All findings in our study are reported in accordance with the PRIO-harms guidelines (Bougioukas et al., 2018; Bougioukas et al., 2019). We published our predefined protocol after registering the aims and methods of this overview of systematic reviews at PROSPERO (CRD42020216754) (Pyrgidis et al., 2021).

Two authors (NP, MT) systematically searched PubMed, Web of Science, The Cochrane Library and Scopus databases from inception to June 12, 2021. They also perused the reference lists of all identified studies, as well as potential sources of grey literature, including conference abstracts published in relevant journals and websites for healthcare evidence such as epistemonikos.org. The detailed search strategy is depicted in Supplementary Data S1.

We included systematic reviews with or without meta-analysis in adults with ED performing heterosexual activity that: i) compared the efficacy and safety of any dose of PDE5 inhibitors with another PDE5 inhibitor, with placebo or with other effective treatments; ii) provided outcomes of interest deriving from randomized controlled trials (RCTs); iii) explored the use of any approved PDE5 inhibitor by the EMA and FDA (avanafil, sildenafil, tadalafil, and vardenafil) alone or in combination with other treatment modalities both in the general male population, as well as in specific patient groups and; iv) were conducted based on the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA statement. On the other hand, we excluded: i) systematic reviews or meta-analyses assessing the efficacy and safety of PDE5 inhibitors for indications not relevant to erectile function and; ii) narrative reviews, editorials or letters to the editor.

When we identified both systematic reviews and meta-analyses addressing similar outcomes, the meta-analyses were only included, given that they included more primary studies. Similarly, when we identified systematic reviews or meta-analyses and network meta-analyses addressing similar outcomes, the network meta-analyses were only included, given that they included more primary studies. When we identified studies with the same design (systematic reviews or meta-analyses or network meta-analyses) addressing similar outcomes, ideally the most recent study or, otherwise, the most methodologically rigorous study, based on quality assessment, was included (Cooper and Koenka, 2012). Therefore, for each outcome, we included only one study, preferably a network meta-analysis followed by a meta-analysis and a systematic review.

Two authors (NP, MT) independently performed a three-step parallel review of title, abstract and full text of all retrieved records based on our predetermined selection criteria. All records excluded at the level of full text evaluation were saved and presented in Supplementary Data S2. Any discrepancies throughout the screening process were resolved by consensus. Data extraction was undertaken independently in a Microsoft Excel spreadsheet. The two authors tabulated information regarding study characteristics, intervention details and outcomes. They also performed a pilot test before data extraction to ensure the coherence of the procedure (Higgins et al., 2019). In studies evaluating the effect of PDE5 inhibitors on erectile function with multiple questionnaires, data regarding the International Index of Erectile Function (IIEF)–Erectile Function Domain or IIEF-5 were only extracted.

We employed the AMSTAR 2 tool to assess the quality of all included systematic reviews, meta-analyses or network meta-analyses (Shea et al., 2017). The extent of overlapping among included studies was estimated based on the corrected covered area (CCA) and was presented with novel graphical approaches (Pieper et al., 2014; Bougioukas et al., 2021). The strength of evidence for each meta-analytic effect was determined based on the Grading of Recommendations Assessment, Development and Evaluation (GRADE) (Guyatt et al., 2008). In particular, we evaluated the strength of evidence of each meta-analytic effect based on the corresponding results from the systematic review.

The primary outcome of our overview was the mean change in the erectile function after PDE5 inhibitor intake in the general population measured with the IIEF-5 or the IIEF–Erectile Function Domain. Secondary outcomes included: i) mean change in the erectile function after PDE5 inhibitor intake in specific patient groups based on data availability measured with the IIEF and; ii) severe adverse events and dropouts after PDE5 inhibitor intake both in the general population, as well as in specific patient groups based on data availability.

We performed a descriptive analysis of all included studies. For studies performing meta-analyses or network meta-analyses, we constructed forest plots with the corresponding confidence interval (CI) for all relevant meta-analytic effects. In particular, continuous effect estimates were presented in the form of standardized mean difference (SMD) or weighted mean difference (WMD), while categorical effect estimates were presented in the form of odds ratio (OR). When we identified studies that performed meta-analyses with a fixed effects model, we reanalyzed all outcomes using the DerSimonian and Laird random effects model.

For each outcome, heterogeneity was evaluated with the I² and publication bias was estimated with the Egger’s statistical test. Due to the plethora of primary studies included in each meta-analysis, all relevant measures were presented as they were reported in each study without reviewing the corresponding primary studies (Aromataris et al., 2015). All analyses were undertaken using Microsoft Excel (Version 16.42) and R statistical software (version 3.6.3).

