Bibliometric and Two-Sample Mendelian Randomization Analyses of the Causal Relationship Between Hypertension and Erectile Dysfunction

Summary

This study integrates bibliometric and two-sample Mendelian randomization analyses to systematically examine research trends and assess the causal relationship between hypertension and erectile dysfunction, identifying key contributors, emerging topics, and genetic evidence linking hypertension to increased erectile dysfunction risk.

Abstract

Hypertension (HT)-related erectile dysfunction (ED), a condition secondary to HT, is characterized by a persistent inability to achieve or maintain an erection sufficient for satisfactory sexual performance. This study aimed to systematically analyze and visualize publications related to HT-related ED using bibliometrics to identify key topics, research hotspots, and knowledge gaps, and to explore the underlying causal relationship between HT and ED using Mendelian randomization (MR). Literature was retrieved from the Web of Science Core Collection, and publishing trends were profiled by country, institution, author, journal, and collaboration networks. For the MR analysis, single-nucleotide polymorphisms considerably associated with HT were selected, followed by pruning, filtering, and adjustment for potential confounders. The inverse variance weighting method was used as the primary analysis, complemented by MR–Egger and weighted median approaches, along with sensitivity analyses including heterogeneity and pleiotropy assessments to ensure robustness. Between 2001 and 2025, 1,661 articles were published in 1,099 journals, reflecting the evolving global research status and future directions. Academic institutions in Europe and North America have played a dominant role. The most prolific country, institution, journal, and author are the United States, University of São Paulo, Journal of Sexual Medicine, and Faix, A., respectively. Common keywords include nitric oxide, metabolic syndrome, and endothelial dysfunction. MR results confirmed that HT has a significant positive causal effect on ED risk. Research hotspots primarily involve impotence, oral sildenafil, vardenafil, safety, inhaled nitric oxide, international index, predictor, late onset hypogonadism, testosterone, obesity, oxidative stress, and phosphodiesterase 5 inhibitor. Two-sample MR provides robust causal evidence that supplements observational findings, offering a comprehensive framework for understanding HT-related ED.

Introduction

Hypertension (HT) is the predominant risk factor driving cardiovascular disease (CVD) development and mortality¹. CVD leads to nearly 17 million deaths globally each year, of which approximately 55% are attributed to complications of HT^2,3. Erectile dysfunction (ED) is characterized by persistent inability to achieve or sustain a penile erection adequate for satisfactory sexual activity⁴. Surveys have shown that, among patients with HT, the prevalence of ED ranges from 36% to 45%, which is significantly higher than that in non-hypertensive populations⁵. High blood pressure and ED exhibit a clear correlation⁶. High blood pressure can negatively impact sexual function by reducing blood flow, potentially leading to decreased libido, difficulty with arousal, and impaired orgasm. This sexual dysfunction can also affect a patient’s adherence to HT treatment^7,8,9.

The term “bibliometrics” was first coined by Alan Pritchard in 1969¹⁰. As an analytical approach based on bibliometric indicators, it enables the quantitative evaluation of research performance within a specific field^11,12. The CiteSpace clustering software and the VOSviewer visualization tool are widely used for bibliometric analysis that facilitate the examination of research trends through clustering and mapping techniques, presenting findings as visual knowledge structures^13,14,15. Using these tools, the literature on HT-related ED can be systematically visualized and analyzed over recent decades. In this study, the Web of Science (WOS) database was selected as the data source, and the clustering software, visualization tool, and Microsoft Excel were used to identify and characterize developmental trends and emerging research hotspots in this field.

To further examine the potential causal association between HT and ED, Mendelian randomization (MR) analysis utilizing genome-wide association study (GWAS) data was conducted. This method applies genetic variants as instrumental variables (IVs) to emulate the conditions of a natural randomized controlled trial. Using a two-sample MR framework, the causal effect of HT on ED was systematically evaluated. The MR analysis provides insight into etiological mechanisms and potential early diagnostic value, while the bibliometric analysis identifies key contributors, evolving research frontiers, and future directions in this rapidly developing field. By integrating bibliometric mapping with genetic causal inference, the present study offers a dual-perspective framework: bibliometric analysis delineates the knowledge landscape, identifies research gaps, and generates hypotheses regarding the HT–ED relationship, while two-sample MR provides independent genetic evidence to formally test the causal hypothesis that emerges from the observational literature.

Protocol

This study used publicly available data from the WOS database (https://www.webofscience.com/wos/woscc/basic-search) and did not involve human participants; therefore, ethical approval was not required.

Data Sources and Collection
The Web of Science (WOS) database is a widely recognized and frequently used resource for scientific and bibliometric research. It includes data from approximately 9,000 high-impact journals and over 12,000 academic conference proceedings, offering a comprehensive representation of global research across scientific, technological, medical, and related fields^16,17,18. This database was selected for the present bibliometric analysis because of its rigorous journal selection process, consistent indexing of high-quality peer-reviewed literature, and its provision of comprehensive citation data—including cited references—which are essential for co-citation and bibliographic coupling analyses in tools such as CiteSpace and VOSviewer.

