Prostate cancer (PC) is the second most frequent (29·4 per 100 000) cancer diagnosis in men worldwide. This frequency greatly varies depending on the human development index of each country. PC incidence is almost three times higher in countries with high human development indices (57·7 per 100 000) compared with those with intermediate (19·8 per 100 000) and low (22·9 per 100 000) indices(Reference Sung, Ferlay and Siegel1). This, along with the increased risk associated with migration from low- to high-risk countries(Reference Lee, Demissie and Lu2), suggests that in addition to the well-known risk factors for PC (age, ethnicity and family history of PC), certain lifestyle characteristics, such as diet, may play an important role in prostate carcinogenesis(Reference Lee, Demissie and Lu2,Reference Rawla3) .
Chronic inflammation and oxidative stress are two of the mechanisms proposed for the development and progression of PC(Reference Federico, Morgillo and Tuccillo4,Reference De Marzo, Marchi and Epstein5) . The antioxidant potential of a diet depends on the balance between the dietary intake of vitamins A, C and E, phenolic compounds, carotenoids, n-3 fatty acids, Zn and Se and the intake of pro-oxidants, such as Cu, Fe and n-6 fatty acids(Reference Galaris, Barbouti and Pantopoulos6–Reference Liu, Li and Hou9). However, to date, the association between the antioxidant potential of the diet and PC has focused on evaluating the individual(Reference Li, Wu and Zhuang10–Reference Etminan, FitzGerald and Gleave12) or combined(Reference Pascual-Geler, Robles-Fernandez and Monteagudo13) antioxidant capacity of certain nutrients or has been assessed according to the total antioxidant capacity of the diet(Reference Russnes, Wilson and Epstein14–Reference Vance, Wang and Su17), with inconclusive results. Additionally, few studies have considered the histological differentiation or the clinical stage of PC.
The most frequently evaluated individual antioxidants are lycopene(Reference Li, Wu and Zhuang10), vitamin C(Reference Bai, Qu and Jiang11) and Se(Reference Etminan, FitzGerald and Gleave12). An umbrella and systematic review concluded that higher consumption of both lycopene(Reference Li, Wu and Zhuang10) and vitamin C(Reference Bai, Qu and Jiang11), respectively, is associated with a lower risk of PC. In contrast, higher Se consumption has only been associated with a decreased risk of PC in cohort studies(Reference Etminan, FitzGerald and Gleave12). Furthermore, a higher antioxidant score of the diet, estimated in a case–control study based on the combined consumption of vitamins A, C and E, Se and Zn, was not associated with lower PC risk(Reference Pascual-Geler, Robles-Fernandez and Monteagudo13). Studies on the total antioxidant capacity of the diet, conducted in the USA(Reference Russnes, Wilson and Epstein14,Reference Vance, Wang and Su17) , Sweden(Reference Russnes, Möller and Wilson15) and Iran(Reference Ghanavati, Clark and Bahrami16), used values of the Antioxidant Food Table database, which was developed in Norway(Reference Carlsen, Halvorsen and Holte18) using the ferric reducing antioxidant power assay. Among them, only a prospective cohort study conducted in the USA found that a diet with high antioxidant capacity was associated with a reduced risk of PC, particularly advanced PC(Reference Russnes, Wilson and Epstein14).
In Mexico, PC is the most frequently diagnosed cancer (39·1 per 100 000) and the leading cause of cancer-related death among men(Reference Sung, Ferlay and Siegel1). In the last decades, there has been a dietary transition characterised by a decrease in the consumption of natural foods, which are rich in antioxidants, and an increased intake of refined grains, fast food and sugary drinks(Reference Aburto, Pedraza and Sánchez-Pimienta19). Approximately half of Mexican adults do not eat fruits (50 %) and vegetables (58 %)(Reference Gaona-Pineda, Martínez-Tapia and Arango-Angarita20), and a high proportion of adult men have intakes of vitamins D (96 %), E (90 %) and A (70 %) below the estimated average requirement(Reference Pedroza-Tobías, Hernández-Barrera and López-Olmedo21).
