Treatment of moderate to severe psoriasis remains a challenge to date. Extensive psoriasis often requires complex treatment options such as UV light or various systemic therapies.¹ The latter, however, are often not well tolerated and are associated with safety concerns restricting their long-term use.²

A milestone in the treatment of psoriasis was the introduction of psoralen–UV-A photochemotherapy (oral PUVA) by Parrish and coworkers,³ which used the photosensitizing effect of methoxsalen and consecutive irradiation with UV-A. Oral PUVA became the therapeutic standard, especially for severe psoriasis. However, it may be accompanied by various systemic adverse effects, such as nausea, vomiting, headache, or hepatotoxic effects caused by methoxsalen, as well as the risk of photocarcinogenesis,^4,5 cataract formation, and a generalized photosensitization lasting for about 24 hours requiring photoprotection. Thus, the need for topical photosensitization procedures was recognized.

Fischer and Alsins⁶ developed the so-called bath PUVA, in which psoralen derivates such as trimethoxypsoralen or methoxsalen are dissolved in a warm-water bath. Delivery of psoralens by bath prevents systemic adverse effects associated with oral PUVA. Bath PUVA has the advantage of selective and shorter photosensitization,⁷ leading to a significantly lower cumulative UV-A exposure.⁸ Furthermore, it avoids typical variations in therapeutic effect due to large interindividual differences in the gastrointestinal tract absorption of psoralens.^9,10 A large Scandinavian retrospective analysis demonstrated that bath PUVA with trimethoxypsoralen bears only a low risk of long-term carcinogenicity.^11,12 In recent years, bath PUVA has increasingly replaced oral PUVA in Germany and other European countries. In contrast to Scandinavia, in central Europe methoxsalen is commonly used instead of trimethoxypsoralen as a photosensitizer.

The favorable effects of sun exposure and sea water (climatotherapy) for the treatment of psoriasis, especially at the Dead Sea area, have been known for decades.^13,14 Because climatotherapy is bound to the natural environment, artificial regimens have been developed in an attempt to mimic the natural climatic conditions. For this purpose, patients are immersed in saltwater (SW) baths during (simultaneous application) or before (sequential application) UV-B irradiation. Some studies have reported superiority of sequential SW phototherapy (SW UV-B) over UV-B alone.^15,16 However, the effect of salt concentration and mineral composition on clinical outcome is unclear.¹⁷

Although thousands of patients with psoriasis have been treated with bath PUVA or SW UV-B, rigorous randomized controlled clinical trials examining the effect of bath PUVA or SW UV-B compared with tap-water phototherapy (TW UV-B) are still lacking. Few studies with limited numbers of patients have compared the efficacy of bath PUVA and SW UV-B vs TW UV-B or UV-B irradiation alone.¹⁸

After establishing the reference (TW UV-B or UV-B irradiation alone), we conducted the following trial to evaluate the benefit of bath PUVA and SW UV-B compared with the selected reference and compared with each other in a large clinical trial in patients with psoriasis.

The study was a multisite, unblinded, randomized controlled clinical trial in a 4-group parallel configuration in patients with stable, moderate to severe psoriasis.

Primary objectives of the study were organized in a hierarchical order. In a first step—reference selection—a reference therapy was to be selected on the basis of 2 treatment options: TW UV-B or UV-B irradiation alone. The TW UV-B would be selected as reference for subsequent confirmatory analysis in the case of statistical superiority; if it was not superior, then UV-B would be the standard. We assumed no difference between TW UV-B and UV-B.

In a second step—confirmatory analysis—SW UV-B and bath PUVA were to be compared with the selected reference and with each other (3 comparisons). We assumed SW UV-B and bath PUVA to be superior to the selected reference, and we assumed no difference between SW UV-B and bath PUVA. The hierarchical design was chosen to reduce statistical adjustments due to multiplicity of analysis.

One computer-generated randomization list with block lengths of 12 was prepared for each trial site by the responsible biometrician of the study (A.F.). Patients were centrally randomized to receive UV-B, TW UV-B, SW UV-B, or bath PUVA. Allocation was concealed until eligibility was checked and informed consent was given.

