Introduction

Urinary bladder carcinoma (BCa) ranks ninth among the most common cancers and comprises 3% of all malignancies worldwide [1]. BCa is uncommon before 50 years of age but its risk increases sharply thereafter, and males are affected more than females [2]. In addition to smoking, and occupational exposure to aromatic amines [3, 4], chronic inflammation is a recognized risk factor for BCa. Inflammation is self-limiting host defense mechanism against biological, physical and chemical agents. However, its chronicity may lead to cellular damage causing various diseases including cancer [5]. Urinary tract infection (UTI) refers to the presence of microbial pathogens in urethra or bladder (lower urinary tract) or ureter and pelvis of kidney (upper urinary tract). The most common causative agent for both uncomplicated and complicated UTI is uropathogenic Escherichia coli [6]. Both host and bacterial factors influence the probability whether asymptomatic colonization resolves spontaneously or progresses to symptomatic infection [7]. Human genetic variations particularly those affecting the immune response have been associated with recurrent UTI [7,8,9]. Furthermore, UTI has a propensity to recur and individuals with a history of one or more UTI episodes have 70% increased risk of recurrence [7, 10]. However, it is uncertain whether chronic and/or recurrent UTI plays any role in causation of urothelial carcinoma of bladder [11, 12].

Recently, epidemiological evidence for an association between chronic UTI and BCa was qualitatively summarized [11, 12], which showed that both men and women with chronic UTI had an increased BCa risk [13,14,15,16,17,18,19,20,21,22]. However, some other studies did not support this notion [23,24,25]. Studies on the number and duration of UTI episodes in relation to BCa diagnosis showed that positive association tended to weaken not only with an increasing time lag between UTI and BCa diagnosis [19, 21, 23], but also with a decreasing number of UTI episodes [17, 19]. In contrast, two other studies reported a decreased BCa risk with an increasing number of kidney and bladder infections with the lowest risk for more than three UTI episodes [16, 26]. However, these studies were conducted rather in small samples and/or as sub-group analyses [12], and not surprisingly, therefore, published evidence is largely inconsistent, with wide ranging effect size estimates. This study therefore, quantitatively assessed the magnitude and direction of an overall association between chronic UTI and BCa risk using the techniques of meta-analysis.

Methods

Search strategy

This study was undertaken based on a protocol developed in accordance with Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) [27], and Meta-analysis of Observational Studies in Epidemiology (MOOSE) statements [28]. A literature search of electronic databases including Medline (PubMed), Embase, Ovid, Science Direct, Web of Science and Cochrane Library was conducted for case–control and cohort studies examining the association between chronic UTI and BCa risk and published through June 2016. Key terms combined in Boolean search included bladder or urinary bladder or transitional cell carcinoma (or urothelial carcinoma) or squamous cell carcinoma, adenocarcinoma, cancer or carcinoma or tumor, chronic or recurrent or persistent or continuous or long duration or frequent infection or disease or bacterial urinary tract infection or lower urinary tract or cystitis. Furthermore, the reference lists of retrieved articles were reviewed for undetected relevant studies. Only studies published in the English language were considered.

Inclusion criteria

The studies meeting the following criteria were included in the meta-analysis: (1) observational studies using case–control or cohort design; (2) exposure of interest chronic and/or recurrent bacterial infection of bladder (defined as the infection of the urinary tract that either does not respond to treatment or keeps recurring) in any gender; (3) outcome was histopathology-confirmed bladder carcinoma/cancer (i.e., including mainly urothelial carcinoma (transitional cell carcinoma), squamous cell carcinoma or adenocarcinoma); (4) provided measure of association as odds ratio (OR), relative risk (RR) or hazard ratio (HR) and their 95% confidence intervals (CI) obtained through stratified analysis or multivariable regression analysis. If the data were used in more than one study, the most recent complete article was included.

Data extraction

Data were extracted from each eligible study using standard data collection sheet which included the following variables; the first author’s name, publication year, country wherein study was conducted, participants’ gender (i.e., either one or both genders), study design, sample size, the effect estimates as OR, RR or HR, their 95% CIs and confounding variables accounted for in planning and/or in analysis.

Methodological quality assessment

To assess the methodological quality of selected case–control and cohort studies, the Cochrane Collaboration-recommended 9-point Newcastle–Ottawa Scale (NOS) for nonrandomized studies was used [29]. With NOS a maximum total of 9 points is possible and a study attaining a score of 6 or more is regarded as a high quality study [30]. The quality of case–control studies was assessed in the following aspects: adequate definition of bladder cancer cases, representativeness of cases, definition and selection of controls, control for the most important factor or the second most important factor, exposure assessment, same method of exposure ascertainment in cases and controls and their non-response rates (Table 1). The quality of selected cohort studies was assessed in following aspects: representativeness of exposed and non-exposed cohorts, ascertainment of exposure, documentation that the outcome was absent in members of both the cohorts at the beginning of follow-up, comparability of cohorts based on design or analysis and unbiased assessment of outcome (Table 1).

