Shared Inherited Genetics of Benign Prostatic Hyperplasia and Prostate Cancer

1. Introduction

Benign prostatic hyperplasia (BPH) and prostate cancer (PCa) are two of the most common diseases in men, and the incidence for both conditions increases considerably with age [1], [2]. These two pathological processes negatively impact quality of life and result in considerable healthcare expense. BPH is a histological diagnosis characterized by a proliferation of both stromal and epithelial cells in the transitional zone of the prostate [3]. This proliferation can lead to bladder outlet obstruction and subsequent lower urinary tract symptoms (LUTS). PCa, on the contrary, is a malignant adenocarcinoma primarily found in the peripheral zone of the prostate and, prior to metastasis, is typically asymptomatic, being primarily detected by prostate-specific antigen (PSA) screening [4].

Despite the major differences in cellular growth patterns and stereotyped locations within the prostate, a link between BPH and PCa has been hypothesized, studied, and reported [5], [6]. Their co-occurrence was first documented from autopsy studies in 1957 and 1992 [7], [8]. Since then, conflicting results were reported in many studies [9], [10], [11], [12], [13], [14], [15]. To date, no consensus has been reached on their association and causal relationship [6]. Consequently, the current National Cancer Institute website states that “BPH is not linked to cancer and does not increase your risk of getting PCa” [16].

A major cause for the inconclusive findings is the inherent detection bias of observational studies in diagnosing these two diseases. Patients treated by urologists for one of these diseases are more likely to be examined thoroughly and are therefore more likely to be diagnosed for the other disease [17]. This detection bias is particularly prominent because the likelihood of diagnosing BPH and PCa increases with heightened prostate examinations and diagnostic evaluations that include PSA measurements [18], [19]. Conversely, patients diagnosed with PCa on initial evaluation who are subsequently treated for PCa may never be given a diagnosis of symptomatic BPH, even though histological BPH may be present in the transition zone at the time of prostatectomy. Furthermore, patients treated with androgen deprivation therapy and/or radiation are unlikely to be given a diagnosis of BPH as the cause of any subsequent LUTS.

Inherited single nucleotide polymorphisms (SNPs) offer an alternative approach to assess the association between BPH and PCa. Genome-wide SNPs can be used to estimate the polygenic heritability of each disease and the genetic correlation between the two diseases [20]. This alternative approach does not directly test phenotypic co-occurrence of the two diseases. Instead, it tests the correlation between heritability of individual diseases, which is estimated only from the respective diseases and is therefore not susceptible to the detection bias. Furthermore, established risk-associated SNPs for each disease make it possible to perform a cross-disease association [21], [22].

The primary hypothesis of this study is that BPH and PCa are linked, and that this association is partially contributed by shared inherited genetics via the same genes (causal or pleotropic effects) and/or different genes in linkage disequilibrium (LD). This hypothesis was tested in a large population-based cohort.

2. Patients and methods

3. Results

Among 214 717 White men in the UKB, 24 623 (11%) and 14 311 (6.7%) had a diagnosis of BPH and PCa, respectively (Table 1). Diagnoses of these two diseases were significantly correlated (χ² = 1862.80, p < 0.001). Specifically, 3231 (1.5%) men had a diagnosis of both BPH and PCa, which was significantly higher than the expected number of 1332 (0.62%) men, assuming that the diagnosis of these diseases were independent (χ² = 797.96, p < 0.001). Having a diagnosis of PCa was associated with an OR (95% CI) of 1.58 (1.51–1.65) for also having a diagnosis of BPH (p < 0.001). Conversely, having a diagnosis of BPH was associated with an OR (95% CI) of 1.57 (1.50–1.64) for PCa diagnosis (p < 0.001). These ORs were estimated adjusting for age at recruitment and genetic background.

Genetic susceptibility to BPH and PCa was estimated based on polymorphic SNPs across the genome (minor allele frequency >1%). SNP-based heritability or h² (95% CI) was 0.09 (0.07–0.11) for BPH (p < 0.001) and 0.16 (0.12–0.20) for PCa (p < 0.001; Fig. 1 and Supplementary Fig. 1). Furthermore, a significant and positive genetic correlation between the two diseases was found (r_g [95% CI]: 0.16 [0.03–0.28], p = 0.01).

