A drug target for erectile dysfunction to help improve fertility, sexual activity, and wellbeing: mendelian randomisation study

Abstract Objective To investigate the association of genetically proxied (using a surrogate biomarker) inhibition of phosphodiesterase 5 (PDE5), an established drug target for erectile dysfunction, with fertility, sexual behaviour, and subjective wellbeing. Design Two sample cis-mendelian randomisation study. Setting Summary data on genetic associations obtained from the International Consortium for Blood Pressure and UK Biobank. Participants Individuals of European ancestry from the International Consortium for Blood Pressure (n=757 601) for estimating PDE5 inhibition (using the surrogate biomarker of diastolic blood pressure reduction), and UK Biobank (n=211 840) for estimating the fertility, sexual behaviour, and subjective wellbeing outcomes in male participants. Intervention Genetically proxied PDE5 inhibition. Main outcome measures Number of children fathered, number of sexual partners, probability of never having had sexual intercourse, and subjective wellbeing. Results Genetically proxied PDE5 inhibition was associated with male participants having 0.28 (95% confidence interval 0.16 to 0.39) more children (false discovery rate corrected P<0.001). This association was not identified in female participants. No evidence was found of an association between genetically proxied PDE5 inhibition and number of sexual partners, probability of never having had sexual intercourse, or self-reported wellbeing in male participants. Conclusions The findings of this study provide genetic support for PDE5 inhibition potentially increasing the number of children fathered by male individuals. Absence of this association in female participants supports increased propensity for sustained and robust penile erections as a potential underlying mechanism. Further studies are required to confirm this, however, and these findings should not promote indiscriminate use of PDE5 inhibitors, which can also have harmful adverse effects.


UK Biobank self-reported data
In the UK Biobank (UKB), information on wellbeing (UKB identifier, ID 4526) was ascertained through a single question asked in follow-up in 2009: "In general how happy are you?", and then to choose either: "Extremely happy", "very happy", "Moderately happy", "Moderately unhappy", "Very unhappy", "extremely unhappy", "Do not know", or "Prefer not to answer".Information on number of sexual partners (UKB ID 2149) was ascertained through a single question asked at recruitment: "About how many sexual partners have you had in your lifetime?".The odds of being a virgin was estimated by recoding this question into a binary indicator so that people who had had one or more sexual partner were coded as 1 and leaving those who had had none as coded as zero.

Data sources used for instrument selection and validation
Expression quantitative trait loci (eQTL) data were taken from the 2018 eQTLGen Consortium genome-wide association study (GWAS) (OpenGWAS ID: eqtl-a-ENSG00000138735) of whole blood PDE5A expression (1).This study measured gene expression in 31,684 male and female participants of European ancestry.
Single-nucleotide (SNP) genetic associations with erectile dysfunction were estimated as the inverse-variance weighted meta-analysis of SNP effects form two GWASs.First, the Bovijn et al (2018) GWAS of erectile dysfunction (OpenGWAS ID: ebi-a-GCST006956).This GWAS had 6,175 European cases and 217,630 European controls (2).Second, we used the FinnGen round 8 GWAS of erectile dysfunction (OpenGWAS ID: finn-b-ERECTILE_DYSFUNCTION).This GWAS had 2,038 medical record inferred cases, and 157,478 controls.Finngen is a population cohort study of male and females of Finnish ancestry individuals living in Finland (3).Information of pulmonary arterial hypertension was extracted from the FinnGen round 8 GWAS of this trait in the OpenGWAS project (OpenGWAS ID: finn-b-I9_HYPTENSPUL).This GWAS had 213 medical record inferred cases, and 355,864 controls.

Genome-wide association study in females
The female only GWASs were conducted using the same methods as the male only GWASs.
To estimate variant-outcome associations in females, we used the female subset of sex-stratified GWASs for subjective wellbeing (N=89,815), number of sexual partners: (N=235,926), and the odds of being a virgin: 251,078 females).The estimate corresponding to fertility in females was the number of children they had birthed.This GWAS was performed using UKB data (OpenGWAS ID: ukb-b-1209, N=250,782) (4).

Further details on Mendelian randomization
Mendelian randomisation (MR) is a type of instrumental variable analysis that makes three core assumptions: 1) that the instrument (in this context a genetic variant) is strongly associated with the exposure, 2) that the instrument causes the outcome only via the exposure, and 3) that there is no instrument-outcome confounding.The F statistic of a variant describes the strength of its association with a trait.Bias due to violations of the first assumption of MR is inversely proportional to the F statistic for the variant-exposure association.We therefore evaluate this assumption by calculating the F statistic as the square of the variant-exposure association divided by the square of the standard error of this association.The second two assumptions cannot be proven empirically.In a cis-MR setting, which considers genetic variants at the gene for the protein being studied, the second assumption is more plausible since the exposure of interest, PDE5 inhibition, is very proximal to the gene, and there are therefore likely fewer pathways through which a pleiotropic effect could violate this assumption.

Two-step cis-Mendelian randomization
The product of coefficients methods of conducting a mediation analysis states that that the association between an exposure and an outcome mediated by a third variable is the association of the exposure with the third variable multiplied by the association of the third variable with the outcome.The difference of coefficients method for conducting a mediation analysis state that the mediated effect can be estimated as the total effect of the exposure on the outcome minus the direct effect of the exposure on the outcome (i.e., the effect not mediated by the third variable).It follows, that we can estimate the direct effect by subtracting the mediated effect from the total effect.Twostep cis MR leverages this to adjust variant-outcome estimates for potential sources of bias, by treating the source of bias as a mediator (5).For example, if the variant causes the outcome due to its association with a trait not related to the exposure, this effect can be accounted for by: 1) estimating the variant-trait association (from a GWAS of the trait), 2) estimating the trait-outcome association (using MR), 3) estimating the variant-outcome association mediated by the trait by multiplying the estimated in 1) by the estimate in 2), and 4) adjusting the variant-outcome GWAS summary statistics by subtracting from them the estimates in 3).
We performed two-step cis-MR to adjust our main MR estimates for potentially confounding traits that associate with the variants employed as instruments.The 95% confidence intervals in both steps of the two-step cis-MR were estimated using bootstrap standard errors with 100,000 repetitions.'ukb-b-7376'), and white blood cell count (OpenGWAS ID: ieu-b-30) were extracted from existing UKB GWASs (4,6).Myeliod white cell count, granulocyte count, and sum basophil and neutrophil counts were extracted from the Astle et al (2016) GWAS of that trait (OpenGWAS ID: ebi-a-GCST004626, ebi-a-GCST004614, and ebi-a-GCST004620 respectively) (7).This GWAS had approximately 170,00 male and female participants, mostly recruited form UK Biobank subsamples.Coronary artery disease summary data were taken from the van der Harst et al (2017) GWAS (OpenGWAS ID: ebi-a-GCST005195).This had 122,733 cases and 424,528 controls (male and female, of European ancestry) recruited from the UKB and CARDIoGRAMplusC4D (8).

Assessment of assumptions
Describe any methods or prior knowledge used to assess the assumptions or justify their validity

Sensitivity analyses and additional analyses
Describe any sensitivity analyses or additional analyses performed (e.g.comparison of effect estimates from different approaches, independent replication, bias analytic techniques, validation of instruments, simulations)

Data and data sharing
Provide the data used to perform all analyses or report where and how the data can be accessed, and reference these sources in the article.Provide the statistical code