The study obtained ethics committee approval from the University of Ulm. In addition, all participating trial sites obtained approval from the ethics committees of the respective medical associations at the state level.

Patients with psoriasis without substantial changes during the past month and a Psoriasis Area and Severity Index (PASI) of greater than 7 or an involvement of the total body surface area of 15% or more were considered for study inclusion. Written informed consent was obtained from each patient before randomization.

Patients with erythrodermic, pustular, or isolated palmoplantar psoriasis were excluded, as were patients younger than 18 years, pregnant or breastfeeding women, and patients with malignant hypertension, coronary heart disease, heart failure, arrhythmia, or a history of malignancies. Furthermore, patients currently using photosensitizing agents or medications negatively affecting psoriasis, eg, β-blockers, angiotensin-converting enzyme inhibitors, lithium carbonate, or indomethacin, and patients who were not legally competent, were also excluded.

The washout period for systemic antipsoriatics was 4 weeks, and for topical antipsoriatic agents, 2 weeks. Patients who had received phototherapy up to 4 weeks before study entry were not eligible.

One hundred two outpatient dermatologic clinics and outpatient dermatologic departments of hospitals took part in the study, recruited patients, and collected data.

The following interventions were administered: UV-B without preceding bath; tap-water bath with subsequent UV-B (TW UV-B); SW bath (concentration, 25%) with subsequent UV-B (SW UV-B); and methoxsalen bath (concentration, 0.5 mg/L) with subsequent UV-A (bath PUVA). For UV-B, TW UV-B, and SW UV-B, broadband UV-B (280-320 nm), selective UV phototherapy (SUP) (300-320 nm), and narrowband UV-B (311 nm) were accepted as irradiation sources. Each trial site had to use only 1 source of UV-B. For bath PUVA, broadband UVA (320-400 nm) was the only accepted irradiation source. Patients were phototested before phototherapy. The UV doses were individually adapted to erythemal response.

For patients allocated to UV-B, TW UV-B, or SW UV-B, minimal erythemal dose (MED) was assessed. The MED was defined as the dose to produce a just-detectable erythema with sharp borders within 24 hours. The MED was tested by exposing 6 uninvolved and untanned body sites of an area of 2 cm² to increasing UV-B doses. Patients allocated to TW UV-B or SW UV-B soaked their test body sites for 20 minutes in tap water or 25% salt water immediately before phototesting. Patients assigned to UV-B were phototested without preceding bath. For broadband UV-B and SUP, doses of 0.01, 0.02, 0.04, 0.06, 0.08, and 0.01 J/cm² were administered to the 6 body sites; for narrowband UV-B, doses were 0.2, 0.4, 0.6, 0.8, 1.0, and 1.2 J/cm². The MED was judged visually immediately and 24 hours after irradiation.

For patients allocated to bath PUVA, minimal phototoxic dose (MPD) was assessed. The MPD was defined as the dose to produce a just-detectable erythema with sharp borders within 72 hours. The MPD was tested similarly to MED, with the exception that the body sites were exposed to increasing UV-A doses after they were soaked in warm water containing 0.5 mg of methoxsalen per liter for 20 minutes (bath PUVA). Doses of 0.5, 1.0, 1.5, 2.0, 2.5, and 3.0 J/cm² were administered. The MPD was judged visually immediately and 48 and 72 hours after irradiation.

After completion of phototesting, patients assigned to TW UV-B, SW UV-B, or bath PUVA took a 20-minute whole-body bath in tap water, water with 25% sodium chloride, or water with 0.5 mg of methoxsalen per liter at 37°C, respectively. Because highly concentrated salt water causes damage to water drains, polyethylene foil baths were allowed.¹⁹ Within 20 minutes after the bath, patients were irradiated with UV-B or UV-A. The starting dose for all UV-B spectra was 50% of MED, and for UVA, 30% of MPD. The UV doses were individually adjusted according to a predefined irradiation schedule (Table 1). Patients were treated 4 times a week until remission or for a maximum of 8 weeks (32 sessions). Remission was defined as a reduction of PASI to less than 3.

