Photosensitizing medications and photocarcinogenesis
Conflicts of interest:
Photosensitivity is an exaggerated or abnormal response to ultraviolet (UV) or visible light exposure. Many current medications are known photosensitizers; however, the effects of the sensitization can be subclinical and go unnoticed by the person affected. While some of these drugs are used for short and defined periods, others are used indefinitely for the treatment of chronic disease. The question of whether either of these practices translates into an increased risk of skin cancer is an important one. Numerous medications have real, distinct and well-elucidated mechanisms that potentiate the development of skin cancer, while with some medications the mechanism for the observed carcinogenesis remains unclear. In this article we will discuss the clinical, mechanistic and epidemiological evidence supporting photochemical genotoxicity and carcinogenesis.
Photosensitivity is an exaggerated or abnormal response to ultraviolet (UV) or visible light exposure. Many current medications are known photosensitizers; however, the effects of the sensitization can be subclinical and go unnoticed by the person affected. The photosensitizing properties may be predicted sometimes by the chemical structure of the drugs. Ring compounds with alternating single and double bonds are known to absorb UV radiation well. For photosensitivity to occur the drug absorbs UV radiation of distinct wavelengths. Individual drugs have their own distinct absorption profiles. The activating wavelengths are typically in the UVA spectrum (320–400 nm), though less frequently in shorter wavelengths. While some of these drugs are used for short and defined periods, others are used indefinitely for the treatment of chronic disease. The question of whether either of these practices translates into an increased risk of skin cancer is an important one. While some epidemiological studies have tested this association and other studies have looked at drugs individually, this area remains largely unexplored.
Numerous medications have real, distinct and well-elucidated mechanisms that potentiate the development of skin cancer, while with some medications the mechanism for the observed carcinogenesis remains unclear. In this article we will discuss the clinical, mechanistic and epidemiological evidence supporting photochemical genotoxicity and carcinogenesis.
Clinical and animal studies
Photosensitizing psoralen is administered systemically or topically to patients in conjunction with suberythemagenic doses of UVA (PUVA) and is a well-established, highly effective treatment for many inflammatory skin diseases including psoriasis , mycosis fungoides  and vitiligo . With systemic PUVA there is a well-established increased risk of squamous cell carcinoma (SCC). This was initially predicted in animal models [4, 5] and borne out to be a clinically relevant association through a long-running prospective study known as the PUVA follow-up study [6-9]. Stern et al. followed 1380 patients who were first treated with systemic PUVA for psoriasis in 1975 and 1976. He observed a dose-dependent increased risk of SCC manifest within 5 years of first exposure . Between 5 and 10 years post initial exposure a modest increase in basal cell carcinomas (BCC) was also observed . It was not until more than 15 years had elapsed from the time of first exposure that an increase in incidence of malignant melanoma was observed . So established is the photocarcinogenic effect of 8 methoxypsoralen (MOP), it is used in studies on this topic as a positive control . Topical PUVA studies in humans with 5-MOP, 8-MOP and trioxsalen lack the dose-related risk, though cohorts are smaller and cumulative dosing smaller, making comparison difficult and the power of the studies perhaps insufficient to detect smaller risks [11-14].
Azathioprine, a thiopurine analogue in use clinically as an immunosuppressant, causes photosensitivity selectively to UVA in human skin . Long-term treatment with azathioprine is seen in multiple inflammatory diseases including Crohn's disease, autoimmune conditions and in the setting of solid organ transplantation. The incidence of SCC is increased in transplant recipients by an order of 65 to 250 times [16, 17], and the risk appears to be higher with immunosuppressive regimens that incorporate azathioprine .
Voriconazole, a well-established photosensitizer , is a first line treatment for invasive aspergillosis . Studies looking at the association between voriconazole use and nonmelanoma skin cancer (NMSC) have had conflicting findings. While several studies have supported an association [21-24], a large (n = 467) retrospective US cohort study of adult lung and heart/lung transplant recipients found that while an association was present at the level of crude analysis, when they adjusted for confounding factors, such as patient gender, history of chronic obstructive lung disease (possibly a surrogate marker for smoking status) and history of immune disorder, the association was diminished and not significant . In contrast, a smaller US study (n = 91) of lung transplant recipients at Emory University, Georgia, found that having adjusted for all variables found on univariate analysis to be significant risk factors for NMSC (longer time since transplantation, Fitzpatrick skin type and sun exposure history) voriconazole use persisted as a significant risk factor for the development of skin cancer . There is one case series in which voriconazole use was related to increased risk of melanoma in situ .
