85, 1.41)	0.93 (0.71, 1.21)	1.16 (0.85, 1.58)	0.965
Fiber, g per 1000 kcal	Q1 (<5.2)	Q2 (5.2–<6.3)	Q3 (6.3–<7.4)	Q4 (7.4–<8.9)	Q5 (≥8.9)
Nonadvanced PCa	1	1.05 (1.00, 1.11)	1.02 (0.96, 1.07)	1.01 (0.96, 1.07)	0.97 (0.91, 1.03)	0.084
Advanced PCa	1	1.00 (0.87, 1.15)	1.11 (0.96, 1.27)	1.09 (0.95, 1.27)	1.26 (1.08, 1.46)	0.001
PCa mortality	1	0.90 (0.70, 1.16)	1.04 (0.82, 1.34)	1.16 (0.91, 1.49)	1.16 (0.90, 1.51)	0.072
Vegetable fat, % of kcal	Q1 (<8.8)	Q2 (8.8–< 12.0)	Q3 (12.0–< 15.0)	Q4 (15.0–<18.8)	Q5 (≥18.8)
Nonadvanced PCa	1	1.00 (0.94, 1.06)	1.00 (0.94, 1.07)	1.01 (0.94, 1.08)	0.96 (0.89, 1.03)	0.223
Advanced PCa	1	1.09 (0.94, 1.26)	1.02 (0.86, 1.20)	1.17 (0.99, 1.39)	1.16 (0.96, 1.40)	0.110
PCa mortality	1	0.94 (0.73, 1.22)	1.00 (0.76, 1.32)	1.05 (0.79, 1.41)	1.08 (0.79, 1.49)	0.448
Vegetable protein, % of kcal	Q1 (<3.4)	Q2 (3.4–<4.1)	Q3 (4.1–<4.7)	Q4 (4.7–< 5.7)	Q5 (≥5.7)
Nonadvanced PCa	1	1.01 (0.96, 1.07)	1.01 (0.95, 1.07)	1.00 (0.94, 1.06)	0.98 (0.92, 1.05)	0.484
Advanced PCa	1	1.03 (0.89, 1.19)	1.13 (0.97, 1.30)	1.11 (0.95, 1.30)	1.24 (1.04, 1.47)	0.013
PCa mortality	1	0.96 (0.74, 1.23)	1.04 (0.80, 1.35)	1.01 (0.77, 1.33)	1.29 (0.97, 1.73)	0.040

¹Model was adjusted for age, race/ethnicity, education, marital status, smoking history, waist circumference, attained height, father’s occupation, family history of PCa, screening prostate-specific antigen and digital rectal examination history, diabetes history, adolescent energy intake, adult alcohol intake, adolescent and adult BMI, adolescent and adult physical activity, adult dietary intake of the exposures of interest, and adolescent intake of red meat and dairy. PCa, prostate cancer.

²P value for pizza was calculated by comparing ≥1/mo to ≤11/y.

³Nonstarch vegetables include tomato soup, lettuce, tomatoes, broccoli, carrots, and other vegetables (peas, corns, and green beans).

Table 3 presents fully adjusted HRs for risks of nonadvanced and advanced PCa, as well as PCa mortality by increasing frequency of fruit intake during adolescence. Similar to the analysis of vegetable intake, none of the fruits or nutrients considered were associated consistently with all PCa outcomes. However, a few were associated with individual outcomes. Greater adolescent intake of fruit juice was associated with an increased risk of nonadvanced PCa (P-trend = 0.002) but a decreased risk of advanced PCa (P-trend = 0.019) and reduced PCa mortality (P-trend = 0.025). Greater adolescent intake of total fruit (P-trend = 0.014) was also associated with a significantly increased risk of nonadvanced PCa but not with risk of advanced PCa or PCa mortality. Finally, in contrast to these 2 patterns, greater adolescent apple intake was associated with a suggestive increased risk of advanced PCa (P-trend = 0.059) but not with risk of nonadvanced PCa or PCa mortality. No clear linear trends were observed for oranges, canned fruit, vitamin C, or vitamin A with any of the PCa outcomes evaluated (Table 3 and Supplemental Tables 1–2).

TABLE 3.

