Macronutrient intake and risk of urothelial cell carcinoma in the European prospective investigation into cancer and nutrition
Previous studies have suggested that dietary factors may be important in the development of bladder cancer. We examined macronutrient intake in relation to risk of urothelial cell carcinoma among 469,339 men and women in the European Prospective Investigation into Cancer and Nutrition. Associations were examined using Cox regression, stratified by sex, age at recruitment and centre and further adjusted for smoking status and duration, body mass index and total energy intake. After an average of 11.3 years of follow-up, 1,416 new cases of urothelial cell carcinoma were identified. After allowing for measurement error, a 3% increase in the consumption of energy intake from animal protein was associated with a 15% higher risk (95% confidence interval [CI]: 3–30%; ptrend = 0.01) and a 2% increase in energy from plant protein intake was associated with a 23% lower risk (95% CI: 36–7%, ptrend = 0.006). Dietary intake of fat, carbohydrate, fibre or calcium was not associated with risk. These findings suggest that animal and/or plant protein may affect the risk of urothelial cell carcinoma, and examination of these associations in other studies is needed.
Bladder cancer is the eight most common cancer in the world, although there is wide variation in incidence and mortality rates between countries, being high in North America, Europe and Northern Africa, and low in Asia.1 Bladder cancer is more common in men than women, largely because of differences in smoking habits and historical occupational exposure to carcinogens (e.g., aromatic amines), although a large proportion of bladder cancer incidence remains unexplained. It has been suggested that dietary factors may be important in bladder cancer development, as many diet-related metabolites come into direct contact with the bladder epithelium during excretion.2 In particular, it has been hypothesised that a diet high in meat or fat might increase risk, with a meta-analysis of six case–control studies and one cohort study finding a significant 37% increased risk for high versus low fat intake.3 However, very few epidemiological studies have collected sufficiently detailed dietary information to examine the intake of different types of dietary fat or different sources of protein intake in relation to the risk of bladder cancer.
The aim of our study is to examine the association between macronutrient intake and risk of bladder cancer in the European Prospective Investigation into Cancer and Nutrition (EPIC), a large multicentre cohort study including participants from across Europe with a wide range of nutrient intakes.
Material and Methods
The EPIC is a prospective study conducted in 23 centres in ten European countries: Denmark, France, Germany, Greece, Italy, The Netherlands, Norway, Spain, Sweden and the United Kingdom (UK). Full details of recruitment procedures and collection of baseline data have been described elsewhere.4 In brief, participants were recruited between 1992 and 2000, usually from the general population in each country. Exceptions were the cohort in France (based on female members of the teachers' health insurance organization), the cohorts in Utrecht (Netherlands) and Florence (Italy) (based on women attending breast cancer screening), parts of the cohorts in Italy and Spain (based on blood donors and their spouses) and the Oxford health-conscious subcohort (recruited throughout the UK to enrol a large number of vegetarians). Eligible subjects were invited to participate in the study, and those who accepted gave informed consent and completed questionnaires on their diet, lifestyle and medical history. Participants were predominantly of Caucasian origin.
Participants were not eligible for this analysis if they had previously been registered as having cancer at the time of completing the baseline questionnaire, if they had missing or inconsistent data for the variables of interest or if they were in the top or bottom 1% of the distribution of the ratio of energy intake to energy requirement. Following these exclusions, complete data on diet and follow-up were available for 469,339 participants out of the original 517,330 participants.
Dietary intake during the year before enrolment was measured by country-specific validated questionnaires designed to capture local dietary habits and provide high compliance, and included between 88 and 266 food items. Most centres used a self-administered dietary questionnaire, although in centres in Greece, Spain and in some centres in Italy, the questionnaire was administered by interviewers. In Malmö (Sweden), a questionnaire method combined with a food record was used. To correct for estimation errors and to make dietary exposures comparable across all EPIC centres, a single standardised and computerised 24-hr dietary recall interview was carried out in a random sample (8%) of the cohort.5
Estimated daily nutrient intakes (g/day) were calculated by multiplying the nutrient content of each food of a specific portion size by the frequency of consumption using national food tables from each country, compiled in the EPIC Nutrient Database.6 For this analysis, intake of animal protein was calculated as protein derived from total meat and meat products (which included the subcategories red meat, poultry and processed meat), fish and shellfish (which included the subcategories white fish and fatty fish), dairy products (which included the subcategories milk and milk beverages, yoghurt and cheese) and eggs. Plant protein was calculated as total protein minus animal protein. Umeå (Sweden) was not included in the analyses of animal or plant protein because the relevant data were not available in the central database.