This study aims to identify antioxidants’ consumption patterns (retinol, vitamins C and E, α- and beta-carotene, lycopene, lutein and zeaxanthin, Zn, Mn, Se, caffeic acid, ferulic acid, sinapic acid, luteolin, apigenin, myricetin, kaempferol, gallocatechin, epicatechin, epigallocatechin, epicatechin-3-O-gallate, epigallocatechin-3-O-gallate and n-3) and pro-oxidants (Cu, Fe, quercetin, (+)-catechin and n-6). Moreover, the association of consumption patterns with PC and its histological differentiation was assessed in residents of Mexico City.
Methods
This study comprises a secondary analysis based on a case–control study conducted between November 2011 and August 2014, involving males aged 42–94 years with no prior history of cancer and at least 1 year of residence in Mexico City(Reference Vázquez-Salas, Torres-Sánchez and López-Carrillo22). Cases (n 468) were men with a first histologically confirmed diagnosis of PC at any clinical stage, identified from four tertiary-level and two secondary-level hospitals. Based on histological differentiation (Gleason score), cases were classified as well-differentiated (Gleason ≤ 6), moderately differentiated (Gleason = 7) and poorly differentiated (Gleason ≥ 8). For each case, two population controls (n 920) were matched within ±5 years of the case’s age. Eligible controls were men with no prior diagnosis of PC, no symptoms suggestive of PC (e.g. dysuria, haematuria) and a recent prostate-specific antigen result below 4 ng/ml. To ensure a representative control selection, thirty-three basic geostatistical areas across Mexico City were randomly selected using the most recent National Count of Population and Housing (2005). In each area, ten blocks were sampled, and household visits began at the northeast corner of each block. Households were selected consecutively, and if multiple eligible men were present, one was selected at random. If a participant was not available at the time, up to three follow-up attempts were made before considering an alternative control(Reference Vázquez-Salas, Torres-Sánchez and López-Carrillo22).
Participation rates were 85·9 % (n 402) for cases and 87·5 % (n 805) for controls (Fig. 1). Men who did not agree to participate completed a questionnaire that collected information about their age, educational level, marital status and place of birth. Before signing the informed consent form, each participant was provided with information about the study’s characteristics and their involvement. The study was conducted following the principles established by the Declaration of Helsinki and was approved by the ethics committee of the National Institute of Public Health (INSP, CI:980) and by each of the participating hospitals.
Flow chart of the study population. PC, prostate cancer.
Interview
Data were collected through direct, structured interviews conducted by trained personnel blinded to the study hypotheses. Cases were interviewed in the hospital, and controls were identified and interviewed in their homes within 1 month of case identification(Reference Vázquez-Salas, Torres-Sánchez and López-Carrillo22). The questionnaire included demographic variables such as age, education level (elementary or less, middle school, high school and bachelor’s or postgraduate), region of birth (Mexico City: Mexico City; south: Campeche, Chiapas, Guerrero, Oaxaca, Quintana Roo and Yucatán; central-western: Aguascalientes, Colima, Guanajuato, Jalisco and Michoacán; central-Eastern: Hidalgo; Estado de México, Morelos, Puebla, Querétaro and Tlaxcala; north: Chihuahua, Coahuila, Durango, San Luis Potosí, Zacatecas, Baja California, Baja California Sur, Sinaloa, Sonora, Nayarit, Nuevo León and Tamaulipas; and east: Veracruz and Tabasco) and marital status (single v. not single). Participants were also asked about their personal history of chronic and sexually transmitted diseases (yes or no) and first-degree family history of PC (yes or no), as well as their smoking history, lifetime weight changes and physical activity. The interviews were conducted by trained personnel who were unaware of the specific hypotheses of the study.