Fully adjusted HRs and 95% CIs of prostate cancer by fruit intake at 12–13 y of age in the NIH-AARP Diet and Health Study, 1996–2011¹

Food item	HR (95% CI) by intake frequency					P-trend
Oranges	≤11/y	2/mo	1–2/wk	≥3/wk
Nonadvanced PCa	1	1.04 (0.99, 1.08)	1.05 (1.00, 1.10)	1.05 (0.99, 1.11)		0.143
Advanced PCa	1	1.05 (0.93, 1.17)	1.05 (0.93, 1.19)	1.12 (0.97, 1.29)		0.140
PCa mortality	1	1.05 (0.87, 1.26)	0.91 (0.74, 1.12)	0.99 (0.79, 1.25)		0.672
Apples	≤11/y	2/mo	1–2/wk	≥3/wk
Nonadvanced PCa	1	1.01 (0.96, 1.06)	1.03 (0.98, 1.08)	0.98 (0.93, 1.04)		0.454
Advanced PCa	1	1.10 (0.97, 1.26)	1.02 (0.90, 1.17)	1.19 (1.03, 1.38)		0.059
PCa mortality	1	1.02 (0.82, 1.26)	1.05 (0.85, 1.31)	0.99 (0.77, 1.26)		0.861
Fruit juice	≤11/y	2/mo	1–2/wk	≥3/wk
Nonadvanced PCa	1	0.99 (0.95, 1.04)	1.04 (0.99, 1.09)	1.07 (1.02, 1.12)		0.002
Advanced PCa	1	1.04 (0.92, 1.17)	0.98 (0.86, 1.11)	0.89 (0.79, 1.00)		0.019
PCa mortality	1	0.97 (0.80, 1.19)	0.90 (0.73, 1.11)	0.80 (0.66, 0.98)		0.025
Total fruit	≤2/wk	3–4/wk	5–6/wk	1/d	≥2/d
Nonadvanced PCa	1	1.01 (0.96, 1.07)	1.01 (0.95, 1.07)	1.08 (1.03, 1.14)	1.06 (1.00, 1.12)	0.014
Advanced PCa	1	0.92 (0.81, 1.05)	0.90 (0.78, 1.04)	0.99 (0.87, 1.13)	0.96 (0.83, 1.11)	0.980
PCa mortality	1	0.94 (0.76, 1.16)	0.85 (0.67, 1.08)	0.83 (0.66, 1.04)	0.85 (0.67, 1.09)	0.157
Vitamin C, mg per 1000 kcal	Q1 (<20.9)	Q2 (20.9–<29.3)	Q3 (29.3–<41.9)	Q4 (41.9–<61.9)	Q5 (≥61.9)
Nonadvanced PCa	1	0.96 (0.91, 1.01)	0.98 (0.93, 1.03)	1.01 (0.96, 1.07)	1.03 (0.98, 1.09)	0.026
Advanced PCa	1	1.11 (0.97, 1.27)	1.07 (0.93, 1.22)	1.01 (0.88, 1.16)	1.05 (0.91, 1.21)	0.917
PCa mortality	1	1.14 (0.92, 1.42)	1.00 (0.80, 1.25)	0.89 (0.70, 1.12)	0.95 (0.75, 1.20)	0.182

¹Model was adjusted for age, race/ethnicity, education, marital status, smoking history, waist circumference, attained height, father’s occupation, family history of PCa, screening prostate-specific antigen and digital rectal examination history, diabetes history, adolescent energy intake, adult alcohol intake, adolescent and adult BMI, adolescent and adult physical activity, adult dietary intake of the exposures of interest, and adolescent intake of red meat and dairy. PCa, prostate cancer.

Table 4 presents fully adjusted HRs by increasing frequency of grain intake during adolescence. Greater adolescent intake of dark bread was associated with an increased risk of nonadvanced PCa (P-trend = 0.035) but not with risk of advanced PCa or PCa mortality. No associations were observed for intakes of white bread or carbohydrates (Table 4, Supplemental Tables 1–2, and Supplemental Table 3). There was no evidence of effect modification by calendar time, and no clear patterns were observed in men <55 y of age or in heavily screened men for any of the plant products considered.

TABLE 4.

Fully adjusted HRs and 95% CIs of prostate cancer by grains intake at 12–13 y of age in the NIH-AARP Diet and Health Study, 1996–2011¹

Food item	HR (95% CI) by intake frequency					P-trend
Dark bread	≤11/y	2/mo	1–2/wk	≥3/wk
Nonadvanced PCa	1	1.03 (0.98, 1.07)	1.07 (1.02, 1.13)	1.06 (1.01, 1.11)		0.035
Advanced PCa	1	1.05 (0.94, 1.18)	0.88 (0.77, 1.00)	0.95 (0.84, 1.07)		0.314
PCa mortality	1	1.09 (0.90, 1.32)	0.82 (0.65, 1.03)	0.92 (0.75, 1.13)		0.322
White bread	≤2/wk	3–4/wk	5–6/wk	1/d	2/d
Nonadvanced PCa	1	1.02 (0.97, 1.08)	1.00 (0.95, 1.06)	0.99 (0.94, 1.05)	0.96 (0.91, 1.02)	0.082
Advanced PCa	1	1.06 (0.91, 1.22)	1.05 (0.91, 1.21)	1.13 (0.97, 1.31)	1.12 (0.96, 1.30)	0.188
PCa mortality	1	1.18 (0.92, 1.53)	1.41 (1.11, 1.80)	1.22 (0.94, 1.58)	1.20 (0.92, 1.57)	0.578

¹Model was adjusted for age, race/ethnicity, education, marital status, smoking history, waist circumference, attained height, father’s occupation, family history of PCa, screening prostate-specific antigen and digital rectal examination history, diabetes history, adolescent energy intake, adult alcohol intake, adolescent and adult body mass index, adolescent and adult physical activity, adult dietary intake of the exposures of interest, and adolescent intake of red meat and dairy. PCa, prostate cancer.

