Body Mass Index, Diet-Related Factors, and Bladder Cancer Prognosis: A Systematic Review and Meta-Analysis

Search strategy

A systematic search of the literature up to May 2017 in the PubMed and Embase databases was conducted by using a combination of MeSH and Emtree terms and related key words in title and abstract. The complete search strategy is attached as supplementary material (Appendix A). No limits were applied on publication date or type of study during this search. Reference lists of included articles were checked to identify potentially relevant studies that were missed in the initial search.

We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) criteria (Fig. 1) [11] for evaluating records identified during the literature search. To warrant integrity, the retrieved references were checked independently by two authors (E.W. and A.V.), and disagreements were discussed and resolved.

Fig.1

PRISMA flow diagram for the study selection process.

Study eligibility

Studies were considered relevant to this review if they reported on observational or intervention studies in UBC survivors with respect to BMI, dietary factors or supplements, or physical activity before, at the time of or after diagnosis in relation to recurrence, progression, cancer-specific mortality (CSM), or all-cause mortality (ACM). Only articles written in the English language were considered for inclusion. When multiple articles reported on the same study, only the most recently published article was taken into consideration unless there were differences in the endpoints or subgroups reported. Animal studies, and investigation of BMI, diet or physical activity with intermediate endpoints of disease or with tumor characteristics only without addressing recurrence, progression, or mortality were excluded.

Data extraction

Information extracted from the selected studies included, wherever available, the first author, year of publication, country, sample size and samplecharacteristics, years of diagnosis, follow-up time, outcome, variable of interest, risk estimates with 95% confidence intervals (CI) or p-values, and factors that were adjusted for in the analysis. When necessary, authors were contacted for additional data.

Statistical analysis

Meta-analysis was performed using Review Manager 5.3 [12]. The inverse variance technique was used for meta-analysis of hazard ratios. Due to the clinical heterogeneity inherent in our data, random-effects models were used for all meta-analyses. Statistical heterogeneity was assessed with I² values. When no meta-analysis could be performed, we narratively described the study results separately for BMI, dietary factors, dietary supplements andphysical activity.

Body mass index

Twelve historical cohort studies [5, 6, 16, 18, 19, 22, 23, 25, 29, 33, 39, 40] and one prospective cohort study [38] investigated the association of BMI with the prognostic endpoints of interest (Table 1). Nine studies assessed BMI at time of diagnosis or at time of surgery [6, 16, 18, 19, 22, 23, 25, 38, 39] and four studies did not report on the timing of BMI assessment [5, 29, 33, 39].

Three studies including 892 T1 high grade [5], 338 Ta, T1, Tis [38], and 403 Ta, T1 NMIBC patients [39] reported or provided sufficient data to be subjected to meta-analysis. Both overweight (pooled hazard ratio (HR) 1.29, 95% CI 1.05–1.58, p = 0.02, I² = 0%; Fig. 2A) as well as obesity (pooled HR 1.82, 95% CI 1.12–2.95, p = 0.02, I² = 79%; Fig. 2B) were statistically significantly associated with increased risk of recurrence when compared to normal weight. Based on two studies [5, 39], neither overweight (pooled HR 1.03, 95% CI 0.63–1.70, p = 0.91, I² = 0%; Fig. 2C) nor obesity (pooled HR 1.90, 95% CI 0.93–3.88, p = 0.08, I² = 51%; Fig. 2D) were statistically significantly associated with risk of progression.

Fig.2

BMI and NMIBC prognosis: (A) Overweight vs. normal weight and risk of recurrence; (B) Obesity vs. normal weight and risk of recurrence; (C) Overweight vs. normal weight and risk of progressoin; (D) Obesity vs. normal weight and risk of progression.

