Introduction

Bladder cancer is the sixth leading cancer in the USA, with an estimated 81,190 new cases and 17,240 deaths in 2018. Over 75% of all patients are still alive after 5 years [1]. Moreover, bladder cancer has high recurrence and is the most expensive malignancy to treat, accounting for > 3% of all cancer-related medical payments in the USA [2]. At present the better established risk factors associated with developing bladder cancer include smoking, age, male sex, occupation, and to a lesser extent obesity and physical inactivity [3,4,5]. Since most of the metabolites of ingested food come into direct contact with the bladder mucosa, diet might also play a role in the development of bladder cancer [6].

Previous studies of diet-related bladder cancer risk factors have tended to focus on single food items [7, 8]. For example, the Multiethnic Cohort (MEC) study, which included a total of 185,885 participants and 1137 incident bladder cancer cases, reported a hazard ratio (HR) of 0.40 (95% CI 0.23–0.69) comparing highest and lowest quartiles of vegetable intake [9]. Also, the Los Angeles Bladder Cancer (case–control) Study involving 3246 participants, including 1660 cases, reported a positive association between intake of red meat (salami, pastrami and beef) and bladder cancer risk (comparing highest and lowest quintile: OR 1.33, 95% CI 1.02–1.74) [10]. Emerging evidence suggests that total dietary patterns may provide stronger evidence for diet–disease associations than individual dietary items [11].

The Mediterranean diet has been reported to be effective for preventing non-communicable diseases [12,13,14,15] and reducing overall mortality and the incidence of several cancers [16, 17]. It is generally characterized by a high consumption of fruits, vegetables, legumes and cereals, moderate-to-high consumption of fish, moderate consumption of alcohol (mostly wine), low-to-moderate consumption of milk and dairy products, and low consumption of meat and meat products [18]. The diet distinguishes itself from other dietary recommendations and indices such as the Healthy Eating Index [19], the World Cancer Research Fund and American Institute for Cancer Research (WCRF/AICR) diet recommendations [20] and the Diet Inflammatory Index [21], by its higher levels of dietary fat, mainly monounsaturated fat from olive oil, and higher alcohol consumption, mainly from wine, although alcohol is a risk factor for several cancers [22,23,24,25,26].

To date, few studies [27, 28] have investigated the association between Mediterranean diet and bladder cancer. The European Prospective Investigation into Cancer and Nutrition (EPIC) cohort study, including 477,312 participants (of which 1425 were incident cases), found an inverse but non-significant association comparing a high with a low Mediterranean diet score (MDS) and urothelial cell carcinoma (UCC) overall (HR 0.84 [95% CI 0.69, 1.03]), and for risk of aggressive (HR 0.88 [95% CI 0.61, 1.28]) and non-aggressive disease (HR 0.78 [95% CI 0.54, 1.14]). The association was statistically significant for current smokers (HR 0.66 [95% CI 0.47, 0.93]) [27]. Researchers from the Melbourne Collaborative Cohort Study (MCCS), which included 37,442 participants at time of recruitment (379 incident cases), reported an inverse association for both sexes between the MDS and invasive UCC (HR 0.86 [95% CI 0.74, 1.00]) [28].

Our primary aim was to build on the results of the EPIC cohort study and the MCCS, and to investigate prospectively the potential association between Mediterranean diet and the risk of developing bladder cancer, by aggregating data from 13 cohort studies in a pooled analysis using a meta-analysis approach. Our secondary aims were to examine heterogeneity in any association by sex and disease sub-type (non-muscle-invasive and muscle-invasive bladder cancer).

Materials and methods

Study population

Data were analysed from the Bladder cancer Epidemiology and Nutritional Determinants (BLEND) study. BLEND is a large international nutritional consortium, which included 16 cohort studies conducted in several countries. Thirteen of the 16 cohort studies had sufficient information on food items to be eligible for inclusion in our study on adherence to the Mediterranean diet and the risk of developing bladder cancer. Studies originated from centres in Denmark [29], France [30], Germany [31], Greece [32], Italy [32, 33], The Netherlands [34], Norway [35], Spain [32], Sweden [36,37,38] United Kingdom [39, 40], the USA [41], and Australia [42, 43].