Discussion

In this large prospective analysis of adolescent plant product intake and PCa risk, we observed considerable variation in associations across plant products and PCa outcomes. Overall, none of the individual plant products or groups examined were associated with all PCa outcomes, but a few were associated with individual outcomes. Greater adolescent intakes of tomatoes and nonstarch vegetables were associated with reduced risk of nonadvanced PCa, and greater intakes of broccoli and fruit juice were associated with reduced risk of advanced PCa or PCa mortality. These results support the hypothesis that a diet rich in plant products during adolescence may protect against PCa development. However, other findings for these same or different plant products support the opposite hypothesis. Greater adolescent intakes of fruit juice, total fruit, and dark bread were associated with increased risk of nonadvanced PCa, and greater intakes of legumes, fiber, and vegetable protein were associated with increased risk of advanced PCa and/or PCa mortality. Thus, overall, our results do not provide clear or consistent evidence for a role of adolescent plant product intake in PCa development.

Only a few previous studies have investigated adolescent plant product intake in relation to PCa to which to compare our findings. With respect to vegetable and fruit intake, our variable findings for adolescent plant product intake with nonadvanced PCa are somewhat consistent with those from a US cohort study that observed no association for a vegetarian lifestyle before age 15 y and overall PCa risk, comparing 1 or 2 Adventist and practicing vegetarian parents to none (RR: 0.80; 95% CI: 0.51, 1.26) (44). Our null findings for carrot intake are also consistent with those from a Swedish case-control study that observed no association for adolescent carrot intake (OR: 0.9; 95% CI: 0.6, 1.4) (26). With respect to grain intake, although our null findings for adolescent dark bread intake differ from those from a previous Icelandic case-control study (OR: 0.76; 95% CI: 0.59, 0.98 comparing >1 to ≤1 portions/d of rye bread for total PCa and OR: 0.47; 95% CI: 0.27, 0.84 for advanced PCa) (29), the much higher intake of whole grains in the Icelandic study may explain this discrepancy. Thus, where our intake overlapped with previous studies, our findings appear to be similar to the very limited literature to date.

Despite this general similarity to the adolescent diet literature, many of our findings were unexpected and differ from hypothesized associations. The only somewhat expected findings were for broccoli and tomato intake, that is, reduced PCa mortality and risk of possibly advanced PCa for broccoli intake and a reduced risk of nonadvanced PCa for tomato intake (soup and fresh tomato). A protective role for broccoli intake is plausible because cruciferous vegetables are rich in sulforaphane and indole-3 carbinol, phytochemicals that can neutralize DNA-damaging reactive oxygen species and thus possibly prevent or delay cancer development (45). Inverse associations have also been observed for adult cruciferous vegetable intake with risk of PCa (46–48), especially advanced PCa (48), in several epidemiologic studies.

Similar to broccoli intake, a protective role is also plausible for tomato intake because tomatoes contain lycopene and other carotenoids that can protect against antioxidant-mediated DNA damage (49) and because inverse associations have been observed between adult intake of tomatoes and PCa risk in several previous studies [RR for tomato intake: 0.86; 95% CI: 0.75, 0.98; P = 0.019; lycopene was linearly associated with a reduced risk of PCa with a threshold between 9 and 21 mg/d (50–52)]. Our study found that adolescent tomato intake was protective for nonadvanced PCa but not appreciably protective for advanced PCa and PCa mortality. These findings are consistent with those from the Health Professionals Follow-up Study, in which a stronger protective association was observed for adult tomato sauce intake with low- than high-grade PCa and with total PCa than PCa mortality (53). This finding may indicate that adolescent tomato intake plays a greater role in the early rather than later stages of PCa development.