Dietary information
Dietary intake was assessed using a semi-quantitative FFQ consisting of 127 items, previously validated in studies of the Mexican population(Reference Hernández-Ramírez, Galván-Portillo and Ward23). Cases were asked about their dietary habits during the 3 years before diagnosis, while controls reported their dietary habits for the 3 years preceding the interview. Consumption frequency ranged from ‘never’ to ‘six or more times per day’ and was converted into daily grams or millilitres equivalents using standardised portion sizes based on the National Nutrition Survey 1999(Reference Galván-Portillo, Vázquez-Salas and Hernández-Pérez24). Daily intake of macro- and micronutrients was estimated by multiplying the reported consumption (g/d) by the nutrient content of each taxonomically identified food item using the Food Processor Nutrition Analysis and Fitness Software (version 10.11.0, 2011; ESHA Research, Inc.), which includes data for the traditional Mexican dishes included in the questionnaire. Flavonoid and polyphenol intakes were estimated by multiplying daily consumption by the average individual flavonoid content (mg/100 g) across fifty-seven taxonomically identified food items. These values were derived from the USDA database of flavonoid content in selected foods (Release 3.1) and the Phenol-Explorer database (version 3.6)(Reference Galván-Portillo, Vázquez-Salas and Hernández-Pérez24).
Based on the antioxidant and pro-oxidant potential described in the literature, the following nutrients were selected as antioxidants(Reference Fang, Yang and Wu25): retinol, vitamins C and E, α- and beta-carotene, lycopene, lutein, zeaxanthin, Zn, Mn, Se, phenolic acids (caffeic, ferulic and synaptic acid), flavonoids (flavones: luteolin, apigenin; flavonols: myricetin, kaempferol; and flavanols: (+)-gallocatechin, (-)-epicatechin, (-)-epigallocatechin, (-)-epicatechin-3-O-gallate, (-)-epigallocatechin-3-O-gallate) and n-3 fatty acids. Cu, Fe and the n-6:n-3 fatty acid ratio were selected as pro-oxidants, as suggested by previous studies(Reference Gonzalez, Schemmel and Dugan26,Reference Li, Du and Zhou27) . Quercetin and (+)-catechin were considered pro-oxidants due to their dose-dependent pro-oxidant effects(Reference Procházková, Boušová and Wilhelmová28) and because of the previous findings of our group regarding an increased risk of PC in males with a high intake of these compounds(Reference Galván-Portillo, Vázquez-Salas and Hernández-Pérez24).
The n-6:n-3 ratio was calculated by dividing the daily intake of n-6 by n-3 fatty acids. The residual method proposed by Willett was used for the energy adjustment of the nutrient intake(Reference Willett and Stampfer29).
Antioxidants and pro-oxidants dietary consumption patterns
We identified three dietary consumption patterns of antioxidant and pro-oxidant nutrients using a K-means approach and considering each nutrient intake as energy residuals. This technique groups participants into mutually exclusive groups based on their estimated intake of evaluated nutrients. To determine the optimal number of clusters, we applied the Elbow method (online Supplementary Fig. 1(a)). Three clusters provided a balance between explanatory power and model simplicity, as indicated by the point of inflection in the curve. Additionally, the data visualisation according to the three selected clusters shows clear and well-separated groups (online Supplementary Fig. 1(b)). Thus, retaining three clusters ensures a balance between interpretability, statistical robustness and practical utility in defining consumption patterns. The three consumption patterns – (1) pro-oxidant (PRO), (2) moderate antioxidants intake/low pro-oxidants intake (MaLp) and (3) high antioxidants and pro-oxidants intake (HaHp) – were named based on the dietary intake observed among controls. For each pattern, the intake medians were estimated, along with the graphical distribution of the standardised means of intake for each nutrient and the proportion of adherence to recommended intake levels. The prevalence of adherence to the recommendations was estimated based on intake equal to or greater than the estimated average requirement for adult men: vitamin A (625 µg retinol equivalents), vitamin C (75 mg), vitamin E (12 mg), Zn (9·4 mg), Mn (2·3 mg), Se (0·045 mg), Cu (0·70 mg) and Fe (6 mg)(30,31) .