The nondietary questions covered education and socioeconomic status, occupation, history of previous illness or surgical operations, lifetime history of consumption of tobacco and alcoholic beverages and physical activity. Height and weight were measured at recruitment, except for participants in France, Norway and Oxford (UK), among whom height and weight were self-reported.4
Incident cases of bladder cancer were identified by population cancer registries (Denmark, Italy, The Netherlands, Norway, Spain, Sweden and the UK) or by active follow-up (France, Germany and Greece) that involves a combination of methods, including health insurance records, cancer and pathology registries and direct contact of participants or next-of-kin. Mortality data were obtained from cancer or mortality registries at the regional or national level, and censoring dates for the end of follow-up ranged from December 2004 to December 2008, depending on the centre. For France, Germany and Greece, the censoring date was considered to be the date of the last known contact or date of diagnosis or date of death, whichever came first.
Participants were followed from study entry until a diagnosis of first primary urothelial cell carcinoma (code C67 according to the International Classification of Diseases-Oncology [ICD-O 3rd edition]), any other cancer, death, emigration or end of the follow-up period. Only urothelial cell papillomas and transitional cell carcinomas (morphology codes 812-813), further referred to as urothelial cell carcinoma, were included in the analysis, as these comprise more than 90% of bladder tumours. Bladder cancer cases with other morphology codes or with missing behaviour codes were censored at the time of diagnosis. The cancer diagnosis was confirmed by histology for 80% of the cases, by clinical examination for 9%, by self-report, tomography scan, surgery or cytology/haematology for 9%, by autopsy or death certificate for 1% and was missing for less than 1% of cases.
Information on stage (based on tumour-node-metastasis staging) and grade of the tumour (based on the World Health Organisation Grading system) was extracted from pathology reports, where available. Nonaggressive urothelial cell carcinoma (i.e., a relatively low risk of progression) was defined as all Ta Grade 1 (well differentiated) and Ta Grade 2 (moderately differentiated); aggressive urothelial cell carcinoma (i.e., a high risk of progression) was defined as T1 and higher or carcinoma in situ or Grade 3 (poorly differentiated).
Age-standardised rates in each country were calculated by using a weighted average of the incidence rates in each of four attained age groups (50–54, 55–59, 60–64 and 65–69 years), with weights based on the European standard population.
Cox proportional hazard regression was used to analyse the association between macronutrient intake and risk of urothelial cell carcinoma. Age was used as the primary time variable in all models with entry time defined as age at recruitment and exit time as age at bladder cancer diagnosis or censoring. All results were stratified by sex, age at recruitment (<40, 40–44, 45–49, 50–54, 55–59, 60–64 and ≥65 years) and centre to control for any systematic differences that may occur due to follow-up procedures, questionnaire design and mode of detection between centres. All energy-providing macronutrients were adjusted for energy using the nutrient density method,7 whereby the percentage of energy from each macronutrient plus total energy intake are included in the same model (separate models were performed for each macronutrient); this can be interpreted as the effect of substituting nutrient intake for other sources of energy while keeping total energy intake constant. Analyses were also performed based on the standard multivariable model (absolute macronutrient intake plus total energy intake) and the energy partition model (energy from each macronutrient plus energy from other nutrients).7 All results were similar to those of the nutrient density method and are not presented here. Nonenergy providing nutrients (i.e., fibre and calcium) were adjusted for energy using the standard multivariable method. Macronutrient intake was analysed as categorical variables, based on sex-specific quintiles of the distribution among the noncases across all EPIC centres combined, and the linear test for trend was obtained by including the median value of each category as a continuous variable in the regression model.
In addition to energy intake, potential confounding variables included in the final model were smoking status (nine categories: never, former cigarette smokers who quit >20, 11–20 and ≤10 years ago, current occasional cigarette smokers or those who smoke other forms of tobacco, current smokers who smoke 1–15, 16–25 and >26 cigarettes per day, current and former smokers with an unknown value of intensity or time since cessation, respectively), smoking duration (eight categories: never or occasional cigarette smoker, ≤10, 11–20, 21–30, 31–40, 41–50 and >50 years, unknown) and body mass index (BMI, continuous). Other variables including physical activity, educational level, marital status, alcohol intake, self-reported history of diabetes or cardiovascular problems were not associated with risk and were not included in the final model.
