Introduction

Consumption of fruits and vegetables is an important part of a healthy diet and has been linked to a reduced risk of cancer [1]. A report on diet and bladder cancer (BC) from the WCRF Continuous Update Project concluded that high intakes of fruit and vegetables could decrease the BC risk [2]. In addition, results of a large systematic review suggested that the consumption of citrus fruits and cruciferous vegetables decreases BC risk [3], and another review suggested that a diet rich in vegetables and fruits might be protective against BC [4].

Previous studies have often lacked adequate statistical power to detect associations between dietary factors and BC risk. Aside from total fruit and total vegetable consumption, associations between subgroups of fruits and vegetables and BC risk have been reported sporadically and may be subject to publication bias. As fruits and vegetables are heterogeneous with respect to phytochemical content [5], associations with BC risk may differ between types of fruits and vegetables. The use of individual data from multiple case–control studies could substantially increase statistical power, and covariates can be standardized across studies.

The aim of this large-scale pooled study was to investigate the association between total fruit and total vegetable consumption and specific subgroups of fruits and vegetables and BC risk using data of 5,648 BC cases and 10,517 controls from 11 studies which were included in the Bladder cancer Epidemiology and Nutritional Determinants (BLEND) consortium.

Methods

Study population

Data were analyzed from the BLEND study. BLEND is a large international nutritional consortium on BC, which includes nineteen case–control studies originating from countries all over the world [6]. Eleven of the nineteen case–control studies had sufficient information (i.e., data on usual dietary intake, method of dietary assessment, geographical region, ethnicity, gender, smoking status, disease status, age at BC diagnosis and/or age at enrollment of the study) to be eligible for inclusion in our study on the influence of total fruit and total vegetable consumption and specific subgroups of fruits and vegetables on BC risk. Studies originated from the USA, Belgium, Sweden, Italy, Canada and China. Each participating study has been approved by a local ethics committee. Informed consent was obtained from all individual participants included in each study.

Data collection

Details on the methodology of the BLEND consortium have been described elsewhere [6]. Briefly, BC cases were ascertained by using medical record review or linkage with a cancer registry. Both non-muscle invasive BC (NMIBC) and muscle invasive BC (MIBC) were considered as outcomes.

For each study, participants were asked to report the frequency of fruit and vegetable items consumed during the preceding one or two years before study enrollment. All studies made use of a validated self-administered food frequency questionnaire (FFQ) or a questionnaire administered by a trained interviewer regarding the frequency of fruit and vegetable consumption. Summary details of the FFQs used in the included studies can be found in Appendix 1, Table 5. Total fruit consumption was computed as the sum of all fruit items in each study, and total vegetable consumption as the sum of all vegetable items provided. The following subgroups of fruits were defined: citrus fruits, pome fruits, soft fruits, stone fruits, tropical fruits, fruit mixtures, and fruit products (Table 1). For vegetables, the defined subgroups were: leaf vegetables, brassica, stalk vegetables, shoot vegetables, tubers, onion-family vegetables, root vegetables, fruit vegetables, pod seeds, fungi, seaweeds, vegetables mixtures, and vegetables products (Table 1). No analyses were performed for fruit mixtures, fruit products, onion-family vegetables, root vegetables, fruit vegetables, pod seeds, fungi, seaweeds, vegetable mixtures, or vegetable products because of the limited number of participants consuming fruits or vegetables from these subgroups.

Table 1 Characteristics of vegetable subgroups investigated in the pooled case–control analyses
Full size table

Statistical analysis

Participants who reported a history of cancer other than nonmelanoma skin cancer prior to study entry, or had missing data on age at study entry, gender, smoking status, pack years, or fruit and vegetable consumption, were excluded from the analyses. In the analyses, fruit and vegetable consumption were categorized into three consumption levels (low intakes/moderate intakes/high intakes), corresponding to study-specific marginal tertiles. The investigated fruit subgroups (citrus fruits, pome fruits, soft fruits, stone fruits, and tropical fruits) and vegetable subgroups (leaf vegetables, brassica, stalk vegetables, shoot vegetables, and tubers) were also modeled as study-specific marginal tertiles. Although the tertile approach does not take into account true differences in the distribution of population intakes across studies, reported intakes may differ across studies based on country-specific portions sizes and differences in FFQs used, and is therefore our preferred approach [6].

