A meta-regression evaluating the effectiveness and prognostic factors of oral phosphodiesterase type 5 inhibitors for the treatment of erectile dysfunction

Abstract

The effectiveness of phosphodiesterase type 5 inhibitors (PDE5-Is) for erectile dysfunction (ED) varies considerably among trials, but available studies investigating the factors that affect the effectiveness are few and findings are not consistent. A systematic search was performed in PubMed, Cochrane Library, and EMBASE to identify randomized controlled trials comparing PDE5-Is with placebo for the treatment of ED. The methodological quality of included studies was assessed by the Cochrane Collaboration's tool for assessing risk of bias. The associations between prespecified study-level factors and effectiveness were tested by a random effects meta-regression model. This study included 93 trials with 26 139 patients. When all PDE5-Is were grouped together, Caucasian ethnicity was associated with 15.636% (95% confidence interval [CI]: 0.858% to 32.579%) increase in risk ratio (RR) for Global Assessment Questionnaire question-1 (GAQ-1), and 1.473 (95% CI: 0.406 to 2.338) score increase in mean difference (MD) for posttreatment International Index of Erectile Function-Erectile Function domain score (IIEF-EF), compared to Asian ethnicity. A one-score increase in baseline IIEF-EF was associated with −5.635% (95% CI: −9.120% to −2.017%) reduction in RR for GAQ-1, and −0.229 (95% CI: −0.425 to −0.042) score decrease in MD for posttreatment IIEF-EF. In conclusion, PDE5-Is are more effective in Caucasians than Asians, and in patients with more severe ED.

INTRODUCTION

Erectile dysfunction (ED) is the persistent inability to achieve or maintain an erection sufficient for satisfactory sexual intercourse.1 It is a common disorder reported to affect as many as 152 million men worldwide and this figure was predicted to hit 322 million by 2025.2 As to individual patients, ED may cause serious distress and social stigma, with marked effects on self-esteem, family relationships and quality of life.3,4,5,6 Fortunately, the treatment of ED has been advanced by the introduction of oral treatments like phosphodiesterase type 5 inhibitors (PDE5-Is).7,8

Oral PDE5-Is are currently the first-line therapy for ED.7,8 Our previous meta-analysis has demonstrated that PDE5-Is are more effective than placebo, and tadalafil seems to be the most effective agent followed by vardenafil.9 However, data from that study also revealed that the range of the effective rate (measured by Global Assessment Questionnaire question-1 [GAQ-1]) among included trials was quite wide, varying from 49.5% to 94%. This implied that the effectiveness of PDE5-Is might be influenced by certain factors.

As a multifactorial condition, ED is known to be associated with age, ethnicity, and comorbid conditions.10,11,12 Some studies suggested that the effectiveness of PDE5-Is seem to be better in Caucasians than Asians,13 but similar among black American, Hispanic American men and Caucasians.14,15 Other factors like body mass index (BMI), disease severity/duration, etiology, and comorbidity may also influence the treatment effects, but the evidence is rather controversial.16,17,18,19,20,21,22 Therefore, we carried out this meta-regression to explore the factors affecting the effectiveness of PDE5-Is for the treatment of ED.

MATERIALS AND METHODS

Literature search

We carried out an electronic search of Cochrane Library (Issue 4, 2012), PubMed (1966 to April 2012), and EMBASE (1984 to April 2012) using the following keywords: phosphodiesterase-5 inhibitor, tadalafil, sildenafil, vardenafil, lodenafil, mirodenafil, udenafil, avanafil, erectile dysfunction, impotence, randomized controlled trial, etc. No limitations were set on publication status or language. We also searched the MetaRigister and WHO International Clinical Trials Registry Platform, and manually checked the reference lists of the included studies to identify further trials.

Studies selection

We included randomized controlled trials (RCT) comparing oral PDE5-Is with placebo for the treatment of ED. Studies were included where they recruited ED patients, irrespective etiology, disease severity or duration. The primary outcomes of this study were GAQ-1 and the International Index of Erectile Function-Erectile Function domain score (IIEF-EF). Two authors independently determined the study eligibility by examining the titles, abstracts, and full articles progressively.