Table 1. Irradiation Schedule

All patients were free to use an emollient cream as needed. The use of systemic or topical antipsoriatic medications, such as cyclosporine, retinoids, fumarates, corticosteroids, vitamin D₃ analogue, or tar, was not allowed.

Individual criteria for discontinuation of study participation were worsening of PASI by more than 20%, severe phototoxic reactions with blistering, and withdrawal of informed consent.

The primary outcome was incidence of therapeutic success, defined as a reduction of PASI or involved body surface area by 50% or more during the intervention period. The PASI²⁰ is an internationally accepted, clinician-reported, psoriasis-specific outcome measure based on the involved body surface area, as well as a semiquantitative estimation of erythema, infiltration, and scaling. A 50% reduction in PASI is commonly considered a clinically important outcome.²¹ Higher scores indicate more severe lesions and/or greater skin involvement.

Secondary outcomes were the self-administered PASI (S-PASI),^22,23 the short form of the Questionnaire on Experience With Skin Complaints (SF-QES),²⁴ and global ratings of disease severity, treatment effect, and tolerability both by patients and observers.

The S-PASI was operationalized in the same way as the PASI (reduction of S-PASI or involved body surface area by 50% or more). The S-PASI proved to be a reliable, valid, and responsive outcome measure.²² The S-PASI is scored and interpreted in the same way as the PASI.²³

The SF-QES consists of 4 subscales scored from 0 to 4. One scale has an inverse effect on the underlying construct.²⁵ Total score, defined as the sum of all subscales, can range from −4 to +12. Higher scores indicate a higher level of stigmatization feelings.

Global ratings were assessed by 100-mm visual analog scales, higher scores indicating a more severe disease status, better treatment effect, and higher tolerability.

The PASI, S-PASI, and global ratings of disease severity were assessed at baseline; after 2, 4, and 6 weeks; and at the end of the treatment period (maximum, 8 weeks), as were global ratings of treatment effect and tolerability. The SF-QES was recorded at baseline and at the end of the treatment period.

At each visit, erythemal response, adverse events, and irradiation dose were recorded. At the end of treatment, the cumulative UV dose was calculated (in joules per square centimeter). Doses for narrowband UV-B were calculated separately from those for broadband UV-B and SUP because of a different irradiation schedule (Table 1). Erythemal response was graded on a 6-step scale ranging from no erythema (grade 0) to blistering (grade V). All grade IV (edema) and grade V erythemas were considered phototoxic reactions. In addition, grade III erythema (defined as “fiery red” erythema) was considered a phototoxic reaction if, in consequence, the supervising investigator decided to reduce the dose or to postpone treatment.

Before the start of the study, all responsible investigators from all trial sites met to standardize study processes and to train in the use of the PASI. In addition, each trial site was visited by a monitor before the study, who checked the phototherapy units and the eligibility of the involved study staff and who trained personnel again concerning study processes and the PASI. During the study, on-site monitoring was performed as needed and a telephone hotline was installed. A final visit took place at all sites.

Patients were not told to which intervention they were randomized. Nevertheless, only patients assigned to TW UV-B or bath PUVA could in fact be considered to be blinded. The bath solutions of TW UV-B and bath PUVA do not differ in physical appearance, taste, or color. Patients assigned to SW UV-B or UV-B could not be blinded. A highly concentrated saltwater bath can be easily identified by taste or buoyancy.

Blinding of PASI raters was intended. Both the staff and the patients were instructed not to inform PASI raters about the treatment allocation. Success of rater blinding was evaluated at the end of the intervention period.

For UV-B, we assumed an incidence for therapeutic success of 75% during the intervention period.¹⁴ An increase of at least 10% compared with UV-B only was judged clinically relevant and should be detectable within the study.

Confirmatory analysis consisted of 3 pairwise comparisons. To control for multiple testing, sample size estimation was based on a Bonferroni-adjusted α level of .017 (.05/3). Given a 3-sample parallel-group design, a binary outcome, balanced groups, and a 2-sided test, 300 patients per group had to be included to ensure a power of 0.8.

Sample size estimation for reference selection was based on an α level of .05 owing to the hierarchical design of the analysis. With 300 patients in each of the groups for reference selection, the resulting power was 0.87. In total, study enrollment of 1200 patients was planned.

All patients who had received at least 1 treatment were included. Missing values at the end of treatment were replaced according to the last-observation-carried-forward strategy. If data were completely missing, treatment failure was assumed.

Hypotheses were tested by Fisher exact test. In addition, numbers needed to treat and their corresponding confidence intervals were calculated to control the predefined multiple significance level of 5% of the confirmatory analysis. P values were adjusted according to the Holm-Shaffer closure test procedure (α/3, α, α).²⁶

Secondary analyses were performed by Fisher exact test, analysis of variance, or other common descriptive statistics. Results of inferential statistics should be interpreted descriptively. No statistical adjustments were performed. In contrast to the primary analysis, no replacement strategies for missing values were applied for secondary analyses. Two sensitivity analyses were performed. One used a modified PASI criterion (≥75% reduction in PASI or involved body surface area), and the other was a per-protocol analysis based on a set of prespecified criteria (see next section).

In a safety and tolerability analysis, patients who had received another treatment as specified by the allocation schedule were assigned to the intervention they had actually received.

For a patient to be included into the per-protocol analysis, all of the following criteria had to be fulfilled: (1) conformance with all eligibility criteria, (2) correct treatment combination of bath solution and UV spectrum, (3) total number of treatments within ±20% of target, (4) starting dose within ±20% of target, (5) deviations from incremental regimen not exceeding 50% of applications, (6) total treatment period not longer than 91 days (13 weeks), (7) participants to receive the treatment as specified by the allocation schedule, and (8) no missing data on the primary outcome measure.

During the intervention period, which started on April 30, 2001, and ended on November 11, 2003, 1241 patients were randomized. Eighty-two patients withdrew consent before treatment started and were excluded from analysis. The remainder (n = 1159) were considered for primary analysis.

Details of the participants' progress through the phases of the study are shown in the Figure. The median number of recruited participants per trial site was 10 (25th-75th percentile, 6-16).

Baseline characteristics of the enrolled sample are shown in Table 2. More than half of the study participants were male (59.6%), were employed (58.3%), were classified as having skin type III according to Fitzpatrick criteria (63.1%), reported a current flare of less than 1 year (72.6%), and had already had experience with phototherapy (86.4%). Less than half were current smokers (38.6%), had had at least 1 inpatient treatment due to psoriasis (41.7%), and had already taken systemic medication for psoriasis (30.2%). Mean age was 47 years (SD, 14 years), mean body mass index (calculated as weight in kilograms divided by height in meters squared) was 27 (SD, 5), and mean duration of disease was 21 years (SD, 13 years). Median PASI was 17 (25th-75th percentile, 12-23), and median percentage of involved body surface area was 24% (25th-75th percentile, 16%-35%).

There was a slightly higher proportion of men in the bath PUVA group than in the other groups (65.2% vs 57.0%-58.5%). In the UV-B group, there was a slightly higher proportion of persons with a current flare less than 1 year (75.7% vs 71.4%-72.6%), as well as a slightly higher proportion of persons who had already taken systemic medication for psoriasis (34.2% vs 27.6%-30.2%). Baseline characteristics did not differ between groups in any other aspect.

Six participants had neither a PASI of 7 or greater nor an affected body surface area of 15% or greater and therefore violated the inclusion criteria. Twelve patients violated a single exclusion criterion; none violated 2 or more. Forty-one patients received another treatment at the trial site as assigned by the study center, and 43 patients received a treatment combination other than that defined in the protocol (eg, tap-water bath plus subsequent UV-A).

Initial UV dose differed by more than 20% in 41 cases. Of 385 nonresponders, 185 (48.1%) underwent fewer treatment sessions than planned (<26 sessions). The incremental regimen was not completely performed in accordance with the protocol in 62.0% (686 of 1106). In 30.8% (211 of 686) the deviations exceeded 50%. Deviations from the incremental regimen did not differ between groups. In 14.3% (163 of 1136) total treatment period lasted longer than 91 days (13 weeks).

At the end of the intervention period, PASI raters stated that they knew the treatment assignment in 58.2% of cases (587/1008).

In 28 cases, the primary outcome measure was set to treatment failure. In 23 cases the case record forms were lost, and in 5 cases only baseline data were available (Figure).

Incidence of therapeutic success was significantly higher (P<.001) in TW UV-B than in UV-B. According to the protocol, TW UV-B was chosen as the reference.

In the subsequent confirmatory analysis, incidence of therapeutic success was significantly higher (P<.001) in both the SW UV-B and bath PUVA groups than in TW UV-B. Bath PUVA was not statistically superior to SW UV-B (P = .34) (Table 3).

Secondary analysis based on S-PASI and sensitivity analysis using a modified PASI criterion (reduction of PASI or involved body surface area ≥75%) confirmed the results of the primary analysis (Table 3).

The median reduction of PASI was 44% for UV-B (25th-75th percentile, 18%-72%; n = 264), 62% for TW UV-B (25th-75th percentile, 33%-82%; n = 272), 76% for SW UV-B (25th-75th percentile, 51%-91%; n = 291), and 84% for bath PUVA (25th-75th percentile, 57%-93%, n = 297).

Less than half of the participants considered in the primary analysis (47.7% [553/1159]) could be included in the per-protocol analysis. In contrast to the primary analysis, incidence of therapeutic success did not differ significantly between SW UV-B and TW UV-B (84.8% vs 81.2%) (Table 3).

Change scores on the SF-QES did not differ significantly between groups (P = .47). Overall median improvement was −0.14 (25th-75th percentile, −1.27 to 0.00; n = 1107).

Descriptive statistics of all global ratings are shown in Table 4. Change in disease severity and treatment effect were rated similarly by clinicians and patients. Tolerability of treatment was rated somewhat higher by clinicians than patients. Patients reported a better effect of study interventions compared with treatments they had received in the past except for UV-B only. Tolerability of all study interventions was judged better than other treatments received in the past.

Overall mean cumulative UV-B dose was 5.4 J/cm² (SD, 5.9 J/cm²) for patients treated with broadband UV-B or SUP and 31.4 J/cm² (SD, 17.7 J/cm²) for patients treated with narrowband UV-B. There were no significant differences in total UV-B exposure between the UV-B groups for patients treated with broadband UV-B or SUP (P = .46) or for patients treated with narrowband UV-B (P = .79). Cumulative UV-A dose was 63.6 J/cm² (SD, 36.9 J/cm²). Table 5 shows the descriptive statistics for total UV exposure by spectrum and intervention.

According to information recorded by the investigators in the treatment protocol, phototoxic reactions were observed in 12.3% (31 of 252), 7.1% (19 of 266), 11.6% (33 of 284), and 5.4% (16 of 297) of patients treated with UV-B, TW UV-B, SW UV-B, and bath PUVA, respectively. Seven patients developed edema and 4 developed blisters. Nine of the patients with grade IV or V erythema were treated according to the irradiation schedule (Table 1). All grade IV or V erythemas were reversible and occurred throughout in all intervention groups (5 patients in the bath PUVA group, 3 in the SW UV-B group, 2 in the TW UV-B group, and 1 in the UV-B group).

Eighty-five adverse events were reported in 72 study participants. In 46 cases the investigators assumed a possible, probable, or definite relationship with the study intervention. The most common adverse events that were judged treatment related were phototoxic reactions (28 cases); itching, tensing, or burning of skin (4 cases); flare of skin lesions (3 cases); and herpes zoster (3 cases). In all but 3 cases adverse reactions had resolved before the end of treatment.

Investigators reported 13 serious adverse events in 13 cases. Seven of them met the criteria for serious adverse events (inpatient hospitalization, n = 6; death, n = 1) (Table 6). In 2 participants with a correctly classified serious adverse event, the investigators assumed a relationship between study intervention and the event (alcohol problem, malignant melanoma). Contrary to the investigator's statement, the principal investigators (R.S. and T.B.) considered the occurrence of an alcohol problem during the intervention period an event unrelated to the study intervention. In the second case, a malignant melanoma was diagnosed after the eighth irradiation. Queries at the trial site disclosed that an “atypical nevus” had already been detected by a dermatologist 17 months before study enrollment. Furthermore, it is very unlikely that the tumor had progressed during the short period of 2 weeks. One death was reported during the intervention period. This 38-year-old man died suddenly and unexpectedly of renal failure. No preexisting disease or drug use was known, especially no cardiac or pulmonary diseases, which could have decompensated during balneophototherapy. A causal relationship with the study intervention was therefore excluded by the principal investigator.