Physical activity, smoking and weight change trajectories
The history of physical activity(Reference Vázquez-Salas, Torres-Sánchez and Galván-Portillo32), smoking(Reference Jiménez-Mendoza, Vázquez-Salas and Barrientos-Gutierrez33) and weight changes(Reference Hernández-Pérez, Torres-Sánchez and Hernández-Alcaráz34) were previously modelled as life trajectories. Regarding physical activity, calculated in metabolic equivalents per h/week/year for four life stages (15–18, 19–29, > 30 years and the last 3 years before the interview), three patterns were identified: a low pattern characterised by an average of 27·5 metabolic equivalents per h/week/year; a moderate pattern with an average of 88·4 metabolic equivalents per h/week/year; and a high pattern, with an average of 186·8 metabolic equivalents per h/week/year. Smoking history included never smokers and two smoking patterns. The low pattern was characterised by a median consumption of 4 cigarettes/d, an age of onset of 17 years and an average smoking duration of 35 years. The high smoking pattern had a median consumption of 15·5 cigarettes/d, an age of onset of 15 years and an average duration of 48 years(Reference Jiménez-Mendoza, Vázquez-Salas and Barrientos-Gutierrez33).
The lifetime trajectories of weight changes were identified through self-reported body silhouettes at different ages (5, 10, 20, 30, 40 and 50 years)(Reference Hernández-Pérez, Torres-Sánchez and Hernández-Alcaráz34). The trajectory with the lowest weight change corresponded to individuals who experienced weight gain during adolescence, with no significant changes thereafter. The consistently high weight trajectory included individuals with consistently high weight throughout their lives, and the weight gain trajectory was characterised by a consistent increase in weight over their lifetime, reaching overweight status between the ages of 40 and 50 years.
Eight cases and eleven controls were excluded due to implausible energy intakes (< 3347·2 or > 18 828 kJ/d)(Reference Willett35). Additionally, one participant was excluded due to missing values in the weight change trajectories. The final sample consisted of 394 cases and 793 controls (Fig. 1).
Statistical analysis
The distribution of selected characteristics was compared between cases and controls using the Student’s t test for continuous variables or the χ 2 test for categorical variables. To ensure that the results are useful for recommendation purposes, we grouped some foods based on their antioxidant and pro-oxidant content (online Supplementary Table 1) or used the pre-existing categories from the FFQ. Using the Kruskal–Wallis test, we compared the median consumption (in grams) of these food groups and/or individual foods according to each consumption pattern. Subsequently, to estimate the contribution percentage of each food group or individual food, we used independent logistic regression models for the MaLp and HaHp patterns, using the PRO pattern as the reference category. The percentage contribution was calculated as the difference in pseudo R2 between the saturated model and the model excluding the food group or individual food.
To estimate the association between dietary consumption patterns of antioxidants and pro-oxidants with PC and the degree of histological differentiation, we utilised independent unconditional logistic regression models. These models were adjusted for potential confounders identified through a directed acyclic graph (online Supplementary Fig. 2): age, education, region of birth, first-degree family history of PC and trajectories of physical activity, smoking and weight changes over the lifespan.
Given that tobacco consumption is considered pro-oxidant, we evaluated a potential effect modification between smoking history and dietary consumption patterns. Due to sample size constraints, the final models were stratified only by ever-smokers (low and high smoking trajectories) and non-smokers. To assess the robustness of our findings, we conducted three sensitivity analyses targeting potential pre-diagnostic behavioural changes. First, to examine potential age-related variation in dietary associations, we carried out a stratified analysis by age at diagnosis (< 65 and ≥ 65 years). Second, we excluded participants who reported weight loss due to dietary changes 2 years before diagnosis or interview. Lastly, we excluded individuals who attributed their cancer to dietary factors to evaluate the influence of perceived causality and potential recall bias.
All statistical analyses were performed using Stata/MP 16.0. Statistical significance was established with a P-value < 0·05.
Results
Table 1 shows the distribution of selected characteristics among cases and controls. As a result of the matching, the average age of cases and controls was similar (68 (sd 8·4) v. 67 (sd 8·9), P = 0·17). Compared with controls, a lower proportion of cases were born in Mexico City (51·1 v. 65·9 %, P < 0·01) and had a higher level of education (20·6 v. 11·7 %, P < 0·01), greater prevalence of personal history of chronic diseases (62·7 v. 45·2 %, P < 0·01) and sexually transmitted diseases (26·7 v. 10·9 %, P < 0·01). Additionally, a higher proportion of cases than controls reported a family history of PC (10·4 v. 2·5 %, P < 0·01) and having experienced a weight gain trajectory (50·8 v. 44·4 %, P = 0·02) throughout their lives. The absence of physical activity (14·7 v. 9·3 %, P < 0·01) and energy intake was significantly higher among cases than among controls. At the time of diagnosis, most cases were classified as moderately differentiated (35·5 %) or poorly differentiated (38·6 %) (data not shown in table).