When examining established risk-associated SNPs for BPH and PCa, significant cross-disease genetic associations were found. Among the 250 risk-associated SNPs of PCa or BPH from previous GWAS studies, 49 were significantly associated with the risk of the other disease at p < 0.05, which was significantly more than expected by chance (N = 12, p < 0.001 [χ² test]). Specifically, among the 239 established independent PCa risk–associated SNPs, 43 were associated with BPH diagnosis at p < 0.05 (Fig. 1 and Supplementary Table 1). Reciprocally, among the 15 established independent BPH risk–associated SNPs, ten were associated with PCa diagnosis at p < 0.05. Four established GWAS-significant risk-associated SNPs of PCa and BPH overlapped.

We also evaluated the cumulative effect of SNPs on disease risk. In addition to highly significant associations between disease-specific GRS and their respective disease risks, significant cross-disease associations were also found (Table 2). GRS based on the 15 established BPH risk–associated SNPs was positively associated with PCa diagnosis (OR = 1.26, p < 0.001). Similarly, GRS based on the 239 established PCa risk–associated SNPs (GRS_PCa) was significantly associated with BPH diagnosis (OR = 1.03, p < 0.001).

Furthermore, GRS was also associated with specific phenotypes of BPH and PCa (Table 2). Notably, GRS_BPH was inversely associated with lethal PCa risk when comparing lethal PCa cases versus nonlethal PCa cases (any PCa patients who did not die from the disease, OR = 0.58, p = 0.002). Upon examining the results by GRS_BPH deciles (Supplementary Table 2), while no trend of lethal PCa prevalence with GRS_BPH deciles was found (p = 0.1), the prevalence of lethal PCa in men of the highest GRS_BPH decile was 0.23%, noticeably lower than that of the remaining deciles (0.33%, p = 0.02). In contrast, the prevalence of nonlethal PCa increased slightly with higher GRS_BPH deciles (p_trend = 0.01), with the highest rate (6.6%) found in men of the top GRS_BPH decile. As for GRS_PCa, no association with lethal PCa was found in a case-case analysis (OR = 0.99, p = 0.8). The prevalence of both lethal and nonlethal PCa increased by each decile at a similar scale (Supplementary Table 2).

In light of the genetic correlation and cross-disease association of GRS between these two diseases, we next explored the clinical utility of using two GRS values for predicting the diagnosis of BPH, PCa, and both diseases (Table 3 and Fig. 2). For example, using a GRS of 1.5 as a cutoff value (ie, a 1.5-fold increased risk over the general population), men with both high GRS_BPH and high GRS_PCa had a considerably higher prevalence of these two diseases than men with both low GRS_BPH and low GRS_PCa (16% vs 5.1% for PCa, 17% vs 11% for BPH, and 3.8% vs 1.2% for both BPH and PCa). Similarly, men with either high GRS_BPH or high GRS_PCa had a higher prevalence of these two diseases. Men with low GRS_BPH and high GRS_PCa had a higher prevalence of lethal PCa (0.68%) than those with both low GRS_BPH and low GRS_PCa (0.25%) and especially those with high GRS_BPH but low GRS_PCa (0.11%). Regarding surgical intervention for BPH, men with high GRS_BPH underwent TURP more frequently than those with low GRS_BPH, regardless of GRS_PCa status (6.5–7.4% for high GRS_BPH vs 3.3–3.6% for low GRS_BPH).

4. Discussion

The primary goal of this study is to address the longstanding controversy surrounding the association between BPH and PCa [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15]. This clinically important question has been complicated by an inherent detection bias in epidemiological studies where patients diagnosed with one disease are typically examined more thoroughly by urologists and therefore have a higher chance of being diagnosed with the other disease [17]. It is practically impossible to estimate the degree (partially or total) of the detection bias contributing to the observed association of BPH and PCa in traditional studies. Utilizing a large population-based cohort of over 200 000 men with diagnostic information for both diseases and genome-wide SNP data, we tested the association using both traditional epidemiological and alternative inherited genetic approaches (Fig. 1). We (1) demonstrated a statistical association of phenotypic diagnoses of BPH and PCa (p < 0.001) and (2) revealed a polygenic inherited basis for each of these two diseases (h² of 0.09 and 0.16 for BPH and PCa, respectively), and (3) more importantly, the present analysis suggests that these two diseases are genetically correlated and that they share part of a polygenic background (r_g = 0.16) and risk-associated SNPs. These findings provide strong statistical evidence that diagnoses of BPH and PCa are linked, and the excess concurrence of these two diseases was in part contributed by inherited genetics and not entirely driven by the detection bias.