Our literature search yielded 686 potentially eligible unique systematic reviews. Ultimately, we included 23 systematic reviews with 563 primary studies and 154,796 participants (Shabsigh et al., 2007; Vecchio et al., 2010; Xiao et al., 2012; Taylor et al., 2013; Yuan et al., 2013; Schmidt et al., 2014; Chen et al., 2015; García-Perdomo et al., 2017; Tan et al., 2017; Tian et al., 2017; Allen et al., 2019; D’Andrea et al., 2019; Lai et al., 2019; Li et al., 2019; Liao et al., 2019; Liu et al., 2019; Munk et al., 2019; Zhou et al., 2019; Adamou et al., 2020; Madeira et al., 2020; Wang et al., 2020; Zhu et al., 2020; Mykoniatis et al., 2021). Of them, 20 were meta-analyses or network meta-analyses that included a total of 258 relevant meta-analytic effects (Shabsigh et al., 2007; Vecchio et al., 2010; Xiao et al., 2012; Taylor et al., 2013; Yuan et al., 2013; Schmidt et al., 2014; Chen et al., 2015; García-Perdomo et al., 2017; Tan et al., 2017; Tian et al., 2017; D’Andrea et al., 2019; Lai et al., 2019; Li et al., 2019; Liao et al., 2019; Liu et al., 2019; Zhou et al., 2019; Adamou et al., 2020; Madeira et al., 2020; Wang et al., 2020; Zhu et al., 2020; Mykoniatis et al., 2021). The step-by-step screening procedure is illustrated in Supplementary Data S2 and Supplementary Data S3.

The baseline characteristics of the included systematic reviews are presented in Table 1. Overall, a total of 12 studies were funded. Accordingly, the mean overall AMSTAR 2 score of 11 ± 2.9 indicated that the methodological quality of most of the included systematic reviews was generally of sufficient standard (Supplementary Data S4). Across the included studies, the Egger’s test was only reported for 4 meta-analytic effects and no publication bias was detected. The non-reporting of publication bias was predominantly attributed to the low number of studies (<10) included for each meta-analytic effect. Overall, heterogeneity was available for 74 meta-analytic effects, of which 33 displayed substantial heterogeneity I² (>50%). The mean heterogeneity was 36% (range: 0–99%). Furthermore, the overall CCA was 3%, while the CCA for the primary outcome was 37.4%. The corresponding graphical analyses can be seen in Figure 1 and in Supplementary Data S5.

Three network meta-analyses have addressed the efficacy of all approved PDE5 inhibitors on ED compared to each other or to placebo in the general population. Yuan et al. (2013) explored, for the first time, the efficacy of all approved PDE5 inhibitors without accounting for the different dosages and formulations of each PDE5 inhibitors. Subsequently, Chen et al. (2015) performed a trade-off network meta-analysis to account for these differences and to combine outcomes deriving from the four most important ED questionnaires, namely the IIEF-5 or IIEF–Erectile Function Domain, the Sexual Encounter Profile (SEP) question 2 (SEP-2) and 3 (SEP-3), as well as the Global Assessment Questionnaire question 1 (GAQ-1). However, since most included studies did not provide data on one or more of the four questionnaires, the authors applied multiple imputations to perform these analyses (Chen et al., 2015). In an attempt to harmonize clinical outcomes and heterogeneity, Madeira et al. (2020) recently performed a network meta-analysis including studies that assessed ED only with the IIEF-5 or IIEF–Erectile Function Domain.

Combining the outcomes of all available dosages, Yuan et al. (2013) suggested that tadalafil may be considered the most effective compound in the general population, followed by vardenafil. Hence, based on the network meta-analysis by Chen et al. (2015) sildenafil 50 mg (SMD: 0.47, 95% CI: 0.34–0.59) followed by sildenafil 100 mg (SMD: 0.46, 95% CI: 0.35–0.56) were considered the treatments of choice when compared to placebo. In the most recent network meta-analysis by Madeira et al. (2020) sildenafil 25 mg (WMD: 13.08, 95% CI: 10.1–16.06) seemed to be statistically superior to all interventions in improving the IIEF compared to placebo, but studies with low-dose sildenafil are lacking. Still, comparing among different PDE5 inhibitors, sildenafil 50 mg or sildenafil 100 mg seem to be the most effective compounds in the general population. The latter derived, however, predominantly from indirect comparisons among different PDE5 inhibitors (Yuan et al., 2013; Madeira et al., 2020). Of interest, low-dose daily tadalafil may be more effective for the management of ED than high-dose on-demand tadalafil (WMD: 1.24, 95% CI: 0.03–2.44) (Zhou et al., 2019). Overall, in the available studies, the strength of evidence for most pairwise comparisons between different dosages and formulations of PDE5 inhibitors was considered low or very low due to the lack of high-quality RCTs comparing different types of PDE5 inhibitors. All relevant meta-analytic effects are summarized in Figure 2 and Supplementary Data S6.