To minimize the potential impact of database updates on record consistency, all search and retrieval operations were completed within a single day (December 20, 2025). The Web of Science Core Collection was accessed via institutional subscription through Beijing University of Chinese Medicine using the standard Clarivate web interface. The search period covered publications from January 1, 2001, to December 20, 2025. The year 2001 was chosen as the starting point because it follows the landmark approval of sildenafil in 1998 and the subsequent establishment of oral phosphodiesterase type 5 inhibitors as first-line therapy for ED. This timeframe ensures comprehensive coverage of the modern research era while excluding earlier, less standardized studies.

The search was performed in the Web of Science Core Collection using the Topic (TS) field, which includes the title, abstract, author keywords, and Keywords Plus. All citation indexes within the Core Collection (i.e., SCI-Expanded, SSCI, A&HCI, CPCI-S, CPCI-SSH, ESCI, CCR-Expanded, and IC) were included without any restrictions. All Clarivate default search settings were retained without modification, including any automatic term mapping or expansion settings. The complete search query was defined as: TS = (hypertension AND (impotence OR “erectile dysfunction”)). Only documents classified as “Article” or “Review” were included. The retrieved records were exported in plain text format, with “Full Record and Cited References” selected as the output content. Due to the Web of Science platform’s export limit of 500 records per batch, the 1,661 records were exported in four batches. Batch 1 contained records 1–500, batch 2 contained records 501–1000, batch 3 contained records 1001–1500, and batch 4 contained records 1501–1661. Prior to analysis, the four plain text files were merged into a single dataset using the deduplication and merging function in CiteSpace (version R6.1.3) (Data→Import/Export→Remove Duplicates), thereby removing any potential duplicate records while preserving the complete dataset for subsequent bibliometric analysis.

Bibliometric Analysis and Software
Bibliometric analysis was conducted using a combination of specialized visualization and statistical tools.

CiteSpace (version R6.1.3) is a Java-based application widely used to visualize and analyze trends and patterns in scientific literature¹⁹. It was developed by Dr. Chen Chaomei in 2004²⁰. The software was operated on a Microsoft Windows 10 (64-bit) system with Java Runtime Environment version 8. Built on principles of scientometrics, data analysis, and information visualization, it reveals knowledge structures by examining patterns, distributions, and relationships within the literature. In this study, CiteSpace was applied for keyword clustering and burst detection. The time slicing was set to one year per slice across the 2001–2025 period. For network construction and pruning, the g-index was used with a scaling factor of k = 25. Keyword clustering employed the log-likelihood ratio algorithm to generate cluster labels. To identify research hotspots, citation burst detection was conducted with the number of states set to 2, a default ratio of a₁/a₀ = 2.0, and a minimum burst duration of 2 years. The gamma parameter was defined as 1.04 for keyword burst detection and 0.97 for reference burst detection.

VOSviewer (version 1.6.18) is a bibliometric analysis tool designed for knowledge mapping and visualization²¹. It supports various analytical approaches, including literature analysis, co-occurrence analysis, and bibliographic coupling. In the present study, VOSviewer was used to generate visual representations of countries/regions, authors, institutions, cited journals, and keywords, as well as to produce density maps. For these visualizations, minimum occurrence thresholds were applied to maintain clarity in the graphical representation. Specifically, a threshold of at least 1 publication was set for the analysis of countries/regions and authors, a threshold of at least 4 publications was set for institutions, a threshold of at least 2 publications was set for journals, and a threshold of at least 6 occurrences was set for keyword co-occurrence analysis. For co-citation analyses (cited journals, co-cited authors, and co-cited references), a minimum threshold of 10 citations was applied. Default association strength normalization was used for all network constructions. For all VOSviewer analyses in this study, including co-authorship networks (countries/regions, institutions, authors), journal co-citation analysis, and keyword co-occurrence analysis, the full counting method was applied.

This study aimed to describe key characteristics of the literature, including countries/regions, institutions, journals, highly cited articles, co-citation networks, and frequently occurring keywords. In addition to noun phrases extracted from titles and abstracts, keywords provided in publications were also analyzed to identify trends in keyword occurrence and citation patterns. Keywords with a minimum occurrence threshold of 6 were included in the co-occurrence and clustering analyses. No manual removal of keywords was performed to avoid introducing subjective bias; all terms meeting the occurrence threshold were retained for objective analysis. All analytical procedures were independently verified by two researchers to ensure the accuracy and reproducibility of the network layouts and citation metrics.

MR Analysis
A two-sample MR method was used to evaluate the potential causal link between HT and ED. This approach is founded on three principal assumptions: (1) the genetic instruments are significantly associated with the exposure (relevance); (2) the instruments are not related to confounding variables influencing the exposure–outcome association (independence); and (3) the instruments affect the outcome solely through the exposure (exclusion restriction).