The analytic approach was a two-stage process. First, study-specific odds ratios (ORs) were calculated using unconditional logistic regression models (with low consumption levels as the reference group). The majority of the included case–control studies matched on age and gender only. Although matched methods (e.g., from conditional logistic regression) are robust to matching distortion, unmatched methods like unconditional logistic regression appear to be viable options for loose-matching data, e.g., data matched on a small number of demographic variables like age and gender. After that, the estimates were pooled using a random effects meta-analysis approach to calculate an overall estimate and 95% confidence interval (CI). Adjustments were made for the following potential confounders: gender (male, female), age at study entry (< 45 years, 45–49 years, 50–54 years, 55–59 years, 60–64 years, 65–69 years, 70–74 years, > 75 years), smoking status (never smoker/former smoker/current smoker), and pack years (< 9 years, 9–17 years, 18–30 years, 31–46 years, > 47 years). Interactions between age and total fruit and vegetable consumption, gender and total fruit and vegetable consumption, and smoking status and total fruit and vegetable consumption were tested and showed no significant interactions. Nonetheless, besides the overall analysis, subgroup analyses were performed on the two main study-specific matching factors gender and age, as recommended by the study of Smith-Warner et al. [7]. Small study effects were analyzed by funnel plots. A sensitivity analysis was performed, leaving out the only study conducted in Asia. All statistical analyses were performed using Stata software version 14. A two-sided p value < 0.05 was considered statistically significant.

Results

Of the nineteen case–control studies, six studies were excluded for providing no data on fruit or vegetable consumption, and two studies were excluded for providing no data on portion sizes. The eleven included case–control studies [8,9,10,11,12,13,14,15,16,17,18] originated from Belgium, Italy, Sweden, China, Canada, and the USA. In total, out of 17,012 eligible participants, 871 were excluded for having missing data on fruit or vegetable intakes (n = 595), age at study entry (n = 8), or pack years (n = 268) (Fig. 1).

Fig. 1
figure 1

Flow diagram of exclusion criteria for participants included in the pooled case–control analyses on fruit and vegetable intakes

Full size image

Baseline characteristics of the included studies are presented in Table 2. A total of 5637 BC cases and 10,504 controls were analyzed. Most of these included participants were male (68%) and Caucasian (89%). The mean age for BC was 60 years across all studies. Even though all studies made use of an FFQ to measure usual fruit and vegetable consumption (over the preceding one or two years before study enrollment), the number of fruit and vegetable items described in each questionnaire varied widely.

Table 2 Characteristics of the 11 eligible case–control studies according to disease status, gender, age, and smoking status
Full size table

Total fruits

Among the eleven included studies, nine demonstrated a trend towards reduced BC risk with higher intake of total fruit, although the association was not always statistically significant (Table 3, Fig. 2). Remarkably, one study reported a non-significant increased risk (Table 3). However, upon pooled analysis, a reduction in BC risk associated with higher total fruit intake was observed. In the overall pooled analysis, greater intakes of fruit showed a reduced BC risk compared to the lowest intakes of fruit (OR 0.79; 95% CI 0.68–0.91) (Table 3). The tests for heterogeneity showed moderate heterogeneity for the association between total fruit and BC (I2 = 42.8%).

Table 3 Odds ratios of bladder cancer by total fruit intake for all participants
Full size table
Fig. 2
figure 2

Funnel plot of overall fruit analyses results

Full size image

In the subgroup analysis on gender, similar results were observed in both men and women, showing a pooled decreased BC risk among those consuming the highest intakes of fruit (ORmen = 0.83; 95% CI 0.71–0.95) (Table 4), and ORwomen = 0.60; 95% CI 0.40–0.80 (Table 5)). Whilst the heterogeneity among men for the might not be considerable (I2 = 25.7%), the heterogeneity among women may represent moderate heterogeneity (I2 = 43.3%) (Appendix 2, Tables 6, 7).