Murine models investigating the phototoxicity of fleroxacin, a fluroquinolone antibiotic, demonstrated a dose-dependent phototoxic effect, and it was noted that tumors, malignant and benign, developed in the mice that had severe phototoxic reactions . A further murine study comparing the photocarcinogenic potential of several of the fluoroquinolone group (nalidixic acid, lomefloxacin, fleroxacin, ciprofloxacin and ofloxacin) in the setting of chronic suberythemagenic UVA exposure, with 8-MOP as a positive control, found that tumor development was significantly enhanced in all groups given a test substance compared with mice exposed to UVA alone. While the majority of tumors in the fluroquinolone-treated groups were benign, the lomefloxacin plus UVA group developed a strikingly high rate of SCC which in the majority of cases were cystic, large and deeply invasive .
Exogenous compounds have been shown to potentiate UV-mediated DNA damage through a number of mechanisms. The psoralens utilized in PUVA, 8-MOP and 5-MOP, are furocoumarins. These compounds form monoadducts, through an oxygen-independent photoreaction and the target site of their action is the 5,6 double bond of thymidine. The monoadduct formed incorporates either the furan moiety or the pyrone ring. The pyrone-side monoadduct is not photoreactive, but the furan-side monoadduct is, and its photoactivation by UVA results in the development of tumorigenic DNA interstrand cross-links through cycloaddition (for a recent review see Cadet et al. ). These so-called PUVA mutations prevent adequate DNA repair. Studies show that patients harbor both psoralen-related mutations as well as the UV-signature mutations resulting from UV radiation alone .
The photochemical genotoxicity associated with fluoroquinolones is complex (for recent reviews see Cadet et al. , Lhiaubet-Vallet et al.  and de Guidi et al. ). The formation of cyclobutane pyrimidine dimers in double-stranded DNA by triplet-triplet energy transfer is at least one of the mechanisms underlying lomefloxacin-sensitized tumorigenesis . Fluoroquinolone genotoxicity can also be mediated through Type 1 and Type 2 oxidative photosensitization reactions. These reactions give rise to the production of a distinct collection of degradation products and therefore the predominant reaction type can be determined. In Type 1 reactions, the reaction type associated with lomefloxacin, one-electron transfer (hydrogen atom extraction) mediated by the triplet-excited photosensitizer occurs. Type 2 reactions involve the generation of singlet oxygen, and this represents the predominant mechanism for DNA damage associated with ofloxacin and norfloxacin [34-36]. Irrespective of the mechanism, consequences include DNA damage and apoptosis, and cell type determines relative vulnerability, with keratinocytes being most vulnerable to UVA-mediated lomefloxacin damage, melanocytes most resistant and fibroblasts falling somewhere in between .
The efficacy of azathioprine as an immunosuppressant is dependent upon its metabolism to 6-thioguanine (6-TG) and the incorporation of this compound into cellular DNA. While DNA itself is a poor absorber of UVA, the presence of the 6-TG compound converts DNA into a UVA chromophore, absorbing maximally at a wavelength of 342 nm. The absorbed UVA is then a source for the generation of mutagenic highly reactive oxygen species, which leads to DNA mutations not recognized by the DNA repair mechanisms, resulting in cumulative DNA damage . The finding that the photosensitivity and DNA damage is largely reversed on switching to mycophenolate mofetil  supports the assertion that the increased risk of skin cancer is not simply a function of immunosuppression but rather is related to the photosensitizing effect of azathioprine.
Eight percent of people are found to have phototoxicity to voriconazole . As voriconazole itself does not absorb UVA or UVB, the phototoxicity is most likely related to its main metabolite voriconazole N-oxide, an effective chomophore. In vitro testing has found the absorption spectrum of the N-oxide metabolite to have two peaks; one in the UVC spectrum and one in the UVB spectrum, 264 nm and 313 nm, respectively . This is at odds with an in vivo study that found that voriconazole-related photosensitivity is predominantly UVA mediated . The authors of the in vivo study postulated that the difference is possibly related to spectral shift following interaction with the cytochrome P450 pathway [19, 42, 43]. Polymorphisms in the cytochrome P450 pathway, specifically CYP2C19, can result in impaired metabolism and are known to alter plasma voriconazole levels significantly . This may contribute to variability in photosensitivity from patient to patient; however, we can find no study directly addressing this question. It has been postulated that some of the observed photosensitivity seen with voriconazole is related to its interaction with retinoid metabolism , but the relevance of this interaction is contested .