In contrast to our findings for broccoli and tomato intakes, our positive trends for greater adolescent intake of other vegetables and related macronutrients (i.e., legumes, fiber, and vegetable protein) with increased risk of advanced PCa and possibly PCa mortality were unexpected. They also differ from the generally protective findings observed for adolescent vegetable protein and fiber intakes with risks of other cancers, such as breast cancer (or breast cancer risk factors), in several previous studies (54–56). They are consistent, however, with positive results observed for adolescent fiber intake with rectal cancer (but not colon or pancreatic cancer) in a previous AARP Study analysis (30, 57). Although these variable findings are generally difficult to explain, one possible reason could be differences in major sources of fiber intake across studies based on differences in the detail of their FFQs. In the AARP analysis, the most common contributors to adolescent vegetable protein and fiber intakes were legumes and white bread (vegetable protein: 27% from legumes and 14% from white bread; fiber: 20% from legumes and 10% from white bread). In contrast, common contributors to these nutrients in other studies that assessed diet in the 1960s to 1980s were grains, including breakfast cereals and vegetables (55, 56). In addition, in our study, the highest quintile of energy-adjusted adolescent fiber intake was only 16.3 g/d, whereas in other studies and analyses, it was considerably higher [Nurses’ Health Study II = 28.9 g/d for adolescent fiber intake and AARP Study = 36.4 g/d for adult fiber intake using a >100-item FFQ and >25 g/d in most of the highest fiber intake categories in a meta-analysis reporting no association between adulthood fiber and PCa risk (56, 58)]. Although these differences in absolute fiber intake might not in and of themselves explain our findings because relative fiber intake rankings could still have been preserved in our study, the fact that our highest contributors to fiber and vegetable intakes were also different from other studies speaks to the possibility that these variables captured different foods and nutrients than in previous studies with longer FFQs. This difference may be important as a previous study of breast cancer observed inverse findings with adolescent fiber intake from vegetables and fruit but not from grains and legumes (56).

Other unexpected findings in our analyses were those related to adolescent fruit intake. Specifically, we found that adolescent intakes of fruit juice and total fruit were associated with increased risk of nonadvanced PCa, whereas fruit juice was associated with a decreased risk of advanced PCa and reduced PCa mortality. These findings stand in contrast to the inverse findings observed for adolescent fruit intake with breast cancer in previous studies (59), as well as the null findings observed for adolescent fruit intake with colon, rectal, and pancreatic cancer in previous AARP analyses (30, 57). In general, these conflicting findings are difficult to explain. One possible reason for differences in findings across PCa outcomes could be detection bias. However, no differences were observed in baseline PCa screening history by adolescent fruit intake or after adjustment and stratification by baseline PCa screening. Therefore, any residual detection bias would have to be explained by later differences in screening by adolescent intake of fruit but not other plant products, which seems unlikely. Other possible explanations include residual confounding by geographic or economic correlates of adolescent fruit intake or chance.

Despite our unexpected findings, a major strength of this study was its use of prospective data. These data enabled us to establish a temporal relation between adolescent diet and PCa onset. In addition, the large sample size of the AARP Study and its detailed collection of data on participant demographic, lifestyle, and medical characteristics allowed us to investigate associations by PCa type, control for a wide range of confounders, and perform stratified analyses. There are important limitations to note as well. First, adolescent diet was assessed by self-report several decades after adolescence; thus, measurement error likely occurred. However, any measurement error should have been nondifferential by exposure status, leading to an attenuation of our findings toward the null rather than to a false strengthening of our findings away from the null. Furthermore, results in younger men, whose recall of adolescent diet should have been better than that of older men, were generally similar to those in older men. Second, the adolescent FFQ asked about participants’ diet at ages 12–13 y, which is earlier than when the prostate grows and develops rapidly in most boys [ages 14–15 y (9, 41–43)]. Although we assumed that boys’ diets would be unlikely to change in composition from ages 12–13 to 14–15 y, this might not have been the case for boys who entered and exited puberty during different world events, such as World War II. However, no effect modification was observed by calendar time, possibly suggesting that diet recalled for early puberty also reflected diet later in puberty. Third, adolescent diet was assessed by a relatively short FFQ (i.e., 37 rather than >100 items). Although this FFQ included items on 15 plant products, it still missed a number of other plant products, such as cereal, rice, bananas, melon, peppers, zucchini, and eggplant, which may have contributed to some of our unexpected findings. Fourth, we did not collect information on portion size on the adolescent FFQ. However, imputation of portion size using data collected near in time to participants’ adolescence should have helped to reduce this concern. Finally, PCa screening information was collected only before baseline. Therefore, we could not control for possible differences in screening after baseline, allowing for the potential of residual detection bias.

To summarize, in one of the few studies to investigate adolescent plant product intake and PCa risk, we found inconsistent evidence to suggest that adolescent intake of plant products is associated with PCa risk and mortality. In addition, even though some of our findings support a protective role for individual plant products (e.g., broccoli and tomatoes), positive findings for other plant products tempered our conclusions and raised the possibility of chance as an explanation for our findings. Studies with detailed, real-time, or more recently recalled adolescent diet and lifestyle data, as well as longitudinal PCa screening data, may be necessary to answer this challenging life course question.

Supplementary Material

Supplemental Tables 1–3 and Supplemental Figure 1 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.

Abbreviations used:

DRE

digital rectal examination

PCa

prostate cancer

PSA

prostate-specific antigen