Table 1 Characteristics of the case–control and cohort studies included in the meta-analysis to quantify the association between chronic urinary tract infection and bladder carcinoma risk
Full size table

Statistical analysis

Meta-analysis was carried out using random-effects models that took into account within and between studies’ heterogeneity [31]. Heterogeneity across the included studies was evaluated using Cochran Q statistic and was considered statistically significant if the associated p value was < 0.05. I2 metric was used to estimate the proportion of total variation across the included studies due to heterogeneity rather than by chance and was interpreted as absent (I2: 0–25%), low (I2: 25.1–50%), moderate (I2: 50.1–75%) or high (I2: 75.1–100%) [32]. A pre-defined set of sub-group analyses were conducted to examine between-group differences in random-effects summary estimates of BCa risks by chronic UTI across the studies. Sub-groups were based on study location (USA, Europe and others), study period (before and/or during 1990 vs. after 1990) and exposure definition (cystitis, chronic UBI or chronic/recurrent UTI). A random-effects meta-regression analysis was conducted for quantitative factors including sample size, proportion of male cases and NOS score to examine their potential modulating effects on the relationship between Chronic UTI and BCa risk. Potential publication bias was assessed using funnel plot, Egger’s regression asymmetry test and Begg’s rank correlation test [32, 33].

Sensitivity analysis

An influence analysis was conducted quantitatively by examining the sensitivity of heterogeneity metrics to exclusion of studies one at a time and in combination of two or more thereof. Accordingly, a sequential algorithm was used to identify which of the studies singly or in combination minimizes I2, and \(\tau^{2}\), and Higgins’ H statistics less than their threshold values for low, moderate, or high heterogeneity [32, 34]. The objective of this uncertainty analysis was to assess the robustness of conclusions concerning the size and the direction of adjusted effect estimates yielded by the meta-analysis [35].

Results

Selection process

A total of 2740 potentially relevant articles identified by the search of PubMed (522), Embase (372), web of Science (122), Ovid (468), Cochrane Library (258) and Science Direct (998). Duplicate articles (294) articles were excluded. Additionally, 2419 studies were excluded by screening the titles and/or abstracts. Remaining 27 articles were retrieved for full text review. From the reference lists of the retrieved articles, four relevant articles were identified for inclusion in the meta-analysis. Subsequently, after detailed evaluation,10 articles were excluded for various reasons (Fig. 1). Finally, remaining 21 studies (18 case–control and 3 cohort studies) were included in the meta-analysis.

Fig. 1
figure 1

Flow chart depicting the selection of case control and cohort studies for meta-analysis of association between chronic urinary tract infection and bladder carcinoma

Full size image

Study characteristics

Eighteen case–control [12,13,14,15,16,17,18,19,20, 22,23,24,25,26, 36,37,38,39], and three cohort studies [21, 40, 41] met the inclusion criteria. Of 18 case–control studies, 9 were published from USA [14,15,16,17, 20, 22, 25, 26, 37], 2 each from Germany [13, 18] and Italy [19, 23], and 1 each from UK [42], Denmark [24], Iran [38], Turkey [36], and Netherlands [12]. Seven of 18 case–control studies were pair-matched. The included 18 case–control studies widely varied in their sample sizes (cases/controls) from the smallest (173/173) to the largest (4915/21718). Only one study exclusively included female cases and controls, whereas, remaining 17 studies included both males and females among cases and controls. The proportion (%) of male cases ranged from zero to 90% in 18 case–control studies. Eight of 18 case–control studies presented the OR and its 95% CI, while remaining 10 studies presented RR and its 95% CI as a measure of association between chronic UTI and BCa. The number and type of adjusted confounding effects varied across the 18 case–control studies (Table 1a).

Three cohort studies included in this meta-analysis were reported from Sweden [40], Taiwan [21], and USA [41]. These three cohort studies reported a total of 2,606,179 person-years at risk and 1105 BCa cases developed during the follow-up period [21, 40, 41]. One of the cohort studies, retrospectively tracked the BCa patients (12,195; men = 9326; women = 2869), who reportedly either had hematuria (81.4% men; 60.5% women or UTI (18.6% men; 39.5% women) as a presenting claim at the index visit within one year (follow up of 1 year) preceding the diagnosis of BCa [41]. One cohort study presented the standardized incidence ratio [40]. while remaining two cohort studies presented adjusted hazard ratio (HR) along with 95% CIs as a measure of association between chronic UTI and BCa [21, 41]. Follow-up period in three cohort studies ranged from 1 to 25 years (Table 1b).