New studies show that patients with BRAF positive mutations who have metastatic melanoma may benefit from inhibition of the BRAF mutated pathway . Vemurafenib is an inhibitor of V600E mutated BRAF. Vemurafenib-induced photosensitivity is UVA mediated  and may relate to aminolevulinic acid synthase2 controlled upregulation of erythropoiesis and a consequent increase in zinc-protoporphyrins . The observed high rate of SCCs and keratoacanthoma seen with the use of vemurafenib is attributed to paradoxical upregulation of the mitogen-activated protein kinase (MAPK) pathway in the setting of upstream RAS mutations [49-51]. We can find no evidence to date that the photosensitivity observed with vemurafenib is a significant contributor but would contend that oxidative-stress-associated DNA damage would be a likely additive mechanism as in other UVA sensitizers, and further mechanistic studies are warranted.
Epidemiological studies have been undertaken to address the association between photosensitizing medications and skin cancer, and such large studies pick up increased risk at lower levels. A study in the Jutland county of Denmark (n = 40 656) looked exclusively at the use of photosensitizing diuretics and incidence of SCC, BCC and melanoma . They found an increased risk of SCC in patients who been prescribed an amiloride-hydrochlorthiazide combination or amiloride alone [incidence rate ratio (IRR) of 1.79, 95% confidence interval (CI) 1.45–2.21 and IRR of 2.26, 95% CI 0.94–5.43, respectively). For melanoma they had similar findings, with an increased risk with an amiloride-hydrochlorthiazide combination therapy or amiloride alone (IRR of 1.43, 95% CI 1.09–1.88 and IRR of 1.21, 95% CI 0.39–3.74, respectively). Use of indapamide, a drug that absorbs in the UVB spectrum , increased the incidence of melanoma but had no effect on incidence of SCC or BCC, suggesting that the absorption spectrum of the photosensitizing drug may play a role in determining the type of the skin cancer that develops. Amiloride, in contrast, absorbs maximally in the UVA spectrum. This supports previous evidence implicating UVB absorption in the development of melanoma .
A large epidemiological study including all Danish residents over the age of 15 (n = 4 761 749) looked at the association between the long- and short-term use of photosensitizing medications and incidence of SCC, BCC, melanoma and Merkel cell carcinoma . In their study long-term (LT) medications were medications typically used in the treatment of chronic conditions (the assignment did not correspond to use in a particular patient rather the typical pattern of use) and included diuretics, nonsteroidal anti-inflammatory drugs (NSAIDs), cardiovascular medications, oral hypoglycaemic agents, anticonvulsants and the cytotoxic agent methotrexate. Short-term (ST) medications were medications typically used for a defined, short duration and included in this grouping were antimicrobials, antimalarials and the systemic retinoids acitretin and isotretinoin. Two of the LT medications, methyldopa and furosemide, were associated with a ≥ 20% increased risk of skin cancer (BCC and SCC, respectively), and increasing duration of use was associated with increasing risk, further supporting the validity of the association. With ST medications they found a ≥ 20% increased risk of skin cancer in users of any ST medication compared with nonusers and individually an increasing risk of SCC with increasing use was identified with use of doxycycline, sulfamethazole with trimethoprim, acitretin, isotretinoin, an increased risk of BCC seen with ciprofloxacin, ketoconazole, sulfamethazole with trimethoprim and tetracycline and an increased risk of melanoma with hydroxychloroquine. The term ‘short-term’ medication referred to classes of drugs that are typically used for short durations such as antimicrobials; however, included in this category were medications such as tetracyclines, which in settings such as acne vulgaris can be used for extended periods.