Table 4 Odds ratios of bladder cancer by total vegetable intake for all participants
Full size table

Similarly, when analyzing the data according to age (i.e., 60 years and under), consistent results were observed. In the overall pooled analysis, participants younger than 60 years of age with high fruit consumption exhibited a significant decrease in BC risk (OR 0.76; 95% CI 0.64–0.88), with no considerable statistical heterogeneity observed in this group (I2 = 0.0%) (Appendix 3, Table 10). Likewise, a decreased BC risk was observed in participants 60 years or older consuming the highest intakes of fruit (OR 0.80; 95% CI 0.69–0.91), with negligible heterogeneity for this association (I2 = 5.5%) (Appendix 3, Table 11). Additionally, the sensitivity analysis, excluding the Chinese study [12], yielded similar results compared to the overall analysis (OR 0.83; 95% CI 0.72–0.93).

Citrus fruits, pome fruits, soft fruits, stone fruits and tropical fruits

No associations between high consumption of soft fruits or stone fruits and BC risk were found (Table 3). Notably, a trend indicating a lower BC risk was observed for soft fruits in four out of the five studies (Table 3), with the exception of the Stockholm case–control study, where an increased BC risk was observed. For citrus fruits, an overall decreased BC risk was noted with high consumption (OR 0.81; 95% CI 0.65–0.98). Interestingly, a similar trend of lower BC risk for citrus fruits was evident in eight out of the ten studies (Table 3), except for the Belgian case–control study on BC and the Stockholm case–control study, which reported an increased BC risk. However, the overall decreased BC risk results for citrus fruit revealed considerable heterogeneity (I2 = 66.2). Similar patterns were observed in the stratified analysis by gender and age (i.e., > 60 years and under) (Appendix 2, Tables 6, 7, Appendix 3, Tables 10, 11). Among women participants and participants aged 60 years or more, an overall decreased BC risk was observed with high consumption of citrus fruits (OR 0.56; 95% CI 0.32–0.80, and OR 0.80; 95% CI 0.62–0.99, respectively). However, both analyses showed considerable heterogeneity (I2 = 62.5%, and I2 = 53.2%, respectively). Conversely, among participants younger than 60 years, an association between the highest intakes of citrus fruits and BC risk was observed (OR 0.77; 95% CI 0.64–0.91) with no considerable statistical heterogeneity (I2 = 0.0%). In men, no association was found between greater intakes of citrus fruits and BC risk (Appendix 2, Table 6).

Overall, greater consumption of pome fruits was associated with a decreased BC risk (OR 0.76; 95% CI 0.65–0.87) and statistical heterogeneity was not considerable (I2 = 0.0%). In the subgroup analyses, pome fruit consumption (highest versus lowest intakes) was associated with a lower BC risk in men (OR 0.77, 95% CI 0.58–0.97), women (OR 0.58, 95% CI 0.42–0.73), < 60 years (OR 0.52, 95% CI 0.31–0.72), and ≥ 60 years (OR 0.85, 95% CI 0.71–0.99), with low to moderate heterogeneity (I2 = 40.2%, I2 = 0.0%, I2 = 15.3%, and I2 = 0.0, respectively) (Appendix 2, Tables 6, 7, Appendix 3, Tables 10, 11).

High consumption of tropical fruits was associated with a decreased BC risk in the overall analysis (OR 0.84; 95% CI 0.73–0.94, I2 = 20.5%). Again, it should be noted that a trend towards a decreased risk was shown in seven out of the nine studies, while the New Hampshire and Molecular epidemiology of BC studies showed a trend towards an increased BC risk. In the subgroup analyses on age, both age groups (< 60 and ≥ 60 years) showed associations between the highest intakes of tropical fruits and BC risk (OR 0.84; 95% CI 0.70–0.98 and OR 0.83; 95% CI 0.72–0.95, respectively) and no heterogeneity was observed for these associations (Appendix 3, Tables 10 and 11).