Our study is, to our knowledge, the first rigorous clinical trial demonstrating that both bath PUVA and SW UV-B are superior to TW UV-B, and that TW UV-B is superior to UV-B. Not only are the effects statistically significant, but the numbers needed to treat and definition of primary outcome indicate quantitative and clinical relevance. Confidence intervals show good precision of estimates. The hierarchical design of our study also implies that bath PUVA, SW UV-B, and even TW UV-B are superior to UV-B, the treatment option that is currently reimbursed by the statutory health insurance in Germany.

We know of only 3 other randomized controlled studies addressing SW balneophototherapy for psoriasis. Gambichler et al¹⁸ were not able to detect a significant difference in clearance of the psoriatic lesions between the sites soaked in SW and TW (P>.50). On the other hand, Dawe and coworkers²⁷ did observe a faster initial clearance and a slightly greater fall in psoriasis severity for patients treated with Dead Sea soaks and narrowband UV-B than for patients treated with narrowband UV-B alone. Nevertheless, they considered this benefit not sufficient to justify the additional effort of balneophototherapy. In contrast to our study, both groups compared only the treatment of limbs in a half-sided manner in limited numbers of patients (ie, for each of those patients, a left side–right side comparison was made). Thus, effects of whole-body balneophototherapy might have been underestimated. A third study compared saline spa water or combined water and UV-B vs conventional UV-B for psoriasis in 71 patients.

In contrast to our results, Léauté-Labrèze et al²⁸ found no beneficial effect of bathing in enhancing phototherapy. Nevertheless, in their study only 21 to 24 patients per group were randomized; thus, the number of patients included might have been too small to detect significant differences.

Eighty-two patients withdrew consent before treatment started and were excluded from analysis. Their withdrawal was definitively not a consequence of treatment response. Most of the withdrawals were assigned to UV-B, indicating that participant preferences might have played a role. Sensitivity analyses were robust against various imputation strategies for missing values with 1 exception. If missing values in the TW UV-B group were set on failure (n = 24) and missing values in the UV-B group were set on success (n = 31), no significant effect could be found in favor of TW UV-B compared with UV-B (P = .11). However, according to a recent systematic review,²⁹ the impact of participant preferences on outcome of randomized controlled trials is generally low. Thus, we are firmly convinced that our strategy to exclude participants without any study intervention is the most valid approach to analyze our trial data.³⁰

Secondary analyses with S-PASI or a modified PASI criterion confirm the results of the primary analysis, whereas in the per-protocol analysis no significant difference between SW UV-B and TW UV-B could be established (Table 3).

Not unlike many other multisite trials, this study had to be subject to some pragmatic conditions. At several trial sites just 1 investigator was available, who had to perform both adaptation of treatment regimens and PASI scoring. Therefore, it is not surprising that at the end of the intervention period more than half of PASI observers stated they had known the treatment allocation. Furthermore, because of the nature of the intervention, neither UV-B nor SW UV-B could be masked.

Two findings in particular might indicate that observer bias has probably played a minor role, if any. These findings are, first, the high level of homogeneity of ratings between patients and PASI raters and, second, the superiority of TW UV-B compared with UV-B. The latter finding was surprising and not in keeping with the economic interests of the participating investigators.

In the primary analysis, SW UV-B was statistically superior to TW UV-B, while in the per-protocol analysis no statistically significant effect in favor of SW UV-B was found. We hypothesized that this finding might possibly be associated with differences in protocol adherence between groups. Although, in a post hoc multiple logistic regression model, protocol adherence proved to be a strong independent predictor of outcome (odds ratio, 14.5; 95% confidence interval, 8.8-23.0), no significant effect of the interaction between protocol adherence and treatment allocation on outcome could be demonstrated (P = .32). Any other predictor of outcome or a combination of predictors is also a likely candidate to have affected structural equality between groups in the per-protocol set and thus may have introduced selection bias.

