The global burden of lower urinary tract symptoms suggestive of benign prostatic hyperplasia: A systematic review and meta-analysis

Benign prostatic hyperplasia is a common non-malignant condition among older men, but the epidemiology is poorly characterised. We summarised and determined the global prevalence of benign prostatic hyperplasia. A systematic search on PubMed, EMBASE and CENTRAL was performed up until 31st July 2016. Studies that described the epidemiology of benign prostatic hyperplasia were included and cumulative plots of prevalence estimates were calculated. A total of 31 prevalence rate estimates from 25 countries were identified. The combined prevalence estimates showed that the lifetime prevalence of BPH was 26.2% (95% CI: 22.8–29.6%). We found that there was an increasing prevalence of BPH with age. However, we found no significant difference between (a) rural, urban or mixed sites, (b) different countries, (c) respondent representativeness. (d) sample size or (e) study quality. We also found no significant change in the prevalence over the past 20 years. While there is substantial variation between sites estimates, results suggest that nearly 1 in 4 men will suffer from BPH over their lifetime. The study revealed there are significant gaps in knowledge, which provides opportunities for future research to further enrich the epidemiological landscape with data.

Epidemiology. In general, the prevalence of BPH increases with increasing age, with the highest prevalence in participants aged 70 and above. The median point prevalence was 25.2% and the 10% and 90% quartiles ranged from 19.0% to 37.9%. The highest prevalence of BPH was reported by Naslund and colleagues 31 who surveyed patients from their clinic from 6 US states in 2007 while the lowest prevalence was found in Da and colleagues in Shanghai, China 39 . Meta-analysis of 30 studies using a random effects model yielded a summary prevalence of 26.2% (16,437/76,246 individuals; 95% CI: 22.8-29.8%). However, a high level of heterogeneity was observed (I 2 = 99.2%, Q-value = 3493.89, τ = 0.01, p < 0.01). Serial exclusion of each study in the  Continued sensitivity analysis demonstrated that no individual study influenced the overall prevalence by more than 1% (Supplementary Table S2).
To provide a range of BPH prevalence estimates due to the methodologically diverse studies, estimates were stratified according to diagnostic criteria and study level characteristics. When evaluated by BPH diagnostic criteria, summary prevalence estimates ranged from 26.2% for studies that used only objective measurements (798/2,837 individuals, 95% CI, 22.8-29.6%, Q = 3.60, τ 2 = 0.001, I 2 = 44.5%) to 28.8% when using only subjective questionnaires such as the AUA-SI or the IPSS ( 29.4%]; Q = 0.85, P = 0.36), but this was not statistically significant. Similarly, no statistically significant difference in prevalence estimates were noted when stratified between rural, urban or mixed populations (Q = 0.58, p = 0.90), comparing respondent representativeness (Q = 0.04, p = 0.85), cohort sample size (Q = 0.01, p = 0.99) or study quality (Q = 0.22, p = 0.64). However, the prevalence rate was much lower in the study conducted among Africans compared to those conducted among Asian or Caucasians (Q = 101.34, p < 0.01). In the meta-regression analysis, none of the covariates analysed were significantly associated were associated with heterogeneity of prevalence estimates ( Table 2).
Age specific prevalence. Of the total 31 studies, only 25 studies reported age-specific stratified data for analysis. Grouped summary data showed that there was an increasing prevalence of LUTS/BPH with age, with a pooled prevalence of 14.8%, 20.0%, 29.1%, 36.8% and 38.4% for age groups of 40-49 years, 50-59 years, 60-69 years, 70-79 years and 80 years and above respectively (Fig. 2), but there was a high level of heterogeneity.

Sensitivity analyses. Sensitivity analyses showed that there was very little difference in prevalence estimates
when studies were excluded sequentially, or stratified by sample size and study quality. However, visual inspection of funnel plot was asymmetrical, suggesting that there was some evidence for bias due to small-study effects ( Supplementary Fig. S1).