Nutrient intakes were calibrated using a multivariable linear model in which the 24-hr diet recall values from the calibration subsample of participants were regressed on the main dietary questionnaire values.8 These models also included total energy intake, study centre, participants' height and weight and age at recruitment as covariates, country-specific interaction terms with energy intake and nutrient, and were weighted by the day of the week and season of the year of the 24-hr diet recall. The calibration model was used to compute individual predicted values for each of the dietary exposures of interest. Cox regression models were performed by using the predicted (calibrated) values on a continuous scale for all nutrients of interest.
We examined whether the association between calibrated nutrient intake and risk of urothelial cell carcinoma was modified by sex, age at diagnosis (<65 and ≥65 years), smoking status (never, former and current), BMI (<25 and ≥25 kg/m2), time between recruitment and diagnosis (<2 and ≥2 years) and tumour aggressiveness (nonaggressive and aggressive), by including interaction terms in the disease models related to a continuous increment of calibrated nutrient intake across the categories of interest. Heterogeneity between countries in the association of nutrient intake with risk of urothelial cell carcinoma was assessed using chi-square tests. All p values presented are two tailed, and p values below 0.05 were considered statistically significant. Analyses were performed using Stata v11 (Statacorp, College Station, TX).
After an average of 11.3 years of follow-up, 1,416 participants (1,016 men and 400 women) were diagnosed with a first incident urothelial cell carcinoma among the 469,339 participants included in our study, with ∼5.3 million person-years of follow-up (Table 1). The median age at diagnosis of bladder cancer was 66 years (range: 26–89 years) in men and 64 years (range: 34–83 years) in women. Information of tumour stage and grade was available for 786 (59%) of the cases; of these, 402 (51%) were defined as nonaggressive and 384 (49%) as aggressive.
Table 1. Distribution of cases and controls by subcohort in EPIC
Age-standardised incidence rates were approximately fivefold higher in men than in women, being 55.4 and 11.4 per 100,000 person-years, respectively. The highest rates were in The Netherlands and Italy for men and in Germany for women, whereas Greece had the lowest incidence rates (Table 1). Overall, 80% of urothelial cell carcinomas were among either former or current smokers (data not shown).
Table 2 shows the mean intakes of macronutrients, based on the 24-hr recall data, for each country. The range of nutrient intakes between countries varied from 1.3-fold for total energy and total fat, twofold for dairy protein and monounsaturated fat and up to 3.5-fold for animal protein. Energy intake from total protein and animal protein was highest in Spain and lowest in the health-conscious cohort in the UK. Energy intake from plant protein was lowest in Sweden and Denmark and highest in the health-conscious cohort in the UK. Energy from carbohydrate intake was highest in the UK and Italy and lowest in Greece and Spain. Energy intake from total fat was highest in Greece and lowest in Italy, although intake of saturated fat was highest in Sweden. Monounsaturated fat intake was highest in Greece compared to other countries, whereas polyunsaturated fat intake was highest in the health-conscious cohort in the UK and lowest in Italy. Dietary fibre intake was lowest in Sweden and highest in the health-conscious cohort in the UK, whereas calcium intake was highest in Denmark and lowest in Italy.
Table 2. Mean energy and macronutrient intakes by country in EPIC, based on 24-hr recall data1
Categorical analyses of the association of macronutrient intake with the risk of urothelial cell carcinoma showed that, after minimal adjustment for age, sex, centre and total energy intake, animal protein intake was associated with an increased risk (hazard ratio [HR] for the highest vs. lowest quintile = 1.18, 95% confidence interval [CI]: 0.98–1.44; ptrend = 0.03); this risk estimate was attenuated after additional adjustment for detailed measures of smoking and BMI (HR for the highest vs. lowest quintile = 1.12, 95% CI: 0.92–1.36; ptrend = 0.14; Table 3). Energy intake from dairy protein was also associated with an increased risk (fully adjusted HR for the highest vs. lowest quintile = 1.27, 95% CI: 1.07–1.51; ptrend = 0.04; Table 3). In contrast, energy intake from plant protein was associated with a lower risk of urothelial cell carcinoma (age- and sex-adjusted HR for the highest vs. lowest fifth of intake = 0.76, 95% CI: 0.62–0.93; ptrend = 0.004), although this was also attenuated after further adjustment for smoking and BMI (HR for the highest vs. lowest fifth of intake = 0.89, 95% CI: 0.73–1.09; ptrend = 0.22; Table 3). Energy intake from total protein, total fat or its components, carbohydrate, fibre or calcium, was not significantly associated with risk of urothelial cell carcinoma (Table 3).