A shared inherited risk between BPH and PCa may arise from several potential sources, including (1) LD of different genes for BPH and PCa in the same chromosomal region; (2) pleiotropy where same genes or variants affect both BPH and PCa; (3) causal effect where genes cause a disease first, which in turn cause the other disease; and (4) biases from population stratification. While the likelihood for the last source is low because this study was based on White men from a population-based cohort and the analysis adjusted for population stratification, we cannot differentiate between the sources of LD, pleiotropy, and causality. Therefore, we can only conclude that the shared inherited risk is directly or indirectly associated with both BPH and PCa, but does not necessarily cause these two diseases directly.

Although a Mendelian randomization (MR) analysis is a well-established method to interrogate the causal effect of an exposure to an outcome [25], the validity of key assumptions underlying an MR analysis (independent and exclusion) in our study is difficult to justify [26]. The independent assumption states that there are no unmeasured confounders of the associations between genetic variants and outcome. The exclusion restriction assumption requires that the genetic variants affect the outcome only through their effect on the risk factor of interest. Therefore, it is difficult to interpret MR results even if these are statistically significant. However, as an exploratory effort, we performed two MR analyses to assess a possible causal relationship between BPH and PCa using 32 BPH risk–associated SNPs, and between PCa and BPH using 100 PCa risk–associated SNPs (Supplementary Table 3). The causal association was first tested based on a one-sample MR analysis using the inverse-variance weighted method. Although a significant causal association was found in both MR analyses, significant heterogeneity was also found in both analyses. Results from additional robust methods for sensitivity analysis (weighted median and MR-Egger) were inconsistent. Therefore, the interpretation of the causal relationship between the two diseases via MR remains inconclusive.

Furthermore, as a comparison, we performed a similar analysis to assess the association and genetic correlation between BPH and another urological cancer (bladder cancer). While the diagnoses of these two diseases were highly correlated (χ² = 1615.00, p < 0.001), there was no significant genetic correlation between the two diseases (r_g = 0.01, p = 0.93). These results suggest that unlike BPH and PCa, where the observed link is partially explained by shared inherited risk, there is no evidence that BPH and bladder cancer share an inherited genetic risk; instead, the observed link between BPH and bladder cancer may be caused largely or completely by the detection bias.

The finding of a shared inherited risk between BPH and PCa from this study may have clinical utility and implications for understanding the etiology. For example, identifying commonality of genes in the chromosomal regions that are associated with both BPH and PCa may help better understand the etiology for these two diseases. Toward this effort, we performed a preliminary pathway analysis for the 65 nearest genes in the 51 independent regions associated with both BPH and PCa using the Kyoto Encyclopedia of Genes and Genomes (Supplementary Table 4). These genes are significantly enriched in only one biological pathway, the PCa pathway (p = 0.007, Benjamini correction; Supplementary Table 5). A major caveat of this analysis is that the nearest genes may or may not account for the observed genetic associations, a common challenge for understanding the biological mechanisms of GWAS findings.

The genetic correlation between the two diseases, especially the cross-disease genetic association, also suggests that GRS_BPH and GRS_PCa may be used in the clinic to stratify the risk for these two diseases. For example, men with either high GRS_BPH or GRS_PCa, or both have higher risks for these two diseases, alone or both, and men with high GRS_BPH are more likely to undergo surgical intervention. Men identified to have a high risk for PCa may benefit from more aggressive screening, while men at a high risk of BPH may warrant earlier referral to urological subspecialists and consideration of early intervention for LUTS with surgery or minimally invasive procedures [27], [28], [29], [30].