In a substantial proportion of patients the irradiation dose was not applied strictly according to the irradiation schedule, probably increasing external validity, but at the expense of underestimation of the absolute effects of the treatments.

Since the participating trial sites were equipped with different types of irradiation devices, it was not possible to restrict UV-B irradiation to only 1 type of irradiation source. Therefore, broadband UV-B, SUP, and narrowband UV-B were allowed. To minimize bias, any participating trial site was restricted to the use of a single UV-B spectrum throughout the trial. Therefore, potentially effect-distorting comparisons, such as TW–broadband UV-B vs SW–narrowband UV-B or narrowband–UV-B vs TW–broadband UV-B, were avoided.³¹

Narrowband UV-B was found to be superior to broadband UV-B in several studies.^31,32 In our study, however, narrowband UV-B was not a significant predictor of treatment response in a post hoc multiple logistic regression model using backward elimination (probability parameters: Pin, 0.05; Pout, 0.10).

The irradiation schedule was based on patients' individual erythemal response (Table 1). Therefore, patients of different trial sites should have received equally erythemogenic doses despite different light sources.

Phototoxic reactions are known to appear in 10% to 30% of phototherapy treatments.³³ In our study, phototoxic reactions occurred in 99 (9.0%) of the 1099 documented cases, and 1.0% of all patients (11/1099) developed grade IV or V erythema. Participants rated the tolerability of all interventions as better than that of previously received treatments, indicating that all tested interventions in our study are comparably safe and well-tolerated treatment modalities. Nevertheless, long-term UV exposure due to repeated courses of UV treatment might be associated with an increased risk of photocarcinogenesis.^34,35

At baseline, our study participants had a median PASI of 17 (25th-75th percentile, 12-23). Patients with this disease severity are also likely candidates for systemic antipsoriatics.³⁶ Furthermore, the therapeutic benefit of SW UV-B (mean PASI reduction from 18.9 to 6.27) or bath PUVA (mean PASI reduction from 19.8 to 5.36) after 8 weeks in our study is comparable to the results obtained by Heydendael et al³⁶ for methotrexate (PASI reduction from 13.4 to 5.0) and cyclosporine (14.0 to 3.8). Compared with SW UV-B or bath PUVA, systemic antipsoriatics bear a higher risk of severe adverse effects, such as hepatotoxicity, nephrotoxicity, hypertension, immunosuppression, gastrointestinal adverse effects, and teratogenicity. This might result in a better benefit-risk ratio for SW UV-B or bath PUVA than for systemic antipsoriatics, but data on direct comparisons are still lacking.

In our study, no significant difference was found between bath PUVA and SW UV-B. Higher concentrations of methoxsalen may result in a better therapeutic benefit, but these concentrations bear the risk of systemic resorption of methoxsalen and therefore of systemic adverse effects.³⁷

The mode of action of saltwater balneophototherapy is still subject to research. It has been shown that the photosensitivity to UV-B is markedly increased (lower MED) after exposure to tap-water or saltwater baths, but no concentration-dependent effect was found.³⁸ Other authors found a decrease in MED after baths in tap water or low-concentration sodium chloride solution (<4%), but no decrease in MED after a bath in 26% sodium chloride solution.¹⁵ In our study, both MED and cumulative UV doses did not differ between UV-B, TW UV-B, and SW UV-B. This implies that different irradiation doses due to different levels of photosensitivity are unlikely to be responsible for the better therapeutic effects of TW UV-B or SW UV-B. As a possible mode of action of salt water baths prior to UV-B irradiation, it has been proposed that leukocyte elastase might be eluted by the bathing procedure.^17,39

Further research should be directed to large-scale observational trials to assess the safety of repeated courses of UV treatment, as well as the benefit, risk, and costs of systemic antipsoriatics compared with SW UV-B and bath PUVA in the long term. In addition, dose-finding studies for SW UV-B and bath PUVA seem to be reasonable.