A population-based case-control study in New Hampshire, United States, examined whether use of photosensitizing medications increased the risk of NMSC . Five thousand seventy-two participants were enrolled in the study, of which 1906 were controls, 1567 had a history of BCC and 1599 had a history of SCC. Medications assessed included antimicrobials (tetracyclines, sulphonamides and fluoroquinolones), cardiovascular medications (thiazide, potassium sparing and loop diuretics, calcium channel blockers, alpha-adrenergic agonists, antiarrhythmics and sulfonylureas) chemotherapeutics (methotrexate and tamoxifen), NSAIDs, psychiatric medications (benzodiazepines and tricyclic antidepressants) and retinoids (topical tretinoin and oral isotretinoin). They found a modest increase in risk of SCC (odds ratio (OR) 1.2, 95% CI 1.0–1.4) and BCC (OR 1.2, 95% CI 0.9–1.5) in those who had used photosensitizing medications. The association with SCC was pronounced in patients who reported a tendency to sunburn (OR 1.5; 95% CI 1.1–2.0) with no increase in risk in those who reported a tendency to tan (OR of 1.0; 95% CI 0.8–1.3). While a tendency to burn did not predict a higher risk for BCC in those who had utilized photosensitizing medications it did predict a higher risk of multiple BCC (OR 1.7; 95% CI 1.0–3.0). A history of ever having utilized a photosensitizing antimicrobial was associated with an increased risk of BCC (OR 1.9, 95% CI 1.3–2.8) and was particularly associated with early-onset (< 50 years) BCC (OR 2.1, 95% CI 1.3–3.5). Increased duration of use (> 1 year) was associated with increased risk. Tetracycline use showed both increased risk of BCC and early onset BCC, and the most common indication for use was skin rash or acne vulgaris. The authors postulated that perhaps the use of tetracyclines increased risk of BCC because tetracyclines are often used in teenage years for treatment of acne vulgaris, an age group in which sun exposure is associated with risk of BCC . An increased risk of SCC but not BCC was seen with photosensitizing cardiovascular medications.
Although the findings from epidemiological data are conflicting on several points cumulatively the evidence does suggest that concern for increased risk of skin cancer with the use of photosensitizing medications is well founded. The relationship appears to be multifactorial with host factors such as sun sensitivity , age at time of use  along with variables such as duration of treatment [55, 56] and absorption spectrum  of the medication all potentially playing a role in determining whether a patient who takes a photosensitizing drug develops a skin cancer, and if they do; the age at onset  and the skin cancer type [52, 55, 56].
Predicting photochemical carcinogenesis
The prediction of chemical photogenotoxicity is an area of intense clinical and industrial interest and recommendations on testing have been reviewed by an international working group . Photosafety testing is generally required for compounds that absorb in the 290–700 nm range and the European Medicines Agency had until recently recommended parallel preclinical testing for photoallergy, phototoxicity and photogenotoxicity with a positive result in photogenotoxicity testing being considered a marker for photocarcinogenic potential . In contrast, the US Food and Drug Administration did not require specific preclinical photogenotoxicity or photoallergy testing. They recommended short-term photoirritation testing in animals, and if positive the product literature should state a warning of potential photogenotoxicity . Testing for photochemical clastogenicity with the photo-chromosome aberration assay has been a widely used in vitro method to predict photogenotoxicty. With growing concern regarding reports of pseudophotoclastogenicity and the poor correlation between positive in vitro test results and in vivo results, the conclusion of the international working group was that photoclastogenicity testing could no longer be recommended for regulatory testing of photogenotoxic potential. In fact, the group concluded that due to difficulty with interpreting the clinical relevance of current photogenotoxicity test results, they do not add meaningfully to data derived from phototoxicity testing, photochemical reactivity testing and known class effects, and hence routine photogenotoxicity testing could not be recommended as a standard component of photosafety testing . The International Conference on Harmonisation (ICH) addressed the question of photosafety testing in the ICH framework and similarly concluded that currently available tests for photogenotoxicity and photocarcinogenicity are not useful in assessing human pharmaceuticals .
Given the high incidence of skin cancer, particularly NMSC, even a modest increase in incidence would have very significant economic and clinical repercussions. Photosensitizing medications are used ubiquitously, and effective patient education practices need to be in place to alleviate risk. In high-risk patients receiving photosensitizing medications, such as solid-organ-transplant patients taking immunosuppressive agents including azathioprine and in addition voriconazole, it is advisable for a dermatologist to be involved early in their care to advise on photoprotection and to aid in the early identification and prompt management of skin cancer. Given that much of the photosensitivity stems from UVA photosensitivity most reliance should be given to behavioral change limiting UV exposure, the wearing of UV protective clothing including brimmed hats and sunscreens that offer broad-spectrum photoprotection. Reliance on sunscreen protection alone is likely to be ineffective given the cost and the evidence that in general use, sunscreens are inadequately applied .