Total vegetables

Although high total vegetable consumption was associated with BC risk (OR 0.82; 95% CI 0.70–0.94), heterogeneity was observed (I2 = 52.9%) (Table 4, Fig. 3). In men, the highest intakes of total vegetables were associated with a decreased BC risk with no considerable heterogeneity (OR 0.80; 95% CI 0.71–0.88, I2 = 1.0%) (Appendix 2, Table 8). Similar results were found for participants ≥ 60 years (OR 0.81; 95% CI 0.71–0.91, I2 = 0.0%) (Appendix 3, Table 13). Greater intakes of total vegetables among participants < 60 years were significantly associated with a decreased BC risk (OR 0.70; 95% CI 0.52–0.88). However, substantial heterogeneity was observed for this association (I2 = 47.8%) (Appendix 2, Table 3). The sensitivity analysis did not change the result (OR 0.82; 95% CI 0.69–0.95).

Fig. 3
figure 3

Funnel plot of overall vegetable analyses results

Full size image

Leaf vegetables, brassica, stalk vegetables, shoot vegetables, and tubers

No associations between high consumptions of brassica, stalk vegetables, and tubers, and BC risk were observed (Table 4). Although subject to substantial heterogeneity, associations were found for the overall high intake of leaf vegetables and decreased BC risk (OR 0.82; 95% CI 0.68–0.96). This trend was observed for nine out of the eleven studies, while the South and East China ca–co study and NESCC study showed a trend towards an increased BC risk (Table 4).

High intakes of shoot vegetables significantly decreased the BC risk with no observed substantial heterogeneity (OR 0.87; 95% 0.78–0.96, I2 = 0.0%) (Table 4). In the subgroup analyses, significant associations for leaf vegetables (OR 0.70; 95% CI 0.56–0.84) and shoot vegetables (OR 0.83; 95% CI 0.68–0.97) were found in participants < 60 years with no substantial heterogeneity (I2 = 20.0%, and I2 = 0.0%, respectively) (Appendix 3, Table 12). Although associations were found for leaf vegetables and BC risk in men, women, and participants ≥ 60 years, substantial heterogeneity was observed for these associations (Appendix 2, Tables 8, 9, Appendix 3, Table 13).

Discussion

In this large pooled analysis of eleven case–control studies, comprising 5637 cases and 10,504 controls, significant inverse associations were found between high fruit and vegetable consumption and BC risk. There were inverse associations between total fruit, citrus fruit, pome fruit, and tropical fruit consumption and BC risk. No associations were found for high consumption of soft fruits or stone fruits. For vegetable consumption, inverse associations were found between total vegetable, leaf vegetable, and shoot vegetable consumption and a BC risk. Brassica, stalk vegetables, and tubers were not associated with BC risk.

It seems plausible that substances in fruits and vegetables including minerals, phytochemicals, and antioxidant nutrients, have potentially anticarcinogenic properties to protect against the development of cancer [1, 19]. Throughout Europe and the USA, apples and pears are the most consumed pome fruits. Apples contain a wide range of phytochemicals which have been found to have strong antioxidant properties and the ability to inhibit cancer cell proliferation [20]. Hence, results from an Italian case–control study and from the EPIC study, which included data from ten European countries, confirm that greater consumption of apples and pears decreases the BC risk (OR 0.63; 95% CI 0.39–0.99, and OR 0.90; 95% CI 0.82–0.98, respectively) [21, 22]. In addition, scientific evidence suggests that vitamin C, an essential nutrient abundant in citrus fruits, exerts anticancer effects in the bladder through diverse pathways. These include a malignancy-inhibiting shift in the transcriptome, as well as elevating levels of 5-hydroxymethylcytosine [23].

Brassica vegetables also contain high levels of phytochemicals, such as glucosinolates and isothiocyanates, and are therefore expected to lower the BC risk [19, 24]. This, however, was not confirmed in this study. Leafy vegetables contain high concentrations of carotenoids that could potentially protect against the damage to DNA by scouring free radicals [25]. Hence, results of a meta-analysis indicated that per 0.2 serving increment of daily green leafy vegetable consumption, the BC risk decreases with 2% [26]. We observed evidence of an inverse association between consumption of leafy vegetables and BC risk, which is consistent with these potential biological mechanisms.

