No Causal Link between Phosphodiesterase Type 5 Inhibition and Melanoma

Ever since the introduction of sildenafil 20 years ago, phosphodiesterase type 5 (PDE5) inhibitors have become the mainstay therapy in the treatment of penile erectile dysfunction (ED), which can be severely limiting in an estimated 5% to 20% of men worldwide [1]. PDE5 inhibitors work by blocking the PDE5 enzymatic activity that degrades cyclic guanosine monophosphate (cGMP), a smooth muscle relaxant found in the corpus cavernosum of the penis, thereby promoting increased penile smooth muscle relaxation, blood flow, and subsequent enhancement of the male erectile response [2].

The link between PDE5 and melanoma was first alluded to in 1993 when Drees et al [3] isolated a cGMP-specific PDE isoenzyme in B16 mouse malignant melanoma cells. Subsequent studies confirmed PDE5 expression and activity in melanoma-associated antigens in human malignant melanoma cells [4]. Furthermore, PDE5 inhibitors were shown to stimulate melanin synthesis [5], which significantly promoted the development of melanomas [6]. And just recently, a histopathologic study found a unique correlation between markedly dilated vessels and melanoma growth in two patients with long-standing sildenafil use, hinting at a potential physiologic mechanism [7].

Possibly, the most detailed mechanism was proposed by Arozarena et al [8], who discovered that the melanoma-associated ^V600EBRAF mutation promoted melanoma metastasis by downregulating PDE5 protein expression. Their findings were particularly compelling given that BRAF mutations are implicated in over 50% of melanomas [9]. In their study, inhibition of PDE5 elevated cGMP and cytosolic Ca²⁺ levels, which stimulated actin-myosin contractility and melanoma cell invasion. Interestingly, the response was specific to BRAF mutant melanoma cells, as the same effect was not seen in NRAS mutant melanoma cells. These investigators, however, did not find evidence that pharmacologic PDE5 inhibition (i.e., sildenafil, tadalafil, or vardenafil) further augmented melanoma invasiveness in BRAF mutant melanoma cells, as PDE5 protein was already downregulated in these cells.

A second signaling pathway was further elucidated by Dhayade et al [10], who confirmed that activation of the cGMP pathway by PDE5 inhibition promoted melanoma cell growth and migration in a p44/42 MAPK-dependent manner. In addition, they found that sildenafil administration in mice promoted the growth of existing melanomas, which supports their theory that release of the ‘PDE5 brake’ on the cGMP growth-promoting pathway generates more proliferative melanoma tumors [11].

Paradoxically, there also exists evidence that PDE5 inhibition may have a protective role in preventing melanoma formation by promoting antitumor immunity and manipulating the tumor microenvironment [12]. Myeloid-derived suppressor cells (MDSCs) enhance tumor survival and proliferation through a multitude of different pathways, including suppression of antitumor T cell activity [13]. Three clinical trials found that oral administration of tadalafil, a PDE5 inhibitor, reduced MDSCs and restored antitumor T cell immunity in patients with head and neck squamous cell carcinoma [14, 15] and metastatic melanoma [16]. Although overall increase in survival was not demonstrated, PDE5 inhibition may represent a novel adjunct to immunotherapy.

Given the widespread popularity of PDE5 inhibitors, there has been growing interest in the recent decade to further clarify the potential risks and long-term side effects of PDE5 inhibitor use. Li et al [17] was the first group to report an association between sildenafil use in USA males and increased risk of developing melanoma. Since then, eight more studies have attempted to address the same clinical question with varying results [18, 19, 20, 21, 22, 23, 24]. Therefore, we conducted a systematic review of the current literature to investigate the association between PDE5 inhibitor use and malignant melanoma. We also sought to determine if under-expression of the PDE5A gene translates into worse clinical outcomes for melanoma patients.