Distribution of selected characteristics between cases and controls (Mean values and standard deviations; numbers and percentages)
* P-value for the χ 2 test; † There are five missing observations. Mexico City: Mexico City; South: Campeche, Chiapas, Guerrero, Oaxaca, Quintana Roo and Yucatán. Central-western: Aguascalientes, Colima, Guanajuato, Jalisco and Michoacán. Central-Eastern: Hidalgo; Estado de México, Morelos, Puebla, Querétaro and Tlaxcala. North: Chihuahua, Coahuila, Durango, San Luis Potosí, Zacatecas, Baja California, Baja California Sur, Sinaloa, Sonora, Nayarit, Nuevo León and Tamaulipas. East: Veracruz y Tabasco; ‡Hypertension, diabetes or dyslipidaemia; § History of at least one sexually transmitted infection (STD) during lifetime; || Family history of prostate cancer in first-degree relatives; ¶None: men who have never smoked; Low: median consumption of 4 cigarettes/d, onset age 17, average smoking duration of 35 years; High: median consumption of 15·5 cigarettes/d, onset age 15, average smoking duration of 48 years; **Minor weight changes: subjects who experienced weight gain during adolescence but maintained normal weight throughout life; Consistently high weight: individuals with consistently elevated weight throughout life; Weight increase: subjects who experienced weight gain over their lifetime, reaching overweight between ages 40 and 50 years; ††None: men who do not engage in any physical activity throughout their life; Low: individuals with consistently low physical activity; Moderate: individuals with moderate physical activity throughout life; High: individuals with consistently high physical activity.
Table 2 and Fig. 2 show the characteristics of the three dietary consumption patterns. The HaHp pattern was characterised by significantly higher consumption of all the evaluated nutrients, except for vitamin A, ferulic acid and n-6, for which intake was only significantly greater than that observed in the MaLp pattern. Meanwhile, antioxidant consumption in the MaLp pattern was lower than that in the HaHp pattern and higher than in the PRO pattern. The intake of vitamin E, Zn, Mn, Se and n-3 was lower compared with the other two patterns. Additionally, the consumption of pro-oxidants (Fe and n-6) was significantly lower than in the other two patterns. Finally, the PRO pattern was characterised by the lowest antioxidant consumption and the highest n-6:n-3 ratio. Regarding the percentage of adherence to the estimated average requirement, the HaHp pattern showed the highest adherence, while the PRO pattern was characterised by lower adherence for most of the antioxidants (online Supplementary Fig. 3).
Antioxidant and pro-oxidant consumption patterns characterisation among controls (Median values and interquartile ranges)
IQR, interquartile range; PRO, pro-oxidant; MaLp, moderate in antioxidants/low in pro-oxidants; HaHp, high in antioxidants and pro-oxidants.
Kruskal–Wallis test: *Significantly different from PRO pattern (P < 0·05). †Significantly different from MaLp pattern (P < 0·05).
Average standardised intake of antioxidant and pro-oxidant patterns among controls. PRO, pro-oxidant; MaLp, moderate in antioxidants/low in pro-oxidants; HaHp, high in antioxidants and pro-oxidants.