Six studies were performed in NMIBC and muscle-invasive bladder cancer (MIBC) patients combined, which results are more difficult to interpret because of the difference in relevant disease outcomes (i.e., risk of recurrence and CSM, respectively). Four studies investigated BMI in relation to recurrence. Three studies reported sufficient data to be subjected to meta-analysis [18, 19, 23]. Neither overweight (pooled HR 0.87, 95% CI 0.67–1.14, p = 0.31, I² = 60%; Fig. 3A), nor obesity (pooled HR 1.12, 95% CI 0.54–2.31, p = 0.76, I² = 95%; Fig. 3B) were statistically significantly associated with risk of recurrence. One study did not find an association for continuous BMI [16]. Four studies investigated CSM, three of them reporting sufficient data to be included in a meta-analysis [18, 19, 23]. Neither overweight (pooled HR 0.82, 95% CI 0.65–1.02, p = 0.08, I² = 45%; Fig. 3C), nor obesity (pooled HR 0.98, 95% CI 0.46–2.10, p = 0.97, I² = 94%; Fig. 3D) were statistically significantly associated with CSM. One study did not find an association for continuous BMI [16]. Four studies investigated ACM [6, 16, 18, 33]. One study reported an association of overweight and obesity compared to normal weight with higher risk of ACM (HR 1.40, 95% CI 1.23–1.57 and HR 1.81, CI 95% CI 1.60–2.05, respectively) [18], while three studies did not find an association [6, 16, 33]. Progression was not investigated in any of these studies.

Fig.3

BMI and prognosis in all stages series: (A) Overweight vs. normal weight and risk of recurrence; (B) Obesity vs. normal weight and risk of recurrence; (C) Overweight vs. normal weight and risk of CSM; (D) Obesity vs. normal weight and risk of CSM.

Four historical cohort studies investigated clinical outcomes exclusively in MIBC patients. Three studies investigated ACM [22, 29, 40], two of them reporting a lower ACM for higher BMI (26.3–29.1 compared to <23.6: HR 0.80, 95% CI 0.69–0.93 and ≥31.2 compared to <31.2: HR 0.46, 95% CI 0.26–0.83, respectively) [29, 40] and the other not showing an association for ACM and CSM [22]. One study found no differences in risk of progression or ACM when comparing overweight or obesity with underweight [25].

Dietary factors

Four small cohort studies evaluated the association of dietary factors with UBC outcomes (Table 2). These studies examined the effects of vitamin A intake [27], intake of different (alcoholic) drinks [37], fluid intake [9], and fruit and vegetable intake [36].

In a study of 102 NMIBC patients, patients with a prediagnosis dietary intake of vitamin A below the median (not specified) were found to have a 1.8 times higher recurrence rate per 1000 person-months (p = 0.02), but no statistically significantly increased risk of recurrence (RR 1.34, 95% CI 0.5–3.34) compared to patients having a dietary intake of vitamin A above the median [27].

Three studies were performed in NMIBC and MIBC patients combined [9, 36, 37]. A Japanese study in 258 patients found an inverse association for ever versus never prediagnostic consumption of alcoholic beverages with 5-year ACM (HR 0.46, 95% CI 0.26–0.79) [37]. No associations fornon-alcoholic beverages and artificial sweeteners were found. Daily post-diagnostic total fluid intake measured using repeated fluid intake questionnaires was not associated with risk of recurrence in a US study in 267 patients [9]. A US study in 239 patients investigated self-reported prediagnostic intake of fruits and vegetables in relation to CSM and ACM [36]. Patients consuming ≥1 compared to those consuming <1 serving per month of raw broccoli had a lower CSM (HR 0.43, 95% CI 0.25–0.74) and ACM (HR 0.57, 95% CI 0.39–0.83, data not shown). Consumption of 1–3 servings of raw cruciferous vegetables per month compared to <1 serving per month was inversely associated with ACM (HR 0.67, 95% CI 0.46–0.97), but not with CSM. No association was found for consumption of more than three servings. Consumption of total fruit, total vegetables or individual cooked cruciferous vegetables was not associated with CSM or ACM.

Dietary supplements

Fourteen dietary supplementation studies in UBC patients were identified (Table 3). In these randomized controlled trials, the effects of the retinoids etretinate [13, 32, 35] and fenretinide [20, 34], vitamin B6 (pyridoxine) [17, 31], vitamin E (tocopherol) [26], multivitamins [24, 30], a Lactobacillus casei preparation [14, 28, 41], and selenium [21] were investigated. Thirteen of the 14 studies were performed in NMIBC patients [13, 14, 17, 20, 21, 26, 28, 30–32, 34, 35, 41] and one trial was conducted in NMIBC and MIBC patients combined [24].