Data collection and coding

Details on the methodology of the BLEND consortium have been described elsewhere [44]. Briefly, the primary data from all included studies were incorporated into one dataset. All data provided were checked and converted from daily, monthly, or yearly food intake to weekly intake, and intakes by portion were also converted to intake by grams. Data on bladder cancer diagnosis were mainly ascertained by self-reported questionnaires. Dietary data, collected using food frequency questionnaires in all studies, were recoded using the Eurocode 2 food coding system [45]. In addition to information on dietary intake, the BLEND data also included study characteristics (design, method of dietary assessment, recall time of dietary intake and geographical region), participant demographics (age, sex, and ethnicity), bladder cancer pathology (non-muscle-invasive and muscle-invasive disease), and smoking status (current/former/never) all measured at baseline.

Mediterranean diet score

To measure the degree of adherence to the Mediterranean diet, we used a nine-point scale that was constructed by Trichopoulou et al. [46]. Nine food items were included, namely, consumption of (1) cereals, (2) fruits and nuts, (3) vegetables, (4) legumes, (5) fish, (6) meat, (7) dairy products, (8) fats, and (9) alcohol/ethanol. For each component, a value of 0 or 1 was assigned using its sex-specific median for each study as a cutoff value. For the presumed beneficial components (vegetables, legumes, fruits and nuts, cereals, and fish), a value of 0 was assigned to those consuming less than the median cutoff, and a value of 1 was assigned to those consuming as much as the median cutoff or more. For the presumed detrimental components (meat and dairy products), a value of 1 was assigned to those consuming less than the median cutoff, and a value of 0 was assigned to those consuming as much as the median cutoff or more. For alcohol, a value of 1 was assigned to men consuming between 70 and 350 g per week and to women consuming between 35 and 175 g per week. We assumed that one portion of alcohol of any type contained a standard amount of 10 g of ethanol. For fat intake, we calculated the ratio of fats from plant sources to total fat and assigned a value of 0 to those consuming less than the median cutoff, and a value of 1 to those consuming as much as the median cutoff or more. We used the ratio of plant-to-total fat because we hypothesized that the effect of dietary fat may depend on its source and not solely on the quantity consumed. For example, monounsaturated fat is present in both olive oil and animal products, and by just summing up the total amount of monounsaturated fat consumed it may not take into account the potentially different biological responses related to dietary source.

The MDS ranged from 0 (minimal adherence) to 9 (maximal adherence). Scores between 0 and 3 were classified as “low adherence”, scores of 4 and 5 were classified as “medium adherence”, and scores of 6 or higher were classified as “high adherence”.

Statistical analysis

Cox proportional hazard models using age at recruitment as the starting point on the time scale were used to calculate HRs and 95% confidence intervals (95% CI) for developing bladder cancer, comparing medium and high adherence with low adherence. The MDS was also analysed as a continuous variable (0–9). The proportional hazards assumption was examined through Schoenfeld residuals [47]. When considering all included participants, the assumption of proportional hazards was violated and therefore we compared the association between MDS score and risk of bladder cancer in all subjects younger than 70 years and in those older than 70 years to assess to what degree the HR changed over time. The Cox regression models were all adjusted for total energy intake in kilocalories (by applying a restricted cubic spline), sex and smoking status (never, former or current smoker). Furthermore, survival time was estimated by subtracting age at exit by age at entry in the cohort as T0, thereby correcting for age in the analysis and also the study sample from which the cases originated was corrected for by introducing study ID as a random effect. Analyses were stratified on sex and disease sub-type (non-muscle-invasive or muscle-invasive disease). To test for residual confounding by smoking, the association between MDS score and risk of bladder cancer was also investigated while stratifying for smoking status (ever/never).

Additionally, unstratified analyses were repeated to determine the effect of both alcohol and fats as two distinctive features of Mediterranean diet, with alterations to the estimation of the MDS in an exploratory analysis. To test the effect of alcohol on the MDS, we excluded the alcohol component from the diet score. For fats, we repeated the analysis by excluding fats from the diet score altogether and by replacing the lipid ratio (fats from plant sources divided by total fats) with only olive oil intake. All statistical analyses were performed using Stata/SE 14.2 [48].

Results

Dietary data from 646,222 study participants, including 3639 incident cases and 642,583 non-cases were analysed. Disease sub-type was known for 2425 cases, of which 945 (39%) were muscle-invasive bladder cancer (MIBC) and 1480 (61%) were non-muscle-invasive bladder cancer (NMIBC). Compared with non-cases, bladder cancer cases were more likely to be male (74%) and to be current or former smokers (79%). Of all cases, 22% originated from Scandinavian countries, 12% from Mediterranean regions, and 42% from other countries in Western Europe. The remaining 24% of the cases were living in the USA (10%) or Australia (14%); the Australian study (MCCS) oversampled people born in Greece or Italy [42, 43] (Table 1).