In specific patient populations, the efficacy of PDE5 inhibitors is also well documented. In particular, the available evidence suggests that, in patients with diabetes, all PDE5 inhibitors are superior to placebo (Liao et al., 2019). Comparing among different formulations of PDE5 inhibitors, vardenafil and sildenafil on-demand display the highest efficacy (Liao et al., 2019). In patients with prostatectomy-induced ED, PDE5 inhibitors at any dosage and formulation are superior to placebo and are also recommended as part of penile rehabilitation strategies (Tian et al., 2017). Similarly, their superiority is observed in individuals with hypertension, chronic kidney disease, renal transplantation, spinal cord injury, multiple sclerosis or other neurogenic disorders, obstructive sleep apnea and antidepressant- or antipsychotic-related ED (Shabsigh et al., 2007; Vecchio et al., 2010; Xiao et al., 2012; Taylor et al., 2013; García-Perdomo et al., 2017; Allen et al., 2019; Li et al., 2019; Liu et al., 2019). Of note, in patients with psychogenic ED, the current evidence indicates that no significant differences are demonstrated between PDE5 inhibitor monotherapy and psychological interventions (WMD: −0.28, 95% CI: −1.19–0.64) (Schmidt et al., 2014). Nevertheless, the level of evidence for most outcomes based on the GRADE approach was deemed low or very low. The corresponding meta-analytic effects are presented in Figure 3.

Compared to placebo, all dosages and formulations of PDE5 inhibitors present more treatment-related adverse events based on the three network meta-analyses by Yuan et al. (2013), Chen et al. (2015), and Madeira et al. (2020). Hence, the ORs for developing severe adverse events, compared to placebo, do not reach statistical significance for all dosages of PDE5 inhibitors. The within group safety analysis demonstrated that there are no major differences among different PDE5 inhibitors in the general population. Still, sildenafil 100 mg seems to be associated with more treatment-related adverse events and discontinuations due to adverse events (Madeira et al., 2020). Of note, PDE5 inhibitors do not seem to significantly improve endothelial function compared to placebo both in the general population and in patients with diabetes (D’Andrea et al., 2019). Similarly, PDE5 inhibitors do not affect semen parameters (Tan et al., 2017). The relevant comparisons are available in Supplementary Data S7.

The addition of testosterone in patients with ED and hypogonadism treated with PDE5 inhibitors seems to further improve erectile function (Zhu et al., 2020). In particular, among patients receiving transdermal or oral testosterone due to hypogonadism, sildenafil on-demand at a dose of 50 or 100 mg, as well as 5 mg daily tadalafil display a safe and effective profile. Still, no clear recommendations about the optimal type and dose of treatment can be made. On the other hand, although PDE5 inhibitors improve erectile function in patients with lower urinary tract symptoms (WMD: −4.77, 95% CI: −6.40–−3.14), the addition of a-blockers does not further ameliorate symptoms (WMD: −0.8, 95% CI: −1.65–0.06) (Adamou et al., 2020). In the Chinese general population, the combination of Chinese herbal medicine and tadalafil compared to tadalafil monotherapy may provide beneficial outcomes on ED (WMD: −2.67, 95% CI: −3.15 to −2.19) (Wang et al., 2020). In patients with psychogenic ED, the addition of acupuncture to tadalafil or sildenafil seems to further improve erectile function (Lai et al., 2019), while the addition of PDE5 inhibitors to psychological interventions does not improve symptoms compared to psychological interventions alone (Schmidt et al., 2014). Interestingly, adding to a PDE5 inhibitor: a second PDE5 inhibitor, low-intensity shockwave therapy, a vacuum erectile device or antioxidants is effective in patients with ED, while folic acid, metformin and angiotensin-converting enzyme inhibitors may be effective in some patients with ED but the evidence is still scarce (Munk et al., 2019). The effect of all available combination therapies with PDE5 inhibitors is presented in Figure 4.

The findings of the present study should be interpreted with respect to some limitations. Due to the plethora of primary studies and due to the complexity of some meta-analytic effects, we did not review the primary studies included in each systematic review and did not consider further primary studies published after these systematic reviews. Therefore, for all measures and outcomes, we relied on the information presented in each included systematic review. Even though we included only systematic reviews and meta-analyses of RCTs to provide high level of evidence, the impact of observational, real-world data on the safety and efficacy of PDE5 inhibitors remains uncaptured. Importantly, some systematic reviews and meta-analyses were of low methodological quality, while others displayed high quality and significantly contributed to our results. Based on the previous notion, it should be stressed that, for each outcome, the most methodologically rigorous study based on study design, protocol existence, duplication of data extraction, application of recommended tools and performance of sophisticated statistical analyses was only included. Still, despite our objective criteria, some important studies might have been excluded. Moreover, it was beyond the scope of this review to explore the role of PDE5 inhibitors in disorders not relevant to ED such as premature ejaculation, priapism or pulmonary hypertension. Finally, it should be stressed that all overviews of systematic reviews suffer from the limitation that they summarize a broad topic and much of the information provided in individual studies or systematic reviews cannot be reported in detail.