The food groups or items contributing most to the HaHp and MaLp patterns were green vegetables, maize tortillas, seeds and fruits (online Supplementary Table 2), with some differences regarding the proportion of consumers and median consumption (Table 3). Compared with the other two patterns, the HaHp pattern had a higher percentage of consumers of seeds and green vegetables, as well as a lower proportion of consumers of sugary beverages and Mexican food. However, the median consumption of Mexican food in this pattern was higher than that reported in the MaLp pattern (Table 3). In the MaLp pattern, a greater proportion of consumers of antioxidant-rich fruits, liver, fish and seafood was observed. Yet, it is worth noting that the median consumption of liver and seafood was lower than that of the other patterns, while the consumption of antioxidant-rich fruits (7·10 v. 11·42 g/d) was significantly lower than that in the HaHp pattern. MaLp pattern also had the lowest proportion of consumers of saturated fats and junk food. Finally, the PRO pattern was characterised by a higher proportion of consumers of sugary beverages and junk food, with a significantly lower consumption of antioxidant-rich vegetables and fruits.
Food intake according to antioxidant and pro-oxidant patterns among consumer controls (Percentages; median values and interquartile ranges)
IQR, interquartile range; PRO, pro-oxidant; MaLp, moderate in antioxidants/low in pro-oxidants; HaHp, high in antioxidants and pro-oxidants.
Significant differences (P-value < 0·05) are shown in bold.
*χ 2 test; **Kruskal–Wallis test. †Significantly different from the pro-oxidant consumption pattern (P < 0·05). ‡ Significantly different from the moderate in antioxidants/low in pro-oxidants consumption pattern (P < 0·05).
Table 4 shows the association between the identified dietary patterns and PC. Compared with the PRO pattern, and independent of age, the MaLp (OR: 0·70; 95 % CI 0·53, 0·93) and HaHp (OR: 0·72; 95 % CI 0·52, 0·99) patterns were associated with lower odds of having PC. The magnitude of the association was similar for both patterns and remained significant even after adjusting for age, region of birth, education, family history of PC, physical activity trajectories, smoking and weight changes throughout life. MaLp patterns consumption was mainly associated (OR MaLp v. PRO: 0·50; 95 % CI 0·30, 0·84) with well-differentiated PC (Gleason ≤6). When stratifying by smoking history (online Supplementary Table 3), the MaLp pattern (OR: 0·66; 95 % CI 0·46, 0·95) and HaHp pattern (OR: 0·62; 95 % CI 0·40, 0·96) were associated with lower odds of PC only among ever-smokers. In the sensitivity analyses, results remained in the same direction, primarily among older participants (online Supplementary Table 5).
Association between dietary antioxidant and pro-oxidant consumption patterns and PC and PC histological differentiation in Mexican men (Odds ratios and 95 % confidence intervals)
PC, prostate cancer; PRO, pro-oxidant; MaLp, moderate in antioxidants/low in pro-oxidants; HaHp, high in antioxidants and pro-oxidants.
Significant associations (P-value < 0·05) are shown in bold.
Gleason ≤ 6: well-differentiated; Gleason = 7: moderately differentiated; Gleason ≥ 8: poorly differentiated.
*Model adjusted by age.
†Model adjusted by age, region of birth, education, family history of prostate cancer, physical activity trajectories, weight change trajectories and life-course smoking.
Discussion
In this study, we identified three consumption patterns: PRO, MaLp and HaHp. Comparing the MaLp and HaHp patterns, differences were observed not only in the quantity and type of antioxidants but also in pro-oxidant consumption, which was lower in the MaLp pattern. Both patterns were similarly associated with lower odds of having PC, compared with the PRO pattern, and no differences were observed based on the histological differentiation. The differences in the characterisation of these two patterns and the similarity in the magnitude of the association highlight the importance of the predominance of antioxidants over pro-oxidants as a protective factor against PC.
Our results are consistent with studies that have evaluated the overall antioxidant capacity of the diet(Reference Russnes, Wilson and Epstein14–Reference Vance, Wang and Su17), particularly with a prospective cohort study conducted in the USA that observed a decreased risk of PC among men with a diet high in antioxidant capacity(Reference Russnes, Wilson and Epstein14). However, their measurement approach does not consider the differences between antioxidant intake, let alone the antioxidant/pro-oxidant imbalance. Comparing our results with studies that have attempted to measure the antioxidant and pro-oxidant balance using oxidative balance scores(Reference Geybels, Verhage and van Schooten36–Reference Lakkur, Goodman and Bostick39) is not possible, as these scores only consider some dietary components and include lifestyle characteristics.