Table 1 Characteristics of the 13 eligible studies according to subject status, sex, age, TNM stage, and smoking status

The overall HR estimates for bladder cancer associated with MDS, after adjustment for total energy intake, smoking status, and sex, are presented in Table 2. A total of 6,577,179 person years, including 3581 cases, were analysed. Overall, high adherence to the Mediterranean diet was associated with a decrease in bladder cancer risk compared with low adherence (HRhigh 0.85 [95% CI 0.77, 0.93]). A decreased bladder cancer risk was also found for medium compared with low adherence to the Mediterranean diet (HRmedium 0.91 [95% CI 0.85, 0.99]). In addition, an inverse linear association was found between a one-unit increase in adherence to the Mediterranean diet and risk of developing bladder cancer (HRcontinuous 0.96 [95% CI 0.94, 0.98]). Although the proportional hazards assumption was violated, the results were similar when considering only those younger than 70 years at entry in the study (HRhigh 0.80, [95% CI 0.72, 0.89], HRmedium 0.90, [95% CI 0.83, 0.98]) separately from those older than 70 years at entry in the study (HRhigh 0.86, [95% CI 0.57, 1.29], HRmedium 0.82, [95% CI 0.60, 1.14]), indicating that the presented HRs in Tables 2 and 3 were probably not heavily influenced by this violation. Furthermore, residual confounding by smoking seemed minimal as the results in never smokers (HRhigh 0.84, [95% CI 0.68, 1.04], HRmedium 0.84, [95% CI 0.71, 0.99]) were similar to those in ever smokers (HRhigh 0.80, [95% CI 0.71, 0.89], HRmedium 0.90, [95% CI 0.83, 0.98]).

Table 2 Pooled HR and 95% CI for the association between adherence to the Mediterranean diet and risk of developing bladder cancer for all bladder cancer, by sex, and by disease sub-type
Table 3 Pooled HR and 95% CI of the analyses exploring the effects of alcohol and fats on the MDS score

Results remained consistently below 1.00 for non-muscle-invasive (HRhigh 0.86 [95% CI 0.74, 0.99]) and muscle-invasive (HRhigh 0.89 [95% CI 0.74, 1.07]) patients after stratification on disease sub-type (Table 2).

Results for men (HRhigh 0.86 [95% CI 0.77–0.96], HRmedium 0.89 [95% CI 0.82, 0.97]) and women (HRhigh 0.90 [95% CI 0.74–1.10], HRmedium 0.84 [95% CI 0.73, 0.98]) were comparable and in line with the overall estimates. Although total person-time was higher for women, the total number of cases was much higher for men (Table 2).When stratified on both disease sub-type and sex, HRs were consistently below 1.00, except for high compared with low adherence to the Mediterranean diet and risk of muscle-invasive disease for women (HRhigh 1.05 [95% CI 0.68, 1.60]) (Table 2).

In the exploratory analysis, we obtained similar results after excluding either fats (HRhigh 0.88 [95% CI 0.78, 0.99], HRmedium 0.92 [95% CI 0.85, 0.99]) or alcohol (HRhigh 0.93 [95% CI 0.83, 1.04], HRmedium 0.93 [95% CI 0.86, 1.00]) from the diet score. Also, consistent results were found in the relation between adherence to the Mediterranean diet and bladder cancer risk when we replaced the lipid ratio (fats from plant sources divided by total fats) with olive oil intake only (HRhigh 0.82 [95% CI 0.74, 0.90], HRmedium 0.91 [95% CI 0.84, 0.98]).

Discussion

Main findings

We investigated the association between adherence to the Mediterranean diet and bladder cancer risk and observed an overall inverse association between a high adherence to the Mediterranean diet and the risk of developing bladder cancer. Analyses stratified by sex and disease sub-type showed similar results, indicating that the association is unlikely to be confounded by factors that might differ between these subgroups.

Previously published results from studies that have investigated the association between adherence to the Mediterranean diet and the risk of developing bladder cancer are in line with our findings. Although not statistically significant, Buckland et al. [27] reported inverse associations between adherence to the Mediterranean diet and risk of overall, aggressive or non-aggressive bladder cancer for men and women. In contrast to our association between Mediterranean dietary adherence and non-muscle-invasive disease, Dugué et al. [28], based on the MCCS, found a weak inverse association between adherence to the Mediterranean diet and urothelial cell carcinoma only. It is worth mentioning that these two studies [27, 28] used different dietary fat assessment measures for the Mediterranean diet. Buckland et al. also used a different grading score for determining dietary adherence.