GWAS summary statistics were obtained from the IEU OpenGWAS platform (https://gwas.mrcieu.ac.uk/; accessed on December 20, 2025). No additional local database snapshot was generated; reproducibility is supported by the use of stable GWAS identifiers, FinnGen release versioning, and archived analysis code. Genetic instruments for HT were obtained from the FinnGen dataset (FinnGen Biobank, release 5; finn-b-I9_HYPTENS_EXNONE), defined as hypertensive diseases excluding secondary HT, including 55,917 cases and 162,837 controls of European (Finnish) ancestry. Outcome data for ED were obtained from the FinnGen dataset (release 5; finn-b-ERECTILE_DYSFUNCTION), comprising 1,154 cases and 94,024 controls of European ancestry. Single-nucleotide polymorphisms (SNPs) significantly associated with HT at the genome-wide threshold (P < 5 × 10⁻⁸) were first selected as candidate instrumental variables. Linkage disequilibrium (LD) clumping was then applied to exclude correlated variants, using a threshold of r² < 0.001 and a clumping window of 10,000 kb. To reduce the possibility of weak-instrument bias, the F-statistic was calculated for each retained SNP using the Wald ratio formula: F = (β_exposure / SE_exposure)², where β_exposure and SE_exposure represent the SNP–HT association estimate and its standard error, respectively. Only variants with F > 10 were included in the final MR analysis.

To minimize confounding effects, the retained instrumental variables were evaluated using LDtrait (LDlink) for reported associations with smoking and alcohol consumption; no instruments required exclusion on this basis (Supplementary Table 1). Harmonization of exposure and outcome data was conducted using the harmonise_data() function in the TwoSampleMR package (version 0.6.2), which aligned effect alleles across datasets, inferred strand orientation for palindromic SNPs based on allele frequency (minor allele frequency < 0.3), and excluded palindromic SNPs with ambiguous alignment.

The inverse variance weighting (IVW) method was employed as the primary analytical approach, while MR–Egger, weighted median, and mode-based methods were used as supplementary analyses. The main IVW estimate was calculated using the multiplicative random-effects model implemented in the mr_ivw() function of the TwoSampleMR package with default parameters. Effect sizes were expressed as odds ratios (ORs) with 95% confidence intervals (CIs) following exponentiation of the beta coefficients. MR-PRESSO analysis was conducted to assess horizontal pleiotropy through the global test and to identify potential outliers; the distortion test was applied when outliers were detected. A significance threshold of P < 0.05 was used. No SNPs were excluded prior to the final analysis, as no outliers were identified by MR-PRESSO. The results were visualized using forest plots, funnel plots, scatter plots, and leave-one-out analyses. The analytical code is publicly available at GitHub, release v1.0: https://github.com/tengfeitcm/Two-Sample-Mendelian-Randomization-/releases/tag/v1.0, ensuring reproducibility of the exact analysis version used in this study.

Results

Bibliometric Analysis Results
Time trends in publications and citations:
Annual publication output serves as a key indicator of research development and, to some extent, reflects the progression of knowledge within a field. As of December 20, 2025, a total of 1,661 publications related to HT-associated ED were identified in the Web of Science database (Figure 1). The yearly distribution of published articles is presented in Figure 2.

Discussion

By providing a comprehensive and systematic overview of the research topics, trends, and global landscape of HT-related ED, this study offers a rapid and preliminary understanding of the field. In the context of the big data era, it is increasingly important for researchers to recognize the developmental trajectory of their respective research areas. Compared with systematic reviews or meta-analyses, bibliometric analysis applies specialized visualization tools to comprehensively evaluate existing literature, allowing in...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
CiteSpace	Drexel University, USA	R6.1.3	RRID:SCR_025121
CMplot	Yin L, et al. rMVP: a memory-efficient, visualization-enhanced, and parallel-accelerated tool for genome-wide association study. Genom Proteom Bioinform. 2021;19(4):619-28.	4.5.1	NA
FinnGen dataset	FinnGen Consortium	Release 5	RRID:SCR_022254
gwasglue	MRCIEU, University of Bristol	0.0.0.9000	NA
IEU OpenGWAS platform	MRCIEU, University of Bristol	NA	NA
LDlink (LDtrait)	National Cancer Institute, NIH, USA	NA	RRID:SCR_011403
Microsoft Excel	Microsoft Corporation, USA	2021	RRID:SCR_016137
MR-PRESSO	Verbanck et al.	1	NA
R software	R Foundation for Statistical Computing, Austria	4.4.0	RRID:SCR_001905
TwoSampleMR	MRCIEU, University of Bristol	0.6.3	RRID:SCR_019010
VariantAnnotation	Bioconductor	1.50.0	RRID:SCR_000074
VOSviewer	Leiden University, Netherlands	1.6.18	RRID:SCR_023516
Web of Science Core Collection	Clarivate Analytics, USA	Accessed on December 20, 2025	RRID:SCR_005051