While confirming the previous null result that GRS_PCa does not differentiate aggressiveness of the disease in PCa patients [21], [31], [32], we obtained a novel finding that GRS_BPH is inversely associated with lethal PCa in a case-case analysis. Specifically, men with the highest GRS_BPH decile have a lower risk for lethal PCa but a similar risk for nonlethal PCa when comparing with men in the remaining deciles. This finding is important and plausible considering that (1) GRS_BPH is positively associated with prostate volume [22] and (2) prostate volume is inversely associated with aggressive PCa [33]. However, additional studies are needed to confirm this novel finding before exploring its clinical utility. Furthermore, more studies are warranted to understand specific SNPs and genes underlying this association and their biological effects on lethal PCa. Large PCa patient cohorts with germline data, PSA measurements, magnetic resonance imaging findings, prostate volume, Gleason grade at the time of biopsy and prostatectomy, location of tumor and/or histological BPH in peripheral and transition zones, as well as long-term disease follow-up are crucial to confirm and understand the association.

Several limitations of this study are noted. First, because ∼95% of patients in the UKB are White, our analyses and conclusions were limited to this group only. The generalizability of our finding in other ethnic and racial populations needs to be evaluated. Second, it is recognized that the UKB cohort was included in a previous GWAS that identified BPH risk–associated SNPs [22]. Although this may inflate the performance of GRS_BPH for predicting BPH risk, it has limited impact on the key findings of genetic correlation. Third, the lack of detailed clinical variables relevant to BPH and PCa in this population-based biobank, such as PSA, prostate volume, Gleason grade, and quantification of LUTS, hinders more granular analysis and understanding of clinical and genetic associations. Some findings from this study, especially the inverse association between GRS_BPH and lethal PCa, should be considered preliminary. A comprehensive analysis in large and clinically well-characterized urological patient cohorts is needed. Finally, we did not perform a functional analysis to investigate the biological mechanism of the genetic associations for both BPH and PCa. This is in part due to the nature of this genetic study where the primary goal is to identify chromosomal regions with statistical evidence for association. More importantly, we recognize the substantial challenges in these functional studies. The chromosomal region for each genetic association is typically large (>500 kb) and many of these regions are outside of coding genes. Nevertheless, we feel that genetic association studies provide critical data relevant to human diseases. Geneticists and biologists can collaborate to understand biological mechanisms of genetic associations.

5. Conclusions

In conclusion, utilizing genome-wide SNP data from a large population-based cohort, we demonstrated that BPH and PCa share common polygenic inherited risk. This novel genetic result suggests that the excess of co-occurrence of these two diseases is not completely driven by the detection bias. The current conclusion that BPH and PCa are not related, as stated by the National Cancer Institute and other authoritative agencies, may be reconsidered.



Author contributions: Jianfeng Xu had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Xu, Helfand, Glaser.

Acquisition of data: Xu.

Analysis and interpretation of data: Shi, Wei, Lanman, Xu, Helfand, Glaser.

Drafting of the manuscript: Xu, Helfand, Glaser.

Critical revision of the manuscript for important intellectual content: Glaser, Shi, Wei, Lanman, Ladson-Gary, Vickman, Franco, Crawford, Zheng, Hayward, Isaacs, Helfand, Xu.

Statistical analysis: Shi, Xu.

Obtaining funding: None.

Administrative, technical, or material support: None.

Supervision: Xu, Helfand.

Other: None.



Financial disclosures: Jianfeng Xu certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: NorthShore affiliated authors—NorthShore University HealthSystem has an agreement with GoPath Laboratories for genetic tests of polygenic risk score. Brian Helfand discloses a relationship with GoPath Laboratories in which he is a consultant and speaker, and relationship with Ambry Genetics in which he is a speaker.



Funding/Support and role of the sponsor: None.



Ethics statement: The UKB was approved by North West – Haydock Research Ethics Committee (REC reference: 16/NW/0274; IRAS project ID: 200778). Data from the UK Biobank was accessed through a Material Transfer Agreement under application reference number 50295. This study was performed in accordance with the Declaration of Helsinki.



Data sharing: The data used in this study are available in the UK Biobank, a publicly available repository. Data were accessed through a Material Transfer Agreement under application reference number 50295. For additional information, please feel free to contact the corresponding author Jianfeng Xu, DrPH.



Acknowledgments: We are grateful to the Ellrodt-Schweighauser family for establishing Endowed Chair of Cancer Genomic Research (Xu), and Chez and Melman families for establishing Endowed Chairs of Personalized Prostate Cancer Care (Helfand), as well as the Rob Brooks Fund for Personalized Prostate Cancer Care at NorthShore University HealthSystem.

Associate Editor: Véronique Phé

Appendix A. Supplementary data

The following are the Supplementary data to this article: