Nonsurgical Repigmentation Therapies in Vitiligo: Meta-analysis of the Literature

Objective  To assess the effectiveness and safety of nonsurgical repigmentation therapies in localized and generalized vitiligo by means of a meta-analysis.

Data Sources  Computerized searches of bibliographic databases, a complementary manual literature search, and contacts with researchers and pharmaceutical firms.

Study Selection  Predefined selection criteria were applied to both randomized and nonrandomized controlled trials.

Data Extraction  Two investigators independently assessed the articles for inclusion. When there was a disagreement, a third investigator was consulted.

Data Synthesis  Sixty-three studies were found on therapies for localized vitiligo. Of these, 10 of 11 randomized controlled trials and 29 of 110 patient series were included. One hundred seventeen studies on therapies for generalized vitiligo were found. Of these, 10 of 22 randomized controlled trials and 46 of 231 patient series were included. Among randomized controlled trials on localized vitiligo, the pooled odds ratio vs placebo was significant for topical class 3 corticosteroids (14.32; 95% confidence interval [CI], 2.45-83.72). In the patient series, topical class 3 and class 4 corticosteroids carried the highest mean success rates (56% [95% CI, 50%-62%] and 55% [95% CI, 49%-61%], respectively). Side effects were reported mostly with topical psoralen and intralesional and class 4 corticosteroids. In the randomized controlled trials on generalized vitiligo, the odds ratio vs placebo was significant for oral methoxsalen plus sunlight (23.37; 95% CI, 1.33-409.93), oral psoralen plus sunlight (19.87; 95% CI, 2.37-166.32), and oral trioxsalen plus sunlight (3.75; 95% CI, 1.24-11.29). In the series, the highest mean success rates were achieved with narrowband UV-B (63%; 95% CI, 50%-76%), broadband UV-B (57%; 95% CI, 29%-82%), and oral methoxsalen plus UV-A therapy (51%; 95% CI, 46%-56%). Oral methoxsalen plus UV-A was associated with the highest rates of side effects. No side effects were reported with UV-B therapy.

Conclusions  Class 3 corticosteroids and UV-B therapy are the most effective and safest therapies for localized and for generalized vitiligo, respectively.

VITILIGO IS an acquired skin disorder characterized by sharply demarcated depigmented lesions with variable size and shape that have the tendency to expand over time. It is estimated that about 1% of the world population is affected by the disease, regardless of age, sex, and skin color.¹

There are several possible approaches in the management of vitiligo. Some patients can be reassured simply by explaining the nature of the skin condition and by giving advice on the use of camouflage products and sun-protective measures. In others, especially in the dark-skinned population, vitiligo may cause disfigurement that can lead to serious impairment of the quality of life.² In certain cultures, patients with vitiligo are still regarded as social outcasts. For these groups of patients, active treatment can be considered.

Nonsurgical repigmentation therapies represent the first-line active treatment modality in vitiligo. Currently best studied and therefore most applied are oral and topical psoralen plus UV-A (PUVA), phenylalanine plus UV-A, oral and topical khellin plus UV-A, UV-B narrowband and broadband therapy, and corticosteroids (oral, topical, and intralesional).^3-8

There are many studies reporting on the clinical effectiveness and safety of the various treatments in vitiligo. Guidelines and review articles^4-7 have been published on how and in which patients these treatments should be applied. However, these recommendations were based on data obtained from only a limited number of references and on personal and institutional preferences. No methodological information was given on how these guidelines had been developed. Because none of these articles included a systematic literature search, it is likely that relevant studies were missed. It has been shown that review articles in journals or in textbooks, even if they were recently published, contain unsound, imprecise, and biased information.^9-12

According to the principles of evidence-based medicine, the best medical care is provided by integrating the "best external evidence, as obtained from clinically relevant research studies into the daily practice."^13,14(p71) ¹⁵Combining all relevant studies in a meta-analysis can increase the power and precision of estimates on effectiveness and side effects profiles.^16,17 The development of guidelines for the choice of the most effective and safest therapy in vitiligo should be based also on the available evidence in the literature.^18,19 We therefore performed a meta-analysis of the available literature on the most widely used forms of nonsurgical repigmentation therapy, with regard to both effectiveness and safety.

The computerized bibliographic databases MEDLINE (National Library of Medicine, Bethesda, Md) and EMBASE (Elsevier Science BV, Amsterdam, the Netherlands) were screened for clinical trials from January 1966 to December 1997. No language restrictions were applied. As main key words (including analogs and derivatives), we used "vitiligo," "phototherapy," "PUVA therapy," "ultraviolet therapy," "phenylalanine," "khellin," "glucocorticosteroids synthetic," and "anti-inflammatory agents." Other sources were abstract books of symposia and congresses, theses, textbooks, monographs, reviews, editorials, letters to the editor, free or rapid communications, and the reference lists from all the articles retrieved. Also, 21 leading authorities in the field of vitiligo and 9 pharmaceutical companies were contacted to provide us with any additional published and unpublished references.

The studies were divided into those describing therapies for localized vitiligo and those reporting on therapies for generalized vitiligo. Localized vitiligo was defined as vitiligo affecting less than 20% of the total body surface.⁷

Two investigators (M.D.N. and W.W.) independently assessed the articles for inclusion. When there was a disagreement, a third investigator (P.M.M.B.) was consulted.

Included were both randomized controlled trials (RCTs) and nonrandomized controlled trials on oral and topical psoralen plus sunlight or artificial UV-A (both sources were termed UV-A), including methoxsalen, trioxsalen, bergapten, and unsubstituted psoralen (PS); UV-B broadband and narrowband; phenylalanine plus sunlight or UV-A; oral and topical khellin plus sunlight or UV-A; and corticosteroids (oral, topical, and intralesional). Topical corticosteroids were divided into class 3 ("potent corticosteroids"), which included the drugs betamethasone valerate and halometasone, and class 4 ("very potent corticosteroids"), which included clobetasol propionate. With regard to intralesional administration of corticosteroids, studies using triamcinolone acetonide were included. In studies on photoradiation, both natural sunlight and artificial UV-A were used in the same patient series. These UV sources were regarded as equally effective, provided that sun exposure was performed between 11 AM and 2 PM.² Excluded were double publications, combination therapies, studies that used obsolete drug(s) or dosage schemes, studies that reported on fewer than 5 patients, and studies with insufficient data on effectiveness.

Nonrandomized controlled trials were also analyzed as patient series. Because comparative or placebo-controlled trials can contain a description of at least 2 patient series, the total number of patient series could exceed the total number of studies included.

In case of double publications, the most elaborate publication was selected.

Treatment was regarded as successful when more than 75% repigmentation was achieved. The odds ratio (OR) was used as the effect measure in this meta-analysis because it is regarded as the most satisfactory metric with which to combine across trials with discrete outcomes.²⁰ The OR describes the odds of a successful event (ie, >75% repigmentation) in a patient receiving the active therapy relative to a patient receving the placebo. As a result, an OR greater than 1 indicates greater effectiveness of the active therapy compared with the placebo.

To obtain a more precise estimate of the treatment effect, ORs were combined across similar trials. A random-effects model according to DerSimonian and Laird²¹ was used to pool these ORs and to calculate their 95% confidence intervals (CIs). The random-effects model was preferred over the fixed-effects model because study populations and treatment outcomes in these trials were expected to be statistically heterogeneous, although all RCTs met the specific selection criteria. Calculations were done with Meta-Analyst software for MS-DOS (Joseph Lau, Boston, Mass).

Sample size–weighted averages were calculated for each modality by dividing the total number of patients who achieved more than 75% repigmentation by the total number of patients in the included series. The 95% CIs of these averages were calculated with the computer program Confidence Interval Analysis for MS-DOS (British Medical Journal, London, England) by means of the exact method.

The treatment duration (range and mean values) was also determined for each treatment modality. Treatment-specific side effects were estimated by dividing the number of patients with side effects by the total number of patients studied. For each study, sample size–weighted averages of these frequencies and their 95% CIs were calculated with the same software as was used for calculation of the success rates.

Fifteen of the 21 leading authorities and 7 of the 9 pharmaceutical firms contacted replied and provided us with further relevant references. No study required a third reviewer to resolve disagreements about selection.

Literature Search. With regard to therapies for localized vitiligo, 63 studies were found. Eleven RCTs were identified, of which 10 met our eligibility criteria (Table 1).^22-29 One study on class 3 corticosteroids was excluded because it was a double publication (same trial published in 2 languages). The method of randomization was not reported in any of the studies found.

Effectiveness. Pooled ORs showed nonsignificant differences between class 4 corticosteroids (OR, 1.00; 95% CI, 0.16-6.21), intralesional corticosteroids (OR, 1.42; 95% CI, 0.31-7.78), topical khellin (2%-3%) plus UV-A (OR, 1.18; 95% CI, 0.38-3.62), or topical khellin (5%) plus UV-A (OR, 1.00; 95% CI, 0.39-2.54) and their respective placebos. Class 3 corticosteroids had a pooled OR of 14.32 (95% CI, 2.45-83.72) vs placebo (Table 2).

Literature Search. A total of 110 patient series could be identified, of which only 29 series could be included, reporting on 993 patients (Table 3). Thirteen series with topical methoxsalen plus UV-A were excluded because obsolete drug compositions were used (such as methoxsalen plus 8-isoamyleneoxypsoralen) or because 1% concentration instead of the currently advised 0.1% concentration of methoxsalen solution was used. Reasons for exclusion of patient series are listed in Table 3. In cases in which several criteria of exclusion were applicable, only the most important one was listed.

Effectiveness.Figure 1 shows that class 3 corticosteroids had the highest percentage of patients achieving more than 75% repigmentation (56%; 95% CI, 50%-62%), followed by class 4 corticosteroids (55%; 95% CI, 49%-61%).

Side Effects. Of the 29 included series, 28 (97%) reported side effects. Of the topical PUVA group, methoxsalen had the highest proportion of patients developing phototoxic reactions (58%; 95% CI, 51%-65%), followed by trioxsalen (39%; 95% CI, 23%-56%) and PS (25%; 95% CI, 12%-38%). Atrophy was the most common side effect with local corticosteroids, occurring mostly in patients receiving intralesional corticosteroids (33%; 95% CI, 22%-43%), followed by patients using class 4 corticosteroids (14%; 95% CI, 10%-18%) and class 3 corticosteroids (2%; 95% CI, 1%-5%). Other less common reported side effects of corticosteroids were telangiectasia, corticosteroid-induced acne, and hypertrichosis. Studies on topical khellin (2%, 3%, and 5%) plus UV-A did not report any side effects (Table 4).

The mean treatment duration did not vary much, from 5 to 8 months (Table 4). One study reported the prolonged but intermittent use of class 3 corticosteroids for a period of 21 months.³⁰ In general, the duration of treatment strongly depended on the reponse; when there was no reponse after 2 to 3 months, therapy was discontinued.

Literature Search. A total of 117 studies on therapies for generalized vitiligo were found. Twenty-two RCTs were identified, of which 10 were included (Table 5).^31-38 Eight studies (from 4 publications) were excluded because they described combination therapies.^31,33,39,40 Of these 8 studies, 4 RCTs were found on oral corticosteroids combined with oral PUVA. Four other studies (from 2 publications) were excluded because of insufficient data on outcome measures ("mean achieved percentage repigmentation" was used as outcome measure).^39,40 No RCTs were found on oral bergapten plus UV-A, broadband UV-B, or narrowband UV-B. The method of randomization was not reported in any of the studies found.

Effectiveness. The OR vs placebo was significant for oral methoxsalen plus sunlight (OR, 23.37; 95% CI, 1.33-409.93), oral PS plus sunlight (pooled OR, 19.87; 95% CI, 2.37-166.32), and oral trioxsalen plus sunlight (pooled OR, 3.75; 95% CI, 1.24-11.29). For therapies that used phenylalanine plus UV-A and oral khellin plus sunlight, no significant differences were found between the active drug and the placebo (Table 6).

Literature Search. A total of 231 patient series were identified. Of these, 46 series were included, reporting on 1866 patients (Table 7).

The exclusion criterion "obsolete drug or dosage scheme" was applied for studies that used daily instead of the currently advised twice- to thrice-weekly intake of psoralen.⁴¹ Daily psoralen intake followed by daily UV exposure is not recommended because of the risks of cumulative phototoxic events.^3,41 Furthermore, studies on oral PUVA performed in the 1950s were excluded because of the use of an obsolete drug containing methoxsalen plus 8-isoamy-leneoxypsoralen. Other reasons for exclusion are listed in Table 7. In cases in which several criteria of exclusion were applicable, only the most important one was listed.

For the effectiveness and safety analysis, studies on oral corticosteroids were further divided into studies that described daily intake of corticosteroids and studies on oral minipulse therapy.

Effectiveness.Figure 2 shows that phototherapy with narrowband UV-B had the highest percentage of patients who achieved more than 75% repigmentation (63%; 95% CI, 50%-76%), followed by broadband UV-B (57%; 95% CI, 29%-82%). Only 1 study of each was included. Oral methoxsalen plus UV-A and oral bergapten plus UV-A had success rates of 51% (95% CI, 46%-56%) and 43% (95% CI, 38%-48%), respectively. The differences between the mean success rates of narrowband UV-B, broadband UV-B, and oral methoxsalen plus UV-A were not significant.

Side Effects. Of the 46 included series, 40 (87%) reported side effects (Table 8 and Table 9). Of the oral photochemotherapeutic modalities, methoxsalen had the highest proportion of patients who developed nausea (and vomiting) as a side effect (29%; 95% CI, 24%-35%), followed by khellin (9%; 95% CI, 2%-16%) and PS (8%; 95% CI, 2%-15%). Nausea was caused least often by bergapten and trioxsalen (1%; 95% CI, 0%-2%; and 2%; 95% CI, 0%-3%, respectively). Phototoxic reactions were seen mostly in patients who ingested methoxsalen (25%; 95% CI, 20%-30%), followed by bergapten (6%; 95% CI, 4%-8%). In patients using trioxsalen and PS plus UV-A, phototoxic reactions rarely occurred (1%; 95% CI, 0%-3%; and 0%, respectively). Abnormal results of liver function tests were observed mostly in patients using khellin (17%; 95% CI, 8%-26%). Miscellaneous systemic reactions (headache, dizziness) were reported mostly in patients taking oral PS (24%; 95% CI, 14%-33%). Pruritus, as one of the miscellaneous cutaneous effects, was most frequently seen with the use of methoxsalen (31%; 95% CI, 26%-37%). An increased contrast between normal and depigmented skin was also reported mostly with the use of methoxsalen (10%; 95% CI, 6%-13%). Studies on phenylalanine with UV-A, broadband UV-B, and narrowband UV-B did not report any systemic or local side effects. In 1 patient series with oral khellin, 7 (28%) of the 25 patients dropped out of the study because of (asymptomatic) elevation of liver transaminase levels. In the remaining 7 photochemotherapeutic modalities, the side effects reported were all mild and transient and did not lead to discontinuation of the therapy in any of the patients described. The mean treatment duration of photochemotherapies varied between 9 months for phenylalanine plus UV-A and 24 months for oral PS plus UV-A.

In comparison with daily intake of corticosteroids, oral minipulse therapy with corticosteroids was associated with a lower mean proportion of patients with systemic and cutaneous side effects (Table 9). The differences among the rates of various side effects between daily therapy and oral minipulse therapy were, however, not significant. The treatment duration for oral minipulse therapy varied between 6 and 24 months, depending on the number of courses given (each course lasted 5 weeks).

Analysis of the available RCTs on therapies for localized vitiligo showed that only class 3 corticosteroids had significantly better results than placebo. Analysis based on patient series revealed that topical class 3 corticosteroids and class 4 corticosteroids had the highest mean proportion of patients with more than 75% repigmentation. Treatment-specific side effects were reported mostly with topical psoralen and intralesional and class 4 corticosteroids. In generalized vitiligo, the OR vs placebo was significant for oral methoxsalen plus UV-A and oral PS plus UV-A. In the series, the highest mean success rates were achieved with narrowband UV-B, broadband UV-B, and oral methoxsalen plus UV-A therapy. Oral methoxsalen plus UV-A was associated with the highest rates of side effects. No side effects were reported with narrowband or broadband UV-B therapy.

Databases are known to be only partially successful in identifying relevant studies.¹⁷ Therefore, other sources were screened manually to optimize the results of the search. In addition, leading authorities and pharmaceutical firms were contacted to minimize publication bias.⁴² However, publication bias cannot be excluded in our data set despite extensive efforts to gain all of the performed studies. Only a few studies reporting "negative findings" of a therapy were found.^28,37

We have separated treatment modalities indicated for patients with localized vitiligo from those indicated for patients with generalized vitiligo. Nevertheless, selection bias is likely to influence our findings since, between studies, patients may have suffered from different severity of disease. Smaller lesions may have responded better to therapy than larger ones. Some studies using oral PUVA also included patients with only localized vitiligo lesions.

It is likely that the way of assessing repigmentation also differed between studies. Most studies did not mention which and how many observers were involved in scoring of the repigmentation. For this reason, we have excluded studies that used ratings such as "repigmentation or no repigmentation" as outcome measures without giving percentages of the observed repigmentation. Perhaps, in the future, computer image analysis may reflect more accurately the percentage of repigmentation vs depigmentation in patients with vitiligo.⁴³

Other factors that may have influenced treatment outcome are differences between patient populations in age and skin type. Although some authors had the impression that younger and dark-skinned patients gained better results, no definitive controlled study was performed to support this observation. It was also disappointing that patient compliance was not reported in any of the studies that used phototherapy. Monitoring of compliance is relevant, since phototherapy in vitiligo is time consuming and requires continous high motivation of the patients. Furthermore, the issue of whether the treatment had improved quality of life was not addressed in any of the studies.

We therefore propose that future clinical trials in vitiligo should evaluate treatment outcome in relation to such variables as localization and duration of the lesions, skin type, age, quality of life, and compliance of the patients. In addition, we support the performance of follow-up studies on the permanency of therapy-induced repigmentation, in particular for photoradiation. So far, only 3 studies have addressed this issue. However, all 3 studies described PUVA therapy with the use of a daily dosage scheme, which is now considered obsolete.^44-46

Only a few RCTs that included both localized and generalized vitiligo could be included in this analysis. No RCT was found on topical psoralen, oral bergapten, and narrowband and broadband UV-B therapy. Therefore, caution is needed in the interpretation of the data. In RCTs on therapies for generalized vitiligo, only a small num-ber of the patients treated with placebo had achieved successful repigmentation. A possible explanation may be that in vitiligo, spontaneous repigmentation rarely occurs and is never complete. Some even believe that "spontaneous" repigmentation does not exist and is in fact UV-induced repigmentation.⁸ In therapies for localized vitiligo, repigmentation observed in patients treated with placebos can therefore also be attributed to either sun exposure (studies with class 4 corticosteroid and intralesional corticosteroid) or combined UV-A irradiation (studies with topical 2%-5% khellin plus UV-A).

Analysis of RCTs and of the patient series showed that class 3 corticosteroid was the most effective and safest therapy for localized vitiligo. We do not advise topical class 4 corticosteroids and intralesional corticosteroids because of the significantly higher rates of atrophy. Topical psoralen plus UV-A was shown to be associated with a high risk of phototoxic events (erythema, blistering) and is therefore to be prescribed under strict precautions.⁷ Despite the absence of side effects with the use of topical khellin plus UV-A, we do not recommend this option because the drug is not shown to be effective in the analysis of RCTs and the analysis of patient series.

The choice of therapy in generalized vitiligo can also be made after consideration of the reported effectiveness and safety profiles. Most studies on photoradiation concluded that more than 75% repigmentation could be achieved only when patients were treated on a regular basis and for longer periods. For photoradiation, at least 1 year of continous treatment was required. These observations indicate that only well-motivated, highly compliant patients are suitable for photoradiation. "Complete" repigmentation was rarely reported with the use of oral photochemotherapeutic modalities. Most authors attributed this to the observation that certain skin areas, such as the distal dorsal surfaces of the hands and feet, tips of fingers and toes, areas of bony prominences, palms, soles, nipples, and lips scarcely repigment at all. Estimation of the response judged by overall body repigmentation may have led to lower response rates because the above-mentioned nonresponding areas were included in the calculations. Broadband UV-B and narrowband UV-B were described in only 1 study each.^47,48 Narrowband UV-B therapy is considered a novel phototherapy in vitiligo. More studies are needed to confirm the good results and to establish its long-term side effects.

Analysis of the side-effect profiles in the patient series showed that treatment with psoralen (both orally and topically) is not easy to perform. Severe phototoxic reactions can be avoided by careful monitoring of UV exposures. Nausea can be avoided by concomitant ingestion of low-fat food or milk.³ The use of oral khellin was associated with elevated transaminase levels in a high proportion of the patients. Moreover, in 1 series, 28% of the patients terminated the therapy because of this adverse effect. Therefore, we do not advise khellin as 1 of the drugs of choice. On the other hand, no side effects were reported with the use of phenylalanine plus UV-A, broadband UV-B, or narrowband UV-B. Figure 2 shows that phenylalanine plus UV-A is a relatively ineffective therapy. However, one may prefer UV-B (narrowband or broadband) therapy over oral methoxsalen plus UV-A therapy when comparing the side effects. Other advantages of UV-B therapy over oral PUVA include no need for oral intake of psoralens, shorter treatment sessions, and applicability in pregnant women and children younger than 12 years.⁴⁸

Guidelines on the use of photochemotherapeutic modalities in vitiligo should also give recommendations on maximum treatment duration. Both PUVA and UV-B (broadband and narrowband) therapies need not be continous, since these modalities are well known for their carcinogenic properties.^49-51 In patients with psoriasis, long-term PUVA therapy was found to be associated with an increased risk for skin cancer, especially squamous cell carcinoma.⁵² To date, only 2 patients with vitiligo have been described with squamous cell carcinoma after prolonged PUVA therapy.^53,54 This relatively low incidence may be explained by the fact that, in contrast to patients with psoriasis, those with vitiligo receive lower cumulative PUVA dosages, do not expose themselves to extra sun rays, and do not use tar preparations, cystostatic drugs (methotrexate), or immunosuppressive drugs (cyclosporine). Nevertheless, we suggest that guidelines for maximum cumulative PUVA doses in vitiligo should follow those recommended for psoriasis.⁵⁵ With regard to UV-B therapy (both broadband and narrowband), there are presently still insufficient epidemiological data available on humans to provide evidence-based advice regarding a safe maximum dose. However, by means of a dose-response model it has been calculated that long-term narrowband UV-B therapy may carry substantially less risk for skin cancer than PUVA therapy.⁴⁹

The use of oral corticosteroids is still considered controversial by many authors,^4-7 since systemic side effects (such as moon face and weight gain) are associated with this therapy.^56-58 Our results are in agreement with those of these authors, since the drug is relatively ineffective and is related with a high rate of side effects.

On the basis of the results of this review, the following recommendations can be made concerning the choice of the most effective and safest therapy. For patients with localized vitiligo, class 3 corticosteroid is advised as first-choice therapy. When patients exhibit generalized vitiligo, UV-B (narrowband or broadband) therapy or oral methoxsalen plus UV-A is recommendend. This review may be used as a basis for the development of evidence-based guidelines for the management of vitiligo.

Accepted for publication August 6, 1998.

This work was supported by a grant from the Commission for Guidelines for Clinical Practice, Academic Medical Centre, University of Amsterdam, Amsterdam, the Netherlands.

Presented at the Second Stichting Nederlands Instituut voor Pigmentstoornissen Symposium, Amsterdam, November 21, 1997.

We thank Heléne Dyserinck, clinical librarian, for her assistance with the bibliographic database searches.

A complete list of all of the studies identified is available on request from the authors.

Corresponding author: M. D. Njoo, MD, Netherlands Institute for Pigmentary Disorders, Instituut voor Wetenschappelÿk Onderzoek Building, Academic Medical Center, Meibergdreef 35, 1105 AZ Amsterdam, the Netherlands (e-mail: snip-ww@knoware.nl).