Effectiveness of 5‐fluorouracil treatment for actinic keratosis – a systematic review of randomized controlled trials

Actinic keratosis (AK) lesions present as dry, rough, yellow–brown, scaly plaques which may become thickened and horny. Most AKs are caused by chronic exposure to ultraviolet (UV) radiation and are therefore most common in middle-aged and elderly fair-skinned individuals. Regions with a higher UV exposure show a higher prevalence, and the incidence rate increases with age.1 Around 15–25% of lesions resolve spontaneously over a 12-month period, and the risk of a single lesion progressing to an invasive squamous cell carcinoma (SCC) is in the range 0.25–20% per year.2,3 Opinion differs about the classification of AKs – some argue that they should be classified as in situ SCC because they are histopathologically indistinguishable from SCCs,3 whereas others argue that the classification of in situ SCC, although histopathologically correct, is liable to misinterpretation by consumers, leading to unnecessary and excessive concerns generated by a diagnosis of cancer and additional costs to already overburdened healthcare systems.4 Nevertheless, the inability to predict which AK lesions will transform into invasive SCCs means that the treatment of all AKs is indicated. Topical 5-fluorouracil (5-FU) is a well-established treatment for AK. 5-FU is an antineoplastic antimetabolite, which interferes with the synthesis of DNA and RNA, provoking unbalanced cell growth and death. 5-FU enables field therapy for patients with multiple AKs, and may also promote the healing of subclinical lesions.5 Treatment is lengthy (about 4 weeks) and application site reactions, ranging from redness, soreness and weeping to shallow ulceration and crusting, are a consequence of treatment. Despite the widespread use of 5-FU to treat AK, the evidence supporting its use has not been reviewed systematically or appraised critically, unlike that for imiquimod.6–8 We therefore reviewed systematically the published literature to address this gap, and to identify areas in which further research is needed. In January 2008, the Medline (1966 to present) database (via PubMed) was searched for randomized controlled trials (RCTs) of treatment of AK with 5-FU. The EMBASE database and The Cochrane Central Register of Controlled Trials were also searched, as were the reference lists of the included studies. Studies were included in this review if they reported RCTs on humans comparing the treatment of AK with 5-FU, placebo, or another active treatment, or investigated different 5-FU dosage regimens. Participants in the included studies all had clinical symptoms of AK – a histologic diagnosis was not required. Papers were excluded if they were not written in English or if they were case reports or observational studies. Each included study was reviewed by two authors (DA and SM) and the following data were extracted (if available) and recorded: publication citation; details of the participants (number of participants in each treatment group, gender, age, and number and location of lesions); method of diagnosis of AK; details of the intervention regimen (dosing frequency, duration of treatment, assessment intervals, and length of follow-up); study design; number of patients achieving complete clearance of lesions; changes in the mean number of lesions; other outcome data; number of patients withdrawing from the study as a result of adverse events; and authors’ declared conflicts of interest, sponsorship, and acknowledgment of editorial freedom. Each study was evaluated for the following areas of potential bias: random sequence generation; allocation concealment; blinding during the study; objectiveness of outcome assessment; completeness of follow-up; use of intention-to-treat analysis; occasions of selective reporting; and objectiveness of outcome measures.9 Where possible, the weighted mean difference (WMD) and 95% confidence intervals (CIs) were calculated for studies reporting changes in the mean number of lesions per participant and mean lesion area per patient, and the odds ratios (OR) and 95% CIs for studies reporting the number of patients achieving 100% clearance.10 The use of different outcome measures in most trials precluded a pooled estimate of the treatment effect for each outcome measure across trials. In all, 29 studies were identified through database and hand searching of reference lists. Of these, 16 were rejected (Fig. 1) and the remaining 13 RCTs were included. Number of papers located and selected on fluorouracil treatment of actinic keratosis One study compared 5% 5-FU with facial resurfacing with either carbon dioxide (CO2) laser or 30% trichloroacetic acid peel for the treatment of AK and the prevention of nonmelanoma skin cancer over 5 years.11 Only data relating to the treatment of AK lesions, measured 3 months after treatment, were included in this review. Another study compared the efficacy of a 1-week treatment with 0.5% 5-FU followed by cryotherapy on the remaining lesions at 4 weeks post-treatment with cryotherapy alone.12 A reduction in