Data extraction and quality assessment

The following details were extracted by two authors independently: study characteristics, patient characteristics, intervention, control, method, and outcomes. Discrepancies were resolved by discussion. The authors of original studies were consulted for missing information where necessary. The methodological quality of included trials was assessed and reported by the Cochrane Collaboration's tool for assessing risk of bias.23

Data analysis

We initially undertook a random effects meta-analysis of the comparison between PDE5-Is and placebo. Summary effect size was calculated as mean difference (MD), or risk ratio (RR), together with their 95% confidence intervals (CIs). Heterogeneity among studies was assessed with the Chi-square test and the I² -index statistic.23,24 Subgroup analysis was carried out according to ethnicity and drug class.

We then explored associations between the prespecified study-level factors and effectiveness (GAQ-1 and posttreatment IIEF-EF) using a random effects meta-regression model.25,26,27 The factors included in univariate meta-regressions included age, weight, BMI, height, ethnicity (Asians were used as reference group, and the study population of a trial was considered as Asian/Caucasian if over 70% of included patients were Asians/Caucasian), disease severity, disease duration, comorbidity (proportion of diabetes mellitus, hypertension, depression, benign prostatic hyperplasia, and hyperlipidemia), life-style related factors (proportion of smokers, and alcohol drinkers), and ED etiology (proportion of organic, psychological, and mixed ED patients). For those variables identified as being significantly associated with outcomes in the univariate meta-regressions, separate univariate meta-regressions were performed for PDE5-Is type, where a minimum of 10 trials were available. The covariates included in the multivariate meta-regressions were selected based on our background knowledge and the way described previously.26,28 Multivariate meta-regressions were performed for all PDE5-Is together, as well as for individual agents with 10 or more trials.

Sensitivity analyses were carried out according to dosage and quality of included studies. Publication bias was examined by Egger's tests and funnel plots.29 All the data analyses were undertaken in STATA version 12 (StataCorp, College Station, Texas, USA).

RESULTS

Search results and study characteristics

Our search strategy yielded 5938 citations, of which 5709 were excluded after review of titles and abstracts. Of the remaining citations, 229 full texts were screened and 93 studies, with 26 139 patients, were included (Figure 1). The majority of included studies included Caucasians and Asians, accounted for 63.3% (13 816/21 834) and 21.3% (4651/21 834), respectively. Included trials covered 7 different PDE5-Is: sildenafil (38 studies), tadalafil (20), vardenafil (24), udenafil (5), mirodenafil (2), avanafil (2), and lodenafil (2). The methodological quality varies among included studies, and the overall methodological quality is moderate. The baseline characteristics (Supplementary Table 1 ^{(6.9MB, tif)} ) and methodological quality assessment (Supplementary Table 2 ^{(6.3MB, tif)} ) were summarized in the [Supplementary file ^{(61.8KB, pdf)} ].

Figure 1.

Flow chart of study selection. ED: erectile dysfunction; PDE5-Is: phosphodiesterase type 5 inhibitors; RCT: randomized controlled trial.

Supplementary Table 1

The baseline characteristics and reference of included studies

Supplementary Table 2

The methodological quality of included studies

Supplementary file

REFERENCES

Meta-analysis results

Sixty-eight studies including 18 346 patients contributed to the meta-analysis for GAQ-1, and demonstrated that PDE5-Is were more effective than placebo (RR: 2.61; 95% CI: 2.46 to 2.78). Subgroup analysis by ethnicity indicated that PDE5-Is were significantly more effective in Caucasians than Asians (P = 0.0009), but subgroup analysis by drug class suggested no significant difference among PDE5-Is (P = 0.55, Table 1).

Table 1.