Although no previous studies investigated specifically the role of shoot vegetables or tropical fruits in BC risk, these vegetables and fruits contain a wide range of carotenoids, folates, vitamins, and carotene, which may offer protection against the development of BC [27].

In line with our research a meta-analysis conducted in East Asians showed an inverse association between total fruit intake and BC risk [23]. Furthermore, within our BLEND study investigating fruit intake and BC risk among thirteen cohort studies, an inverse association between fruit intake and BC risk was found in women, but not in men. Nor was an association found for any fruit subgroup [28]. These null associations were confirmed in a Japanese cohort study including 1,287,514 person-years of follow-up. In addition, a null association was also found for total vegetable intake and any subgroup vegetable [29]. These observed differences might be due to the difference in study design. Although both case–control and cohort studies are subject to (a different form of) selection bias, and other methodological limitations, such as measurement error, there is an important methodological difference between these different study types. While in case–control studies, the assessment of lifestyle occurs after diagnosis, in cohort studies, the assessment of lifestyle occurs prior to diagnosis. It is therefore thought that case–control studies are more prone to recall bias and provide a lower certainty of evidence than cohort studies [30, 31]. However, since some of our results have low heterogeneity, the role of bias is likely to have minimal influence on these results [32]. In addition, recall bias has been addressed and analyzed for its consequences in many epidemiological/methodological papers, and no clear answer on the magnitude of the effect of this specific type of bias could be drawn.

Several of our results revealed discrepancies among individual study findings and indicated significant heterogeneity in the effects of fruit and vegetable consumption on BC risk. This aligns with previous research highlighting the challenges of comparing diet-disease relationships globally due to variations in dietary habits and assessment methods across populations. For instance, differences in portion size estimations, nutrient databases used, and other factors contribute to result heterogeneity, complicating generalization and interpretation. Therefore, while our analysis focused on individual food items rather than overall dietary patterns, it is essential to consider the broader context of dietary diversity and assessment methods when interpreting these findings.

Strengths and limitations

This study has some limitations. First, the number of fruit and vegetable items described in each FFQ varied widely across the studies. Although it has been reported that fruit and vegetable servings increase with the number of fruit and vegetable items on a questionnaire [33], the total of fruit and vegetable questions on the FFQs did not significantly modify the association between fruit or vegetable consumption and BC risk. In addition, the number of studies included in the fruit and vegetable subgroup analyses varied depending on whether the items comprising a particular fruit or vegetable subgroup were asked on the study-specific FFQs. Consequently, the power to examine associations for some subgroups is more limited compared with that for analyses of total fruit and total vegetables. Second, while case–control studies are valuable in investigating associations, they inherently possess limitations, particularly when assessing dietary factors which may be influenced by participants' cancer status [34]. Furthermore, as mentioned previously, case–control studies are generally considered to have higher risk of recall bias than cohort studies as the selection of the controls may be subject to population stratification [30]. We acknowledge that cohort studies and randomized controlled trials generally provide stronger evidence. Despite this, our analysis accounts for potential biases by addressing low heterogeneity across studies, suggesting minimal impact on our findings [32]. At last, while limited data on other potential BC risk factors such as body mass index and socioeconomic status were available, current literature suggests their contribution to BC risk is relatively small [35,36,37,38]. Besides, this pooled analysis also has several strengths, including the large sample size with harmonized variables across multiple studies, providing high statistical power to examine the role of total fruits and vegetables, subgroups of fruits and vegetables, and the possibility to perform subgroup analyses on gender and age. Moreover, we found low heterogeneity between studies for many of the association reported, the role of bias is likely to have minimal influence on our results.

Conclusion

This comprehensive study provides compelling evidence that the consumption of fruits overall, citrus fruits, pome fruits and tropical fruits reduce the BC risk. Besides, evidence was found for an inverse association between total vegetables and shoot vegetables intake.