AK count was measured prior to cryotherapy, and only these data were included in this review. Table 1 presents details of the included studies, including the number and characteristics of the participants, treatment groups, assessment period, and main outcomes. Eight studies compared 5% 5-FU with other treatments – imiquimod,13,14 cryotherapy,14 diclofenac sodium 3% gel (DFS),15 facial resurfacing,11,16 photodynamic therapy (PDT),17 5% 5-FU augmented with tretinoin,18 and 0.5% 5-FU.5 One study compared different dosing regimens of 5% 5-FU (twice daily for 3 weeks vs. twice daily for 1 day per week for 12 weeks).19 Three studies compared 0.5% 5-FU with placebo,12,20,21 and one compared 0.5% 5-FU with 5-aminolevulinic acid (ALA) PDT, activated with either blue light or pulsed laser light22 (Table 1). Several different outcome measures were used to determine the efficacy of treatment, including absolute and proportional changes in lesion counts per patient, changes in total lesion count, change in lesion area, tolerability, and patient preferences (Table 1). Nine studies reported the proportion of patients achieving 100% clinical clearance of the lesions.5,11–14,17,20–22 The duration of follow-up varied between the studies. As the complete healing of lesions treated with 5% 5-FU may not be evident for up to 2 months following the cessation of active treatment,23 8 weeks post-treatment was considered to be the minimum time period to ascertain short-term benefit. The assessment in seven studies was short, with outcomes being assessed 4 weeks after the cessation of treatment.5,11,13,14,19–21 Three studies assessed outcomes at around 8 weeks,11,13,18 and long-term assessment data were reported in four studies, at 6 months17 and 12 months14,16,19 post-treatment. Treatment with 5% 5-FU resulted in an average reduction of 79.5% (range, 59.2%–100%) in the mean number of lesions, from an average of 24.0 at baseline (range, 10.3–61.8) to 4.0 (range, 0–8.8) at follow-up.5,11,16,18,19 In comparison, treatment with 0.5% 5-FU resulted in an average reduction of 86.1% (range, 77.9%–91.7%) in the mean number of lesions, from an average of 13.9 lesions at baseline to 3.9 lesions at follow-up.20,21 The average number of lesions was reduced by 94.5% (range, 92.9–96.6%) when treated by laser resurfacing,11,16 and 28.0% (range, 21.6–34.4%) when treated with placebo12,20,21 (Table 1). Of the eight studies that reported a reduction in the mean or median number of lesions per patient, only three provided sufficient data to enable assessment of the WMD.11,12,18 There was no difference in the WMD of 5% 5-FU compared with 5% 5-FU augmented with tretinoin (WMD, –0.8; 95% CI, –2.4, 0.8),18 resurfacing with CO2 laser (–3.3; –9.9, 3.3) or 30% trichloroacetic acid peel (–1.1, –6.0, 3.8).11 Treatment with 0.5% 5-FU for 1 week resulted in a significantly greater reduction in the mean number of lesions per patient than did placebo (4.8; 3.1, 6.5).12 Treatment with 5% 5-FU cleared 93.8% (606/646) of lesions at 24 weeks,13 and 98.0% (124/126) of lesions at 4 weeks,15 compared with 65.9% (323/490) of lesions cleared with imiquimod13 and 89% (111/125) of lesions cleared with DFS15 (Table 1). The study comparing 0.5% 5-FU with ALA PDT, activated by either blue light or pulsed laser, reported clearance values of 79%, 80%, and 60% of lesions, respectively, but no raw data were reported.22 Across the studies, an average of 49.0% (range, 0%–96%) of patients treated with 5% 5-FU5,11,13,14,17 and 34.8% (14.9–57.8%) of patients treated with 0.5% 5-FU5,12,20–22 were reported as achieving clearance of 100% of lesions. In comparison, 100% clearance was reported by no patients treated with acid peel11 or ALA PDT activated with red light,17 in 0–4.3% of patients using placebo,12,20,21 in 37.5% (3/8) of patients treated with CO2 laser resurfacing,11 in 50% (6/12) of patients treated with ALA PDT activated by blue light,22 in an average of 54.5% (range, 24–85%) of patients treated with imiquimod,13,14 and in 68% of patients treated with cryotherapy.14 Only four papers provided sufficient data to calculate ORs and 95% CIs for this outcome measure (Table 2).11,12,14,22 Significantly larger numbers of patients achieved clearance of 100% of lesions when treated with 5% 5-FU compared with cryotherapy (OR, 10.8; 95% CI, 1.2, 94.9),14 and 0.5% 5-FU for 1 week compared with placebo (30.0; 1.7, 516.5)12 or ALA PDT activated by pulsed laser light (11.0; 1.1, 114.1).22 Sustained clearance of the treatment area at 12 months after the cessation of treatment was assessed in only one study, and was observed in 33% (8/24) of patients treated with 5% 5-FU, 4% (1/25) of patients treated with cryotherapy, and 73% (19/26) of patients treated with imiquimod.14 Treatment with 5-FU is associated with application site reactions, and most studies assessed the severity of these reactions, although the methods varied between studies making data synthesis difficult.5,12,13,15–17,20–22 Cosmetic outcome was assessed in only one study – at 3 months, there was no difference between the groups treated with 5% 5-FU, cryotherapy, or