Three studies evaluating the effect of the aromatic retinoid etretinate had conflicting results [13, 32, 35]. One small study [13] in 32 patients found a lower recurrence rate in the etretinate group compared to the placebo group, while a study of 96 patients [32] and a trial of 79 patients [35] found no differences in recurrence. Nevertheless, in the last study the mean interval to the second recurrence was longer in the etretinate group (20.3 versus 12.7 months, p = 0.006), resulting in less transurethral resections of the bladder per patient-year (p < 0.01). This may suggest that etretinate needs to be taken for a prolonged period before an effect becomes evident. Fenretinide did not affect recurrence in 99 Italian [20] and 137 US [34] patients.

Supplementation with pyridoxine was found not to affect recurrence or progression in 121 stage I patients [17]. However, when patients with recurrences during the first ten months or followed up less than ten months were excluded, recurrence rate was significantly lower for pyridoxine compared to placebo. Again, this suggests that a longer duration of supplementation is required. A study in 291 patients [31] did not find an effect of pyridoxine on risk of recurrence, progression, or ACM.

A small trial in NMIBC patients in Iran (n = 46) reported that daily intake of 400 IU vitamin E after diagnosis resulted in a lower risk of recurrence (RR = 0.53, 95 CI% 0.11–0.94) compared to a control group [26].

Two trials tested if supplementation with a megadose of multivitamins had an effect on recurrence[24, 30]. In 65 NMIBC and MIBC patients, a reduced risk of recurrence was found in the experimental group that received a supplement with the recommended daily allowance (RDA) plus a megadose of vitamin A, B6, C, E and zinc compared to RDA alone [24]. In contrast, a large multicenter trial in 670 BCG naïve patients found no difference in recurrence-freesurvival [30].

Three studies conducted in Japan investigated the effect of 3 gram/day of the orally administered probiotic Lactobacillus casei preparation on prognosis [14, 28, 41]. A study of 202 patients receiving standard treatment of intravesical instillations with epi-adryamicin randomized to a daily administration of Lactobacillus casei preparation for one year versus a control condition (no Lactobacillus casei preparation), reported that the recurrence rate was lower in the Lactobacillus casei group compared to the control group (HR 0.57, 95% CI 0.35–0.93) [28]. No effect on progression or ACM was found. A study in 138 patients receiving either a Lactobacillus casei preparation or placebo did not find a difference in recurrence-free interval when all patients were included in a univariable analysis [14]. However, when analysis was restricted to78 patients with primary multiple tumors or recurrent single tumors, Lactobacillus casei preparation was associated with a longer recurrence-free interval as compared to placebo (688 days versus 543 days; HR 2.58; p = 0.013). In a study of 58 patients, the recurrence-free survival interval of the group who received the Lactobacillus casei preparation was prolonged to 1.8 times the interval of the control group (350 days versus 195 days; HR 2.41, p = 0.028) [41].

A study in 292 NMIBC patients conducted in Belgium showed that selenium supplementation did not lower the risk of recurrence compared to placebo [21].

Limitations of the studies included

Studies on BMI and dietary factors conducted to date mostly included a heterogeneous studypopulation with different stages and grades of UBC and undergoing different treatments. Risk of recurrence, progression and CSM differs by UBC stage and grade [48]. This should be taken into account when evaluating the association of BMI and dietary factors with these disease endpoints. However, none of the studies included in our review performed subgroup analyses by tumor stage and grade, possibly masking true associations. Also, disease progression was investigated in a minority of the studies and several observational studies only investigated the association with CSM or ACM, which are not preferable endpoints for NMIBC. Since risk of death from NMIBC is relatively low, mortality mainly reflects the general health condition of the participants instead of the consequences of bladder cancer. Thus, a possible association of BMI and dietary factors with disease outcomes in these heterogeneous UBC populations may have been missed.