Association between chronic UTI and bladder carcinoma risk

Case–control and cohort studies

The random-effects model based summary estimate from 18 case–control studies showed that chronic UTI was significantly and independently associated with an increased BCa risk (ORRE = 2.33; 95%CI 1.86, 2.92) (Fig. 2). This meta-analysis of case–control studies detected a significant (p < 0.001) heterogeneity (I2 = 91.7%; 95% UI 88.3, 94.4; Higgins’ H = 3.6; 95% UI 3.0, 4.2). The random-effects model based pooled effect-estimate from three cohort studies showed that patients with chronic UTI had significantly increased BCa risk compared to individuals without chronic UTI (RRRE = 2.88; 95% CI 1.20, 6.89). The summary RRRE from three cohort studies had nearly the same magnitude as that of summary estimate ORRE sought from 18 case–control studies (Fig. 2). Furthermore, a statistically significant (p < 0.001) heterogeneity across the effect-estimates from three cohort studies was also detected (I2 = 96.2%, 95% UI 92.0, 98.2; Higgins’ H statistic = 7.3; 95% CI 4.8, 11.1).

Fig. 2
figure 2

Forest plot depicting the association between chronic urinary tract infection bladder carcinoma in case–control (above) and cohort (below) studies with summary odds ratio (OR) and summary hazard ratio (HR) estimates along with their 95% confidence interval (CI)

Full size image

Sub-group and random-effects meta-regression analyses

To evaluate the possible sources of between-study heterogeneity, 18 case–control studies on chronic UTI and BCa risk were further examined either by sub-group analyses (study location, study period, exposure definition) or by random-effects meta-regression analysis (sample size, proportion of male cases, NOS score). Sub-group analyses showed that the increased BCa risk associated with chronic UTI was independent of studies’ geographic location (p = 0.061), study period (p = 0.440), and exposure definition (p = 0.843) (Table 2). Additionally, no considerable evidence of modulating effect was found by study sample size (p = 0.421), proportion of male cases (p = 0.395) and NOS score (p = 0.583) on the association between chronic UTI and BCa risk.

Table 2 Sub-group analysis of association between chronic urinary tract infection and urinary bladder carcinoma risk by selected adjustment variables from case–control studies pooled in meta-analysis
Full size table

Publication bias

Visual examination of funnel plot from meta-analysis of 18 case–control studies showed little asymmetry. Quantitatively, no evidence of significant publication bias was observed using Egger’s regression asymmetry test (Egger’s p = 0.053) and Begg’s rank correlation test (Begg’s p = 0.192).

Sensitivity analysis

Sensitivity analysis by excluding the studies with the highest or the lowest effect estimates singly or in combination did not show a substantial reduction in value of I2 (approximate proportion of heterogeneity in adjusted summary ORRE for BCa). Furthermore, the direction and magnitude of adjusted summary ORRE did not change meaningfully (Table 3). Additionally, cumulative analyses of 18 case–control and three cohort studies showed no meaningful change in trend of reporting BCa risk over the studies’ period. Thus, all the studies considered in sensitivity analysis were retained in the final meta-analysis.

Table 3 Results of influence analysis with sequential algorithm of case–control studies included in the meta-analysis of association of chronic urinary tract infection and bladder carcinoma risk
Full size table

Discussion

Infection is one of the main contributors to cancer incidence. During 2008, 12.7 million new cancer cases occurred worldwide and of these about 2 million (16%) were attributable to infectious agents. This fraction was higher in less developed (22.9%) than developed (7.4%) regions of the world [43]. BCa is the 13th most common cause of cancer-related deaths worldwide [44]. Epidemiologic studies have inconsistent results regarding UTI as a risk factor for urothelial carcinoma (previously known as transitional cell carcinoma) of the bladder, which accounts for about 90% of the bladder cancers [45, 46]. Therefore, this study quantitatively assessed the magnitude of an overall association between the chronic UTI and BCa risk using the techniques of meta-analysis.

Main findings

The meta-analysis of eligible 18 case–control and 3 cohort studies provided robust evidence for an increased BCa risk among those with chronic/recurrent UTI compared to those without it. This seems to be the first study that has meta-analyzed the association between chronic/recurrent UTI and BCa risk. Therefore, the results from any related previous meta-analysis of case–control and/or cohort studies exclusively addressing this question were unavailable for relative appraisal. Chronic UTI and resultant cystitis have been implicated in the occurrence and progression of BCa [44, 47]. Globally bladder cancer cases attributable to infection have declined to 1.6% over the years [48]. Despite this decline, still a very high population risk of BCa is attributed to chronic UTI (41%; 95% CI 36, 48%) [44].