imiquimod; however, at 12 months, 4% of patients treated with 5% 5-FU or cryotherapy and 81% of patients treated with imiquimod showed an excellent cosmetic outcome (based on scarring, atrophy, and induration).14 This is a very large difference, but it is unclear whether investigators were blind to the treatment groups, introducing a high risk of bias in these results. Only two studies assessed patient preferences for treatment.5,15 In the study comparing 0.5% and 5% 5-FU, 85% (17/20) of patients preferred 0.5% 5-FU, with the remaining three patients preferring 5% 5-FU.5 The study comparing DFS and 5% 5-FU reported that 79% of patients were very or completely satisfied with DFS, compared with 68% with this level of satisfaction with 5% 5-FU.14 The patients were not blind to the treatment in either of these studies, however, limiting the validity of this assessment. Overall, only three studies reported the number of patients withdrawing from the study as a result of adverse events: 1.9% (4/213) of patients using 0.5% 5-FU21,22 and 5.9% (1/17) of patients using 5% 5-FU.17 Only four studies provided any information about the proportion of patients completing the designated treatment period. Table 3 presents the available information on the number of patients ceasing treatment, patient preference for treatments, and withdrawals as a result of adverse events. The majority of studies were poorly reported (Table 4), with only five studies providing any detail about randomization methods5,12,14,16,19 and only two of these adequately describing allocation concealment.12,16 Only one study was double blind,12 and four were single blind (outcome assessment only).5,13,15,19 Although most studies claimed to undertake intention-to-treat analysis, only five included all randomized patients in the analysis.14–16,21,22 Insufficient baseline data were provided in two studies,20,22 preventing the comparison of treatment groups. Eight studies did not report fully on their pre-specified outcomes; commonly, studies reported percentages without giving either the denominator or numerator, or did not provide standard deviations or ranges for the data when reporting means or medians.5,13,15,16,19–22 In summary, most of the included studies were at moderate to high risk of bias, and therefore the results should be interpreted with caution and few, if any, generalizations could be made. Our systematic review provides valuable evidence to indicate that about one-half of patients using 5-FU for the treatment of AK lesions can expect complete clearance; that, overall, an 80% reduction in lesion count can be expected; and that a 90% reduction in total lesion count is likely. Most patients (about two-thirds) can expect to require retreatment after 1 year, and although few patients stop treatment as a result of adverse events, up to one-half may not be able to complete the full treatment course. The quality of the studies providing this evidence is, however, poor. Evidence on alternative treatments (studied head to head with 5-FU) is limited. Most studies were small, and only two12,16 reported sample size calculations to ensure adequate power. The three studies comparing 0.5% 5-FU with placebo12,20,21 involved 528 participants and were adequately powered, based on the rationale provided in one of the studies.12 Of the remaining studies (excluding that which reported a sample size calculation), the number of participants ranged from 17 for the split hand study comparing 5% 5-FU with ALA PDT17 to 75 for the parallel group study comparing 5% 5-FU, cryotherapy, and imiquimod,14 although the latter study appears to have been adequately powered based on the reported outcomes. Two studies contained three parallel groups and involved a total of 23 and 36 patients, respectively, and were clearly underpowered.11,22 The small number of participants in most studies limits generalizations being drawn from the results. This is disappointing as high-quality evidence on comparative treatments is highly desirable. The conflicting results from the two studies comparing 5% 5-FU with imiquimod highlight the need to have adequately powered studies, robust outcome measures, and study designs that are, as far as possible, at low risk of bias. Incomplete or no reporting of randomization,11,13,15,17,18,20–22 allocation concealment,5,11,13,15,17–22 blinding,5,11,13,15–22 raw data,5,13,20–22 and CIs or standard deviations5,16,19–21 led to a considerable loss of information, which limited data synthesis across studies. Had this information been available, meta-analysis might have been possible, leading to much more precise estimations of effect. The duration of follow-up in seven studies was less than 8 weeks – the recommended length of time for the maximum effectiveness of treatment to become apparent. It is possible that a further reduction in the number of lesions after the cessation of follow-up would have occurred. This is of concern in studies comparing 5-FU with a more rapidly acting treatment, such as PDT, and would bias the results against 5-FU. The lack of consistency in outcome