Meta-analysis of phosphodiesterase 5 inhibitors versus placebo for the treatment of erectile dysfunction

Thirty-one studies including 8052 patients contributed to the meta-analysis of IIEF-EF. The pooled MD between PDE5-Is and placebo was 6.58 (95% CI: 5.86 to 7.31). Subgroup analysis by ethnicity indicated that PDE5-Is were significantly more effective in Caucasians than Asians (P = 0.03). Additionally, the test for subgroup difference by drug class was also significant (P = 0.002, Table 1).

Meta-regression results

Univariate meta-regressions

When PDE5-Is were grouped together, meta-regressions demonstrated that both ethnicity and baseline IIEF-EF were significantly associated with GAQ-1 and posttreatment IIEF-EF; ED duration were significantly associated with posttreatment IIEF-EF. Specifically, Caucasian ethnicity was associated with 23.061% (95% CI: 7.104% to 41.396%) increase in RR for GAQ-1, and 1.880 (95% CI: 1.071 to 2.689) score increase in MD for posttreatment IIEF-EF, when compared with Asian ethnicity; a one-score increase in baseline IIEF-EF was associated with 5.138% (95% CI: −8.021% to − 2.166%) reduction in GAQ-1, and 0.329 (95% CI: −0.536 to − 0.121) score decrease posttreatment IIEF-EF. ED duration was significantly associated with GAQ-1 (% reduction in RR: 6.885%; 95% CI: 1.196% to 12.895%) but not with posttreatment IIEF-EF (Change in MD: 0.086; 95% CI: −0.291 to 0.463). Other factors including age, weight, BMI, height, proportion of smokers, proportion of drinkers, and ED etiology showed no significant relationships with the treatment effect (Table 2 and Figure 2).

Table 2.

Univariate meta-regression analysis for GAQ-1 and posttreatment IIEF-EF

Figure 2.

Univariate meta-regression analysis. Log risk ratio of GAQ-1 according ethnicity (a), baseline IIEF-EF (b), and ED duration (c); mean difference of posttreatment IIEF-EF according to ethnicity (d), baseline IIEF-EF (e), and ED duration (f). ED: erectile dysfunction; GAQ-1: Global Assessment Questionnaire question-1; IIEF-EF: International Index of Erectile Function-Erectile Function domain score.

Sildenafil, tadalafil, and vardenafil were included into the separated meta-regressions testing the associations of ethnicity, baseline IIEF-EF and ED duration with the effectiveness. However, the meta-regression for baseline IIEF-EF on tadalafil was not carried out because the range of mean baseline IIEF-EF in related studies was too narrow (from 13 to 16). The regression results indicated that Caucasian ethnicity was positively associated with treatment effectiveness compared to Asian ethnicity; and baseline IIEF-EF was negatively associated with the effectiveness. However, statistical significance was only achieved in the meta-regression of: 1) GAQ-1 with tadalafil by ethnicity (% of change in RR: 56.526%; 95% CI: 14.719% to 113.567%); 2) posttreatment IIEF-EF with sildenafil by ethnicity (Change in MD: 2.574; 95% CI: 0.465 to 4.684); 3) GAQ-1 with vardenafil by baseline IIEF-EF (% of change in RR: −5.499%; 95% CI: −10.619% to − 0.085%); and 4) GAQ-1 with sildenafil by disease duration (% of change in RR: 12.644%; 95% CI: 1.198% to 25.386%) (Table 2).

Multivariate meta-regressions

The ethnicity and baseline IIEF-EF were included in the multivariate meta-regressions. Although univariate meta-regressions found that ED duration was significantly associated with RR for GAQ-1, it was not involved in the multivariate meta-regressions because no study provided data on ED duration for tadalafil, which was the agent showing the strongest association between ethnicity and GAQ-1. Including ED duration in the multivariate meta-regression would remove tadalafil related trials, thus the association between ethnicity and GAQ-1 would be greatly underestimated. Multivariate meta-regressions for individual PDE5-Is were carried out on sildenafil and vardenafil.