In several studies, BMI and diet were not the primary factor of interest, and only one or a few single dietary factors were studied in relation to UBC prognosis. Since dietary components interact with each other in a complex way, it is preferable to take the entire dietary pattern into account instead of single dietary factors. Furthermore, depending on dose, dietary supplements can prevent or promote growth of tumor cells or existing (pre)cancerous lesions. For example, the ATBC study reported an increased instead of expected decreased incidence of lung, prostate and stomach cancer in Finnish male smokers receiving supplements with supraphysiological doses of β-carotene versus placebo [49]. In addition, recent mouse experiments showed that antioxidants may protect melanoma and lung cancer cells and promote growth and the ability to metastasize [50, 51]. UBC survivors could be using dietary supplements more frequently than population controls, as is observed for prostate cancer survivors [52]. Given the lack of evidence for a beneficial effect on prognosis in UBC, and reasoning that the same biology may apply for UBC as for other neoplasms, urologists may consider to advise UBC survivors against using dietary supplements. This advice is in line with the WCRF/AICR cancer prevention recommendation not to use dietary supplements to protect against cancer [53].

The observational studies on BMI were generally based on historical cohorts with information retrospectively collected from medical records of consecutive patient series. This may have led to selection bias and incomplete retrieval of information. Also, almost none of the included studies compared the same BMI categories, and some studies [29, 40] did not use official BMI categories as defined by the World Health Organization to compare patients, possibly explaining the inconsistent results. Furthermore, only one of the studies included other measures to assess obesity or body composition, such as waist-to-hip ratio or densitometry [33]. The observational studies on dietary factors were mostly based on follow-up of existing case-control studies, were small, did not report the type of questionnaire used, and defined only two broad comparison groups (e.g. ever vs. never, >vs. <median, ≥1 vs. <1 servings per month) [27, 36, 37]. None of the studies investigated post-diagnostic BMI and only one study investigated post-diagnostic diet but was restricted to overall fluid intake [9]. Post-diagnostic changes in BMI or diet in relation to prognosis were not addressed, while the most relevant question for patients is whether making lifestyle changes affects their prognosis. Most supplementation studies also had a limited sample size and in some of these studies only a few patients experienced the outcome(s) of interest, highly limiting the statistical power. Some studies did not report whether study participants and investigators were blinded[13, 14, 17, 24, 26, 32, 41] and one study was not blinded [28]. In addition, for both observational and supplementation studies heterogeneity was present in type of exposure, timing of exposure assessment and outcomes of interest which makes it difficult to compare studies and to draw conclusions. Incomplete reporting of patient characteristics and methods aggravated this problem. Also, smoking was only taken into account in some studies. Thus, there is a potential risk of bias in individualstudies.

Ongoing challenges and steps to be taken

Taking into account the shortcomings of the studies published to date, there is an urgent need for well-designed, prospective cohort studies on the association of dietary and other lifestyle factors with UBC prognosis. There are several considerations for the design of such studies. First of all, patients newly diagnosed with UBC should be recruited rapidly so that data collection can start shortly after diagnosis. Ideally, questionnaires should cover the entire diet, and assessment of pre-diagnostic dietary and lifestyle habits should be followed by repeated post-diagnosis assessments to capture changes in lifestyle. In this way, the most relevant question for the patient can be addressed, i.e. whether lifestyle changes are related to UBC prognosis. Furthermore, blood samples should be collected to investigate biomarkers of dietary and lifestyle habits. A representative, population-based sample of UBC patients should be included to assure generalizability, but NMIBC and MIBC patients should be investigated separately and, if possible, by molecular subtype as relevant disease outcomes and prognosis are clearly different for these subgroups [54]. Sample size should be large enough to study associations with sufficient power, and to study progression which is a less frequent outcome. Smoking and perhaps antidiabetic medication use should be taken into account as the association between lifestyle and UBC prognosis may be confounded by these factors, or may be different according to subgroups of these factors. Furthermore, physical activity may be a relevant lifestyle factor for further investigation as well, since it has been suggested to improve immune system functioning and to reduce systemic inflammation, thereby positively influencing the course of disease [55, 56].

Also, studying body composition may be more relevant than BMI, as BMI cannot distinguish between fat mass and skeletal muscle mass, and only low muscle mass but not fat mass has been associated with increased risk of ACM in MIBC patients[33, 57].