Li et al [17] was the first study to report a positive association between self-reported sildenafil use and melanoma in a cohort of 25,848 USA male health professionals. They did note that sildenafil users were on average older, more obese, and more likely to have a history of severe or blistering sunburns, all of which are known risk factors for developing melanoma [27]. However, a significant association persisted even after they adjusted for a variety of common risk factors, such as age, body mass index, smoking, physical activity, hair color, sun exposure, sunburns, number of moles, and family history of melanoma. The primary disadvantage of the study was that the disease outcome was so rare, as only 142 melanoma cases were identified, of which only 14 participants reported as having used sildenafil in the past. The study might also have been subject to information or recall bias, as those who diligently self-reported medication may also engage in more health-seeking behaviors (i.e., skin cancer screenings), which can lead to an increased number of melanoma diagnoses [28].

The findings from Li et al's 2014 study [17] prompted Loeb et al [18] to further investigate a possible causal relationship between PDE5 inhibitor use and melanoma. They conducted a large national nested case-control study using comprehensive population-based data from the Swedish national registries. The registries included detailed information on melanoma staging and location, medications prescribed and filled, socioeconomic factors (i.e., education, income, and marital status), and presence of other comorbidities (i.e., prostate cancer and basal cell carcinoma [BCC]). In this study, researchers included 435 melanoma cases, which was 30 times more than those in Li's original study [17]. Their results were consistent with the 2014 study, as they also found an increased risk for melanoma amongst users.

In order to fulfill Hill's Criteria of Causality [29] and ascertain whether or not PDE5 inhibitors actually ‘caused’ melanoma, however, Loeb et al [18] examined the dose-response relationship and specificity between PDE5 inhibitor use and melanoma. Interestingly, they only found a pronounced risk in men who filled a single prescription, rather than in men who filled 2–5 prescriptions or even ≥6 prescriptions. There was also no evidence to suggest that vardenafil or tadalafil, both of which have longer half-lives compared to sildenafil, further increased the risk of melanoma, arguing against a dose-response relationship. Furthermore, they found a correlation between PDE5 inhibitors and BCC, a disease that has not yet been linked to cGMP signaling, which reduced the specificity of the melanoma association. A major potential confounder unaccounted for in this case was ultraviolet (UV) sun exposure, as it is a known risk factor for both BCC and melanoma [30] (i.e., if the PDE5 inhibitor cohort also had more BCCs, they may also have had more sun exposure). Lastly, they found that PDE5 inhibitor use most significantly increased the prevalence of stage 0 and I melanoma, and ‘not’ the more advanced types (stages II–IV), which contradicts the theory that PDE5 inhibition promotes aggressive melanoma invasion [8, 11].

Matthews et al [20] conducted the largest matched cohort study to date, by analyzing UK national registry data on 706,037 men. They found a slightly elevated risk for melanoma in men with ≥1 PDE5 inhibitor prescriptions. Like Loeb et al [18], they also found a positive association between PDE5 inhibitors and BCC. Although data on UV sun exposure was not collected, post-hoc analysis found overwhelming evidence linking PDE5 inhibitors to prior solar keratoses, a known marker for sun exposure [31]. PDE5 inhibitor users were also more likely to have solar keratoses ‘before’ receiving their first prescription, suggesting that users were more likely to have excessive sun exposure at baseline compared to non-users. There was also no evidence linking cumulative PDE5 inhibitor exposure to melanoma, arguing against a dose-response relationship.

Since then, three more conference abstracts have published positive correlations between PDE5 inhibitor use and melanoma [22, 23, 24], of which only one was eligible for this review. Ma et al [22] conducted a small case-control study that followed 730 men from the Rochester Epidemiology Project (REP) database, a regional hospital network registry. Similar to Li et al's study [17], however, the small subpopulation of 278 melanoma cases limited the strength of this study. While they did not find a statistical difference in melanoma staging, they did find a higher association with lentigo maligna melanoma amongst users. Interestingly, lentigo maligna melanoma is a subtype of melanoma linked to older age and chronic sun exposure [32], further corroborating the theory that sun exposure may be a major confounder.