Table 3. Multivariable HRs (95% CI) for risk of urothelial cell carcinoma associated with macronutrient intake
Table 4 shows the risk estimates for urothelial cell carcinoma associated with consumption of macronutrients when evaluated as continuous variables, before and after calibration. Here, both observed and calibrated measures of energy intake from animal protein and plant protein were associated with risk; a 3% increase in intake of calibrated energy from animal protein (roughly equivalent to a 60 percentile increase) was associated with a 15% increase (HR = 1.15, 95% CI: 1.03–1.30, ptrend = 0.01), and a 2% increase in calibrated energy intake from plant protein (roughly equivalent to a 80 percentile increase) was associated with a 23% lower risk (HR = 0.77, 95% CI: 0.64–0.93, ptrend = 0.006). Dairy protein was not significantly associated with risk. No other nutrients were significantly associated with risk.
Table 4. Multivariable HRs (95% CI) for risk of urothelial cell carcinoma for specified increments in observed (uncalibrated) and calibrated intakes of macronutrients1
The correlation between calibrated intake of energy from animal protein and energy from plant protein was −0.5. Analyses of the association of calibrated intakes of energy from animal protein (as a continuous variable) by thirds of plant protein showed no evidence of statistical heterogeneity (pheterogeneity = 0.47); there was also no heterogeneity in the association of energy intake from plant protein according to thirds of energy intake from animal protein (pheterogeneity = 0.91; data not shown).
The association between calibrated intakes of energy from animal and plant protein and risk of urothelial cell carcinoma according to subgroups of gender, age at diagnosis, smoking status, BMI, time between study recruitment and diagnosis and tumour aggressiveness is shown in Table 5. There was no evidence of statistical heterogeneity between any of the subgroups examined, with the exception of plant protein intake and time between recruitment and bladder cancer diagnosis (pheterogeneity = 0.01; Table 5). There were no significant differences in any associations by country (data not shown).
Table 5. Multivariable HRs and 95% confidence intervals (CIs) of urothelial cell carcinoma by patient characteristics for specified increments in calibrated intakes of animal and plant protein in EPIC1
In our prospective study of 469,339 men and women and over 1,400 incident cases of urothelial cell carcinoma, we found no evidence that intake of energy from total protein, fat or carbohydrate is related to the risk of urothelial cell carcinoma. However, a high intake of energy from animal protein was associated with an increased risk, and a high intake of energy from plant protein was associated with a reduced risk of urothelial cell carcinoma.
There has been much speculation about the role of animal foods and, in particular, meat intake in relation to the risk of bladder cancer, although the epidemiological evidence is inconsistent.3 It has been suggested that the high nitrosamine content of some processed meat products may increase risk, although previous results from EPIC do not support this hypothesis.9 The results from our study raise the possibility that the protein content of animal foods may also be of importance, although a US-based cohort study with over 300 cases found no association with either total or animal protein intake, despite comparable intakes of animal protein in the two populations.10 Case–control studies have also generally found no association between total protein intake and risk of bladder cancer,11–14 although one case–control study in men from the US reported an inverse association for the highest versus the lowest category of total protein intake, but which was restricted to a subgroup of older men (RR = 0.46, 95% CI: 0.26–0.83).15
Although it is difficult to disentangle the effects of a high intake of animal protein from that of a low intake of plant protein, one explanation for our finding that a high intake of animal protein is associated with an increased risk is that animal protein, and in particular, dairy protein, may raise circulating concentrations of insulin-like growth factor-I.16 This peptide hormone has been shown to be associated with an increased risk of cancer of the breast, prostate and colorectum17–19 and also possibly bladder cancer.20 However, it is also possible that the associations found for protein intake with the risk of bladder cancer might reflect residual confounding by smoking, a strong risk factor for bladder cancer.21 Although all statistical models included detailed information on smoking history, current smokers had a higher intake of energy from animal protein and a lower intake of energy from plant protein than never smokers (as shown in Table 5). In addition, although there was no significant heterogeneity in the risk estimates by smoking status, the associations for both animal and plant protein intake were somewhat stronger among current and former smokers compared to never smokers, suggesting that a residual effect of smoking cannot be ruled out. Further examination of these associations in nonsmokers is needed to clarify whether there is a truly independent association of protein content on risk. Intake of animal protein intake is also associated with BMI,22 although its inclusion in the final model made very little difference to the overall risk estimates, suggesting that any residual confounding effect of BMI is likely to be minimal. Finally, it is possible that the association seen with animal and plant protein on risk may be due to other unmeasured dietary factors, such as specific fatty acids or vitamins and minerals. However, there is no consistent evidence that these factors are associated with risk,3, 23, 24 and the associations, if any, are unlikely to be strong enough to fully explain our observations.