Several mechanistic pathways have been proposed by which infectious agents contribute in causation of BCa. Of infectious agents, E. coli is a predominant (70–80%) uropathogen. Additionally, evidence from an experimental study suggested that UTI with E. coli play a key additive and synergistic role during bladder carcinogenesis [49]. Because of recurrent bacterial UTI, chronic inflammation may lead to carcinogenesis and one of the key molecules that link chronic inflammation and cancer is represented by the NF-κB family of transcription factors [50, 51]. Furthermore, it has been argued that once the cystitis is initiated by infectious agents, it can mediate pathogenesis of BCa by stimulating cancer cell growth, invasion and metastasis through the recruitment of inflammatory cells and signaling molecules [52]. Cystitis may also increase absorption of the carcinogens. Furthermore, bacterial flora in the urine may contribute to the production of nitrites that are converted to carcinogenic nitrosamines [53,54,55]. Paradoxically, however, it has been suggested that the patients with recurrent UTI have reduced BCa risk, probably due to anti-cancer effect of antibiotic treatment for bladder infection, higher exposure to non-steroidal anti-inflammatory drugs of patients with history of bladder infection and/or immune response triggered by UTI [26]. Furthermore, it may well be that cystitis of infectious origin may merely represent a complication of cancer during its early growth before clinical diagnosis [17].

Adjustment for confounding

Although sub-group-specific risk estimates showed statistically non-significant differences for study location, study period and exposure definition, yet these differences reflected the existence of some residual confounding. For example, regarding the study location, sub-group-specific summary ORRE turned out to be relatively stronger for seven studies from Europe, whereas, this relationship was attenuated for nine studies conducted in USA than the overall summary ORRE for 18 studies together. Similarly, sub-group-specific summary ORRE was relatively stronger for studies published after 1990 than of those published during or before 1990. Additionally, differences among sub-group-specific estimates by exposure definitions were statistically non-significant, though studies wherein exposure was defined as chronic UTI suggested slightly stronger relationship with BCa. Random-effects meta-regression analyses showed statistically non-significant effect of proportion of male cases, sample size and NOS score on the adjusted summary ORRE from the main analysis.

Heterogeneity analysis

In this meta-analysis, a high between-study heterogeneity across the size of effect estimates was noted both for 18 case–control and 3 cohort studies on a previously described scale [56]. However, this heterogeneity could not be explained in sub-group analyses by study location, study period, exposure definition, sample size, proportion of male cases and/or NOS score. Nonetheless, this observed high between-study heterogeneity might be due to one or more of the life styles, occupational and nutritional factors that might be moderating the associations between chronic UTI and BCa risk across studied populations. Previously, disproportionate prevalence of risk factors such as cigarette smoking, environmental exposure to tobacco smoke, occupational exposures, quality of drinking water, dietary factors, including animal fats, processed meet, insufficient fruit and vegetables consumption, genetic polymorphism and use of drugs (e.g., phenacetin cyclophosphamide) with fluctuating strengths of their associations with BCa have been reported [57]. These known BCa risk factors have not been substantively considered in the studies included in this meta-analysis. Therefore, varying prevalence of these known factors might have interplayed with chronic UTI in increasing the BCa risk in studied populations and contributed to high between-study heterogeneity in this meta-analysis. Evaluation of such factors may be an impetus to future research.

Strengths and limitations of this study

This study seems to be the first one, which meta-analyzed the relationship between chronic UTI and BCa based on 18 case–control and three cohort studies. Furthermore, no significant publication bias could be detected. Additionally, the results were substantially robust on sub-group analyses, meta-regression and sensitivity analysis. Moreover, all the included case–control and cohort studies employed histopathologic evidence as a standard diagnostic tool for BCa.

This study has few limitations. First, there were only three cohort studies, therefore, temporal relationship between chronic UTI and BCa was assumed mainly based on analysis of case–control studies. However, the direction and strength of association between chronic UTI and BCa sought in meta-analyses of two groups of studies under two different study designs, conducted at various times and on different populations were consistent. Second; though the usage of varying exposure definition (i.e., chronic UTI) across the studies did not turn out to be a significant source of between-study heterogeneity, yet a consistent exposure definition and its assessment would have been reassuring about the observed association. Third, all the case–control studies did not report duration of chronic UTI before BCa diagnosis, therefore, median induction period for this exposure could not be ascertained. Therefore, future studies, may examine this variable in relation to BCa risk. Finally, the association between chronic UTI and BCa in majority of the case–control studies included in this meta-analysis came from sub-group analyses. Therefore, the statistical power of the studies to estimate the effect measure of interest could have been an issue. Future research on this question may consider this aspect.

In conclusion, the finding of this meta-analysis of 18 case–control and 3 cohort studies suggested a significantly increased BCa risk among patients with chronic infection of lower urinary tract/cystitis. Nonetheless, due to the presence of inevitable between-study heterogeneity and inconsistent patterns of adjusted confounding effects by the included studies, more data are needed to clarify the role of chronic UTI in causation of BCa and if established, prompt and effective treatment of UTI may minimize a substantial proportion of BCa risk.