measures further limited the opportunities for data synthesis. Also of note was the limited focus on patient-centered outcomes, with few studies systematically or comprehensively measuring patient satisfaction with either the treatment itself or its outcome. This is surprising, given that local site reactions are a natural consequence of treatment, and that the effectiveness of treatment is directly related to the seriousness of this reaction and the duration of treatment. Although few patients actually withdrew from any of the studies as a result of the adverse events associated with 5-FU, four studies reported early cessation of treatment by 4%,21 11%,20 17%,22 and a very high 86% of patients in the study comparing 5% 5-FU with 0.5% 5-FU.5 Several of the other studies did not specify the intended length of the treatment period, but rather stated that patients applied 5-FU for between 2 and 4 weeks, or between 2 and 7 weeks. Nor did these studies provide any detail about the actual length of time that the trial participants used the treatment. The studies by Jorizzo et al.20 and Weiss et al.21 clearly showed that the efficacy of 0.5% 5-FU was linked to the treatment duration, but studies of this nature have not been conducted with 5% 5-FU. Nevertheless, the length of time that patients can tolerate treatment is an important clinical issue, and the lack of information provided in these studies hinders informed decision making by both the clinician and patient. A further methodologic concern was the lack of objective methods of assessment of the number of AK lesions. Most studies simply counted the lesions at baseline and follow-up, and compared the counts. Four studies produced a physical map of the lesions (either by photography or by charting the location of lesions onto a diagram),11,15,19,22 whereas other studies simply counted the lesions at each assessment. A longitudinal study of the natural history of AK found a very high turnover of AK, with large numbers developing, regressing, and recurring over time.24 Because of the labile nature of these lesions, studies investigating treatment options for their eradication need to ensure that they are able to accurately trace the evolution of all lesions included at baseline. None of the included studies could make this claim. We conducted a systematic and comprehensive search but were unable to locate any RCTs comparing 5% 5-FU with placebo, no studies determining an optimal duration of treatment, and only two studies comparing 5-FU with imiquimod (another commonly used topical treatment). The conflicting results from these two studies highlight the need to have adequately powered studies, robust outcome measures, and study designs that are, as far as possible, at low risk of bias. Clearly, 5% 5-FU is considered to be clinically effective – Valeant Pharmaceuticals have reported that more than 30 million prescriptions have been written for Efudex® worldwide over the last 30 years.25 This figure becomes even more startling when it is considered that it is not the treatment of choice in continental Europe for the removal of AK lesions, unlike areas such as Australia, South Africa, and southern USA, where there is a high prevalence of these lesions. Presumably, clinicians in these countries have adopted this treatment, have found it to be clinically effective and efficient, and have had insufficient complaints from patients to consider alternative treatments – an example of “eminence-based treatment” rather than “evidence-based treatment.”26 Nevertheless, there are consequences of the lack of quality RCTs investigating the treatment of AK with 5% 5-FU. There is a lack of good quality empirical evidence to inform clinicians about which of their patients will gain the greatest benefit from this treatment, or the characteristics of AK lesions that are most, or least, likely to be cured. There is also insufficient evidence to enable clinicians to make informed choices between 5-FU and imiquimod, and between 5-FU and other treatment modalities, such as facial resurfacing or PDT. There is insufficient evidence on patient satisfaction to enable consumers to make informed treatment choices. Also, there is virtually no evidence about the long-term effectiveness of this treatment. These gaps clearly need to be addressed, given the high and increasing worldwide prevalence of AKs over recent decades, particularly in regions with high exposure to UV light.1 Topical 5-FU therapy: is not an extremely effective therapy; typically only affects dysplastic areas of skin; appears to be relatively well tolerated in practice; typically produces an excellent cosmetic reaction when used on the face; may be associated with a photosensitive reaction; works best on the legs; prevents new AKs from forming in treated areas; needs to be continued for 4 weeks to obtain the full effect; is also listed on the PBS in Australia for use on Bowen's disease (intraepidermal SCCs); has been utilized in two different strengths. Answers found on next page 1 False 2 True 3 True 4 True 5 True 6 False 7 False 8 True 9 True 10 True