The mechanisms by which higher consumption of characteristic foods of the MaLp and HaHp patterns decreases the likelihood of PC may involve those related to oxidative balance. Oxidative stress, which occurs when there is an imbalance between the production of reactive oxygen species and the body’s ability to neutralise them or repair the damage they cause, has been identified as a key factor in prostate carcinogenesis. Oxidative balance depends on the availability of antioxidant molecules, primarily from diet, and of antioxidant enzymes that use metal ions (Cu, Zn and Mn) to carry out this function. However, when the consumption of these antioxidant elements is low, their redox capacity is diminished(Reference Fang, Yang and Wu25). Furthermore, when there is greater availability of pro-oxidants at the cellular level (n-6 fatty acids, Cu, Fe, quercetin and catechins), the balance between antioxidants and pro-oxidants can tilt towards pro-oxidation(Reference Galaris, Barbouti and Pantopoulos6,Reference Saleh, Adly and Abdelkhaliq7,Reference Procházková, Boušová and Wilhelmová28) , which is associated with cellular damage, mutations and alterations in DNA methylation that favour cancer development and progression(Reference Liu, Li and Hou9,Reference Fahmy, Alhakamy and Rizg40) . This last mechanism could explain the similar association observed between HaHp and MaLp patterns concerning PC.
Smoking is a well-known non-dietary pro-oxidant. It induces oxidative damage and decreases blood levels of antioxidants(Reference Astori, Garavaglia and Colombo41). In this context, probably baseline levels of oxidative stress and available concentration of antioxidants vary significantly among high and lower-intensity smokers. As far as we know, there is no evidence regarding the association between the combined dietary intake of antioxidants and pro-oxidants with PC risk, according to smoking status. Nevertheless, a significant reduction in PC risk was observed only among smokers with low dietary beta-carotene intake who received E vitamin supplementation(Reference Kirsh, Hayes and Mayne42). This finding is similar to ours regarding MaLp and HaHp patterns and PC, but only among ever-smokers. It is important to note that our sample size did not allow us to evaluate this association based on smoking trajectories. However, differences in prooxidative consumption likely vary across patterns, and this association may differ depending on the intensity and frequency of tobacco use.
Some studies have reported that there may be differences between in vivo and in vitro nutrient antioxidants or pro-oxidant activity because in vivo metabolic modifications, as well as the presence of other nutrients, can enhance or reduce the properties of removing or generating reactive oxygen species(Reference Fang, Yang and Wu25,Reference Procházková, Boušová and Wilhelmová28) . In addition, we consume foods rather than individual nutrients, and foods can be sources of both antioxidants and pro-oxidants. Thus, we consider one of the main strengths of this study to be the identification of consumption patterns of antioxidants and pro-oxidants, an approach that considers these interactions and reflects individuals’ diets more accurately. Additionally, unlike other approaches, this method could reflect regional consumption differences and variations in antioxidant and pro-oxidant content according to geographical location.
Furthermore, we believe that the likelihood of our results being a consequence of selection bias is low. Participation rates for both cases and controls exceeded 80 %, with no significant differences in sociodemographic characteristics between participants and non-participants (online Supplementary Table 4). Although evaluating the representativeness of cases is inherently complex due to the absence of a national cancer registry and the lack of a unified healthcare system, the distribution of cases identified in hospitals with (39 %) and without (61 %) social security aligns with previous data from Mexico City(Reference Puentes-Rosas, Sesma and Gómez-Dantés43). Furthermore, the proportion of cases with moderate and poorly differentiated PC is consistent with findings from a prior study conducted among men in Mexico City without social security(Reference Torres-Sánchez, Hernández-Pérez and Escamilla-Nuñez44). Additionally, the proportion of adherence to the recommended intake of vitamins A (22·3 v. 29·7 %), C (58 v. 62·7 %) and E (5·4 v. 9·4 %), as well as Fe (94·6 v. 99 %) among controls, is comparable to that observed in adult men in the 2012 National Health and Nutrition Survey(Reference Pedroza-Tobías, Hernández-Barrera and López-Olmedo21,Reference Sánchez-Pimienta, López-Olmedo and Rodríguez-Ramírez45) .