When PDE5-Is were grouped together, ethnicity and baseline IIEF-EF were significantly associated with RR for GAQ-1 and MD for posttreatment IIEF-EF. This was consistent with the results from univariate meta-regressions. Specifically, the Caucasian ethnicity was associated with 15.636% (95% CI: 0.858% to 32.579%) increase in RR for GAQ-1, and 1.473 (95% CI: 0.406 to 2.338) increase in MD for posttreatment IIEF-EF when compared to Asian ethnicity; A one-score increase in baseline IIEF-EF was associated with 5.635% (95% CI: −9.120% to − 2.017%) reduction in RR for GAQ-1, and 0.229 (95% CI: −0.425 to − 0.042) score decrease in MD for posttreatment IIEF-EF (Table 3).

Table 3.

Multivariate meta-regression analysis for GAQ-1 and posttreatment IIEF-EF

In multivariate meta-regressions for individual PDE5-Is, Caucasian ethnicity was generally positively associated with effectiveness compared to Asian ethnicity, and the baseline IIEF-EF was negatively associated the effectiveness. However, statistical significance was only achieved in the meta-regression of posttreatment IIEF-EF and ethnicity with sildenafil (Change in MD: 2.231; 95% CI: 0.173 to 4.289), and GAQ-1 and baseline IIEF-EF with vardenafil (% change in RR: −5.809%; 95% CI: −10.943% to − 0.379%) (Table 3).

Sensitivity analysis and publication bias

Sensitivity analysis according to drug dosage and the methodological quality of included studies did not greatly change the regression results (Supplementary Table 3 ^{(940KB, tif)} ). Visual inspection suggested asymmetry in the funnel plots for GAQ-1 but not posttreatment IIEF-EF, and this was confirmed by Egger's test (GAQ-1: P =0.001; IIEF-EF: P =0.354) (Figure 3).

Figure 3.

Funnel plots of included studies. GAQ-1: global assessment questionnaire question-1; IIEF-EF: international index of erectile function-erectile function domain score.

Supplementary Table 3

Sensitivity analysis by methodological quality and dosage

DISCUSSION

The principal findings of this study are as follows: (1) PDE5-Is were more effective in improving GAQ-1 and IIEF-EF compared to placebo irrespective of ethnicity, disease severity, and study duration; (2) PDE5-Is as a whole were more effective in Caucasians than Asians, and the treatment effects were better in patients with severer ED; (3) The associations between ethnicity/disease severity and the effectiveness for individual agents were similar to those for all PDE5-Is considered as a whole, though the associations were not always statistically significant; (4) Age, weight, BMI, height, comorbidity, smoking, alcohol consumption, and ED etiology did not appear to be associated with the treatment effects.

As expected, our study indicated that PDE5-Is are more effective in severe ED patients than mild ED patients. This is because there are upper and lower limits regarding the ED symptoms, severe ED patients are, therefore, more likely to have a large magnitude of improvement in ED symptoms compared to patients with less severe ED. For example, the maximum possible improvement in IIEF-EF for a mild ED patients (assume baseline IIEF-EF = 23) is 7, while this figure can be as large as 22 for severe patients with a baseline IIEF-EF of 8. In addition, the beneficial effects of PDE5-Is consist of the effect directly attribute to medicine and a placebo effect. One previous study indicated that the placebo effect is significantly larger in mild ED patients, which in turn lead to better treatment effect in severe patients.16 Our finding is supported by the subgroup analysis in some original trails19,20 and the meta-regression in the previous systematic review.16

The difference of the effectiveness of PDE5-Is between Caucasians and Asians may be attributed to multiple factors. Firstly, when compared with Asians, Caucasian ED patients included in our study were likely to have severed ED symptoms, which in turn lead to better response to treatment. In addition, culture, socioeconomics, and psychosocial issues may influence whether a patient acknowledges he has ED, seeks treatment for ED or responds to prescribed treatments. Lastly, the difference in physical characteristics may also influence the effectiveness. Though the current evidence concerning ethnic variation of mechanisms of PDE5-Is is sparse, some studies suggested that sex hormone levels, which act centrally, influencing libido and nocturnal erectile function, as well as acting on nitric oxide pathways in the corpora cavernosa, are different between Caucasian and Asian ED patients.30,31 Further studies are needed to investigate the mechanisms in future.