We also found that higher intakes of energy from plant protein were associated with a nonsignificant increased risk in those diagnosed within 2 years of recruitment and a reduced risk for those diagnosed later than this. Although this difference could reflect reverse causation, early symptoms of bladder cancer do not generally produce major changes in dietary habits. In addition, the lack of a corresponding effect seen with animal protein suggests that this difference might also be due to chance given the number of subgroup analyses performed.
Dietary fat intake, particularly saturated fat from animal products, is a major component of the Western diet, and ecological studies have found total fat intake to be positively correlated with bladder cancer mortality rates between countries.25, 26 Our finding that intake of energy from total fat was not associated with risk is consistent with results from two prospective studies in the US,10, 11 although a meta-analysis (that included six case–control studies and one cohort study) reported a summary relative risk of 1.37 (95% CI: 1.16–1.62) for high versus low intakes of total fat.3 There are very limited data on the type of dietary fat on the risk of bladder cancer, with case–control studies reporting both positive,13 inverse27 and null associations28 with saturated fat intake. Such inconsistent observations from both epidemiological and experimental studies29, 30 suggest that the association, if any, is small.
It has been suggested that a high intake of dairy products, and in particular, milk intake, may be associated with a reduced risk of bladder cancer.11, 27, 31, 32 However, our findings that intake of energy from saturated fat or intake of calcium was not related to risk and that intake of energy from dairy protein was associated with a small increased risk do not provide strong support for this hypothesis. Although it is possible that other components of dairy products, such as the bacterial composition or lactose content, might be important factors, these results suggest that the macronutrient content is not strongly related to risk. Other studies have also found no association with calcium intake, lending further support to the hypothesis that this nutrient is not related to the risk of bladder cancer.10, 33, 34
Although all EPIC dietary questionnaires have been standardised and validated using multiple 24-hr diet recalls, as well as 24-hr urine samples,35 it is well known that estimating nutrient intake using questionnaires is associated with some degree of random error that tends to attenuate risk estimates as well as systematic errors.36 However, nutrient intakes in our study were calibrated using intake derived from a single 24-hr recall, which serves to reduce between-cohort differences due to systematic exposure measurement error.8 Although this method may not completely account for all measurement errors (because of likely correlated errors between the 24-hr recalls and dietary questionnaires), it is thought that adjustment for total energy may partly remove some errors in estimated nutrient intake, especially for nutrients highly correlated with energy intake.8 Another limitation is that dietary and other lifestyle factors were only measured at recruitment, and we were unable to examine how changes over time in nutrient intake and other factors such as smoking may have influenced the associations.
Our study is the largest to date to examine the association between macronutrient intake and risk of bladder cancer. It is of a prospective design, which eliminates the risk of recall bias, with over 11 years of follow-up and a wide range of dietary intakes between the participating countries. We were also able to examine whether the associations of macronutrient intake and risk were modified by age, sex, smoking history, body size and other factors. We also used various methods for adjusting for energy intake, all of which produced similar results, and all analyses were adjusted for known risk factors of bladder cancer, where possible. Finally, this analysis was restricted to urothelial cell carcinomas, thereby reducing any differential effect of nutrient intake on other histological types of bladder tumour.
In conclusion, although a high intake of energy from animal protein and a low intake of energy from plant protein is associated with an increased risk of urothelial cell carcinoma, the possibility of residual confounding by smoking cannot be excluded. Further research is necessary to investigate these associations in nonsmokers and to study the possible effect of protein composition on bladder carcinogenesis.