Some limitations must also be considered. This is a retrospective study that used an FFQ to measure the dietary intake of antioxidants and pro-oxidants. We do not reject the possibility of exposure misclassification, given that the questionnaire was not specifically designed to accurately assess nutrient intake or account for dietary changes over time. However, neither participants nor interviewers were aware of the study hypothesis, and the findings remained consistent across sensitivity analyses. The stronger protective association observed among older participants likely reflects greater stability in dietary habits within this group, reducing the potential for symptom-driven modifications prior to diagnosis. In addition, the exclusion of individuals who either attributed their PC to dietary factors or reported pre-diagnosis weight loss due to dietary changes helps mitigate concerns regarding recall bias and reverse causality. Taken together, these factors suggest that differential exposure misclassification is unlikely. Other limitations are related to the lack of information on the type of iron (haem or non-haem) and the use of vitamin supplements, although it is unlikely that these have impacted the characterisation of the patterns. The consumption of animal-source foods, the main source of haem iron, was similarly distributed among the identified patterns. Vitamin supplement use was reported by only 23·7 % of men, with B–complex vitamins being the most used (59·1 %). Regarding confounding control, we cannot rule out that our results may be affected by residual confounding. We did not adjust for alcohol consumption, despite its role as a pro-oxidant factor associated with diet and PC. Alcohol intake tends to decrease with age, and we considered that self-reported consumption for the 3 years preceding diagnosis or interview may not accurately reflect long-term intake patterns. Finally, the low frequency of family history of PC among controls did not allow us to evaluate a possible interaction between these antioxidant/pro-oxidant patterns and greater genetic load.
In conclusion, our findings suggest that dietary consumption patterns characterised by higher consumption of antioxidants, regardless of pro-oxidant intake, are associated with lower odds of having PC. These results underscore the importance of the balance between antioxidants and pro-oxidants in the diet as a relevant factor for the prevention of PC and for formulating public health recommendations. Overall, the characteristic consumption of the MaLp and HaHp patterns would be easily achieved and exceeded with the daily consumption recommendations of antioxidant-rich foods established in the recently published EAT-Lancet dietary guidelines(Reference Willett, Rockström and Loken46). Adaptations of these dietary guidelines can be made according to local antioxidant-rich foods. Nevertheless, these findings should be validated through prospective studies; in addition, potential linear or non-linear relationships as well as other interactions should be evaluated using alternative dietary-pattern approaches in a larger sample size.
Acknowledgements
We extend our sincere gratitude to all participants and the following hospitals for their invaluable collaboration in the development of this study: Hospital General de México, Instituto Nacional de Cancerología, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, Hospital de Oncología del Centro Médico Siglo XXI, Hospital General Regional No. 1 ‘Dr. Carlos McGregor Sánchez Navarro’ and Hospital Regional ‘Adolfo López Mateos’. Additionally, we express our appreciation to Elizabeth K. Galván-Miranda for her expertise in English language editing.
The main study was funded by Fondo Sectorial en Salud del Consejo Nacional de Ciencia y Tecnología de México (CONACYT) (funding numbers: 140482 and 272810). For this analysis, AMA received a scholarship from CONAHCYT (grant number: 1229488); nevertheless, this agency did not participate in the study design, statistical analysis or manuscript writing.
A.M-A.: Formal analysis, Data curation, Writing – original draft, Writing – review & editing. J.G.H-P.: Conceptualisation, Methodology, Formal analysis, Data curation, Writing – original draft, Writing – review & editing. M.G-P.: Methodology, Data curation, Writing – original draft, Writing – review & editing. F.R-C. and S.R-R.: Writing – review & editing. L.T-S.: Project administration, Methodology, Conceptualisation, Writing – original draft, Writing – review & editing.
The authors declare that they have no conflict of interest.
This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of the National Institute of Public Health (CI: 980) and by the participating hospitals.
Informed consent was obtained from all participants included in the study.
The datasets used and/or analysed during the present study are available from the corresponding author upon reasonable request.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114525104984.