Contrary to the studies above, three studies in our search reported no association between PDE5 inhibitors and melanoma [19, 21]. Lian et al [19] and Matthews et al [20] both utilized the same UK national registry data for their prospective cohort studies; Lian et al [19], however, implemented more stringent patient eligibility criteria, limiting the inclusion of subjects to males with a diagnosis of ED. They restricted their study population to 1) reduce surveillance bias, as PDE5 inhibitor users tend to have higher education levels, which correlates to more skin cancer screenings [33, 34]; and 2) reduce confounders like sedentary lifestyle behaviors and cardiometabolic comorbidities (i.e., obesity, diabetes, and cardiovascular disease) linked to both ED and melanoma [35, 36]. Lian et al’s analysis [19] was not able to find an association with BCC amongst users, despite failing to control for the potential confounder of sun exposure proposed by Matthews et al [20]. This suggests that confounders unrelated to sun exposure (i.e., lifestyle or health-seeking behaviors) may play a larger role in the PDE5 inhibitor melanoma association.

Lastly, Pottegård et al [21] conducted two independent case-control studies using separate patient populations (Denmark and Northern California), and found no association between PDE5 inhibitors and melanoma in either study. Consistent with Loeb et al’s findings [18] regarding melanoma staging, Pottegård et al [21] did find slightly elevated risk of localized (stage 0 or I) melanoma in their Dutch population, but not with metastatic disease. This association, however, disappeared completely when health-seeking markers (i.e., education and frequency of ambulatory visits) were accounted for, reducing surveillance bias. Although they found no increased melanoma risk amongst users in general, high cumulative use (200–500 tablets) of PDE5 inhibitors increased melanoma risk in their Dutch population. This observation was mutual between Lian et al [19], who found a modest increase in risk with ≥7 prescriptions (median=20 prescriptions). These isolated findings may be a reflection of traits found in a particular patient subgroup (i.e., behavioral traits or polypharmacy) and do not strongly support a dose-response relationship between PDE5 inhibitors and melanoma. Not surprisingly, Pottegård et al [21] found that the increased risk disappeared in Denmark males taking ≥500 tablets.

Our systematic review was limited to observational studies, which are inferior to randomized controlled trials in determining a causative relationship between exposures and outcomes. We also included one of three conference abstracts in our review, which often omit critical information about study methods; therefore making them difficult to assess for bias. There is also a potential for publication bias, as some authors received funding from pharmaceutical companies producing PDE5 inhibitors [18, 20, 21]. Additionally, several studies were limited to western men from countries with a homogenous population (i.e., Denmark or Sweden) [18, 21] or Caucasian men [21], which reduces the applicability of the studies to more diverse populations.

And finally, from the animal models described by Arozarena et al [8] and Dhayade et al [10], one would predict that low PDE5A gene expression in melanoma would impart poor survival. However, our gene expression analysis of GDC data did not support this. Although our analysis provided compelling evidence against the role of PDE5A gene in dictating clinical outcomes in melanoma patients, it does not exclude the possibility that PDE5 may still have a role in the development of new melanomas.

At present, there lacks evidence to support that PDE5 inhibitor use causes increased risk for melanoma, and there is only minimal evidence to support the association between the two. A consistent dose-response relationship has not been well established, and the concurrent association between users and BCC in several studies reduces the specificity of the association with melanoma. The studies that demonstrated a positive correlation often failed to account for major confounders such as sun exposure, health-seeking behaviors, or metabolic comorbidities that are linked to both PDE5 inhibitor use and melanoma. Finally, there is no strong molecular rationale for this association in human melanoma patients as PDE5A gene under-expression does not affect overall survival. Nonetheless, the substantial evidence implicating PDE5 inhibition in the cGMP-mediated melanoma progression remains an intriguing area of research, but should not alter clinical practice.