The univariate meta-regression suggested that disease duration was positively associated with the GAQ-1. This may be because ED tends to be more severe in patients with longer duration, and the treatment effect is better in severer patients. In addition, our meta-regressions suggested that the effectiveness of PDE5-Is was comparable among patients with various ages, and etiologies, which is in agreement with many previous studies.16,17,18,20,21 Regarding the potential influence from comorbidity, evidence has indicated that patients with diabetes mellitus-associated ED are less likely to response to treatment, however, no significant association was identified in our study. This might because our study was based on aggregated data, and it may be influenced by ecological bias. Additionally, the proportion of diabetes patients in the included studies, which is the independent variable in the meta-regression model, may not be a good covariate because the range is very limited (from 0% to 40%).

The separated meta-regressions for individual PDE5-Is were only undertaken on sildenafil, tadalafil, and vardenafil due to the limited data. The results suggested that sildenafil was significantly more effective in Caucasians than Asians, which was in agreement with the study by Mohanty.13 Though not all the separated meta-regressions results for individual PDE5-Is reached statistical significance, most of them were in agreement with the results of the meta-regressions of all PDE5-Is. Small sample size was a potential cause of the lack of statistical significance as the numbers of trials and patients in the meta-regressions with all agents were approximately 3 times larger than those in the meta-regressions of individual agents.

To the best of our knowledge, this study is the most comprehensive meta-regression analysis of PDE5-Is for ED to date. We employed a search strategy with high sensitivity to ensure most trials were included. This enabled us to locate a much higher number of studies and patients compared to previous reviews on this topic. In addition, our findings were robust due to the high consistency in regression results between different regression models and study outcomes. In addition, the regression results for the primary predictors were also confirmed by the subgroup analyses and sensitivity analyses.

Limitations of this study primarily arise from the quality of original trials reviewed. However, sensitivity analysis according to quality of included studies showed no major impact to the results. In addition, ecological bias may be caused in meta-regressions because they include trial-level rather than individual-level demographics as the co-variables.28 In addition, meta-regression can lead to false positive results particularly when several covariates are used.25,26 However, we believe this was not a serious threat to this study because the primary findings were also confirmed by subgroup analysis.

This study contributes to the understanding of the treatment effects of PED5-Is for ED. The results are encouraging, particularly for patients with severer ED as they may get even more benefits from PDE5-Is treatment. The decrease in the price for some PDE5-Is provides even stronger evidence to consider PDE5-Is as the first choice for the treatment of ED. For future clinical trials and meta-analysis, the effect modifiers identified by this study should be adjusted in study design, analysis and interpretation. As PDE5-Is showed different treatment effects between Caucasians and Asians, it would be interesting to investigate if this difference is caused by cultural or genetic factors.

CONCLUSION

PDE5-Is, on the whole, appear more effective in Caucasians than in Asians, and in severe ED patients than mild ED patients. For individual PDE5-Is, these relationships still exist though they do not always reach statistical significance. Age, weight, BMI, height, comorbidity, smoking, alcohol consumption, and ED etiology appear not to be associated with the effectiveness of PED5-Is based on trial-level data. Future studies are needed to confirm our conclusion due to the limitations of included studies and the risk of ecological bias.

AUTHOR CONTRIBUTIONS

JLT, CM and JQY contributed to the study design, coordinated and drafted the manuscript. JQY, ZYY, and XHF collected the data. JQY, CM and SYS performed quality assessment and statistical analyses. All authors revised the manuscript, read and approved the final manuscript.

COMPETING INTERESTS

All authors declare no competing interests.