Despite the limited evidence for a role of the Mediterranean diet in the development of bladder cancer overall, several studies have focused on some key elements of this dietary pattern and found some beneficial effects. For example, it has been shown that the consumption of vegetables and fruits is inversely associated with the risk of bladder cancer [9, 49]. This finding is not unexpected, since both vegetables and fruits contain large quantities of polyphenols, carotenoids, and vitamins C and E, which have antioxidant functions, allowing them to prevent DNA damage by neutralizing reactive oxygen species [50, 51]. Conversely, a positive association with the risk of developing bladder cancer has been reported for high consumption of animal products, such as red and processed meats and animal proteins [52,53,54]. During high-temperature cooking of meat, specific substances which are known to be involved in bladder cancer carcinogenesis are formed [55]. In addition, red meat is rich in iron, which is associated with increased formation of N-nitroso compounds (NOCs). These compounds have been suggested to induce tumours in the bladder [56].

While reportedly lower in saturated and animal fats, the Mediterranean diet is associated with a higher intake of dietary fat (approximately 35% of total energy intake) usually from monounsaturated dietary fat. Another important element of the Mediterranean diet that has been studied as a single food item in the relation with bladder cancer is olive oil. Both Goulas et al. [57] and Brinkman et al. [58] showed that a higher intake of olive oil reduced bladder cancer risk. Traditionally, it has been thought that the monounsaturated fat component of olive oil was at least partly responsible for the Mediterranean diet’s health benefits but, after reviewing our sensitivity analyses using different dietary fats, this does not appear to be the case.

A possible additional explanation for a protective effect of the Mediterranean diet might be the high concentration of polyphenols in olive oil. These dietary factors are well known for their anti-oxidative and anti-inflammatory properties [59, 60]. In addition, polyphenols have been shown to have a beneficial effect on cellular function [61]. Since processes such as deregulated cell proliferation and suppressed cell death often provide a basis for tumour progression, polyphenols in olive oil may help to protect the cells of the bladder membrane against further metastasis [61]. High concentrations of polyphenols can also be found in wine, which is the main source of alcohol consumption in Mediterranean regions. Although it was expected that high concentrations of polyphenols from olive oil and wines could explain the beneficial effect of adhering to the Mediterranean diet on bladder cancer risk, it was not evident from our analyses. Therefore, more detailed analysis on polyphenols and other components of the Mediterranean diet in their relation to bladder cancer risk is needed to help explain the beneficial effect of high adherence.

Although BLEND is the largest known pooled cohort study investigating associations between adherence to the Mediterranean diet and risk of developing bladder cancer, with enough statistical power to permit detailed analyses and to detect smaller effects, it has several limitations. First, limited information was available for other possible risk factors for bladder cancer, such as body mass index, physical inactivity, socioeconomic status, and occupational exposures to carcinogenic chemicals. Adjustments for these factors could have influenced our results. Nevertheless, the current literature suggests only a small proportion of bladder cancer cases can be attributed to these factors [62,63,64]. The study of Buckland et al. [27] found a significantly inverse association for current smokers after stratification for smoking status. We repeated this stratified analysis using our data, and although the inverse association of a high adherence to the MDS and bladder cancer was only statistically significant in ever smokers (HRhigh 0.80, [95% CI 0.71, 0.89]), the stratified HR estimates did not seem to differ substantially between never smokers and ever smokers.

Another limitation of our study includes potential misclassification of frequency of food consumption derived from food frequency questionnaires (FFQs), which could lead to systematic and random error when estimating adherence to the Mediterranean diet within individual studies [65]. Also, we were not able to take into account any possible changes of dietary and lifestyle habits over time, which could lead to misclassification of long-term diet. As previously reported by Dugué et al. [28], using dietary scores does not overcome the limitations inherent to FFQs, but they may help to distinguish between individuals rather than using absolute amounts of specific foods. Lastly, most of the included cohort studies used self-reported questionnaires for the ascertainment of bladder cancer diagnosis. Previous research showed that gathering diagnostic cancer information by the use of self-reported questionnaires could lead to large amounts of false negative findings, that is, cases would be falsely classified as being a non-case [66]. This could have led to underestimation of the true association.

Conclusion

We found evidence that high adherence to the Mediterranean diet was associated with a reduced risk of developing bladder cancer. We could not isolate any particular subgroup of foods (e.g. fats, alcohol) from the MDS that provided a greater benefit over others. This may be because it describes the overall effect of the combined factors of the dietary pattern to be most protective.