Discussion
BPH is a common condition that affects millions of men worldwide. In this study, we systematically reviewed studies to estimate the prevalence of BPH. We found that the pooled prevalence was 26.2% (95% CI: 22.8-29.6%), with estimates differing across studies, because of different BPH definitions, survey methods, response options, geographical locations and sample populations. Despite this, we found that the prevalence of BPH increases as patient age increase, from 14.8% in younger males aged 40 to 36.8% in males aged 80 and above. In most traditional prevalence studies, estimates are usually obtained based upon the population living in a specified area. However, this introduces a selection bias to the study as these boundaries (e.g., cities, county or even nations) may be suboptimal for the detection of variation of disorder between or even within a specific population. Similarly, factors such as age or even migration patterns can influence these estimate. The prevalence of BPH is also thought to vary across difference ethnicities, urban/rural settings and countries. Some of these differences may be attributable to the methodological differences but these variation does exist even in a multinational study which had identical BPH definitions, survey and response format. We also found that there was considerable variability even within studies in the same country. These differences may result from many different sociocultural and environmental factors that can affect prostate health, in addition to genetic factors. Some potential reasons for the marked difference in estimated prevalence could be due to the methodological differences used by different studies. For example, in the study by Trueman and colleagues 25 , the presence of BPH was assessed using a self-administered questionnaire. In contrast, a study by Goh and associates 40 in Korea reported a prevalence rate of 20%. Participants were surveyed by trained investigators and prostate disease was assessed by a physical examination and measurement of prostate volume and serum prostate specific antigen. The current AUA guideline defines LUTS/BPH as the presence of voiding and/or storage symptoms 41 . The absence of a specific, universally accepted operational criterion has led to a diversity of definition and variance in incidence estimates. These methodological differences are in part, likely to account for some of the differences in findings between studies.
The definition of BPH has been problematic due to the variation in case definition used by different studies. In many older studies identified, BPH is commonly described as a chronic urinary symptom experienced by elderly men. However, some studies had defined BPH using radiographically determined prostate enlargement, while others had used the definition of decreased in urinary flow rates, urinary symptoms and in some cases physician-diagnosed BPH. However, the most commonly used measures by which BPH is diagnosed are the   AUA-SI and its internationally validated counterpart, the IPSS. This heterogeneity in BPH definition is thought to account for some differences in BPH prevalence rates.
The significant heterogeneity observed in this study led to a subgroup analysis as well as a meta-regression analyses to determine the potential sources. Effect estimates were similar when studies were grouped according to patient characteristics, suggesting that much of the heterogeneity remains unexplained. As with most meta-analysis of summary data, this study failed to identify the main source of heterogeneity. However, the meta-regression approach has several drawbacks. Firstly, it uses only the average value of a particular characteristic rather than individual values, thus decreasing the power to detect associations. Secondly, the meta-analysis also depended on the availability of published data, and more often than not, most of the methodology is incomplete.
This study has several strengths. To our knowledge, this is the first meta-analysis that attempts to examine and summarise the global prevalence of BPH. This systematic review conforms to the guidelines of Meta-Analysis of Observational Studies in Epidemiology (MOOSE) 42 recommendations. Another strength of this study include a comprehensive and broad search strategy, as well as relevance of finding to clinical practice and research.
This study has some limitations. Firstly, as mentioned, the differences in study methodology and population may have considerable effects on the results. These effect contribute to the substantial variability in reported BPH rates, and it is difficult, if not impossible to separate these effects from the true geographical, cultural, economic and psychosocial differences. Secondly, we also restricted our search to only articles published in English, and thus we may have missed some important data. We did not search "grey literature", as we felt that most of these data would not be sufficiently informative 43 . The current study could not take in consideration other risk factors associated with LUTS/BPH including diet, diabetes, or even body mass index which has substantially changed over the past 3 decades. Similarly, this study could not account for the variation in criteria of LUTS/BPH that has been revised. As such, inclusion of older studies may have led to an underestimation of the prevalence rates.
In summary, the current review provides a benchmark on the prevalence estimates for BPH. However, the wide range of prevalence estimates and case definition suggest that a standard criteria needs to be applied given the importance of understanding the prevalence of BPH and its implication on public health given the increasingly rapid growth of elderly worldwide. Additional research is needed in various areas especially on economic parameters.

Methods
Search strategy and selection criteria. We performed a literature search up until 31 July 2016 using a combination of search terms (Appendix 1) on the following database: Pubmed, EMBASE, Cinahl plus, AMED, CENTRAL and Web of Science for articles describing the prevalence of BPH among males. Keywords used include: prevalence, incidence, prostate enlargement, benign prostate hyperplasia, benign prostatic hypertrophy, bladder outlet obstruction and lower urinary tract symptoms. Two authors independently (SWHL & EMCC) reviewed the records to identify for potentially studies and full text of studies were retrieved if necessary. We also manually search bibliographies of included studies and any relevant review articles for additional references. In the event of multiple publications of identical data, the most informative version of the study was used. Any discrepancies were resolved by open discussion.
Definition. While the term BPH is correctly defined as histopathological hyperplastic changes in the prostate 41 , most studies and clinicians commonly use the term to describe a clinical syndrome that comprised of LUTS, prostatic enlargement and bladder outlet obstructions. In this study, we used the case definition for BPH as stated in the study. In the event that this was not stated, BPH was defined as the presence of moderate to severe LUTS, and used a cut off score of >7 for the American Urological Association Symptom Index (AUA-SI) 44 or the shorter version International Prostate Symptom Score (IPSS). Countries were regarded as industrialised if they fell within the high income definition as defined by the World Health Organisation.
Data extraction and assessment. Two authors separately extracted the studies using a standardised extraction template, including study level characteristics (e.g., urban/rural, study design, year study was conducted, definition of BPH and data collection methods) as well population characteristics (e.g., age-specific rates and ethnicity/race). Study authors were contacted for data clarification if necessary. The methodological quality of each study was judged using a modified version of the Newcastle-Ottawa Scale 45 , which includes 4 criteria namely, sample representativeness and size, comparability between respondents and non-respondents, ascertainment of BPH symptoms and statistical quality. Studies were judged to be low risk of bias if they had a minimum score of 3 points of the maximum 5 points.

Statistical analysis.
We conducted a meta-analysis of incidence data and pooled the estimates and 95% confidence intervals (CI) using the metaprop command developed by the Unit Cancer Epidemiology in Brussel 46 , and used the random-effects model since we expect the presence of heterogeneity. We subsequently conducted a subgroup analysis, and stratified the studies according to study geographic regions, number of cases of BPH, tools used to detect BPH as well as age groups as reported by the study. Potential small study publication bias was assessed using the Begg & Eggers test, as well as visual inspection of the funnel plot. Between studies heterogeneity was assessed using I 2 and Cochran's Q method. In the event of substantial heterogeneity, a random-effects meta-regression analysis was conducted to determine the effects of variables such as population demographics, study characteristics and indicators of error or bias on prevalence estimates. Any factor(s) which was significant in the univariate analysis were included into a multiple regression model. We also performed several sensitivity analyses to assess how our primary estimates differed when we excluded studies sequentially as well as studies with lower methodological quality such as those with poor sampling methods or sample sizes less than 1000 participants. We also assessed the possibility of change in prevalence patterns over time by examining prevalence rates across different study years. All analyses was conducted using Stata version 13.0 (StataCorp, College TX).