Efficacy of combining oral Chinese herbal medicine and NB-UVB in treating psoriasis vulgaris: a systematic review and meta-analysis

The 18 included trials, involving 1416 participants with psoriasis vulgaris, were conducted in hospitals in China and published in Chinese from 2004 to 2014. Characteristics of the included studies are summarized in Table 2. All included studies evaluated oral CHM added to NB-UVB vs NB-UVB alone, and six [8, 13, 14, 18, 19, 22, 24] were three-arm trials that included a comparison with oral CHM alone (not analyzed in this study). A placebo was not used in any included study. Two studies used disease severity as one of the inclusion criteria. Fu et al. [11] recruited psoriasis vulgaris participants with an affected body surface area (BSA) >10 % and Sheng et al. [19] recruited psoriasis vulgaris participants with an affected BSA >10 % and PASI score ≥10. Ten studies reported mean PASI score at baseline. The mean PASI scores were <10 in two studies [10, 17], >20 in one study [16], and ranged from 14.32 to 19.19 in the remaining seven studies [8, 14, 15, 18, 19, 22, 26]. Eight studies recruited psoriasis vulgaris participants targeting specific Chinese medicine patterns [8, 10, 15, 16, 21, 22, 25, 26].

All studies reported that NB-UVB phototherapy was consistently prescribed to intervention groups and control groups. However, there was variation across studies in terms of the initial dosage, dosage increment regimen, treatment frequency and total number of treatment sessions. Most of the studies applied phototherapy two to three times per week for 10–29 treatment sessions in total.

Huang et al. [12] used a single herb extract product; the other 17 studies used multi-ingredient formulas in decoctions [10, 11, 13, 16, 18, 19, 21, 25], pills [12, 17, 24], capsules [20], or granules [14, 22, 26]. CHM was administered daily. No two studies used an identical herbal formula. Although all formulas varied across studies (Table 2), there were similarities in the herb ingredients used. The most frequently used herbs were Rehmannia glutinosa (sheng di huang) in 13 studies [8, 11, 13‐15, 18‐24, 26], Angelica sinensis (dang gui) in 11 studies [10, 11, 13‐15, 19, 20, 23, 24, 26], Smilax glabra (tu fu ling) in nine studies [8, 10, 13, 15, 17, 19, 22, 23, 26], Salvia miltiorrhiza (dan shen) in eight studies [10, 11, 15, 16, 22‐24], and Glycyrrhiza uralensis (gan cao) in eight studies [10, 11, 13, 18, 19, 21, 24, 26]. Furthermore, three herbs, R. glutinosa, A. sinensis, and S. glabra, were used together in six studies [8, 13, 15, 19, 23, 26]. The details of formula and herbs used by included studies can be found in Additional file 1: Table S1.

Sun et al. [20] only reported the total number of phototherapy treatment sessions. The treatment durations of the other 17 studies ranged from 4 to 12 weeks, and ten studies were of 8 weeks (2 months).

Seven studies included follow up with participants after treatment. Four studies followed all included participants for 3 months [15], 6 months [22, 23], and 1 year [25]. Three studies followed participants who achieved PASI-90 for 6 months [10, 18] and 1 year [15].

Seventeen of the 18 studies reported the numbers of participants who achieved PASI-95 or PASI-90, and PASI-60 (at least 95, 90, 60 % reduction in PASI), with reference to the Consensus of Diagnosis and Treatment of Psoriasis Vulgaris with Integrative Medicine [27]. One study reported PASI-95 and PASI-70 [16]. As PASI-50 was taken as the minimum clinically meaningful improvement [6, 28], PASI-60 and PASI-70 were conservatively pooled as PASI-60 to analyze the effect of the combination of oral CHM and NB-UVB. Four studies reported the relapse rates at the end of the follow-up [10, 11, 18, 23], but the definitions of relapse were not stated.

Liu and Sheng [13] reported that one participant in the treatment group and two in the control group were excluded after randomization because of comorbidities, and there were five dropouts in the treatment group and four dropouts in the control group for low compliance. Yan et al. [24] reported one withdrawal in the NB-UVB group owing to AEs, but a phototoxicity reaction was not observed. No dropout was reported in the other 16 studies.

One study was not included in the meta-analysis, because the outcome measures were PASI-25 and PASI-10 [25]. This study reported that the effect rate was much higher in the combination group than in the NB-UVB group in terms of PASI-25 (78.4 vs 8.1 %). The results of the remaining 17 studies were pooled to estimate the efficacy of the combination treatment.

All studies reported AEs (Table 2). Zhong et al. [26] only reported AEs in the combination group, and three studies [18, 21, 24] reported skin reactions but did not mention which group these occurred in. One participant withdrew because of a phototherapy AE [24]; the other withdrawals were not associated with AEs. All AEs reported in the NB-UVB groups were skin reactions, with a mean rate of 31.58 % (8–80 %). AEs reported in the combination groups were skin reactions, gastrointestinal reactions and elevated liver enzymes. The total AE rate (24.42 %) in the combination groups was a little lower than that in NB-UVB groups. In addition, combination treatment lowered the incidence rate of skin reactions including pruritus, skin dryness, erythema, blistering and hyperpigmentation [95/464 (20.47 %) vs 123/428 (28.74 %); RR = 0.66, 95 % CI 0.46–0.96]. The subgroup analysis showed a significant difference in studies administering 8–12 weeks of treatment (eight studies, RR = 0.60, 95 % CI 0.38–0.94), but not in studies with a duration of 4–6 weeks (four studies, RR = 0.77, 95 % CI 0.42–1.43). Mild gastrointestinal reactions, such as nausea, vomiting, abdominal distension and stool change, were reported in the combination groups (21/732, 2.87 %) [12‐14, 16, 17, 26]. Five cases of elevated liver enzymes were reported (5/732, 0.68 %). One case returned to normal within 1 month without additional treatment [12], three returned to normal by reducing the dosage of the CHMs [17], and the outcome was not reported for one [14]. The relationships between the AEs and the interventions were not further discussed in any study. No severe AE was reported.

Risk of bias assessments are presented in Fig. 4. Attempts were made to contact the original authors to obtain further information on 13 studies but the authors of 12 studies could not be reached and the author of one study refused to provide the further information we requested; therefore, risk of bias assessments were based on the published texts. For “sequence generation”, five studies that used random number tables (a table of random numbers is available in statistics textbooks or generated by computer) [14, 16, 17] and drawing lots (including coin-tossing, dice-throwing, or card-shuffling) [13, 19] were assessed as “low risk”, and the others were assessed as “unclear”. All studies were assessed as “unclear” for “allocation concealment” owing to the lack of information. All studies were assessed as “high risk” for “blinding of participants and research personnel”, because all studies used an add-on therapy and a placebo was not used for the CHM in any study. The blinding of outcome assessors was judged “unclear” because information was not provided. One study [24] had one withdrawal in the NB-UVB group, and it was assessed as “low risk” for “incomplete outcome data”, because the proportion of missing data was not enough to have an impact on the effect estimate. The amount of missing outcome data in one study [13] was 12.5 % and ITT analysis was not performed, so this study was assessed as “high risk” for “incomplete outcome data”. The others were assessed as “unclear”. All studies were “low risk” for “selective reporting” because all studies reported the outcomes pre-specified in the methods. All studies were assessed as “low risk” for “other bias”, which refers to the inappropriate influence of funders.

Asymmetrical funnel plots and significant Egger’s test (t = 3.88, P < 0.01) indicated potential publication bias in favor of more positive results for the smaller studies (Fig. 5).

Owing to defects in study design and the risk of publication bias, the quality of evidence was downgraded to low according to the GRADE system (Table 4).